EVALUATION OF THE EFFECT OF SIMPLIFIED AND PATIENT ... · Understanding blood flow patterns in...

ASME Early Career Technical Journal 2011 ASME Early Career Technical Conference, ASME ECTC

November 4 – 5, Atlanta, Georgia USA

EVALUATION OF THE EFFECT OF SIMPLIFIED AND PATIENT-SPECIFIC ARTERIAL GEOMETRY ON HEMODYNAMIC FLOW IN STENOSED CAROTID BIFURCATION

ARTERIES

Rodward L. Hewlin Jr., and John P. Kizito Department of Mechanical Engineering

North Carolina A&T State University Greensboro, NC, USA

[email protected] Tel: (336) 334-7620 ext. 315

ABSTRACT

Numerous CFD studies have been performed on the motivation

to elucidate the role of hemodynamic forces in the development

of atherosclerosis in the coronary arteries and the carotid

bifurcation artery. In order to improve CFD predictions, and to

consider CFD as a clinical diagnostic or treatment planning

tool, there is a need to ensure its accuracy and reliability

through a systematic approach of comparing simplified

geometries to patient-specific geometries. There is also a need

to compare numerical results with clinical data for validation

and verification. In this paper, comparisons of 3-D pulsatile

blood flow through a simplified stenosed carotid bifurcation

arterial geometry and two stenosed patient-specific carotid

bifurcation arterial geometries were simulated using CFD. The

goal was to quantify the efficacy of arterial vessel geometry on

hemodynamic flow parameters (velocity, wall shear stress, and

vorticity) with the same grade of stenosis for all models. The

main conclusion in this work is that dynamic behaviors of

blood through the patient-specific carotid bifurcation arterial

geometries based on numerical approaches show differences

when compared to a simplified carotid bifurcation geometry.

Keywords—atherosclerosis, geometry, hemodynamics,

modeling, patient-specific, simplified, velocity, vorticity, and

wall shear stress

INTRODUCTION

Cardiac, cerebral and peripheral localizations of

atherosclerosis are the leading causes of morbidity and

mortality in the Western part of the world [1]. Factors such as

poor dieting or lack of exercise contribute to the pathogenesis

and progression of atherosclerosis disease. Although these

factors influence disease progression, atherosclerosis usually

involves vascular districts [2]. Clinical tools such as X-ray

contrast angiography, ultrasonography, magnetic resonance

imaging (MRI), and multislice (CT) have been used as a means

of measuring the degree of progression and/or regression of

atherosclerosis. Experimental measurements of fluid velocities

in a scale model of the human carotid bifurcation artery have

been conducted using Laser-Doppler anemometry [3, 4] and

MRI [5-7]. Although MRI does not usually provide data over

an extended area of the artery, 3-D data obtained from MRI

measurements have been used as a means to guide CFD studies.

Recently, wall shear stress (WSS) in the carotid bifurcation

arteries has generated considerable interest as a complementary

explanation for plaque formation [8, 9]. In particular, the

disturbed flow environment, often characterized by low shear

stress, is said to promote atherogenesis [10-12]. The carotid

bifurcation artery where the common carotid artery (CCA)

branches into the external carotid artery (ECA) and internal

carotid artery (ICA) is a common site for atherosclerosis. The

ICA is the most broadly studied vessel in context owing to its

clinical relevance in supporting blood-flow to the brain and also

due to its relative ease of access by high-resolution medical

imaging. Severe plaque buildup in this part of the carotid

bifurcation results in ischemic transient attack (stroke). It is

also recognized that the complex geometry of the carotid

bifurcation artery can affect the blood flow pattern [13]. The

bifurcation angle is considered an important geometric factor

which can affect the flow pattern in the sinus of the ICA, the

atherogenic prone site of the carotid bifurcation artery [14]. It

has also been noted the size of the sinus bulb is related to

gender [15]. Reports have also shown that the average female

bifurcation angle is 51 degrees and the average male bifurcation

angle is 67 degrees [16].

Flow in the carotid bifurcation artery has been investigated

computationally by Nazemi et al. [16], who studied 2D, steady,

pulsatile flow in a carotid arterial segment Perktold et al. [2]

developed a 3-D carotid artery model using a pressure

correction finite element method. Velocity and WSS values in

the carotid sinus were calculated and compared with

experimental data with good agreement. Rindt and Von

ASME 2011 Early Career Technical Journal - Vol. 10 39

Steehoven [17] stimulated blood flow in the carotid bifurcation

and found that decreasing the angle between the CCA and the

carotid sinus could reduce the risk of experiencing axial flow.

Understanding blood flow patterns in different carotid

bifurcations can lead to the identification of people vulnerable

to atherosclerosis and stroke. Moreover, the unique geometry

of each person’s carotid bifurcation artery can affect the blood

flow pattern throughout the artery which affects the risk of

experiencing plaque buildup and high level atherosclerosis.

This paper focuses on a systematic approach to modeling

carotid bifurcation arteries for CFD simulation and examines

the effect of 3-D carotid bifurcation arterial geometry on blood

flow pattern and on WSS. Patient-specific CA models

generated from computed tomography angiography (CTA) are

discussed. The simulated flow profiles in the patient-specific

geometries are compared against a simplified carotid

bifurcation geometry to evaluate the efficacy of modeling

patient-specific geometries against the alternate simplified

geometries.

MATERIALS AND METHODS

The methodology of this paper begins with the generation

of 3-D geometric models of the diseased carotid bifurcation

arteries to be evaluated. In this paper, CFD analyses were

performed on a simplified carotid bifurcation arterial geometry

(SA-1) and patient-specific carotid bifurcation arterial

geometries (CA-1) and (CA-2). The geometries are discretized

and used as computational domains. For blood flow

simulations, the conventional assumptions: flow in the carotid

bifurcation artery is Newtonian, laminar, incompressible, and

isotropic are implemented. By these assumptions, blood can be

modeled by the incompressible Navier-Stokes equations with

blood density specified as 1060 kg/m3 and the corresponding

dynamic viscosity . The commercial software

GambitTM

was used for geometric meshing and FLUENTTM

was used to solve the Navier-Stokes equations in the finite

volume formulation. Grid meshing levels were performed at

three levels: coarse (104,406), coarse (143,948) and medium

(233,045). A sinusoidal velocity waveform boundary condition

was specified at the inlet of the CCA of each arterial geometry

and the ICA and ECA outlets were modeled as pressure outlets

at 12kPa (90mmHg).

In comparison of the velocity at the ICA for geometric

models, a 0.21% difference in velocity was observed for a grid

independence study. For numerical solutions, the velocity

coupling solving technique was implemented. The WSS

vectors are predicted on fluid domain surfaces that represent the

interface boundary between the fluid and the neighboring

tissue. Velocity is monitored at the CCA, ECA, and ICA and

vorticity is calculated at the sinus bulb. CTA images of

stenosed carotid bifurcations were taken from a 50 and 66 year

old male and were processed in Solidworks. Figure 1 presents

a schematic diagram of the Solidworks processing

methodology. The parametric data of the patient-specific

carotid artery geometries (CA-1) and (CA-2) is provided in

Table 1 and obtained from [3].

Figure 1: (a) CTA scan point cloud image, (b) 3D spline

surface generation, (c) lofted and boundary base solid part model of the carotid bifurcation, (d) meshed post-processed carotid artery geometry.

33.2 10 Pa s


Figure 2: CAD schematic of the arterial solid model: (a) CA-1 and (b) CA-2.

The images contain a point cloud, which defines the surface

geometry of the carotid bifurcation by discrete domains. The

images were processed in Solidworks using a curve creation

feature. The curve creation feature united the point cloud with

a 3-D surface spline. This technique was repeated at each

domain until the entire arterial geometry was defined as shown

in Figure 1b. Each 3-D spline domain was then united using a

lofting and boundary boss feature until the final solid product

was obtained as shown in Figure 1c and Figure 2.

RESULTS

Simulations and analyses were performed for three

geometries, the simplified carotid bifurcation arterial geometry

(SA-1) and two patient-specific carotid bifurcation arterial

geometries (CA-1) and (CA-2). The location of the sinus bulb

was monitored for calculating WSS and vorticity as a function

of time. Velocity has been calculated throughout the CCA,

ECA, and ICA of the simplified arterial geometry and patient-

specific arterial geometries at any point in time. Figure 3

shows the velocity waveform monitored at the ICA outflow of

the three geometries at time 0s to 3s. At the beginning of the

cycle, the blood flow rate starts increasing and reaches a

maximum level around 0.45s and then decreases to the lowest

level at 1.3s. Figures 4 and 5 present a plot of the vorticity vs.

time for the SA-1, CA-1, and CA-2 geometries.

Figure 3: ICA velocity waveform for SA-1, CA-1, and CA-

2.

Figure 4: Plot of sinus vorticity vs. time for the CA-1

patient-specific geometry.

Figure 5: Plot of sinus vorticity vs. time for the CA-2

patient-specific geometry and the simplified arterial geometry SA-1.

0 0.5 1 1.5 2 2.5 30

0.1

0.2

0.3

0.4

0.5

Flow Time (sec)

Velo

cit

y (

m/s

)

CA-1 ICA velcity

CA-2 ICA Velocity

SA-1 ICA Velocity

0 0.5 1 1.5 2 2.5 3-70

-60

-50

-40

-30

-20

-10

0

Flow Time (sec)

Vo

rti

cit

y (

1/s

)

0 0.5 1 1.5 2 2.5 3-0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

Flow Time (sec)

Vo

rti

cit

y (

1/s

)

CA-2 Sinus Vorticity

SA-1 Sinus Vorticity

(a) (b)


(a)

(b)

(c)

Figure 6: Axial velocity contour at time T=0.5s (systole), for (a) SA-1, (b) CA-1, and (c) CA-2.

(a)

(b)

(c)

Figure 7: Axial velocity contour at time T=1.0s (diastole), for (a) SA-1, (b) CA-1, and (c) CA-2.


Figure 6 shows the axial velocity contours of the sinus at

time 0.5s (systole), and Figure 7 shows the axial velocity

contours of the sinus at time 1.0s (diastole). The negative

velocity indicates flow reversal zones. As shown in Figures 4

and 5, the maximum flow reversal is observed in the patient-

specific geometry models. For the patient-specific geometries,

it was also observed that the axial velocity is the highest at the

region upstream of the ICA and downstream of the sinus near

the CCA. Regions of flow reversal formed near the sinus wall

are due to the rapid changes in the flow rate, flow viscosity and

the curvature of the sinus. Within these zones, blood flows in

the direction opposite to the mean flow increasing the

probability of plaque deposition.

For the patient-specific geometries, the velocity gradients

near the stenosis are much higher than those near the sinus.

The velocity at the sinus becomes negative and indicates flow

reversal and very low WSS. WSS is an important factor in the

progression and initiation of atherosclerosis. For the simplified

geometry, there is a small region of velocity increase at the

stenosis, and very little flow downstream of the stenosis to the

ICA outlet. The major increase of velocity in the simplified

geometry is experienced in the ECA branch of the artery.

Figure 8 shows the WSS contour for systole (peak velocity) for

the simplified arterial model and patient-specific geometries.

As noted by Malek et al. [12] regions with WSS less than

0.4 Pa are susceptible to atherosclerosis. In the wall shear

stress contours provided in Figure 8, it is shown that the

maximum WSS experienced is 16.3 Pa. The maximum WSS is

shown at the ICA stenosed region of SA-1 and at the stenosis

and bifurcating region of CA-1. A general conclusion from

these WSS patterns is that the sinus bulb is the most

atheroprone site where higher values of WSS appear at the

bifurcation apex and final ends of the ECA and ICA, which

have a smaller diameter. There is a stagnation point at the apex

because shear directions at its sides are different (granting that

the magnitude of the shear would be relatively high).

The stagnation point is the place of plaque buildup.

Another point is that, in general, realistic geometries have

higher values of WSS and the atheroprone site in the sinus bulb

is smaller. When considering WSS, the magnitude of the wall

shear is a function of the geometry of the arterial vessel. This

works proves that when studying hemodynamics of the arterial

vasculature, specifically the carotid bifurcation artery for

diagnostic purposes, it is important to have patient-specific

parametric data for the carotid bifurcation vessel to be

analyzed.

(a)

(b)

(c)

Figure 8: WSS contour at time T=0.5s (systole), for (a)

SA-1, (b) CA-1, and (c) CA-2


CONCLUSION

The focus of this paper was on examining the effect of 3-D

carotid bifurcation arterial geometry on blood flow pattern and

on WSS. Patient-specific CA models were generated from

computed tomography angiography (CTA) scans and modeled

for CFD. The simulated flow profiles in the patient-specific

geometries were compared against a simplified carotid

bifurcation geometry to evaluate the efficacy of modeling

patient-specific geometries against an alternate simplified

geometry. The main conclusion in this work is that dynamic

behaviors of blood through the patient-specific carotid

bifurcation arterial geometries based on numerical approaches

showed differences when compared to a simplified carotid

artery bifurcation geometry.

For the patient-specific geometries, the velocity gradients

near the stenosis were much higher than those near the sinus.

Flow reversal and low WSS was observed in the sinus region of

the patient-specific models. Also, for the simplified geometry,

there was a small region of velocity increase at the stenosis, and

very little flow downstream of the stenosis to the ICA outlet.

The major increase of velocity in the simplified geometry was

experienced in the ECA branch of the artery, while in the

patient-specific geometries, the reverse is true.

Some of the limitations of this work consisted of not

implementing individual patient-specific flow rates/waveform

of the arterial models. This is because these flow rates/waves

were not available. A sine wave velocity profile was used at the

boundary of the CCA. Future studies will incorporate more

realistic boundary conditions obtained from experimental data

from flow phantoms or as an alternate blood flow rate obtained

from volunteers. Also future studies will incorporate the

moving boundary of the arterial wall, instead of using rigid wall

boundary conditions. A future study consisting of a broad

statistical study on the effect of patient-specific arterial

geometry on hemodynamic flow will be conducted.

ACKNOWLEDGMENT This work was supported in part by the Department of

Education Title III grant program to N.C. A&T State

University.

REFERENCES

[1] Cecchi, E., et al. "Role of Hemodynamic Shear Stress in

Cardiovascular Disease." Atherosclerosis 214.2 (2011): 249-56.

Print.

[2] Perktold, K., Resch, M., Peter, R.O., "3-Dimensional

Numerical Analysis of Pulsatile Flow and Wall Shear Stress in

the Carotid Artery Bifurcation," Journal of Biomechanics 24.6

(1991): 409-20. Print.

[3] http://www.charite.de/biofluidmechanik/en/research/

[4] Gijsen, F. J. H., F. N. van de Vosse, and J. D. Janssen. "The

Influence of the Non-Newtonian Properties of Blood on the

Flow in Large Arteries: Steady Flow in a Carotid Bifurcation

Model." Journal of Biomechanics 32.6 (1999): 601-08. Print.

[5] Fayad, Z. A., and V. Fuster. "The Human High-Risk Plaque

and Its Detection by Magnetic Resonance Imaging," American

Journal of Cardiology 88.2A (2001): 42E-45E. Print.

[6] Marshall, I., P. Papathanasopoulou, and K. Wartolowska.

"Carotid Flow Rates and Flow Division at the Bifurcation in

Healthy Volunteers," Physiological Measurement 25.3 (2004):

691-97. Print.

[7] Stary, H. C., et al. "A Definition of the Intima of Human

Arteries and of Its Atherosclerosis-Prone Regions - a Report

from the Committee on Vascular-Lesions of the Council on

Arteriosclerosis, American-Heart-Association." Circulation

85.1 (1992): 391-405. Print.

[8] Cunningham, K. S., and A. I. Gotlieb. "The Role of Shear

Stress in the Pathogenesis of Atherosclerosis," Laboratory

Investigation 85.1 (2005): 9-23. Print.

[9] Libby, P. "Inflammation in Atherosclerosis." Nature

420.6917 (2002): 868-74. Print.

[10] Ku, D. N., et al. "Pulsatile Flow and Atherosclerosis in the

Human Carotid Bifurcation - Positive Correlation between

Plaque Location and Low and Oscillating Shear-Stress,"

Arteriosclerosis 5.3 (1985): 293-302. Print.

[11] Ku, D.N., Giddens, D.P., "Laser Doppler Anemometer

Measurements of Pulsatile Flow in a Model Carotid

Bifurcation," Journal of Biomechanics 20.4 (1987): 407-21.

Print.

[12] Malek, A.M., Alper, S.L., Izumo, S.,. "Hemodynamic

Shear Stress and Its Role in Atherosclerosis," Journal of the

Americal Medical Association 282 (1999): 2035-42. Print.

[13] Lee, S. W., et al. "Geometry of the Carotid Bifurcation

Predicts Its Exposure to Disturbed Flow." Stroke 39.8 (2008):

2341-47. Print.

[14] Nguyen, K. T., et al. "Carotid Geometry Effects on Blood

Flow and on Risk for Vascular Disease." Journal of

Biomechanics 41.1 (2008): 11-19. Print.

[15] Goubergrits, L., Affeld, K., Fernandez-Britto, J. and

Falcon, L. "Investigation of Geometry and Atheroscleorosis in

the Human Carotid Bifurcations." Journal of Mechanics in

Medicine and Biology 3.1 (1999): 31-48. Print.

[16] Nazemi, M., Kleinstreuer, C., Archie, J.P. "Pusatile 2-

Dimensional Flow and Plaque-Formation in a Carotid-Artery

Bifurcation." Journal of Biomechanics 23.10 (1990): 1031-37.

Print.

[17] Rindt, C.C.M., and. Von Steenhoven, A.A "Unsteady Flow

in a Rigid 3-D Model Of The Carotid Artery Bifurcation"

Journal of Biomechanical Engineering-Transactions of The

ASME 118.1 (1996): 90-96


http://www.charite.de/biofluidmechanik/en/research/



EVALUATION OF PULSE JET MIXING USING A SCALAR QUANTITY AND SHEAR STRESS

Ibraheem R. Muhammad and John P. Kizito North Carolina A&T State University

Department of Mechanical Engineering Greensboro, NC, U.S.A

ABSTRACT

This paper uses computational fluid dynamics (CFD) to

investigate the behavior of pulsing, submerged, liquid jets in a

cylindrical mixing tank. The jet pulse was simulated by turning

the jet inlet velocity on and off in a cyclic manner. It was

proposed that a pulsing, liquid jet would provide sufficient

force to dislodge and suspend solid materials. Therefore, it is

necessary to evaluate its performance in different

configurations to determine which gives the best performance

parameters. The shear stress at the bottom of the tank was

calculated and compared for certain flow configurations. In

addition, studies were performed to determine the mixing time,

or time elapse for the tank to become uniform, using a scalar

quantity such as temperature. More specifically, a hot jet

stream was injected into a cooler bulk fluid until the desired

mixing criterion was met. The temperature was calculated at

different locations within the tank and the mixing time was

estimated to be the time elapse for the temperature in the tank

to reach 65% of the equilibrium temperature. The results

showed that the mixing time decreased with increasing number

of jets. However, there was no enhancement in mixing by using

pulsing jets. The wall shear stress at the bottom of the tank

increased as the jet height from the bottom of the tank

decreased.

INTRODUCTION Liquid jet mixers have been used for years in many

industrial processes. Jet mixers operate by withdrawing fluid

from the mixing tank and supplying the fluid back to the tank

through a nozzle at high velocities. As the jet travels in the

tank, the flow field expands as it entrains the bulk fluid. Jet

mixers are advantageous compared to other mixing devices as

they can achieve effective mixing by creating high turbulence

and high shear rates. These mixing devices can be easily

installed and transferred to different vessels, and can be used in

combination with other mixing equipment [1-5]. Jet mixers

have been used in off bottom suspension of solids in such

applications as nuclear waste, food, and pharmaceutical

processing.

To achieve suspension in these systems it is mainly

necessary to provide a force from the jet to overcome the

weight of the particle. Some solid materials contain smaller

particles, which have cohesion forces as the dominant forces

that must be overcome to cause suspension [6]. In these

systems, the material can be described as a Bingham plastic and

it is necessary to overcome a critical shear stress, or yield

stress, at which point the mixture then starts to move and

behave more like a Newtonian fluid [6, 7].

For practical purposes, it is usually not enough to just

dislodge the solid material from the bottom of the tank but to

also maintain the suspension for a certain time period. For

certain cases, particles may not stay suspended, which results in

inaccurate data when fluid samples are taken. In addition, if the

need may arise to move the mixture out of the tank to another

location, equipment failure or malfunction can occur since the

flow was not predicted accurately. Therefore, it is not only

desirable to understand the physics behind eroding or

dislodging stationary solid materials from a surface, but to

understand how they behave once they are suspended.

Previous researchers have studied erosion in unrestrained flows

such as in riverbeds, lakes, and estuaries [7, 8]. However,

erosion in tanks, or confined flows, is different because the

flow environment provides interactions between the solid

material and the solid walls.

There is not extensive literature on pulsing jet mixers for

solid suspension. A pulsing jet can be created using a periodic

discharge from the nozzle by controlling the power input to the

pumping equipment in a cyclic manner. By discharging the jet

in pulses, a unique flow structure is created within the mixing

vessel when compared to continuous jets. It is expected that

pulsing jet mixers will create different flow structures than

continuous jets and have better mixing in some situations due

to increased number of local vortices.

The present paper investigates the mixing behavior of pulse

jet mixer nozzles using the computational fluid dynamics

(CFD) package, FLUENT. The mixing time was calculated by

comparing the temperature throughout the tank as a function of

the inlet temperature, which was initially set at a higher value.

The shear stress on the bottom surface was calculated as well.

The rationale of the present paper is to draw attention to the

potential of pulsing jet mixers in the erosion and suspension of

solid particles in practical applications.


PROBLEM FORMULATION Physical Model

The mixing tank was modeled with a diameter of 0.305 m

and a height of 0.305 m as shown in Figure 1. The different jet

orientations (single, dual, quad, and azimuthal) that were

studied are shown in Figure 1, as well. For the dual and quad

jet mixer systems, the jets were centered and directed outward.

The jets in the multiple jet systems were spaced 2 cm apart

from each other. For the single jet system, the jet was 0.07625

m from the center and directed away from the tank outlet and

towards the center of the tank. The inlet came from a nozzle 8

mm in diameter. The nozzle was positioned at a 45° angle and

is attached to a 1.27 cm (1/2”) diameter piece of piping. The

outlet was a 1.905 cm (3/4”) diameter, overflow hole in the side

wall of the mixing tank and the center of the hole was located

0.29 m from the bottom of the tank. Figure 2 shows an

additional orientation that was studied in which the jets were

directed in an azimuthal fashion. The top view and 3d view of

the system is shown in Figures 2a and 2b, respectively.

Grids

Figure 3 shows the tetrahedral element types that were used

to mesh the volume. Mesh intervals of 10 – 20 mm was used.

It was noticed that there was not much variance in the solution

when the interval size was decreased. The mesh of 15 mm was

chosen because it was the largest interval size in which resulted

in a grid independent solution.

CFD Modeling The CFD simulations were performed using FLUENT, a

commercially available software. The jet flow was modeled by

the standard k- turbulence model with standard wall functions.

The standard k- ɛ has been used to simulate jet mixers because

it was found to be less demanding in terms of computational

time and accuracy requirements were not compromised [9].

The basic equations which describe the flow given an

incompressible, Newtonian fluid with constant properties are

the continuity, momentum, and energy equation, presented in

Equations (1), (2) and (3), respectively [10].

0divV (1)

VpgDt

DV 2 (2)

TTC

k

Dt

DT

p

22

(3)

The transient simulations required a time step of 0.05 s. The

boundary conditions, such as velocity inlet, outlet, and free

shear surface are illustrated in Figure 3. The inlet velocity was

set at 10 m/s when the jet was on. The top surface is open to the

atmosphere and is modeled as a free shear surface. The other

walls were set with the no-slip boundary conditions.

Figure 3. Example of meshed volume and boundary conditions used in simulations.

(a) (b)

Figure 2. Azimuthal oriented jet system including (a) a top view and (b) three-dimensional view.

(a)

(b) (c)

Figure 1. The different jet mixer arrangements including (a) single jet, (b) dual jet, and (c) quad jet systems.


The method used to calculate the mixing time for the

present system was based on a thermal criterion. Initially, the

bulk fluid is at 283° K and the liquid jet enters the tank at a

higher temperature, 350° K. A 65% homogeneity criterion was

used to find the mixing time for the simulations. The mixing

time was calculated as the time between the beginning of the

simulation until the time at which the temperature becomes

65% of the equilibrium temperature, or m = 0.35. This is

mathematically expressed by [11]

,35.0

oTeqT

TeqTm

(1)

where Teq is the equilibrium temperature, To is the initial

temperature of the bulk fluid in the tank, and T is temperature

at any location in the tank at any time. The value, m, is just a

dimensionless parameter which specifies the maximum allowed

deviation of mixing based on the criteria chosen for the study.

The mixing time criterion was chosen to save on

computational time. Temperature probes were placed in

various locations of the mixing tank and the temperature at

those locations were recorded as a function of time for the

duration of the numerical run. When all locations met the

mixing criteria, the tank was considered mixed and the time

elapse was reported as the mixing time. From scaling analysis,

the Peclet number, which compares convection time to

diffusion time in the flow, was determined to be 1.46 x 104.

Since the Peclet number was large, the flow was convection

dominated, thus finding the use of a scalar quantity, like

temperature, to be acceptable.

Initially, the jet mixers were simulated as continuous jets to

determine a baseline for comparison throughout the study. The

mixing time and the wall shear stress on the bottom surface

were calculated and reported. Simulations were then run using

pulsing jet mixers. Table 1 shows the various pulse settings

used in the present study. When the jet was set to on, the inlet

velocity was fixed to 10 m/s and when off the velocity was

fixed to 0 m/s.

The settings used in this system are not indicative of all the

possible settings that can be used, but rather to point in the

direction of which setting may give the best overall mixing. It

was important to keep the pulses such that the time off was

relatively short when measured in terms of particle settling

velocity thus ensuring uniform dispersion of particles.

Table 1: Various pulse settings used in simulations.

Pulse Setting Name Time on (s) Time off (s) Pulse 1 (P1) 2.5 0.5

Pulse 2 (P2) 3.5 0.5

Pulse 3 (P3) 5.0 0.5

Pulse 4 (P4) 5.0 0.25

RESULTS Continuous Jets

Figures 4 and 5 shows the results comparing the various

types of continuous jet configurations. Continuous jets were

modeled using four types of jet configurations (single, dual,

quad, and azimuthal) at two different heights from the bottom

of the tank 0.025 and 0.07625 m. The temperature within the

tank was reported in multiple positions at multiple heights to

ensure the behavior in critical zones of the tanks is accurately

represented. The results showed that to achieve 65%

homogeneity, the mixing time decreased as more jets were

added. In particular, Figure 4 shows the mixing time with the

jets positioned 0.07625 m from the bottom of the tank. Figure

5 shows the results with the jets positioned 0.025 m from the

bottom of the tank.

The results from both heights are summarized in Table 2.

The same trend was observed at both heights. By doubling the

number of jets, the mixing time for a set criteria was decreased

about half. The best jet arrangements at both heights were

shown to be the quad jets and the azimuthal jets which were

expected since they use the most number of jets and introduce

the most convection into the tank as compared to the other

configurations. Specifically, at a height of 0.025 m, the

azimuthal jet had the lowest mixing time, 15.8 s. At a height of

0.07625 m, the quad jet configuration had the lowest mixing

time, 15.4 s. The discrepancy in results between heights was

due to the way the jets were directed and interacted with the

walls.

The temperature contours for the quad jet and azimuthal jet

systems were compared. The temperature contours after 5

seconds of flow time for the quad and azimuthal jets are shown

in Figure 6. Figures 6a and 6b show the profiles in the yz-

plane for the quad and azimuthal jets, respectively. Figures 6c

and 6d show the profiles in the xy- plane for the quad and

azimuthal jets, respectively.

Figure 6 show that the quad jets have more mixing in the

middle of the tank compared to the azimuthal jets. The

phenomena can be attributed to the orientation of the quad jets.

As the quad jets interact with the tank walls, the flow turns over

and cause circulatory patterns. Figure 6b and 6d show the

azimuthal jets oriented such that a swirling flow is created and

the most prominent mixing takes place close to the walls. The

jet discharge from the nozzles for the azimuthal jets is not

shown in the plots because the jets are oriented such that the

nozzles are outside of the plane. The quad jets system has a

significant low mixing zone in the region near the side of the

tank above the edge of the jet circulation off the wall. Lack of

mixing can be clearly seen in the left side of Figure 6c. Low

mixing zones are also observed in Figure 6a, in the upper

vicinity.

The temperature profiles after 10 seconds of flow time for

the quad and azimuthal jets in the yz- and xy- planes are shown

in Figure 7. Specifically, the quad and azimuthal jets in the yz-

plane is shown in Figures 7a and 7b, respectively. The quad

and azimuthal jets in the xy- plane is shown in Figures 7c and

7d. The minimum temperature in the entire tank has increased


from 297º K to 313 º K. Comparison of Figures 6c and 7c,

illustrates that the low mixing zones after 10 seconds are still in

the same general location as those after 5 seconds. Figures 6a

and 7a show that the low mixing zone in the yz plane, though

still in the upper right hand side, shifts from the top after 5

seconds to the side after 10 seconds. Figures 6b and 7b show

that the temperature profiles for the azimuthal jets seem to

become even more uniform after 10 seconds compared to the 5

seconds case. But the low mixing areas are still located in the

center axis of the tank.

Figure 4: Mixing times for continuous jet systems with jets located 0.07625 m from bottom of tank and at a 45° angle.

Figure 5: Mixing times for continuous jet systems with jets located 0.025 m from bottom of the tank and at a 45° angle.

Table 2: Mixing time comparisons for continuous jets of different configurations.

Jet Type H = 0.025 m H = 0.07625 m Single 61.6 s 60.5 s

Dual 29.4 s 29.4 s

Quad 16.1 s 15.4 s

Azimuthal 15.8 s 16.1 s

Figure 7: Temperature profile in the (a) yz-plane of quad jets, (b) yz-plane of azimuthal jets, (c) xy-plane of quad jet, and (d) xy-plane of azimuthal jets at a height of 0.07625 m after 10 seconds flow time.

Figure 6: Temperature profile in yz- plane of quad jets (a), yz-plane of azimuthal jets (b), xy-plane of quad jets (c), and xy-plane of azimuthal jets at a height of 0.07625 m after 5 seconds flow time.


Pulse Jets The results for the mixing times of the different pulse

settings described in Table 1 are plotted in Figure 8. The

different pulse settings were modeled using the quad jet system

positioned at 0.07625 m from the bottom of the tank since the

lowest mixing times were reported for this configuration in

earlier studies. The mixing time results are summarized in

Table 3. Figure 8 shows that the P4 setting had the lowest

mixing time and the P1 setting had the highest mixing time.

The mixing time decreased as the period of jet discharge was

increased and the off period was decreased. The results in

Table 3 show that the P4 and P3 settings had mixing times that

were very close to that of the continuous jet in the same

configuration. The result may be due to there being more local

vortices created in the mixing tank compared to the steady jet

systems. The vortices are created as the jet is pulsed in a cyclic

manner. The mixing time results indicate that it may be

possible to increase mixing with some alternate pulsing settings

or techniques.

Figure 8. Mixing time for different pulse settings in quad jet arrangement with jets 0.07625 m from the bottom of the tank and at an angle of 45°.

Table 3: Mixing times for different pulse jet settings.

Pulse Setting Mixing Time P1 16.8 s

P2 16.3 s

P3 15.6 s

P4 15.5 s

Figure 9 shows the temperature profile for the entire first

pulse cycle (time on + time off) of the quad jet system using the

P1 setting. Figures 9a and 9b show the yz- and xy- planes,

respectively, after the initial on period of 2.5 seconds. Figures

9c and 9d show the yz- and xy- planes, respectively, after the

off period of 0.5 seconds. The jet momentum can be seen

impinging on the bottom of the tank and spreading up the wall.

The results show that the top of the tank is not well mixed

because the total flow time at this snapshot is still short. After

the complete first pulse cycle, the flow has been off for 0.5 s,

but the convection of the jet continues to spread slowly towards

the top of the tank. Figure 10 shows the temperature profile in

the yz and xy plane, respectively, at the end of 3 complete pulse

cycles. These figures show that though the low mixing zones

exist in similar vicinities, the overall profile differs from that of

the continuous jets.

Figure 10. Temperature contour of quad jet system after 3 complete pulse cycles using P1 setting in the (a) yz-plane and (b) xy-plane.

Figure 9. Temperature contour of quad jet system of first cycle using P1 setting in the (a) yz-plane after initial on time of 2.5 s, (b) xy-plane after initial on time, (c) yz-plane after initial off time of 0.5 s, and (d) xy-plane after initial off time.


Figure 11 illustrates the temperature profile of the total first

pulse using the P4 setting. The temperature profile after the

time on phase for the yz and xy planes, respectively, is shown

in Figures 11a and 11b. The temperature profile after the off

phase of the first pulse is shown in Figures 11c and 11d. The

P4 setting had mixing time results just about the same as the

continuous jets. Compared to the P1 setting after a single

pulse, more mixing takes place, which is because a single pulse

cycle using the P1 setting is a total of 3 seconds while that of

the P4 setting is a total of 5.25 seconds. The temperature

profile after 3 complete cycles using the P4 setting is shown in

Figure 12. At this point, the low mixing zones are at a

minimum and because the minimum temperature within the

tank is about 326, the homogeneity criteria set for the study are

almost reached. However, the low mixing zones still appear in

the same locations as with the P1 settings and with the

continuous jets of this configuration. The low mixing zones did

not differ much from case to case, but the temperature profile

varied.

Shear Stress on Bottom Wall Table 4 summarizes the results for the maximum and the

average shear stress on the entire bottom surface for the

continuous jets. As the height from the bottom of the tank was

decreased, the shear stress on the bottom surface increased.

The highest shear stress values were reported for the azimuthal

jets at both heights. At a height of 0.07625 m, the azimuthal

jets had the highest average shear for the surface. The

azimuthal jets were able to cover more surface area with higher

amounts of shear. Similarly, the quad jets had the highest

average shear at the lower height, despite not having the

maximum shear stress values.

Figure 13 shows examples of the shear maps on the bottom

surface after 10 seconds of flow time at a height of 0.07625 m

from the bottom. The shear stress on the bottom wall for the

dual, quad, and azimuthal jets are shown in Figures 13a, 13b,

and 13c, respectively. As previously mentioned, the azimuthal

jets had higher shear stress covering the most surface area on

the bottom wall, as shown in Figure 13c.

The maximum shear stress found for the different pulse

settings and the continuous jets were approximately the same.

The results were expected because the discharge phase of the

jets was long enough that they began to behave like steady jets

while on. The only difference was that for the pulsing jets, the

shear stress dropped to below 1 Pa during the time off period.

The shear stress did not completely go to zero because the

discharge phase of the pulse provided enough momentum to

keep the fluid active during the short off period.

The shear maps can be used to determine the effectiveness

of the jets removing the solids off the bottom of the tank and

thus creating a suspension. For example, in Figure 13a, the

dual jets will not be able to suspend a large portion of solids in

the middle region of the tank. On the other hand, the quad and

azimuthal jets, shown in Figures 13b and 13c, respectively,

have smaller regions where the jets are less effective at the

bottom of the tank. The aim was to minimize these areas

relative to the total base of the tank. Through modification of

Figure 12. Temperature contour of quad jet system after 3 complete pulse cycles using P4 setting in the (a) yz-plane and (b) xy- plane.

Figure 11. Temperature contour of quad jet system of first pulse cycle using P4 setting in the (a) yz-plane after initial on time of 5 s, (b) xy-plane after initial on time, (c) yz- plane after off time of 0.25 s, and (d) xy- plane after off time.


the design or the addition of more jets, these regions can be

further reduced.

Table 4. Maximum and average shear stress on bottom wall using different continuous jet configurations.

H = 0.07625 m H = 0.025 m

No. of jets

Max

Shear

Avg.

Shear on

Surface

Max

Shear

Avg.

Shear on

Surface

Dual Jets 12.36 Pa 2.98 Pa 79.64 Pa 6.79 Pa

Quad Jets 14.19 Pa 4.95 Pa 70.35 Pa 13.39 Pa

Azimuthal

Jets 21.81 Pa 6.58 Pa 80.48 Pa 12.92 Pa

CONCLUSIONS Continuous and pulse jets were modeled to determine the

best configuration for solid suspension processes. The mixing

time was calculated using a 65% homogeneity criterion. The

mixing time decreased as more jets were added. There was no

improvement in mixing times using pulse jet flows over

continuous jets. However, the quad jets using the P4 and P3

pulse setting had mixing times very close to the continuous jets

of the same configuration. The P4 and P3 settings had mixing

times of 15.5 s and 15.6 s, respectively, and the continuous

quad jets had a mixing time of 15.3 s. The results show that

using pulse jets can perhaps lead to a reduction in operational

costs in some applications, as less working fluid and less

energy is used, if properly optimized. However, more studies

must be run to reduce the uncertainties in the data. In addition,

more studies should be run using different pulse settings and

studies should be run for a longer period of time (i.e. higher

mixing time criterion).

The shear stress on the bottom of the tank was calculated

for different jet configurations as well. The azimuthal jets had

the highest shear on the bottom wall for both of the heights

studied. But, the maximum shear cannot be the only factor in

determining overall performance for solid suspension

processes. In the suspension of solids it is ideal to have a mixer

which is able to completely remove the solids from the bottom

surface. Therefore, the distribution of the bottom shear stress is

an important factor. In the current study, the distribution of the

shear stress was visually illustrated using contour color maps

and quantified using the averaging techniques along the bottom

surface. Overall, the quad jet systems provided the highest

average shear stress on the bottom surface at a height of 0.025

m from the bottom of the tank. The quad jet configuration

showed promising results and more studies should be done to

optimize its performance. The results from the current study

provide some insight into the mixing using pulsing jets and its

possible applicability to solid suspension processes.

ACKNOWLEDGMENTS The authors would like to acknowledge the financial

support of the Title III Program at North Carolina A&T State

University, which is administered by the U.S. Department of

Education, Institutional Development and Undergraduate

Education Services.

REFERENCES [1] Bathija, P.R., 1982, " Jet mixing design and applications,"

Chemical Engineering, pp. 89-94.

[2] Manjula, P., P. Kalaichelvi, and K. Dheenathayalan, 2010,

"Development of mixing time correlation for a double jet

mixer," Journal of Chemical Technology & Biotechnology,

85(1), pp. 115-120.

[3] Maruyama, T., Y. Ban, and T. Mizushina, 1982, "Jet

Mixing of Fluids in Tanks. Journal of Chemical Engineering of

Japan, 15(5), pp. 342-348.

[4] Patwardhan, A.W. and S.G. Gaikwad, 2003, "Mixing in

Tanks Agitated by Jets," Chemical Engineering Research and

Design, 81(2): pp. 211-220.

[5] Tatterson, G.B., 2003, Scaleup and Design of Industrial

Mixing Processes, McGraw-Hill, New York, NY, 2nd ed.

[6] Wells, B.E., et al., 2009, "Assessment of Jet Erosion for

Potential Post-Retrieval K-Basin Settled Sludge," Pacific

Northwest National Laboratory: Richland, WA.

[7] Abulnaga, B.E., 2002, Slurry Systems Handbook, McGraw-

Hill, New York, NY.

(a) (b)

(c)

Figure 13. Shear stress map on the bottom surface after 10 seconds of continuous flow time, at a height of 0.07625 m from the bottom, for (a) dual jets, (b) quad jets, (c) and azimuthal jets.


[8] Powell, M.R., Y. Onishi, and R. Shekarriz, Sept. 1997,

"Research on jet mixing of settled sludges in nuclear waste

tanks at Hanford and other DOE sites: A historical perspective,"

PNL-11686.

[9] Patwardhan, A.W., 2002, "CFD modeling of jet mixed

tanks," Chemical Engineering Science, 57(8), pp. 1307-1318.

[10] White, Frank M., 1974, Viscous Fluid Flow, McGraw-

Hill, New York,NY.

[11] H. D. Zughbi and M. A. Rakib, 2004, "Mixing in a fluid

jet agitated tank: effects of jet angle and elevation and number

of jets," Chemical Engineering Science, 59.


ASME Early Career Technical Journal

2011 ASME Early Career Technical Conference, ASME ECTC November 4-5, Atlanta, Georgia USA

MANUFACTURING KNOWLEDGE SHARING: THE UTILISATION OF KNOWLEDGE BASE TECHNOLOGIES IN PLM

Abdul Shakoor, Arshad Mehmood

Loughborough University, UK

Khizar Azam, Riaz Akber Sayyed Department of Mechanical Engineering University of Engineering & Technology

Peshawar Pakistan

KEYWORDS : PLM, NPDI, Knowledge Management, Knowledge Capturing, Knowledge Sharing

ABSTRACT

The maximisation of the prospects for substantial benefits and the preparation of industry to fulfill competitive needs in a globally oriented business environment requires a wide range information-based decision support system. This knowledge sharing framework needs an extensive research to improve inter-operations of cross-functional systems, which subsequently provides the basis for diverse utilisation of knowledge base technologies in manufacturing. The effective knowledge sharing framework in manufacturing industries is used to speed-up the New Product Development & Introduction (NPDI) process and improves the effectiveness of information flow in Product Lifecycle Management (PLM) environment. This paper describes the development of Knowledge Based System (KBS) and to select a process for making circular shapes and to explore the manufacturability of the intended design product. The KBS is an expert system, used to select a suitable process and cutting tools by defining the system’s constraint. The KBS is designed using E2KS; the state of the art software and knowledge capturing tool. The machining examples are used to demonstrate the implementation of the system, which will provide a basis for knowledge sharing in PLM environment. This approach provides a basis for knowledge sharing in manufacturing industries. Future manufacturing industry will benefit from further development and successful implementation of PLM system.

INTRODUCTION

The highly competitive environment and modern business requirements in world of globalisation always drives technological solutions to sustain the competitiveness of the

company. In a globalised economy, companies always face challenges to accelerate the new product development process to shorten time–to-market for early entry into the market and time - to - profit for quick revenue generation. In response to these challenges, companies are under high pressure to meet the demands of the market by launching tailored products to customers for economy of scope, to reduce time-to-volume via mass production for the economy of scale, and to decrease time-to-profit by increasing efficiency of the entire lifecycle for the economy of service [1].

Recently the traditional business model in manufacturing paradigm shifted from make-to-order (MTO), to engineer-to-order (ETO), to configure -to-order (CTO), to design-to-order(DTO) and in near future to innovate-to-order (ITO) as illustrated in Figure 1. The new business models have developed a competitive environment to provide a technical solution to the customer demands. Hence these technologies have changed the traditional mass production (MP) to flexible manufacturing system (FMS) followed by manufacturing knowledge management (MKM) leading to the product customisation (PC), product knowledge management (PKM), and to the product lifecycle management (PLM) as illustrated in Figure 2 [2]. PRODUCT LIFECYCLE MANAGEMENT

Product Lifecycle Management (PLM) is a product

strategic approach based on the product and stakeholder's interaction from inception to the end of life [3]. Different organisations have defined PLM in their own context. The NIST [7] defines PLM as a strategic business approach for the effective management and use of corporate intellectual capitals.


Figure 1. Development of New Business Models

Figure 2. Technological Development

KNOWLEDGE BASE SYSTEM Knowledge Base System (KBS) is considered to be an

important tool for maximising the utilisation of knowledge. KBS aims to follow the experience and knowledge of humans to be represented and used on a computer so as to enhance decision-making ability [8]. This database provides a base for knowledge management. KBS has the potential to accomplish objectives such as to capture, represent , share, reuse , transfer, and maintain the knowledge. It is implemented in a variety of decision making process during last decades, and still remains as centre point for research [4]. Utilisation of the KBS in PLM environment is a novel approach to accelerate the entire product design and development process and enhance innovation, as it

1. Capture and re-use of design and manufacturing

knowledge 2. Digitizing and automating the cyclic and same design

tasks. 3. Encapsulation of design rules using standard parts 4. Enhanced collaboration among design, simulation,

manufacturing, and services and recycling.

This research is mainly focused on the KBS design for manufacturing processes and suggests different ways to share this knowledge in PLM environment, especially for design and manufacturing in PLM system, which has eminence in the information and knowledge domain, needed to support decision process more quicker and more effectively. KBS not only covers the process related knowledge but also the resource knowledge as a prototype in manufacturing industry. This can be extended to capture all related knowledge and information from upstream to downstream applications. MANUFACTURING KNOWLEDGE SHARING SYSTEM DESIGN

There are different techniques to tackle the challenges

faced by manufacturing industry. One of such techniques is the PLM and inter-operatiability of organisations for developing a new product. Successful implementation of a PLM system requires a knowledge base support and a rich information core accessible by all departments and organisations working in a collaborative environment at every stage of PLM. This new innovation in manufacturing knowledge utilisation will put the manufacturing industry into a new paradigm of product design and development. Knowledge base technologies enable the manufacturing industry to exist in the era of knowledge wisdom, information creation and sharing [5]. Manufacturing process in reality is a combination of tangible and intangible transformation to get the end product. Tangibly, we obtain the materialistic form of the end product, but intangibly the knowledge is created with the virtual transformation of data information and management of knowledge. Subsequently this knowledge is utilised, have no boundaries and limits in manufacturing or PLM environment, which is the essence of this research. Figure 3 explains this concept in block diagram.

Raw Materials Products Data Information Information Knowledge

MTO

ETO

CTO

DTO

ITO

MP FMS MKM PC PKM PLM

Business Model changes

Driving Force (Market Dynamics)

Technological Changes

Driving Force (Business Model)

The Manufacturing

Process

Figure 3. Real and Virtual Manufacturing Transformation


METHODS AND TOOLS The KBS design is essentially based on the knowledge

transfer approach, from domain experts directly to systems. However, this KBS has been replaced by the model approach which is based on using conceptual model to model the problem-solving skill of the domain expert [6]. Different tools and methods are applied to organise the data. Various logics and controlling factors are implemented to manage the knowledge, usesd as a resource further downstream in a repetitive environment. Due to these features , knowledge base technologies have become a focus for many manufacturing organisations. Two different tools are used for design the KBS in this research.

• Unified Modelling Language (UML)

• Emergent System’s Engineering Knowledge

Management solution - E2Ks (Software)

The system functionality plays a significant role in KBS. UML is used to organise and arrange the random information and data into a specific order. The UML class diagram is used to structure the process and resource knowledge in research whilst the use-case diagram is used in the graphical representations of systems behaviour and response in terms of system output when an input is provided. E2KS is emerging as state-of-art software used for knowledge capturing in the manufacturing environment, introduced by Emergent Systems for manufacturing knowledge management solutions. E2KS provides rich interface to capture and manipulate information and knowledge for human usage. Furthermore, the captured information can be accessed and used by any tool. For example, E2KS information can be accessed within CAD environment and the CAD information can be validated against the captured E2KS information. E2KS provides software development kit (SDK) based on C++ and Java programs. The E2KS SDK can be used with CAD tools such as NX, CATIA and PLM tools like Teamcenter.

MANUFACTURING KNOWLEDGE

The most important step in this research was to determine what type of manufacturing knowledge is needed to capture and develop the methodology for that. Therefore limiting the scope of research, control convergence methodology was adopted to capture the information and knowledge for making circular type shapes, cutting tools and the materials that can be machined with. The KBS will contain all this information as a source of manufacturing knowledge.

Generally the knowledge base system will respond in two ways. Firstly the required process is identified and followed by the required tool according to the selected process. The base line was compatible with the required tolerance for the manufacturing process under specific conditions. Generally the degree of accuracy expected reasonably from various manufacturing processes under average conditions was a basis for calculation in designing the manufacturing KBS.

WORKING PRINCIPLE

The KBS is working in three steps. The first step in the process is selection under specific requirements given to the system as an Input and in second step the required tool will be selected and in third step the knowledge is shared in PLM environment as shown in Figure 4.

(Material data, Process data, Tool data)

Figure 4. Manufacturing Knowledge Base (Sharing

System) E2KS Process Overview

Step 1 Process Selection: process selection procedure for simple hole making process is shown in Figure 4. The process selection based on three input values, upper tolerance (U.T), and lower tolerance (L.T) and diameter (D) of the hole. The E2KS knowledge base is responded with suitable process. Step 2 Tool Selection: Once the process selected for required product, the KBS will select the required tool in available knowledge resource. Two input parameters are required the flute length (F.L) and diameter (D) of the hole. Generally the

Tool Data

Process Data

Material

Data

Step 1 •Process Selection

Step 2 •Tool Selection

Step 3 •knowledge sharing in

PLM environment

Manufacturing Knowledge

E2KS KBS


overall overview and structure of the KBS and its utilisation is PLM system shown in Appendix A. MANUFACTURING PROCESS SCENARIO TO SHOW THE WORKING OF KBS Scenario 1 (Figure 5) uses a simple case illustrating the hole making process. Scenario 1

• Diameter (D) of the hole: 15 mm • Upper Tolerance (U.T): 50 (µm) • Lower Tolerance (L.T): -50 (µm) • Depth of hole (F.L): 55 mm

The KBS resolves to what process is suitable for required size of hole as the simulation (E2KS Process Look-Up K-PAC) shown in Figure 6.

According to the E2KS Knowledge base drilling is the required process to make the required size hole. Now the drilling tool

K-PAC (knowledge pack) will resolves the required tool. The E2KS simulation is shown in Figure 7. The KBS system shows the two different tool series with required size which both can be used for drilling the required hole size. Along with tool series, the KBS also shows the surface finish that can be achieved with this tool and the materials that can be machined with. Different types of materials can be processed with both series. The tool selection depends on the material that will be processed. This is shown in Figure 7 and it shows how the KBS proceeds with tool selection for required dimensions.

Figure 7. E2KS Tool Selection Process for Scenario 1

Scenario 2

• Diameter (D) of the hole: 4 mm • Upper Tolerance (U.T): 10 (µm) • Lower Tolerance (L.T): -5 (µm) • Depth of hole (F.L): 9 mm

Figure 5. Hole Making Scenario 1

+50 Ø = 15 + -50

Drilling is Required

Figure 6. KBS Process Selection for Scenario 1

Material Type

+10 Ø=4 -5

Required Drill Size

Surface Finish

Material Type

Available Tool

Figure 8. Hole Making Scenario 2


The E2KS process and tool selection for the material in Figure 8, is shown in Figures 9 and 10. The required process is reaming, followed by available tool needs to be selected.

Now as for the reaming operation the tool is available but we cannot make the required hole size directly with reaming. First we have to drill the hole smaller than the required size. For this, first have to consult KBS, either the required drill size is available or not. Let say that for 4mm (D) size and 9 mm (F.L) reamer, required drill is 3.5 mm (D) and 9mm (F.L). Figure 11 shows the available drill size to prepare the hole for reaming operation. The problem occurs if the tools, reamer and drill are of different functionalities, like if the materials are different which can be processed or if there is different in surface finish requirements.

DISCUSSION AND CONCLUSION

The future benefits for the manufacturing industry to remain competitive in the age of highly dynamic digital manufacturing are significant for manufacturing industry if the manufacturing knowledge can be shared within the company and across the whole supply chain. It is very substantial to explore the meaning of sharing to achieve any level of success in knowledge sharing. Different approaches and techniques are used in the manufacturing industry to share the manufacturing knowledge effectively and improve the manufacturability of the system. The KBS is developed for manufacturing industry and used the knowledge capturing tool (E2KS). It demonstrates the effectiveness of knowledge base (KB) technology and approach for capturing and managing the manufacturing knowledge for manufacturing process and required cutting tools. It is also support its value in providing an effective foundation to capture manufacturing concepts and knowledge. The system was kept only to capture the manufacturing knowledge about the processes for making the circular shape designs and to check the capability of the tools. It is a tentative implementation of the proposed system where the library model is implemented in E2KS knowledge base to provide the manufacturing feedback during the product design stage and its manufacturing. Generally the knowledge base approach develops the implementation of the following functionalities;

• E2KS has a variety of option to be used for manufacturing knowledge capturing.

Required D ill Si

Surface Finish

Material Type

Available Tool

Figure 10. Tool and Materials Selection for Scenario 2

Reaming

Figure 9. Process Selection for Scenario 2

Surface Finish

Material Type

Required Reamer size

Available Tool

Figure 11. Drill Tool Selection for Scenario 2


• The different K-PAC like methods K-PACs, look-up K-PACs, Calculation K-PACs etc provides different options to capture the knowledge.

• Relationship between manufacturing and design part can be captured as rules

• The rules can be very complex as multiple logical rules can be established.

• The manufacturing best practice can be captured as method K-PACs

• This E2kS knowledge base provides a foundation for manufacturing capturing and this knowledge can be shared in PLM environment.

The E2Ks knowledge base system partially implemented which can be further developed and integrated with some sophisticated functionalities like:

1) E2KS knowledge base can be shared in PLM environment tools like Teamcentre, etc

2) E2KS offers the integration facility with other CAD software tools like NX, CATIA, so the manufacturing and Design knowledge can be directly captured from CAD data.

3) E2Ks also perform the design realisation activities within the CAD environment by accessing the captured information with knowledge fusion functionality.

The integral E2Ks Knowledge base will provide customer, developer, manufacturer and supplier the capability of knowledge sharing within PLM system and successful implementations. It will improve the product lifecycle collaboration within the enterprise and will manage the product lifecycle activity effectively to accelerate the decision making process by providing the right information at the right time. Indeed the knowledge base will provide the foundation for developing a PLM system and concept. The use of ICT in the manufacturing sector will lay strong frontiers for further development and applications of PLM strategy for the benefit of industries. This research will provide a referring guide for professional researchers to conduct further research. An extensive research is currently underway to explore the inter-operation and knowledge sharing issues across the virtual manufacturing organisation and improve the future inter-operatiability of the system and knowledge sharing procedures.

NOMENCLATURE

PLM: Product Lifecycle Management KBS: Knowledge Base System U.T: Upper Tolerances L.T: Lower Tolerances F.L: Flute length D: Diameter E2Ks: Software CoP: Community of Practice K-PAC: Knowledge Packets

ACKNOWLEDGMENTS The authors acknowledge the financial support of

University of Engineering & Technology (UET) Peshawar and Higher Education Commission (HEC) of Pakistan for this research. REFERENCES

1. Group, A. A. (2001). OLM Strategy, Key to Future Manufacturing Success

2. Guixiu Qiao, C. M. (2004). Manufacturing Information Integration in Product lifeCycle Management. New York: Wiley.

3. Ameri, F., & Dutta, D. ( 2005). Product Lifecycle Management: Closing the Knowledge Loops . Computer-Aided Design & Applications , 577-590

4. Guerra-Zubiaga, D. A. (2004). A manufacturing Model to enable knowledge maintenance in decision Support Systems. Doctoral Thesis . Loughborough University

5. S.Loub, Z., & Fang, X. (2001). KBS-aided Design of Tube bending processes. Engineering Applications of Artificial Intelligence (14), 599-606.

6. R.I.M Young, A. A.-D. (2007). Manufacturing Knowldge Sharing in PLM: a progression towards the use of heavy weigh ontologies. International Journal of Production Research , 45 (7), 1505-1519.

7. Grieves, M. (2006). Product Lifecycle Management : Driving the next Generation of Lean Thinking. New York: McGraw-hill.

8. Stark, J. (2004). Product Life Cycle Management-21st Century Paradigm for Product Realisation. London: Springer.


ANNEX A E2KS Knowledge Base System Overview & KB Utilisation in PLM Environment

Resource Knowledge

3-Axis VMC Knowledge

M/C Knowledge

Tool Knowledge

Material Knowledge

Metal Tool Knowledge

Tolerances Knowledge

Knowledge Capturing Tool

(E2ks)

Drilling Knowledge

Processes Knowledge

Reaming Knowledge

Types of Knowledge

Knowledge Sharing PLM Environment

PLM Environment

PLM tools

Teamcentre, CATIA etc

Facility Knowledge E2KS KBS


ASME Early Career Technical Journal

2011 ASME Early Career Technical Conference, ASME ECTC


MULTIDISCIPLINARY DESIGN OPTIMIZATION OF AEROSPACE VEHICLE – SINGLE

ENGINE ROTORCRAFT

Adeel Khalid Southern Polytechnic State University

Marietta, GA, USA

ABSTRACT Aerospace vehicle design includes several disciplines

with often-conflicting requirements. A formal system design

framework is developed in this research where the designer

coordinates with disciplinary experts to find an overall

optimized design while simultaneously optimizing disciplinary

objectives. The overall system objective function chosen in the

preliminary design is minimum production cost for a light

turbine-training helicopter. Several disciplinary objectives

including specific fuel consumption for propulsion, empty

weight for weights group, and figure of merit for aerodynamics

group are optimized. In addition to disciplinary optimization,

several analyses are performed including vehicle engineering,

dynamic analysis, stability and control, transmission design, and

noise analysis. The design loop starts from the conceptual stage

where the initial sizing of the helicopter is done based on

mission requirements. The initial sizing information is then

passed to disciplinary experts for preliminary design. The

design loop is iterated several times using multidisciplinary

design techniques like All At Once (AAO) and Collaborative

Optimization (CO) approaches. A light training helicopter is

proposed that satisfies all the mission requirements, is

optimized for several disciplines and has minimum production

cost. Merits, demerits, requirements and limitation of the

proposed methodology are discussed.

INTRODUCTION The multidisciplinary nature of the helicopter makes it

hard for the preliminary designer to estimate the actual cost of

the aircraft in the early design stage. There has been a lot of

emphasis on bringing more and more design information early

in the design stage [1, 2, 3, and 4]. In this research, a variety of

disciplinary design analyses are performed at the preliminary

helicopter design stage where the design cycle is repeated in an

iterative fashion using Multidisciplinary Design Optimization

(MDO) to ensure an overall optimized design. All At Once

(AAO) and Collaborative Optimization (CO) techniques

developed by Kroo, Sobieski, Braun et al. [5, 6, 7 and 8] are

used. These methods help integrate various design disciplines

while removing their interdependency. AAO approach solves

the problem by removing disciplinary optimizers and

introducing a multi-objective criterion for the system level

optimizer. CO allows simultaneous use of disciplinary

optimizers thus ensuring that not only an overall optimized

design is obtained, but also the design is best from disciplinary

point of view.

A light turbine-training helicopter is used as baseline

for analysis. The requirements come from the Request for

Proposal (RFP) from AHS 2006 student design competition.

The helicopter is expected to lift two 90 kg people, 20 kg of

miscellaneous equipment, and enough fuel to Hover Out of

Ground Effect (HOGE) for 2 hours, into HOGE at 6,000 ft. on

an ISA + 20oC. The winning design team at Georgia Tech made

use of a variety of software packages and codes for different

disciplines to perform the analyses. The idea is to perform the

required mission with minimum cost. A variety of software

packages and codes for different disciplines are integrated in

this research to perform analyses.

In a traditional design approaches, due to strong

dependency of disciplines on each other, it is not possible to run

several analyses in parallel and therefore the design process is

slowed down significantly. Infact, in several cases, by the end of

the design, it is not possible to complete even one design loop

involving all analyses simultaneously. This process does not

guarantee overall system cost minimization.

In this research, a platform is developed where all the

disciplinary codes, software and analyses are integrated and

several design loops are performed. The overall system design

criterion is chosen to be the minimum production cost. All the

disciplinary and the system level design constraints are

satisfied. Optimized results obtained from AAO and CO

approaches are compared. These methods, when fully

converged, ensure an overall optimized design with several

disciplines involved. A data repository is created where design

information is stored iteratively. As the design matures, new

information is obtained from disciplines and subsequent

analyses are performed. A parallel design helps reduce the

design time from several months to a few hours. The entire

design process is automated in this research. There are no

feedbacks or feed forward loops between different disciplines


and all the disciplinary optimizers are retained. The weight

optimizer minimizes the empty weight of the helicopter, the

propulsion optimizer minimizes the fuel consumption and

emission, and the aerodynamics optimizer maximizes the hover

Figure of Merit. All the other disciplines are of analysis type.

The initial problem setup is time consuming and

tedious, but after the problem is set up and software packages

are integrated, information flow becomes very efficient.

ModelCenter is used as a platform for information transfer and

system level optimization. Individual wrappers are written for

data transfer between ModelCenter and disciplinary codes.

These codes include commercial software, legacy codes and

customized in house programs. Data is transferred in batch

mode for rapid analysis.

The design team estimated the average vehicle cost,

based on the production of 3000 units, to be $203,541.85 per

unit. This estimate was based on sequential disciplinary

analyses. The results obtained from MDO environment with

parallel execution show a reduction in the average unit cost to

$178,175.69. This is a 12.46% reduction in cost while all the

constraints and RFP mission requirements are satisfied. This

study demonstrates an automated framework for preliminary

helicopter design. The framework ensures that detailed

disciplinary analyses are performed in an automated manner in

parallel at the initial design stage while at the same time overall

disciplinary and system level optimized results are obtained.

The platform also allows the designer to add further detailed

disciplinary analyses so that the overall fidelity of the system

can be improved.

DESIGN METHODOLOGY The generic Integrated Product and Process Development

(IPPD) methodology allows the engineers and program

managers to decompose the product and process design

iterations [5]. The product development cycle of IPPD

methodology is further divided into the conceptual and

preliminary design loops. In this research, disciplinary analyses

are identified, linked with each other through a common

platform and preliminary design loop iteration is performed in

an automated fashion. This approach provides an opportunity to

perform system optimization using MDO techniques.

Several requirements exit for a framework to provide an

easy to use and robust MDO environment. Key attributes for the

MDO environment, as listed by Sobieszczanski Sobieski [3],

are computer speed, computer agility, task decomposition,

sensitivity analysis, human interface and data transmission. The

framework requirements for MDO application development

have been outlined in the work of Salas and Townsend [6] as

architectural design, problem formulation, problem execution,

and access to information.

Sobieski [3] indicates that among several tools that

specialize in process integration and exploration, ModelCenter

from Phoenix Integration Inc. focuses on product modeling and

ease of tools and process integration across distribution and

heterogeneous computing environments. ModelCenter is used in

this research because of its flexibility to incorporate several

existing commercial packages e.g. Excel, Matlab, MathCad,

CATIA etc. ModelCenter also facilitates the use of wrappers to

integrate in-house legacy codes.

Two MDO approaches, i.e. All At Once (AAO) and

Collaborative Optimization (CO) are used in this research to

resolve the conflicting objective functions of different

disciplines and the system level optimizer. AAO approach

dictates that all the local design variables and constraints are

moved to the system level. The local disciplinary optimizers are

eliminated. The problem is reduced to a single level scheme

with just one system optimizer. AAO approach with the

capability of parallel execution of disciplines is depicted in

Figure 1. With this approach, the disciplines become

independent of each other. This approach facilitates the

information flow from the repository to the respective

disciplines and back to the repository in an automated fashion.

Figure 1. AAO Approach with a Central Data Repository

When the feedback loops are removed, compatibility

constraints are added to the system level optimizer. The generic

system level problem is defined as follows.

Objective Function: Minimize F

Subject To: BA ggg ,,

Variables: BA XXX ,,

Where

F = Overall Evaluation Criterion (OEC)

g = System level constraints

BA gg , = Compatibility constraints

02

'≤−−= εBBg A

02

'≤−−= εAAg B

X = System level variables

A

B

System

Optimizer

Repositor

y


BA XX , = Disciplinary variables

Where

BA, = Actual output vectors from disciplines A and B

'' ,BA = Intermediate variables

The intermediate variables are copies of true outputs from

disciplinary analyses. These variables are additional

independent variables and are treated like design variables that

the optimizer has to serve. The compatibility constraints ensure

that the difference between the disciplinary outputs and outputs

from system level result in the same values. This approach

eliminates iterations between disciplines. There is no

optimization conflict between disciplines. However the system

level optimizer can become very large for a problem with

several disciplines and several variables. AAO also removes the

control of local variables by local experts or local optimizers.

The expert codes are reduced to simple analysis only mode,

making them servants to the system level optimizer. No problem

solving occurs at the local level. Some of these issues are

addressed by the CO approach.

Collaborative Optimization is a design architecture

developed by Sobieski et al [2, 3], specifically created for large-

scale distributed analysis applications. In this approach, a

problem is decomposed into user-defined number of subspace

optimization problems that are driven towards interdisciplinary

compatibility and the appropriate solution by a system level

coordination process. The fundamental concept behind the

development of Collaborative Optimization architecture is the

belief that disciplinary experts should be able to contribute to

the design process while not having to fully address local

changes imposed by other groups of the system.

To facilitate this decentralized design approach, a problem

is decomposed into sub-problems along domain specific

boundaries. Through subspace optimization, each group is

given control over its own set of local design variables and is

charged with satisfying its own domain specific constraints.

Communication requirements are minimal because knowledge

of other group’s constraints or local design variables is not

required. The objective of each sub-problem is to reach

agreement with the other groups on values of the

interdisciplinary variables. A system level optimizer is

employed to orchestrate this interdisciplinary compatibility

process while optimizing the overall objective. To avoid the

conflict between disciplinary objectives, collaborative

optimization replaces the objective functions of each

disciplinary optimization.

The new objective function attempts to minimize a newly

defined error function, known as J term. These J terms measure

the relative error between the output variables of the

disciplinary tools and corresponding target values. These target

values are set by the system level optimizer, which is configured

to optimize a system level objective function under the

constraint that the J terms in each discipline are kept within a

certain tolerance. Each disciplinary tool is allowed to vary all of

its usual inputs and local variables to minimize its own

objective function. This allows for the disciplinary experts to

focus on their domain specific issues while maintaining

interdisciplinary compatibility. Collaborative Optimization in a

Design Structure Matrix (DSM) format is shown in Figure 2. A

generic Collaborative Optimization problem is given as follows.

Figure 2. Collaborative Optimization Design Architecture


Base

airfield

Training

airfield

0

1 2

3

4

5

6

7

8

10 9

System Level Problem:

Objective Function: Minimize f

Constraints: 0,, ≤CBA JJJ

Variables: ''' ,,, CBAX

Disciplinary Problem A:

Objective Function: Minimize 2

'2

'2

' AACCBBJ AAA −+−+−=

Subject To: 0≤Ag

By Changing: AAA CBX ,,

Similar problems are defined for disciplines B and C. This

decomposition strategy allows for the use of existing

disciplinary analyses without major modification and is also

well suited to parallel execution across a network of

heterogeneous computers.

A complex system like rotorcraft design is composed of

several levels of problems. Each system is divided into further

subsystems. Several models including in-house codes,

commercial software, legacy codes and programs are integrated

in this research using ModelCenter as the common platform.

ROTORCRAFT MDO MODELS AND ANALYSES

Light Turbine Training Helicopter (LTTH) is chosen for the

application of the IPPD methodology using MDO tools.

Detailed disciplinary tools, analyses, software packages, and

programs are identified and integrated using a common

platform. The integration of disciplines using a common

platform enables the transfer of design information from one

discipline to another in an efficient manner. A centralized

database is created where all the latest design information from

all the disciplines is stored. The dynamic platform enables the

application of optimization techniques at the system level.

LTTH baseline information comes from the 23rd annual student

design competition 2006 Request For Proposal (RFP) [5]

published by the American Helicopter Society (AHS). The goal

is to develop a two-place single turbine engine, training

helicopter that is affordable. The mission requirements and

some of the system level constraints are also defined in the RFP.

This includes the capability to lift two 90kg people, 20kg of

miscellaneous equipment, and enough fuel to Hover Out of

Ground Effect (HOGE) for 2 hours, into HOGE at 6,000 ft. on

an International Standard Atmosphere (ISA) + 20oC day. The

winning LTTH design team at Georgia Tech further refined

these requirements and new details are added. Based on the

RFP, a mission profile is generated as shown in Figure 3.

The LTTH design team developed a conceptual baseline

vehicle using the performance requirements stipulated in the

RFP. This was followed by preliminary design, which provides

more detailed analysis in multiple disciplines to identify the

necessary baseline vehicle modifications. Disciplinary analyses

Figure 3. Mission Profile for LTTH design

include aerodynamic performance optimization, structural

design, analysis, and material selection, CAD modeling,

helicopter stability and control analysis, dynamic analysis,

propulsion system design, helicopter training industry research

and cost analysis. The goal of the proposed design can be

summarized as reducing the cost while improving product

quality and value.

Important disciplines involved in the preliminary design

loop of IPPD methodology are identified in this research. All

possible tools, analysis packages, commercial software, legacy

codes and in house programs are utilized and integrated using a

common platform i.e. ModelCenter. Individual optimization

problems are defined for propulsion, economics, weight and

balance and aerodynamics groups. Details of the analyses used

in each discipline are discussed in the following section.

PERFORMANCE ANALYSIS Two methods that are used for performance analysis

are the Fuel Ratio (RF) method and the Georgia Tech

Preliminary Design Program (GTPDP), which is based on

regression of historical data.

Extended RF method for advanced VTOL aircraft is used

to size the vehicle in this research. It is a graphical technique for

the parametric analysis of aircraft for a given design

specification, and provides a preliminary design tool for

investigation of the interrelated effects of significant design

parameters on the gross weight of the aircraft. One of the

biggest limitations of RF method is that it only allows the

designer to minimize the gross weight. Only a few variables are

involved in the analysis. More comprehensive preliminary

design tools like HESCOMP and GTPDP are available for

initial sizing. GTPDP is utilized iteratively in the design loop in

this research.

GTPDP is a preliminary performance analysis program in

nature based on an energy approach and has been written to

yield results quickly and inexpensively [5]. Yet, the results are

of sufficient accuracy that sound understanding of a helicopter

behavior can be obtained and valid comparison can be made.

The mission profile analysis is also performed in GTPDP.

GTDPD is simple and quick preliminary helicopter design code

but the calculations are approximate and it uses simple

approaches and equations.


The initial outputs obtained from the baseline vehicle

analysis in GTPDP are used as inputs for the other disciplinary

analyses. Besides the initial analysis, GTPDP is also used in the

system design iteration to get approximate mission performance

output. GTPDP is integrated with ModelCenter, so information

can be transferred from GTPDP to other disciplines.

VEHICLE ENGINEERING

Besides developing graphical models of vehicle

components, a design package is used for several other analyses

including weight and moment estimates. A state of the art design

package, CATIA V5 is used for the CAD design of vehicle in

this study. LTTH design team developed several detailed

models of components of the baseline helicopter. These

components are then integrated using assembly function.

Isometric and three view baseline vehicle drawings developed

by the design team are shown in Figure 4. With the help of these

parametric models, the vehicle geometry can be updated as

design values change. Some of the data that is extracted from

the models include the moment of inertias, surface areas,

volumes, and masses of individual components. This

information is then passed on to respective disciplines as

needed.

STABILITY AND CONTROL AND TRIM ANALYSIS A variety of analysis tools including the Georgia tech

Unified Simulation Tool (GUST), FlightLab and Matlab trim

analysis code and linear control root locus plots are investigated

for stability and control discipline in this research. GUST, used

at Georgia Tech for UAV research and development, has been

designed to take input parameters from rotor dynamics,

aerodynamics, gear dynamics, and flight controls with sensors –

enabling it to monitor flight characteristics and perform

missions using trajectories in real life scenario [5].

GUST is an instrumental tool in the pilot development

process. FlightLab on the other hand is an industry standard

program. For integration with preliminary design loop, Matlab

based trim analysis and root locus analyses are performed. Pitt

and Peters [9] first harmonic inflow is assumed. The detailed

trim analysis using all the nonlinear equations ensures a

complete and comprehensive helicopter system.

AERODYNAMIC ANALYSIS A combination of preliminary and detailed analyses is

performed in this research. The results obtained from low

fidelity analysis help accelerate the system level iterations. The

high fidelity results obtained from CFD are introduced from

time to time to get better results. A FPDA (Flat Plate Drag Area)

code is used for calculating component drag. Blade element

theory is used to calculate the forces of the blade due to its

rotation through air, and hence the forces and performance of

the entire rotor. Blade element theory combined with Landgrebe

wake model [3, 4] is also used to evaluate the rotor hover

performance. The blade element code with Landgrebe wake

model is integrated with ModelCenter and is used in

conjunction with simplified closed form expression of Blade

element momentum theory model for forward flight. In hover

case, Figure of Merit is maximized by changing the variables

that directly affect hover performance and keeping the

remaining variables fixed at the baseline value. The

optimization problem is given as follows:

Figure 4. Three-view drawing of baseline vehicle


Aerodynamic

Data

Aerodynamic

Data Repository

Optimizer

BEMT

Model

FPDA

Model

NASCART

Model

From System Optimizer

To System Optimizer

Objective Function: Maximize Figure of Merit

Variables:

-12 ≤ Twist angle ≤ -5

2 ≤ No. of blades ≤ 4

1% ≤ Root cutout ≤ 15%

A drag buildup method is used to estimate the equivalent

flat plate drag area of the vehicle. The FORTRAN based drag

estimation code is integrated with ModelCenter. A summary of

Aerodynamic analysis integration is shown in Figure 5.

Figure 5. Summary of Aerodynamic Analyses

PROPULSION ANALYSIS Historical light turbine engines are compared on the

basis of engine weight, specific fuel consumption, compressor

pressure ratios, and mass flow rate. This comparison gives an

approximate idea of the size, weight and compression ratio of

the new engine to be designed. Various analyses packages

available commercially and non-commercially for turbine

engine design are explored. These include the NASA Engine

Performance Program (NEPP), On-Design (ONX) and Off-

Design (OFFX) from Aircraft Engine Design AEDsys [5] and

Gas Turb 10. Gas Turn 10 is a sophisticated program available

for on-design and off-design cycle analysis for turbo shaft

engines. Parametric studies, Monte Carlo simulations and cycle

optimization tasks are completed using this program. An

optimization problem is defined in the propulsion group to

minimize the SFC by changing the size of the engine and

keeping the engine cycle parameters constant. Allison 250-C 20

engine is use for Gas Turb 10 calibration. The optimization

problem is defined as follows.

Objective Function: Minimize SFC

Constraints:

x1 ≤ Mass Flow Rate ≤ x2

y1 ≤ Pressure Ratio ≤ y2

Variables:

Mass Flow Rate

Pressure Ratio

Sequential Quadratic Programming (SQP) is used to

perform the optimization. The results are discussed in the

results section.

ECONOMIC ANALYSIS Economic analysis is of significant importance in this study

because of the fact that the total cost of the vehicle is used as

the Overall Evaluation Criterion (OEC) for the system level

optimizer. All the other disciplines work not only to maximize

their own individual criteria but also minimize the overall cost

of the system. Therefore minimum cost is of interest to every

discipline.

The cost analysis packages used in this research include the

GTPDP cost model [5], Bell cost model and Price H model.

Preliminary cost estimates obtained using GTPDP are group

weights based on historical data. More detailed development,

recurring production, and operating and support cost analyses

are performed using PC based Bell cost model. The model

predicts the cost of each aircraft subsystem by dividing it into

three separate categories: sub-contractor, labor, and materials. A

weight-cost factor is calculated for the baseline design by the

LTTH design team to identify the real cost drivers.

The five areas identified as the most influential are the

power plant, fuselage, flight controls, drive system, and rotor.

These subsystems with their sub-contractors, labor, and material

categories allocated by percentage for the baseline vehicle are

shown in Figure 6. This figure demonstrates the strong

influence of material selection for the fuselage, flight controls,

drive system, and rotor; accounting for over 50% of the total

vehicle production cost. The total average cost for LTTH

baseline design is $200,576. This cost includes both direct and

indirect operating costs. The baseline cost is used as a starting

reference for the objective function of system level design loop.

Direct and indirect costs are also calculated using Bell cost

model. The design team used Price H model to calculate the

new engine cost.

The PC based Bell cost model is integrated with

ModelCenter using a built-in ModelCenter API. The economics

optimization problem is given as follows:

Objective Function: Minimize Avg. Prod. Cost

Constraints: Production units ≥ 3,000

Production rate ≥ 300/yr.

Variables: Number of blades (MR)

Number of blades (TR)


Figure 6. Cost structure for major cost driving systems

RESULTS AND DISCUSSION The application of MDO techniques in the preliminary

rotorcraft design stage involves identification of all the

disciplines involved, their key variables, disciplinary

constraints, objective functions, and optimum results. The goal

is to design a rotorcraft starting from a baseline in minimum

amount of time, at minimum cost while satisfying disciplinary

and system constraints.

Four different MDO problems are solved in this study.

Before performing the AAO and CO analyses, which are

computationally expensive, it is prudent to verify the new

design environment against known design. This provides a level

of confidence in the program design environment and reveals

any inconsistencies within the integration process. The baseline

LTTH is used to verify the analyses that are used. Analyses

performed with the new framework are discussed in problem 1.

There is no optimization involved in any of the disciplines. The

disciplines are reduced to simple analysis modes. This ensures

that the results obtained match with those obtained from AHS

design team and hence prove that the framework developed, its

integration, and connectivity of analyses are correct. In problem

2, individual disciplinary optimizers are included but no system

level optimization is performed. So individual disciplines are

optimized but overall system may not be optimal. In problem 3,

AAO approach is employed. Problem 3 has two parts. In the

first part, three is a single system level objective function, i.e.

cost, that is of interest to every discipline but there are no

disciplinary optimizers. In the second part, an Overall

Evaluation Criterion (OEC) is defined that makes it a Multi-

Objective optimization problem. The OEC is a composite of all

the objective functions of individual disciplines that have

optimizers. The results of problem 3 with the OEC are shown in

Figure 7. Finally, in problem 4, Collaborative Optimization is

performed at the system level; as well as individual disciplinary

optimizations are performed at the disciplinary levels. The

individual disciplinary objective functions are modified from

those of Problem 2 to ensure that there is no conflict between

optimizers. The convergence history and results of problem 4

are shown in Figure 8. The results of all the problems are

compared. Some of the important baseline variables are shown

in Table 1.

Table 1: Results obtained from baseline analysis

Main rotor radius 12.2 ft.

Main rotor tip speed 650 ft./sec

Main rotor chord 0.64 ft.

Main rotor twist angle -10 deg

Main rotor effective hinge offset 0.07

Main rotor vertical distance from C.G 6.64 ft.

Main rotor lateral offset from C.G 0 ft.

Main rotor longitudinal offset from C.G 0.33 ft.

Pre cone angle 2.75 deg

No. of main rotor blades 3

Figure of Merit 0.68

Tail rotor radius 2.25 ft.

Tail rotor twist angle 0 deg

Tail rotor pre cone angle 1.5 deg

Tail rotor blade chord 0.23 ft.

Engine takeoff Horsepower 184 HP

Engine max continuous power 160 HP

Vehicle empty weight 800 lbs.

Vehicle gross weight 1454 lbs.

Vehicle flat plate drag area 7.29 ft2

Flyover noise 90.6 dB

Tail boom length 17 ft.

Average unit production cost $203,541

Direct operating cost $93.48

Figure 7: Convergence history of AAO single objective approach

$178,175

$203,576


Figure 8: Convergence history of AAO with multi-objective approach

CONCLUSIONS Key components involved in the rotorcraft preliminary design

process are identified in this study. Design tools available both

at the conceptual and preliminary design stages are determined.

Capabilities and limitation of these design tools are identified.

These design tools include commercially available software,

legacy codes, and in house programs. Design variables that are

of interest to more than one discipline are linked via a

repository. The architecture enables the transfer of information

between different tools in an integrated manner.

According to IPPD methodology, the detailed mission

analysis is performed at the conceptual stage of the design. The

suggested design methodology allows the integration of various

disciplines in a plug and play fashion. The framework allows

the designer to visualize the effects of change of important

design variables simultaneously at all the disciplines involved.

Objectives and constraints of various disciplines are identified.

Few optimization techniques that facilitate this architecture are

applied in this study. These include individual disciplinary

optimization, All At Once approach, and Collaborative

Optimization.

A light turbine training helicopter is chosen for the proof of

concept. Different optimization problems are compared and

their results are analyzed. It is observed that system

optimization problem requires significant more execution time

than individual design analyses. However the overall design

time is reduced from several months to a few hours with the aid

of the integrated and automated design architecture. Although

the setup time is significant, an overall optimized design is

obtained. It is observed that by making little changes in the

design variables, all the disciplinary constraints are satisfied,

and a helicopter with minimum cost is designed. In summary,

following improvements are observed in the preliminary

rotorcraft design with the use of MDO techniques.

• The design process is automated

• Disciplinary inter-dependency is removed which facilitates

parallel execution of analyses

• Current and updated information is available to all the

disciplines at all times during the design process

• The overall time required to design a rotorcraft is

significantly reduced

Disciplinary and system optimization is possible using

MDO techniques

ACKNOWLEDGEMENTS The author would like to thank the U.S. army for funding

this project through various grants and is especially grateful to

Dr. Daniel Schrage, School of Aerospace Engineering, Georgia

Institute of Technology for supervising the research work that

was done as part of the author’s Ph.D. thesis.

REFERENCE: [1] Mavris, D., Baker, A., Schrage, D., “IPPD through robust

design simulation for an affordable short haul civil tilt rotor”,

Proceedings of the American Helicopter Society 53rd Annual

Forum, Virginia Beach, VA, April 29 – May 1 1997

[2] Braun, P. J. Gage, Kroo, I. M., Sobieski, I., “Implementation

and Performance issues in collaborative optimization”, NASA

Langley Research Center, Proceedings of the 6th

AIAA/USAF/NASA/ISSMO symposium on Multidisciplinary

Analysis and Optimization, September 4-6, 1996

[3] Sobieski, I., Kroo, I. M., “Aircraft design using

Collaborative Optimization”, Proceedings on the AIAA 34th

Aerospace Sciences meeting and exhibit, Reno, NV. January

15-18, 1996

[4] Kirby, M. R., 2001, “A methodology for technical

identification, evaluation, and selection in conceptual and

preliminary aircraft design”, Ph.D. dissertation, Georgia

Institute of Technology

[5] Khalid, S. A., 2006, “Development and implementation of

rotorcraft preliminary design methodology using

multidisciplinary design optimization”, Ph.D. dissertation,

Georgia Institute of Technology

[6] Salsa, A. O., Townsend, J. C., “Framework requirements for

MDO application development”, AIAA paper no. AIAA-98-

4740, 7th AIAA/USAF/NASA/ISSMO symposium on

Multidisciplinary Analysis and Optimization, St. Louis, MO.,

September 2-4, 1998

[7] Braun, R. D., Moore, A. A. Kroo, I. M., “Use of

Collaborative Optimization architecture for launch vehicle

design”, NASA Langley Research Center, Proceedings of the 6th

AIAA / USAF / ISSMO / Symposium on multidisciplinary

analysis and optimization, AIAA Paper No. 96-4018, September

4-6, 1996

[8] Braun, R. D., Moore, A. A., Kroo, I. M., “Collaborative

approach to launch vehicle design”, Journal of Spacecraft and

Rockets Vol. 34, No. 4, July-August 1997

[9] Pitt, D. M. and Peters, D. A., “Theoretical prediction of

dynamic inflow derivatives”, Vertica, Vol. 5, No. 1, 1981

OEC




OPTIMAL LIQUID FILM THICKNESS OF AN EVAPORATING FILM DOWN AN INCLINED PLANE FOR MAXIMUM HEAT FLUX IN SPRAY COOLING

APPLICATIONS

Richard Opoku1and John P. Kizito2 Mechanical Engineering Department, McNair Building

North Carolina A&T State University Greensboro, N.C 27411, USA

[email protected] and [email protected]

ABSTRACT Spray cooling technique as a thermal management

scheme for high heat flux applications has been reported to depend on major parameters such as: the spray droplet size, the spray nozzle type, fluid volumetric flux, spray angle, the surface configuration of the test specimen, working fluid type and the liquid film thickness that forms on the test surface. In this present paper, theoretical investigation has been carried out to determine the optimal non-wavy liquid film thickness to maximize Critical Heat Flux (CHF) in spray cooling applications. A temperature function has been obtained for a heat flux correlation to estimate an optimal liquid film thickness for maximum heat flux of an evaporating thin liquid film flowing down an inclined heated flat surface.

Our results indicate that higher heat fluxes are obtained for vertical plates than inclined planes for the same liquid film thickness and fluid upstream conditions. Maximum Critical Heat Flux (CHF) of 800 W/cm2 was achieved for an optimal liquid film thickness of 400 µm. However, for liquid film thickness above 2000 µm, it is observed that the heat flux is independent of the angle of inclination of the test surface. NOMENCLATURE

u = Velocity in x-direction v = Velocity in y-direction Um = Mean film velocity Ts = Temperature of plate surface To = Free-stream temperature ρo = Free-stream density P = Pressure r = Liquid to vapor density ratio g = Acceleration due to gravity ρl = Density of liquid film

µl = Dynamic viscosity of liquid δ = Liquid film thickness ρv = Density of vapor kl = Thermal conductivity of liquid hfg = Latent heat of vaporization Vn = Evaporating film velocity q = Heat Flux β= Volumetric thermal expansion coefficient µФ`= Viscous dissipation term INTRODUCTION

Previous research works in spray cooling have identified Critical Heat Flux (CHF) to be highly dependent on the liquid film thickness on the test surface. Previous researchers have therefore adopted different mechanisms to obtain thin liquid film on the test surface to enhance heat transfer coefficient and critical heat flux. Zhang et al. [1], experimentally investigated liquid film flow down a heated and/or cooled vertical plate using an infra-red imaging camera. They observed that for a heated plate, the flowing liquid film contracted reducing the wetted surface area; whilst for a cooled plate, the liquid film extended thereby increasing the wetted surface area. The contraction and/or extension of the flowing liquid film on the heated and/or cooled surface were attributed to lateral Marangoni effect due to a heating temperature difference and the liquid flow rate.

Toda [2-3] reported in his experimental investigation that heat flux is a function of the wall superheat and thin liquid film thickness that forms on the test surface. He observed that for a critical thin liquid film thickness on the test surface, direct liquid evaporation into the vapor phase ensues without bubble generation. However, for film thickness beyond the critical film


thickness, nucleate boiling was prominent and heat flux was reduced due to vapor patches on the test surface which decreased the wetted surface area. In a spray cooling experiment by Yang et al. [4-5], film thickness measurement of 85-235 µm was reported. It was reported that the film thickness augmented the heat transfer mechanism; however, the heat flux magnitudes were not reported in their paper. Model simulation of fluid droplet impact on vapor bubble growth and bursting in a thin liquid film of thickness 44.17 µm was conducted by Selvam et al. [6-8]. Their study showed that droplet impingement during bubble nucleation increased mixing of the thin liquid film on the heater surface which also increased the heat flux.

Martinez-Galvan et al. [9], conducted experiments on film thickness measurements using high speed camera equipped with a long distance microscope. In their investigation, they found out that there exists a relation between the variation in the average Nusselt number and the film thickness along the spray cooling boiling curve. They indicated that the heat transfer regimes along that curve are related not only with a variation in the average Nusselt number but also with changes in the film thickness. In the work of Gong et al. [10], micro-conductive probes and confocal optical sensors were used to measure the instantaneous film thickness in an isothermal flow over a silicon wafer to obtain the film thickness profile and the interfacial wave characteristics. The dynamic thickness of an evaporating film on a horizontal silicon wafer surface was recorded using the optical sensor in their experiment. Their results indicated that a critical film thickness (84 µm) initiated film instability on the silicon wafer at heat flux of 56 kW/m2 (5.6 W/cm2).

Bhattacharya et al. [11], theoretically estimated heat transfer involved in spray evaporative cooling from single droplet studies viewpoint with the notion that a spray is equivalent to a multi-droplet array of liquid at low spray flux density. They developed an analytical expression of droplet evaporation time from fundamental heat transfer perspective to estimate strip cooling rate. Their analytical model developed predicts that it is possible to achieve an anomalously high strip cooling rate of Ultra Fast Cooling in a 4 mm thick steel strip by spray evaporative cooling provided the fluid droplet size is reduced to 70 μm. They also observed that smaller droplets were capable of providing the increased cooling load of Ultra Fast Cooling for thicker steel strips. Adiabatic and diabatic film thickness measurement was conducted by Pautsch et al. [12] in a spray cooling experiment using FC-72 as the working fluid. In their experiment, they observed that regions of the test die that exhibited the poorest heat transfer performance had the thickest liquid film. The reduced heat transfer performance was attributed to vapor patches in the thick liquid film.

Review of the literature indicates that the liquid film thickness on the test surface is one of the major parameters that affects critical heat flux and heat transfer coefficient in spray cooling experiments. However, not enough work has been done on the optimal liquid film thickness that will result in a maximum heat flux and enhanced heat transfer coefficient. The

present work is, thus, focused on liquid film thickness on the test surface and how it affects heat flux. MODEL FORMULATION

In this work, a model analysis of a thin liquid film flow over a heated flat surface is considered. The flat plate is subjected to a heat flux ( q ), with liquid flow at free-stream

conditions (To, ρo, µl) on the surface. The plate is inclined at an angle, θ. Figure 1 below shows the representation of the model.

Figure 1. Model representation of thin liquid film

flow down an inclined heated plate The present formulation is based on ultra thin (µm) liquid film flow over an inclined heated surface. The mean non-wavy film thickness (δ) model is used. As the thin liquid film flows on the heated surface, heat transfer from the metal plate to the liquid in contact with the plate is by pure conduction. At the liquid-vapor interface, the heat transfer is by latent energy due to phase change of the liquid into the vapor phase. Thus, evaporating thin liquid film without nucleate boiling is considered in the present analysis. For the present analysis, evaporation of the thin liquid film ensues such that the liquid film recedes at a velocity Vn between the liquid and vapor inter-phase. Thus the wavy hydrodynamic instability which occurs for liquid flow down an inclined plane is mitigated due to the liquid film evaporation. The continuity, momentum and energy conservative laws are applied to obtain the fluid transport properties for the evaporating liquid film. The effect of the liquid film thickness on critical heat flux is the specific objective of the present analysis.


ANALYTICAL CONSIDERATION The conservative transport equations describing the

velocity and temperature field for a laminar liquid flow over a flat plate is considered with the following approximations and boundary conditions:

1. No slip condition at the wall: 0)0()0( ==== yvyu

2. No penetration effects: 0),( =yxv

3. Velocity boundary layer approximations:

2

2

2

2

;x

u

y

u

x

u

y

u

∂∂>>

∂∂

∂∂>>

∂∂

4. The body force in x-momentum equation for an inclined place: )sin(θρ g

5. Viscous dissipation term is negligible: 0=Φμ

6. Constant fluid properties except density which depends on temperature

7. No heat generation inside the fluid: 0=genQ

8. Thermal boundary layer approximations:

2

2

2

2;

x

T

y

T

x

T

y

T

∂

∂>>

∂

∂∂∂

>>∂∂

9. Phase change at the liquid vapor inter-phase is by latent energy:

nVfghlyq ρδ == )(

10. Heat flux at the wall:

0=∂∂−=

yy

Tkq

With the above assumptions and the boundary layer

approximation for the pressure gradient, gvx

P ρ=∂∂

; the x-

momentum equation reduces to:

(1) ))sin((

2

2

l

vg

y

u

μρθρ −

−=∂

∂

The velocity profile in the flow for a film thickness (δ ) is thus obtained as:

(2) 2

1))sin((

)(

−

−=

δμδρθρ y

l

yvgyu

The mean velocity of the falling liquid film is obtained by integration of the velocity profile over the film thickness (δ ). The mean film velocity is obtained as:

(3) 3

2))sin((

l

vgmU

μδρθρ −

=

Using Boussinesq approximation for density on temperature dependence for an evaporating film, a temperature function is obtained as:

(4) 12

3

)sin(

11

1

+

+−= oT

vg

mU

rT l

ρδ

μθβ

where: , ratiodensityr

v

o

ρρ

=

THIN LIQUID FILM EVAPORATION

The current knowledge base has indicated that liquid film evaporation is associated with higher heat fluxes than nucleate boiling. Vapor patches on the surface of test specimen in nucleate boiling heat transfer reduce critical heat fluxes. Thus to maximize critical heat flux in spray cooling, liquid film evaporation is ensured. Using the zero film thickness model/assumption for maximum heat flux, the liquid film instantaneously vaporizes (active thermal flashing) into the vapor state without bubble generation. In such case, the total heat flux is the sum of the sensible and latent energy terms. The total heat flux is thus expressed as:

(5) latentqsensible

qtotalq +=

Noting that the phase change occurs at the liquid-vapor interface ( δ=y ), the flux at the interface following Stefan’s

condition [13] is expressed as:

(6) )(δ

ρ=∂

∂−=yy

TvklknVfghl

Combining the latent energy term and the sensible energy using the temperature function for the thin liquid film evaporation, the total heat flux is obtained as:

( ) (7) 12

3

sin

11

1nVfghl

vg

mUlroTsTlk

q ρρδ

μθβδ

+

+−−−−=


RESULTS AND DISCUSSION We have obtained results with the above analysis using

water as the working fluid Figure 2 shows the effect of the liquid film thickness on the maximum heat flux that can be obtained for different plate inclinations. From Figure 2, it is observed that highest maximum heat flux is obtained for a vertically oriented flat plate with liquid film thickness of about 0.4 mm.

Figure 2. Effect of liquid film thickness on Heat Flux for different angle of inclination

Figure 2 also shows that lower heat fluxes are obtained with decreasing angle of inclination. It is noted that for all angles of inclination, a critical film thickness is reached where a maximum heat flux is obtained. With film thickness higher than the critical film thickness, heat flux is mitigated. Thus for practical applications where maximum heat fluxes are to be obtained, an optimal film thickness should be maintained on the heated surface to ensure efficient film evaporation.

Figure 3 shows the effect of the mean film velocity on

the liquid film thickness on the heated test surface. From Figure 3, it is observed that increasing the film velocity increases the film thickness. For an inclined plate at an angle of 10o, the critical film thickness is obtained at mean velocity of about 0.25 m/s. For vertical and other inclination angles above 10o, maintaining mean velocity 0.3-0.5 m/s will ensure that optimal thin liquid film ensue on the heated surface.

Figure 3. Film thickness versus mean liquid film velocity

CONCLUSION Evaporating thin liquid film on a heated flat surface

has been considered in this study. An expression for heat flux as function of film thickness based on film evaporation has been derived. From the present investigation, it is observed that a maximum heat flux can be obtained when an optimal thin liquid film thickness is maintained on the test surface. Maximum heat flux of 800 W/cm2 was obtained for 0.4 mm liquid film thickness on a vertically oriented heated surface. It is also observed that for the same liquid film thickness, the maximum heat flux decreases with decreasing angle of inclination of the heated test surface.

ACKNOWLEDGEMENT This work was supported by Air Force Research Laboratory (AFRL) through Universal Technology Corporation. REFERENCES

[1]. Zhang Feng, Y.T. Wu, and J. Geng: An investigation of

falling liquid films on a vertical heated/cooled plate. International Journal of Multiphase Flow, 2008. 34(1): p. 13-28.

[2]. Toda S: A study of mist cooling (1st report: investigation of mist cooling). Trans Jpn Soc Mech Eng, 1972. 38:581–8.

[3]. Toda S: A study of mist cooling (2nd report: theory of mist cooling and its fundamental experiments). Trans Jpn Soc Mech Eng, 1973. 39:2160–93.


[4]. Yang J, Pais M, and L. Chow: Critical heat flux limits in secondary gas atomized liquid spray cooling. Exp Heat Transfer, 1993. 6:55–67.

[5]. Yang, J., L. Chow, and M. Pais: Nucleate boiling heat transfer in spray cooling. ASME J Heat Transfer 1996. 118:668–71.

[6]. Selvam R.P., L. Lin, and R. Ponnappan: Computational modeling of spray cooling: current status and future challenges. In: Space technology and applications international forum (STAIF 2005), February 13–17, Albuquerque, NM;, 2005. 746. p. 56–63.

[7]. Selvam RP, L. Lin: Computer modeling of liquid droplet impact on Heat transfer during spray cooling. In: ASME summer heat transfer conference, July 17–22, San Francisco, CA., 2005. HT2005-72569.

[8]. Selvam R., L. Lin, and R. Ponnappan: Direct simulation of spray cooling: effect of vapor bubble growth and liquid droplet impact on heat transfer. International Journal of Heat Mass Transfer, 2006. 49:4265–78.

[9]. Martinez-Galvan E, J.C. Ramos, and R. Anton: Film Thickness and Heat Transfer Measurements in a Spray Cooling System With R134a. Journal of Electronic Packaging, 2011. 133(1).

[10]. Gong S. J, W.M. Ma, and T.N. Dinh: Diagnostic techniques for the dynamics of a thin liquid film under forced flow and evaporating conditions. Microfluidics and Nanofluidics, 2010. 9(6): p. 1077-1089.

[11]. Bhattacharya P, A.N. Samanta, and S. Chakraborty: Spray evaporative cooling to achieve ultra fast cooling in runout table. International Journal of Thermal Sciences, 2009. 48(9): p. 1741-1747.

[12]. Pautsch A.G. and T.A. Shedd: Adiabatic and diabatic measurements of the liquid film thickness during spray cooling with FC-72. International Journal of Heat and Mass Transfer, 2006. 49(15-16): p. 2610-2618.

[13]. Shyy Wei, H.S. Udaykumas, and M.M. Rao: Computational Fluid Dynamics with Moving Boundary. Dover Publications, Inc, Mineola, New York.



November 4–5, Atlanta, Georgia USA

SIMILARITY SOLUTION FOR FLOW OVER A WEDGE WITH PHASE CHANGE

Yacob M. Argaw and John P. Kizito1

North Carolina A&T SU, Mechanical Engineering Department Greensboro, NC, USA

1Assistant Professor, North Carolina A&T State University, Mechanical Engineering Department, 1601 East Market Street, Greensboro, North

Carolina, 27411, [email protected]

ABSTRACT Flow of dry air past a heated wedge, traditionally known

as the Falkner-Skan problem, can be solved using similarity

approach. However, if a thin liquid film layer of thickness ∆δ

covers the surface of the wedge, phase change process becomes

a major mode of heat transfer as the fluid takes a large amount

of heat flux from the heated surface. In this paper, we have

developed a similarity solution for flow over a wedge covered

with liquid film layer using similarity analysis. The liquid

film is treated as secondary phase and is added to the original

energy equation as a source term. Our analysis shows for the

problem to have a similarity solution the wedge angle

parameter, m, and temperature distribution parameter, n,

should be related as 2n=m-1. The temperature profile plot and

the heat flux distribution shows for low Jacob number the heat

transfer rate is enhanced as a result of the phase change.

NOMENCLATURE u,v = velocities in x and y direction (m/s)

U,V = characteristic velocities along x and y

Vi = velocity of liquid to fill the film

ρ = density (kg/m3)

k = thermal conductivity (W/mK)

L,h = characteristic lengths along x and y

T = temperature (K)

S = Source term

Re = Reynolds number

Pr = Prandtl number

Ja = Jacob number

t = time (s)

q = heat flux (W/m2)

T ∞ = ambient temperature (K)

Ts = surface temperature (K)

hfg = latent heat of vaporization (kJ/kg)

Ue = jet free stream velocity, (m/s)

m, β = wedge angle parameters

n = temperature distribution parameter

δ = liquid film layer thickness (m)

η = similarity variable

ψ = stream function

Θ = dimensionless temperature

INTRODUCTION The temperature and velocity flow field for flow over a

wedge, also known as the Falkner-Skan problem, have been a

study topic due to its wide practical applications in

aerodynamics and hydrodynamics fields. Many researchers

like Cebeci et al. [1] Hartree [2] Elbashbeshy et al. [3] Lin et al

[4] and several others have numerically examined the Falkner-

Skan problem using different methods and wide range of

difficulty. Effects of dimensionless parameters like Reynolds

number and Prandtl number; variable viscosity and geometric

wedge angle [3, 5] have been the subject of investigation by

others. Okhotsimskii [6] studied the effect of Jacob number on

the shape and collapse time of a bubble which in turn affects

the heat transfer rate. He found out that when Ja=0 the radius

of the bubble as a function of time assumes a convex profile.

When 0<Ja<2 the curve becomes S-shaped, and when Ja>2

the bubble shape becomes concave. Thus the equation that

describes bubble radius through time changes for these

different ranges of Jacob number. The majority of phase

change studies consider a Jacob number greater than 2, the

present paper also considers Ja>2.

Cooling a wedge surface by a gaseous fluid and

incorporating phase change process to take away a large

amount of heat load maximizes the cooling rate. In particular,

many applications utilize blowing air impingement on a

surface as the easiest way of cooling. We envision a technique

where a constant thickness layer of liquid film on top of the

wedge surface can improve the cooling abilities. The liquid

film will remove heat as it changes its phase to vapor and the

generated vapor will be removed by the air free stream. Phase

change process will not only improve the heat transfer process

but also reduces the surface temperature of the wedge. In the

present paper, we use the Jacob number to measure the effect

of phase change on the flow over a wedge covered by a thin

liquid film. The Jacob number is the ratio of the two modes of

heat transfer used to compare sensible heat to latent heat.


PROBLEM FORMULATION The present study focuses on the effect of phase change on

the overall heat transfer mechanism. Heat from a hot surface

is transferred to the neighboring liquid layer via conduction.

When the temperature of the liquid layer reaches to its

saturation temperature, a portion of liquid evaporates and

removes a large amount of heat from the surface. The

generated vapor is carried away by convection via the moving

fluid.

Free stream of air approaches a wedge surface with

constant velocity U as shown in Figure 1. The wedge angle is

represented as βπ and its half angle is measured as the wedge

is aligned symmetrical to the horizontal axis. Liquid film of

thickness is ∆δ and it covers the surface at all time. Normally,

the liquid film thickness reduces as bubbles are generated and

carried away by the free stream. But in the present problem

formulation, the same amount of liquid that leaves the surface

is added to keep the liquid film thickness constant at ∆δ. The

process is akin to the way high temperature gas turbine blades

are cooled using film cooling technique by liquid aspiration.

Analytical methods to these problems are difficult because

multiple phases are involved and the fluid flow is coupled to

boundary layer and the surface geometry.

We search for a similarity solution given the assumption

that the film layer thickness is kept constant. The fluid

thickness can be constant when the vapor generation is equal

to the aspired fluid for a given heat flux. The other

assumption is that the presence of the liquid layer does not

induce an aerodynamic loss and boundary layer distortion of

the dry air free stream. The dry air free stream instead of

being in contact with the surface is now flows over the liquid

film. Thus the velocity and temperature boundary layers

extend from the surface of the liquid film rather than from the

heated surface. The presence of the liquid film is expected to

enhance the heat transfer process as phase change takes away a

considerable amount of heat. The problem formulation is

summarized in Figure 1 with a schematic which represent the

process.

SIMILARITY SOLUTION APPROACH

Apparently the three general flow equations for the free

stream are transformed to their non-dimensional forms and

then normalized in order to make sure the maximum value in

the velocity and temperature distribution calculation is one of

order unity. For flow over a wedge (Falkner-Skan problem)

the Navier-Stokes equation is modified since the inertial term

is much more dominant than the viscous component.

Therefore continuity, momentum and energy equations of air

flowing over a wedge are given respectively as:

0=∂

∂+

∂

∂

y

v

x

u (1)

2

21

y

u

xd

dP

y

uv

x

uu

∂

∂+−=

∂

∂+

∂

∂ν

ρ (2)

Sy

Tk

y

Tv

x

TuC p +

∂

∂=

∂

∂+

∂

∂2

2

ρ (3)

S is the source term that represents the contribution of the

liquid film layer on the heat removal as it takes a heat

equivalent to the fluid’s latent heat of vaporization. The

source term is constant given that the film thickness is kept at

unvarying level by adding equal amount of liquid that leaves

through phase change.

( )t

ht

hS fg

fg

∂

∂=

∂

∂=

δρ

δρ (4)

Thus the time derivative of the film thickness in Eq. (4) is

the same as rate of pumping additional fluid to the surface, Vi,

which is a value that depends on different parameters. The

velocity, Vi, is to be determined later.

Note that the coordinate system is attached along the

incline plane, as shown in Figure 1. No slip condition

demands a zero velocity at the wall and far from the wall the

velocity become that of the free stream. Thus the conservation

equations, Eq. (1-3), are subjected to the boundary conditions

given in Eq. (5) and Eq. (6).

0,0 === vuy (5)

∞==∞→ TTUuy e;, (6)

Heat flux applied to the wedge is transferred through

conduction across the phase change material and the boundary

layer. Therefore at the wall the heat flux can be written as:

fgiwall hVy

Tkq ρ+

∂

∂= | (7)

For incompressible continuous flow with free stream

velocity distribution Ue, the pressure gradient term can be

related to the free stream velocity. Thus the momentum

equation is modified to the expression in Eq. 8.

2

2

y

u

xd

dUU

y

uv

x

uu e

e∂

∂+=

∂

∂+

∂

∂ν (8)

LetL

xx = ,

U

uu = ,

h

yy = , and

V

vv = (9)

Figure 1. Flow over a wedge with thin liquid film


Normal to the surface the velocity V and the boundary

layer height h are defined in a way that would simplify the

scaling process as follows

2/1Re

Lh = and

2/1Re

UV = (10)

Substituting Eq.9 and Eq. 10 insures that continuity is

conserved. Thus the normalized and non-dimensionalized

momentum equation becomes:

2

2

y

u

dx

dUU

y

uv

x

uu e

e∂

∂+=

∂

∂+

∂

∂ (11)

The free stream velocity is allowed to vary along the

wedge surface, where the exponent, “m”, depends on the

wedge angle related to it as m = β/(2-β). For Falkner-Skan

problem the half angle coefficient can range from zero (flow

over a flat plate) to one (stagnation flow). As a result the value

of m similarly varies between zero and one.

( )m

eCxU = (12)

To find a self-similar solution for the Navier-Stokes

equation the following transformation equations are defined.

( ) fCxx m2

11),( +=ηψ (13)

( )2

11−= mCxyη (14)

The stream line equation is defined in terms of x and a

single variable function f. The derivatives of f are given as:

oU

u

d

dff ==

η' ;

2

2

''ηd

fdf = ;

3

3

'''ηd

fdf = (15)

The constant that multiplies the similarity variables can be

ignored in this analysis since it will disappear at the final

expression of the conservation equations. Apparently, the

transformation of the momentum equation from the x-y plane

to x-η plane is done using chain rule relation. The velocities

and their partial derivatives are converted to the corresponding

similarity expressions as follows:

( ) '2

1

2

1 12

1

yfxm

fxm

v m

m

−

−

−+

+=− (16)

( ) 'fxy

um=

∂

∂=

ψ = 'fUe (17)

( ) ''2

1' 2

33

12

yfxm

fxmyxx

um

m

−+=

∂∂

∂=

∂

∂−

−ψ (18)

''2

13

2

2

fxyy

um

=

∂

∂=

∂

∂−

ψ (19)

( ) '''12

3

3

2

2

fxyy

u m−=∂

∂=

∂

∂ ψ (20)

Then these expressions can be substituted into N-S

momentum equation. After simplification and rearrangement

the non-dimensional normalized momentum equation

becomes:

( ) ''.2

11'''' 2 ff

mfmf

+−−= (20)

Their corresponding boundary conditions are:

( ) 00 =f ( ) 00' =f (21)

( ) 1' max =ηf (22)

The energy term can also be scaled in a similar fashion.

Therefore, normalizing and non-dimensionalizing gives the

following equation.

∗+

∂

∂=

∂

∂+

∂

∂

V

V

U

L

Jay

T

y

Tv

x

Tu i

2

2

21

RePr

1 (23)

The last term, which is the source term due to phase

change, has a Jacob number associated with it. The

normalizing velocity for the rate at which the liquid film is

added on the surface can be defined in a way that would

simplify the energy equation further. Therefore defining

ULV /2=∗ the energy equation becomes:

iVJay

T

y

Tv

x

Tu

1

RePr

12

2

+∂

∂=

∂

∂+

∂

∂ (24)

The non-dimensional temperature is defined as:

s

s

TT

TT

−

−=

∞

θ (25)

Thus temperature distribution can be expanded and

rearranged as

( )( )θ−−−= ∞ 1TTTT ss (26)

( )( )θ−−+=⇒ ∞∞ 1TTTT s (27)

The temperature difference of the heated surface to the

free stream is a function of heat flux and considered to have a

profile given by the following expression, where the term, “n”,

is related to the applied heat flux.

( ) ( )n

sxTT =− ∞ (28)

Therefore partial derivatives of temperature are given as:

( ) '2

1

θ

−=

∂

∂−m

n xxy

T (29)

( )( ) ''1

2

2

θ−−=∂

∂ mn xxy

T (30)

( )( ) ( ) '2

1-1 2

3

1 θθ yxxm

xnx

Tm

nn

−−=

∂

∂−

− (31)

At the wall the applied heat flux has to be equated to the

heat taken away by conduction because of the temperature


gradient along the film plus the latent heat removed by the

phase change process.

ifgVhy

Tkq ρ+

∂

∂−='' (32)

ifg

mn

Vhkxq ρθ +=−

+

''' 2

1

(33)

Therefore solving for iV yields:

fg

mn

ih

xkqV

ρ

θ '.'' 2

1−+

−= (34)

The equation shows the rate at which additional fluid fill

the layer is dependent on the wedge angle parameter “m” and

the temperature distribution parameter “n”. Choosing a

relationship between m and n that would make the dependence

on x to vanish allows a similarity solution to the problem.

Therefore, for ( ) 02/1 =−+ mn and 10 << m a

similarity solution can be found for flow over a wedge covered

by thin liquid film. Once the dependence on x vanishes; since

all the other quantities are constant Vi can be expressed as a

function of yT ∂∂ / only as:

( )'θFVi = (35)

Then these expressions can be substituted into the energy

equation and the latent heat term can be redefined to obtain the

similarity solution. After simplification and rearranging the

general non-dimensionalized form of the energy equation

become

( ){ } '1

''RePr

1''1 θθθθ

Jaffn +−=+− (36)

The corresponding boundary conditions are:

( ) 00 =θ

( ) 10' =θ (37)

( ) 1max =ηθ (38)

RESULTS A Matlab code is written to solve the momentum and

energy ode. The wedge angle parameter m value is fixed to

1/3 (which equivalent to a half wedge angle value of 45o) and

heat flux parameter n =1/3 to satisfy the self-similar variables.

Air is considered as the free stream fluid with Pr =0.72 and

moving at a Reynolds number greater than at least 100. The

solution to the momentum equation gives the well studied

velocity distribution curve as shown in Figure 2. The

convergence of the velocity profile helps to validate the

stability of the code. As the momentum equation shows the

velocity distribution does not depend on the Jacob number.

Thus, the study focuses on solving the energy equation and

determining the thermal boundary layer profile.

The definition of the Jacob number implies that as heat

transfer through phase change in a flow system becomes

dominant the Jacob number gets smaller. For a flow with little

or no Latent heat term the Jacob number is a high quantity.

Figure 3 shows for a higher Jacob number, or heat transfer

predominantly through sensible heat, the dimensionless

temperature distribution converges in a very short boundary

layer thickness. The smaller the non-dimensional position

parameter η, the smaller the thermal boundary layer thickness

becomes. For a Jacob number greater than 30 the temperature

distribution approaches the free stream value at around η=1.0.

However, for flow conditions with smaller Jacob numbers the

boundary layer thickness reaches as high as 2.25, as shown in

Figure 3.

Figure 2. Dimensionless velocity distribution

Figure 3. Dimensionless Temperature distribution


The temperature and non-dimensional heat flux plots for

heat transfer without phase change process are compared in

Figures 3 and 4 respectively. The rate of change of the

dimensionless temperature distribution along the boundary

layer is presented in Figure 4. The peaks of heat flux curves

reduce as the Jacob number decreases. The results show that

the presence of a phase change material blunts the heat flux

intensity.

CONCLUSION Flow over wedges of certain angles superimposed with a

thin liquid film layer that incorporates a phase change

phenomena does have a self-similar solution. The similarity

analysis gives a solution when the wedge angle and the heat

flux rate definition satisfy the relationship

( ) 02/1 =−+ mn for 10 << m . Since this paper does

not study the effect of the wedge angle on the heat flux, the

parameters m and n are selected just to satisfy the condition

specified. Therefore, m=1/3 and n=1/3 are assumed. The

effect of the Jacob number on the heat flux and temperature

distribution is examined. The heat flux profile in the boundary

layer shows for low Jacob number the change in temperature

across the boundary layer thickness is much smaller compared

to a flow without phase change material. The temperature

boundary layer thickness for low Jacob number is much larger

than boundary layer thickness when no phase change material

is involved. The result shows a strong effect of Jacob number

on the heat flux distribution in the boundary layer.

REFERENCES [1] Cebeci, T. and Keller, H. 1971, “Shooting and Parallel

Shooting Methods for Solving the Falkner–Skan Boundary

Layer Equations,” J. Comput. Phys. 7, pp. 289–30

[2] Hartree, D.H., 1937, “On an Equation Occurring in

Falkner and Skan’s Approximate Treatment of the Equations

of the Boundary Layer,” Proc. Camb. Philol. Soc. 33 (Part II)

223–239.

[3] Elbashbeshy, E.M.A., and Dimian, M.F., 2002, “Effect of

Radiation on the Flow and Heat Transfer Over a Wedge with

Variable Viscosity,” Appl. Math. Comput. 132, pp. 445–454.

[4] Lin, H.T, and Lin, L.K, 1987, “Similarity Solutions for

Laminar Forced Convection Heat Transfer from Wedges to

Fluids of Any Prandtl Number,” Int. J. Heat Mass Transfer 30

1111–1118

[5] Graebel, W. P., 2007, “Advanced Fluid Mechanics,”

Elsevier Inc. USA

[6] Okhotsimskii, A.D., 1988, "The Thermal Regime of Vapor

Bubble Collapse at Different Jacob Numbers," Int. J. Heat

Mass Transfer 31(8): 1569-1576.

Figure 4. Temperature gradient along the Boundary Layer




SOLAR POWER GENERATION: METHODS AND COMPARISONS

Rambod Rayegan, Yong X. Tao

University of North Texas Denton, Texas, 76203

ABSTRACT Solar radiation has the highest capacity and the lowest

replenishment time among renewable energies. Available technologies to convert solar radiation to electricity can be studied in two different categories: Photovoltaic (PV) systems and solar thermal power systems. PV systems are employed for both building-scale and utility-scale applications. Solar thermal power systems have been mainly developed for the large scale applications. Low and medium temperature solar thermal power generation is at the research level and it is not commercialized yet. This paper presents a review on different solar power system technologies and pros and cons of each technology.

INTRODUCTION Fossil fuel depletion, atmospheric pollution, global

warming and ozone layer destruction are serious problems experts face in finding more sustainable ways to satisfy the requirements of human life. Today, the total energy consumption in the United States still predominantly comes from fossil fuels, although recent interests in investing in wind and solar electricity have been accelerating. Solar radiation has the highest capacity and the lowest replenishment time among renewable energies. Available technologies to convert solar radiation to electricity can be studied in two different categories: Photovoltaic (PV) systems and solar thermal power systems.

PV systems convert solar radiation into DC electricity using solar panels composed of a number of solar cells. Solar cells work based on photovoltaic effect which is the creation of voltage in a material, upon its exposure to light. PV panels which employ crystalline silicon modules are the main stream of the PV panel market. They are being partially replaced in the market by panels that employ thin film solar cells which are expected to account for 31 percent of the global solar panel market in terms of watts by 2013 [1]. PV systems are employed for both building-scale and utility-scale applications. Building-scale solar power systems, also known as distributed power systems, generate electricity locally for the building. They may be connected to the grid or may be stand- alone systems that need batteries or other electricity storage units.

Solar thermal power generation technology generally refers to a power generation system that involves collecting solar radiation through concentrated collectors to an absorber surface, which will heat a heat transfer fluid. Through a piping system and heat exchangers, the hot fluid will be able to generate vapor to power a turbine by means of a conventional or organic Rankine cycle. A generator connected to the turbine will then generate electricity. In general, such a system can be seen as similar to a coal-burning power plant except that the equipment component for vapor generation with coal combustion is replaced by a solar heating system.

Because of the limitation of solar irradiation and the efficiency of solar collectors, the conventional Rankine cycle is economically feasible only for large scale power plants. The Organic Rankine cycle (ORC) is a substitutive technology that is applicable for small scale power generation. ORC employs low grade heat from different sources such as biomass, geothermal, solar and waste heat of industrial processes. The main difference between ORC and the conventional Rankine cycle is in the working fluid. The boiling point of the working fluid in ORC is much lower than steam, hence there is no need to achieve high temperatures to generate vapor for running a micro-turbine or expander. As a result ORC can be driven at lower temperatures than the Rankine cycles that use water. ORC for use in residential and commercial buildings is at the research level [2-3] and it is not yet commercialized.

Currently, two types of high-temperature solar thermal power systems are in commercial-scale operation. One is parabolic trough technology, and the other is the tower configuration. The parabolic dish and Stirling engine arrangement is another technology that has not been developed enough yet to be available for large-scale operation. Flat plate, evacuated tube and compound parabolic concentrator (CPC) collectors are applicable for low and medium temperature ORCs.

This paper presents a review on photovoltaic and solar thermal power systems for both utility-scale and building-scale applications. The main components of each system and pros and cons of each technology will be discussed.


SOLAR CELLS

The main component of each photovoltaic system is the PV panel composed of a number of solar cells. Solar cells can be studied in two main categories: Crystalline silicon and thin film. Crystalline silicon is the most prevalent bulk material for solar cells and is separated into mono-crystalline silicon (c-Si) and poly-crystalline silicon (poly-Si). c-Si wafer cells are more expensive than poly-Si ones. They are cut from cylindrical ingots. poly-Si cells are made from cast square ingots. They are less expensive to produce than c-si silicon cells but they are less efficient.

A thin-film PV cell is a solar cell made by depositing one or more thin layers of PV material on a substrate [4]. The thickness range of such a layer is wide and varies from a few nanometers to tens of micrometers. The three main types of thin film solar cells are amorphous silicon (a-Si), cadmium telluride (CdTe), and copper indium gallium diselenide (CIGS). Other types include dye-sensitized and organic cells.

a-Si solar cells are fabricated on substrates such as low-cost soda lime glass, stainless steel, and polyimide. The plasma-enhanced chemical vapor deposition is used for the deposition process. A 12% to 13% efficiency range has been demonstrated for small-area laboratory a-Si solar cells now [4]. Some of the major challenges to produce a-Si solar cells are increasing the a-Si solar cell efficiency, reducing the light-induced changes in the devices, developing higher deposition rates for the microcrystalline bottom cell without compromising on the opto-electronic properties of the a-Si tandem devices (which could potentially have negative effect on the solar cell performance), and ultimately reducing the manufacturing cost.

Thin-film CdTe solar cells that have a perfect match with the solar spectrum are one of the most promising thin-film PV devices. Theoretical efficiencies for these devices are about 26%. Laboratory efficiencies of 16.5% for thin-film CdTe solar cell have been demonstrated by NREL scientists [5]. Several deposition processes have been developed for the growth of the absorber layer. Most of deposition processes result in 10% or higher efficiency for thin film CdTe solar cell. Five of these processes have demonstrated prototype power modules: close-space sublimation, electrodeposition, spray, screen printing, and vapor transport deposition. Recent developments are under way to produce CdTe solar modules with increased cell efficiency, standardization of equipment for deposition of the absorber layer, back-contact stability, reduction of absorber layer thickness to less than 1 μm, and controlled film and junction uniformity over a large area.

Thin-film copper indium selenide (CIS) is a direct band-gap semiconductor and has a band-gap of ~0.95 eV. When gallium (Ga) is added to CIS, the band-gap increases to ~1.2 eV, depending on the amount of Ga added to the CIGS film [4]. This material has demonstrated the highest total-area conversion efficiency for any thin-film solar cells. Several major developments for CIS modules are underway to increase

the CIGS solar cell efficiency to scale up the laboratory range of 19.3% to 19.9%, prevent moisture ingress for flexible CIGS modules, decrease absorber layers to less than 1 μm, and investigate CIGS absorber film stoichiometry.

Furthermore, developing approaches for using transparent wide-band gap in CIGS and alloyed CdTe to increase the cell efficiency up to 25% are topics of ongoing research and development [6]. For CIGS’s case, the relative Ga/In and S/Se compositions play the key role in changing the thin film band-gap. CIGS thin films can be prepared by thermal co-evaporation under different uniform and sequential processes to elucidate the film formation and composition control. For CdtTe solar cells, improving cell performance by controlling chemistry and materials processing during film deposition, post-deposition treatments, and contact formation has been the main focus.

UTILITY-SCALE PHOTOVOLTAIC SYSTEMS

There has been a surge of installing large-scale flat-panel PV array systems for utility power productions in the United States and globally during the last five years [4]. The capacity of a single plant has reached as high as 60 MW in Spain. Figure 1 shows the Wyandot Solar Energy Facility, a 12.6 megawatt (DC) solar PV facility located in Upper Sandusky, Ohio. Starting power production in May 2010, the system has 159,200 ground-mounted, thin-film solar panels on a 77-acre plot of land, the largest solar project in Ohio. Projects of similar size have been completed or are under construction in California, Florida, Nevada, Illinois, and other states.

Figure 1. Wyandot Solar Energy Facility in upper Sandusky, Ohio, USA (Courtesy of juwi solar, Inc. [7])

Large-scale ground based PV (LSPV) systems face a

fundamental question of land-use impacts. Denholm and Margolis [8] investigated this issue. When deployed horizontally, the PV land area needed to meet 100% of an average US citizen’s electricity demand is about 100 m2. This requirement roughly doubles to about 200 m2 per person when using 1-axis tracking arrays. By comparison, golf courses and airports each currently occupy about 35 m2 per person in the United States, whereas land used to grow corn for ethanol


production exceeds 200 m2 per person, although this land is concentrated in a fairly small number of states. They also pointed out another factor of disrupting local ecosystems. Deploying tightly packed PV arrays would create the most disruption but would require the least amount of land area. In contrast, the use of pole-mounted 2-axis arrays would require significantly more land but could be substantially less disruptive. Actual shading impacts of ground based PV arrays are also important. Ecosystem impacts of LSPV deployment need to be investigated to arrive at solutions for the growing shade-tolerant native and beneficial species under LSPV arrays. This also includes evaluation of best practices to minimize the use of herbicides and other chemicals and the use of installation and maintenance techniques to provide minimum impacts to the environment.

One of the alternative solutions to provide LSPV power generation systems is using concentrated photovoltaic (CPV) systems, which have a higher energy conversion efficiency than flat panels - currently 40% [9] - with potential up to 50 %. If deployed on a large scale, they will require less land to produce the same amount of energy. Although no proven commercial scale applications have been installed, the commercialization of this technology is accelerating.

Concentrated photovoltaic systems use either parabolic dish mirror systems or a large array of flat Fresnel lenses to focus energy on PV cells [4]. Dish and CPV systems are modular in nature, with single units producing power in the range of 10 to 35 kW. Thus, dish and CPV systems could be used for either distributed or remote generation applications or in large arrays of several hundred or thousand units to produce power on a utility scale. Dish and CPV systems have the potential advantage of mass production of individual units,

similar to the mass production of automobiles or wind turbines, yet they can be integrated into a utility scale solar power plant.

The key technological challenges for increasing the efficiency of CPV systems are optical elements and cell development. Since 2007, Spectrolab, Inc., has developed metamorphic multi-junction solar cells with an efficiency up to 40.7% and lattice-matched three-junction terrestrial cells with 40.1% efficiency. These efforts were partially supported by the High-Performance Photovoltaics program through DOE NREL [9]. The efficiency also benefits from high-band-gap, disordered GaInP top cells, and wide-band-gap tunnel junctions under the terrestrial solar spectrum at high concentration.

Under a contract by the US Department of Energy, SolFocus, Inc., has developed a 1.6-MW system using the technology illustrated in Figure 2 [10]. It possesses the following characteristics:

· Optical efficiency: 74% · Power unit efficiency: 27.0% · Module efficiency: 25.3% · Acceptance angle: greater than 1 deg. · Cell temperature: 50.8°C to 53.9°C · Module degradation: 1.2% per Kelvin The system uses primary and secondary mirrors (Figure

2A) and optical rods to focus the sun’s rays to highly efficient multi-junction PV cells. The combined mirrors, rod, and cells form a power unit, 20 of which are integrated into one panel. Twenty-eight such panels are then mounted and combined on a dual axis tracking support as shown in Figure 2B. Eventually, such a system can be configured to utility- scale capacity.

(A) (B)

Figure 2. Schematic of Concentrated Photovoltaic (CPV) arrays: (A) power unit, and (B) 20-unit panel and 28-panel tracker on a single pole [10]


The target of this development is to provide an operational capacity greater than 3 MW with module manufacturing cost less than $5/W. The developed automated assembly line was validated by a third party to have exceeded the cost target and was expanded to other manufacturing partners. The projected manufacturing capacity is 100 MW by the end of 2011 [11].

Neither the dish-Stirling nor CPV systems use storage or hybrid fossil capabilities to provide a firm resource, although CPV systems could, in principle, make use of battery energy storage. However, present battery storage technology is comparatively inefficient and cost prohibitive. A grid-tie back feed capability has to be in place in order to make economic sense in the short run.

BUILDING-SCALE PHOTOVOLTAIC SYSTEMS

Buildings have a significant impact on energy consumption. Buildings account for 40% of the energy used annually worldwide. Energy efficient design and quality construction can drive the cost of powering a home down by more than 50% [4]. But to become zero energy buildings, i.e., where the net annual electricity consumption from utility is zero, buildings and residences must incorporate some type of on-site energy generation. Photovoltaic systems, as the only commercialized solar electricity production technologies for buildings, are reasonably developed to the extent that numerous companies warranty their building integrated solar products for 20 to 25 years. Most of the recent building solar energy system manufacturers combine reliability, functionality, and aesthetics to remain in the solar market competition. Photovoltaic shingles and slates are examples of these modern products.

There is an enormous potential for deploying solar energy solutions on buildings. Today, there is enough residential and commercial rooftop space to site more than 500 GW of PV capacity, equivalent to placing 4-kW PV systems on more than 125 million homes [12]. Current US electric capacity is about 1000 GW.

A PV power system mainly consists of a few essential elements: PV panels with solar cells, inverters, and disconnects. Solar cells have been explained previously. PV panels produce direct current (DC) electricity in the same way as batteries. Inverters, which can be called the most complex parts of a residential PV system, take DC electricity and convert it to AC for powering typical household appliances.

The National Electrical Code [13] requires that each piece of PV equipment have disconnect switches in addition to the main feeder disconnect, allowing service providers to disconnect the equipment from all sources of power. Disconnects may be circuit breakers or switches. Some inverters have incorporated a disconnect switch into a box attached to the inverter. The number of disconnects in grid-tied systems changes with the system’s complexity. In a simple system, there would be disconnects on both sides of the inverter. In more complex systems, there may be disconnect switches for each string of arrays, sub-array disconnects, main PV disconnects, and for each inverter, sub-panel and AC disconnects.

Because solar energy is not available all the time, the continuous usage of electricity by buildings requires power to be drawn from either an energy storage facility charged by the building solar energy system during the production period or from the conventional electricity grid. Energy storage such as battery banks is expensive and inefficient unless they are used in remote areas where the utility grid is not available. The best solution is for PV systems to have an interconnection with the grid, a scheme of net energy metering that allows a PV system to export excess electricity to the utility and import it back when needed. Essentially, grid-tied systems use the grid as a battery. Most electric utilities in the United States have adopted regulations and guidelines to help design PV systems that operate in parallel with the utility systems. The grid-connected PV systems are the most common and simplest PV systems installed in houses which have been recommended by the Department of Energy [14].

Ground-mounted PV systems may be applicable for buildings with large land holdings mostly located in rural areas. Commonly, solar ground-mounted systems involve steel or aluminum frames attached to a concrete foundation. The lower edge of ground-mounted arrays should be high enough to clear vegetation and accumulated dirt or snow.

Either crystalline or thin-film PV technologies can be utilized in these systems. In cold climates of northerly latitudes, sun-tracking systems are used to maximize the solar energy harvest of ground mounted arrays. It is only worth installing trackers in regions with mostly direct sunlight. In diffuse light, tracking has no noticeable effect on solar gain by PV panels.

BUILDING INTEGRATED PHOTOVOLTAICS (BIPVs) Building-integrated photovoltaics are photovoltaic

materials that are used to replace conventional building materials in parts of the building envelope such as the roof, skylights, or facades [16].

BIPVs are separated into roof-mounted and non-roof-mounted BIPV systems. Roof-mounted BIPVs include cement tile systems, thin-film PV laminate for standing seam metal roofs and shingle systems [4]. Ground-mounted PV systems are more suitable with respect to rooftop systems for high-wind regions.

A cement tile system is the most common type of BIPV system. In this system, some cement tiles are replaced by PV panels that are sized and mounted with the same overall dimensions (Figure 3). Usually, PV panels are lighter than the cement tiles they replace; therefore, structural assessments to add any supporting structure should be conducted before installing PVs. Photovoltaic panels follow the contour of the roof in exactly the same way as the cement tiles. Since PV tiles are replacing roofing materials, they should be compliant with local and national roofing regulations. Electrical connections are made between each tile. In some cases, a single module will replace a set of three or four tiles, reducing the number of connections.

Thin-film PV laminate for standing seam metal roofs is common and easy to install with no additional structural


support required. Solar laminate PV is installed on metal raised-seam roofs between the seams as shown in Figure 4. It is directly attached with glue during installation of the roof without replacing the metal on the roof. The roof ridge, as an easily accessible place for maintenance, is used for electrical wiring and connections. Since they have no glass components, they have higher durability than other types of BIPVs. Thin-film PV laminates are approximately half as efficient as the mono- or multi-crystalline modules, but also currently cost about half as much.

Figure 3. Cement tile BIPV system [15]

Figure 4. Thin-film PV laminate for standing seam metal roofs [15]

Photovoltaics can replace shingles in two ways. In the first method, shingled roofs take advantage of thin-film PV. This lightweight plastic replacement for shingles comes in relatively long strips to replace courses of asphalt shingles and reduce the number of electrical connections. In the second method roof, shingles powered by solar energy (PV shingles) serve to replace ordinary shingles. Electrical lead wires extend from the underside of each shingle and pass through the roof

deck, allowing interior roof space connections. The sun’s heat helps bond the shingles together, forming a weather-resistant seal.

South-facing walls have the potential to generate electricity using BIPVs. They can be covered in PV vertically or can be slanted to act as PV and window shading at the same time. A filtered window or filtered skylight can be made by setting thin-film PV between two sheets of tempered glass. This type of PV system may cost the owner three times more than regular PVs, but it can make a powerful architectural impact.

Mono- or poly-crystalline PV can also be set between sheets of glass to create a dappled effect, blocking the majority of sunlight to make electricity, but allowing shaded light through. This reduces solar gain to the interior of the building while producing electricity.

UTILITY- AND BUILDING-SCALE PHOTOVOLTAIC SYSTEMS COMPARISONS

The cost per watt of installing and operating distributed PV systems is higher than the costs of similar utility systems because of the economics of scale. In the case of a new, large building where the builder might choose to integrate and install PV systems at the time of construction installation, costs may be low, but, in general, there are added costs associated with rooftop systems; especially for retrofits, there can be permitting, architectural, structural, and electrical issues. On the other hand, distributed systems avoid transmission losses through delivering power where it is needed, and residential and commercial systems can be financed along with the rest of a building [4].

Results of NREL-conducted research [15] show that the installation and operation of PV systems for buildings may cost more than twice the installation and operation cost in utility approaches with the same capacity. Figures 5 and 6 illustrate a comparison of rooftop and field PV systems. Concentrating PV has often been presented as a lower-cost approach to utility scale power and could be a major player in this market.

The kilowatt rating in Figure 5 was determined from DC electricity generated at 1000 W/m2 for flat-plate or 850 W/m2 for concentrator systems. The AC rating was assumed to be 85% of the DC STC rating. There are several reasons for lower performance of fixed rooftop systems with respect to CPV systems. In most rooftop systems, PV panels are installed in a horizontal configuration that is not an optimal position for year-round sunlight. The efficiency of a solar cell decreases by increasing its temperature. If the temperature of the panel increases because of poor ventilation from close contact with the roof, performance of the panel is negatively affected. In practical cases, panels may be shaded by a nearby building or tree.

In Figure 6 relative costs, kWh electricity output per installed kWac, and kilowatt-hour electricity output per dollar spent was compared for rooftop and field PV systems. In this analysis, tracked PV systems represent utility-scale solar systems. Results show that the field PV systems generated nearly twice as many kWh per dollar spent, compared with the


rooftop PV systems. Therefore, given a fixed budget, the field approach results in higher electricity generation in comparison to the retail building-based approach. This also implies that public support for utility market incentives would yield a higher rate of return for the investment.

Figure 5. kWhs generated per installed kWac for single-axis tracked flat-plate, concentrator, and fixed rooftop

systems sited in Arizona [15]

Figure 6. Comparison of flat-plate rooftop and field system costs and average annual production for recent

installations in Arizona [15]

SOLAR COLLECTORS

The main component of each solar thermal power generation system is the solar collector that collects heat by absorbing solar radiation. The key parameter to select a solar collector for a solar system is the required temperature for that specific application. Solar collectors can be identified in three main categories: low temperature with the output temperature less than 85 °C, medium temperature with the output temperature below 130-150 °C, and high temperature solar collectors with the output temperature higher than 150 °C [2]. Low and medium temperature solar collectors are mainly used for heating purposes. They can also be employed in Organic Rankine Cycles (ORCs) to generate electricity for building-scale applications. The main types of solar collectors are described below.

A flat plate collector is a low temp solar collector. A typical flat-plate collector is a metal box with a glass or plastic cover (called glazing) on top and a dark-colored absorber plate (typically copper or aluminum) on the bottom. The sides and bottom of the collector are usually insulated to minimize heat loss [17]. Sunlight passes through the glazing and strikes the absorber plate, which heats up, changing solar energy into

heat energy. The heat is transferred to liquid passing through pipes attached to the absorber plate.

An evacuated tube collector is in the category of low- and medium-temperature solar collectors. The collectors are usually made of parallel rows of transparent glass tubes. Each tube contains a glass outer tube and metal absorber tube attached to a fin. The fin is covered with a coating that absorbs solar energy well, but which inhibits radiative heat loss. Air is removed, or evacuated, from the space between the two glass tubes to form a vacuum, which eliminates conductive and convective heat loss [17].

A compound parabolic concentrator (CPC) collector is in the category of medium- and high-temperature collectors. Because they are not focusing (non-imaging), they are a natural candidate to bridge the gap between the lower temperature solar application field of flat-plate collectors to the higher-temperature application field of focusing concentrators [4]. As shown in Figure 7, an ideal CPC collector can concentrate isotropic radiation incident on an aperture “a” within a solid angle θ, with the normal to this aperture without losses. The minimal possible aperture “b” on to which the radiation can be delivered is shown in Figure 7.

(A) (B)

Figure 7. Aperture of a CPC concentrator: (A) flat absorber parallel to the flat entrance aperture a and (B) acceptance angle as a function of half acceptance angle

θ [18] A full CPC can achieve the maximum concentration Cmax

within the angle θ, where Cmax = 1/sin(θ). In practice, cost reduction often yields a design of CPC with truncated configuration to eliminate the upper part of the mirrors. Studies have been underway since 1970 to achieve a high temperature between 150°C and 200°C. The most recent studies utilized evacuated tubes with CPC reflectors.

A parabolic trough collector is a high temperature solar collector. It is constructed as a long parabolic mirror (usually coated silver or polished aluminum) with a receiver tube running its length at the focal point. Sunlight is reflected by the mirror and concentrated on the tube. Heat transfer fluid (usually oil) runs through the tube to absorb the concentrated sunlight. This can increase the temperature of the fluid close to 400°C.


UTILITY-SCALE SOLAR THERMAL POWER SYSTEMS

Currently, two types of major solar thermal power systems have been in utility-scale operation. One is parabolic trough (or trough in short) technology, and the other is the tower configuration [4]. Parabolic dish with Stirling engine is another technology that has not been developed enough yet to be available for large-scale operation. PARABOLIC TROUGH SOLAR POWER TECHNOLOGY

Although many solar technologies have been demonstrated, parabolic trough solar thermal electric power plant technology represents one of the major renewable energy success stories of the last two decades. Trough technology is recognized as one of the lowest cost solar-electric power options available today and has significant potential for further cost reduction [4]. Nine parabolic trough plants, totaling over 350 megawatts (MW) of electric generation, have been in daily operation in the California Mojave Desert for up to 18 years [19]. These plants provide enough solar electricity to meet the residential needs of a city with 350,000 people. Several new parabolic trough plants have been built or are currently under construction or in the early stages of operation in support of solar portfolio standards in Nevada and Arizona and a solar tariff premium in Spain.

Parabolic trough power plants use concentrated sunlight, in place of fossil fuels, to provide the thermal energy required to drive a conventional power plant. These plants use a large field of parabolic trough collectors, made of low-cost, non-optical mirrors, which track the sun during the day and concentrate the solar radiation onto a receiver tube located at the focus of the parabolic shaped mirrors. A heat transfer fluid (HTF) passes through the receiver and is heated to temperatures required to generate steam and drive a conventional Rankine cycle steam power plant. The largest collection of parabolic systems in the world is the Solar Energy Generating Systems (SEGSs) I through IX plants in the Mojave Desert in southern California [19]. The SEGS plants were built between 1985 and 1991. All of the SEGS plants are “hybrids,” using fossil fuel to supplement the solar output during periods of low solar radiation. Each plant is allowed to generate 25% of its energy annually using fossil fuel. With the use of the fossil hybrid capability, the SEGS plants have exceeded 100% capacity factor (based on peak hours) for more than a decade, with greater than 85% from solar operation. The capacity factor of a power plant is the ratio of the actual output of the power plant and its full nameplate capacity over a period of time.

SOLAR POWER TOWER TECHNOLOGY

Power tower systems (also known as heliostat power plants) consist of a field of thousands of sun-tracking mirrors which direct insolation to a receiver atop a tall tower. A molten salt heat-transfer fluid is heated in the receiver and is piped to a ground based steam generator. The steam drives a steam turbine-generator to produce electricity [4]. Because

trough and power tower systems collect heat to drive central turbine generators, they are best suited for large-scale plants, 50 MW or larger [19]. Trough and tower plants, with their large central turbine generators and balance of plant equipment, can take advantage of economies of scale for cost reduction, as cost per kilowatt goes down with increased size. Additionally, these plants can make use of thermal storage or hybrid fossil systems to achieve greater operating flexibility and dispatchability. This provides the ability to produce electricity when needed by the utility system, rather than only when sufficient solar insolation is available to produce electricity, for example, during short cloudy periods or after sunset. This capability has significantly more value to the utility and potentially allows the owner of the solar power plant to receive additional credit, or payment, for the electric generating capacity of the plant.

In the summer of 2009, the first and only commercial power tower plant in North America, the Sierra SunTower Plant, commenced operations and interconnected to the grid. It is a 5 MW, commercial facility located in Lancaster, California. There are several more power tower systems around the world that have been built or were under construction during 2009-2010 with general confidence that uncertainty in the cost, performance, and technical risk of this technology is decreasing. A 2004 predictive analysis [20] shows that, assuming the technology improvements are limited to currently demonstrated or tested systems and a deployment of 2.6 GWe of installed capacity by the year 2020, tower costs could drop to approximately 5.5￠/kWh, or better than trough systems (which is predicted to be 6.2 ￠/kWh). However, the data to confirm this prediction has yet to come.

PARABOLIC DISH AND STIRLING ENGINE TECHNOLOGY

Parabolic dish systems use a dish-shaped arrangement of mirror facets to focus energy onto a receiver at the focal point of the collector. A working fluid such as hydrogen is heated in the receiver and drives a turbine or Stirling engine. Most current dish applications use Stirling engine technology because of its high efficiency [4]. The Sandia National Laboratories (SNL) of the US Department of Energy has been pursuing the aggressive deployment of 25-kW dish-Stirling systems for bulk power. The immediate objectives of the development are to: improve reliability and reduce cost of dish/engine

components and systems; test, evaluate, and improve performance of dish/engine

components and systems; and develop tools for industry to characterize their systems

and components. In 2009, the SNL development team, along with many

industrial partners, continued to operate, maintain, and improve the Stirling Energy Systems six-dish model power plant, cataloging more than 100 development areas. They developed a real-time mirror characterization prototype system for 100% inspection of mirrors on assembly line, engine simulator for development of modern engine control hardware and software. SNL and Infinia together further


conducted optical and systems design of a 3-kW free piston Stirling engine dish system, which is hermetically sealed and requires no maintenance. As of today, no commercially proved units or systems are available for wide deployment yet.

BUILDING-SCALE SOLAR THERMAL POWER SYSTEMS

Because of the limitation of solar irradiation and efficiency of collectors, the conventional Rankine cycle is economically feasible only for large scale power plants. The Organic Rankine cycle (ORC) is a substitutive technology which is applicable for small-scale power generation. The main difference between ORC and the conventional Rankine cycle is in the working fluid. The boiling point of the working fluid in ORC is much lower than steam, hence ORC can be driven at lower temperatures than the Rankine cycles that use water. The selection of working fluid and working conditions of the Organic Rankine Cycle (ORC) has a great effect on the system operation and its energy efficiency and impact on the environment.

In the authors’ previous works [2-3,21], eleven fluids were suggested to be used in solar ORCs designed for low- or medium-temperature solar collectors. The system requirements needed to maintain the electricity demand of a building have been compared for the eleven suggested fluids for two temperature levels of 85°C and 130°C. The building is a one-story commercial building with 395 m2 floor area located in downtown Pensacola, Florida, and is served by grid power.

The simulation results show that the best collector-temperature combination for supplying the building power is the low temperature evacuated tube solar collector. Cyclohexane and Isopentane with respectively 722.54 m2 and 728.16 m2 and Benzene and R245ca with 742.96 m2 required collector area are the best working fluids to be employed in the ORC system in order to maintain the power demand of the building in Pensacola. Isopentane is a more optimal choice for working fluid in comparison to Cyclohexane, Benzene, and R245ca when considering environmental and health issues

The lowest required collector areas to maintain the power demand of the building and the associated collector expenses are still high for a 395 m2 commercial building. The land price to establish the solar field should be added to the solar expenses of the system. Using Compound Parabolic Concentrating (CPC) collectors can be a solution to reduce the required collector area. CPC collector products have not been commercialized for public use as of yet. For this reason the high price of CPC collectors is the main barrier of entry for residential or commercial building applications.

PHOTOVOLTAIC AND SOLAR THERMAL POWER SYSTEMS COMPARISONS

Utility-scale. Among all the large-scale solar power systems, parabolic trough systems and power tower systems have proven to be best suited for large-scale plants of 50 MW or larger. Trough and tower plants, with their large central turbine generators and balance of plant equipment, can take advantage of economies of scale for cost reduction, as cost per

kilowatt goes down with increased size. Additionally, these plants can make use of thermal storage or hybrid fossil systems to achieve greater operating flexibility and dispatchability, accounting for situations in which sufficient solar insolation is unavailable. Large-scale flat panel PV array systems for utility power productions are also widely used because they are easy to scale up or down. When tied to the electric grids, they provide flexible, variable power generation options. However, their costs are higher compared with parabolic trough systems. Concentrating PV systems and parabolic dish systems with Stirling engine technology have relatively high efficiency. Dish and CPV systems are modular in nature, and a single unit could produce power in the range of 10 to 35 kW. They could be used for either distributed or remote generation applications, or in large arrays of several hundred or thousand units for a utility scale application. They also have the potential advantage of mass production of individual units.

The largest group of solar systems in the world is the SEGSs I through IX parabolic trough plants in the Mojave Desert in southern California, built between 1985 and 1991, that have a total capacity of 354 MW. These plants have generally performed well over their 15 to 20 years of operation. There are no operating commercial power tower or dish-Stirling power plants, although some commercial purchase agreements have been in place to pursue those options [1].

Building-scale. Rayegan [2] performed an economic comparison study between the low temperature solar ORC and PV panel system that maintains the electricity demand of a building located in Pensacola of Florida. Table 1 shows the required area and total cost for the suggested solar ORC system (employing low-temperature evacuated tube and Isopentane as working fluid) and PV panel system to maintain the power demand of the building in Pensacola. It can be seen for the suggested ORC system the required collector area to maintain the power demand of the building is more than 60 percent less than required PV panel area to maintain the same amount of power. The total cost to establish the suggested solar ORC system is more than 50 percent less than total cost of running a PV panel system to maintain the power demand of the building in Pensacola.

Table 1. Required area and total cost for the suggested

solar ORC system (employing low-temperature evacuated tube and Isopentane as working fluid) and PV

panel system to maintain the power demand of a building in Pensacola

System Required area [m

2]

Collector/ PV expense

[ x 1000 USD]

ORC package/Inverter

expense [ x 1000 USD]

Total cost

[ x 1000 USD]

Low Temperature Solar ORC

728.16 257.81 75 332.81

PV 1839.00 671.24 33.2 704.44


MAJOR FINDINGS

In this paper, a review of different solar power system technologies and the pros and cons of each technology have been presented. Major findings of the review can be listed as follows: By far, PV-driven applications are dominant in the present

markets in terms of annual productions of cells, panels, and systems and the number of installations in buildings and other applications.

The cost per watt of installing and operating distributed PV systems is higher than the costs of similar utility systems because of the economics of scale. On the other hand, distributed systems avoid transmission losses through delivering power where it is needed.

Among all the large-scale solar power systems, parabolic trough systems and power tower systems have proven to be best suited for large-scale plants of 50 MW or larger.

Large-scale flat panel PV array systems for utility power productions are also widely used because they are easy to scale up or down.

PV panels provide flexible, variable power generation options when tied to the electric grids. However, their costs are much higher compared with parabolic trough systems.

Concentrating PV systems and parabolic dish systems with Stirling engine technology have relatively high efficiency. Dish and CPV systems are modular in nature, and a single unit could produce power in the range of 10 to 35 kW.

Comparison between building-scale photovoltaic and solar thermal power system shows that there is a good potential for commercializing ORC systems for distributed power generation.

REFERENCES [1]http://www.renewableenergyworld.com/rea/news/article/2009/11/thin-films-share-of-solar-panel-market-to-double-by-2013. Retrieved September 1, 2011. [2] Rayegan, R., “Exergoeconomic Analysis of Solar Organic Rankine Cycle for Geothermal Air Conditioned Net Zero Energy Buildings,” Ph.D. Dissertation, Florida International University, 2011. [3] Rayegan, R., Tao, Y.X., “Analysis of Solar Organic Rankine Cycle for a Building in Hot and Humid Climate,” ASME International Mechanical Engineering Congress & Exposition, Denver, Colorado, November 2011. [4] Tao, Y. X., Rayegan, R., “Solar Energy Applications and Comparisons,” in Energy and Power Generation Handbook, Editor: K. R. Rao, Publisher: ASME Press, 2011 [5] Wu, X., 2004, “ High-efficiency polycrystalline CdTe thin-film solar cells,” Solar Energy, Vol.77(6), pp 803-814.

[6] Shafarman, W., McCandless, B., 2008, “Development of a Wide Bandgap Cell for Thin Film Tandem Solar Cells: Final Technical Report,” 6 November 2003 — 5 January 2007. 40 pp., NREL Report No. SR-520-42388. [7] Juwi solar Inc., http://www.juwisolar.com/wyandot-solar. Retrieved September 1, 2011. [8] Denholm, P., Margolis, R. M., 2008, “Impacts of Array Configuration on Land-Use Requirements for Large-Scale Photovoltaic Deployment in the United States,” SOLAR 2008 American Solar Energy Society (ASES), 3-8 May 2008, San Diego, California, pp. 1-7. NREL Report CP-670-42971. [9] King, R. R., 2010, “Ultra-High-Efficiency Multijunction Cell and Receiver Module, Phase 1B: High Performance PV Exploring and Accelerating Ultimate Pathways, Spectrolab, Inc., Subcontract Final Report,” NREL Report SR-520-47602. [10] Horne, S., McDonald, M., Hartsoch, N., Desy, K., 2009, “Reflective Optics CPV Panels Enabling Large Scale, Reliable Generation of Solar Energy Cost Competitive with Fossil Fuels, NREL Report” SR-520- 47310. [11] http://www.solfocus.com/en/news-events/press-releases/2009-09-09.php. Retrieved October 28, 2011. [12] Solar America Initiative–In Focus: The Building Industry, 2007,pp., NREL Report DOE/GO-102007-2389. [13] Wiles, J., 2005, “Photovoltaic Power Systems and the 2005 National Electrical Code: Suggested Practices.” [14] Baechler, M., Gilbride, T., Ruiz, K., Steward, H., Love, P., 2007, “High-Performance Home Technologies: Solar Thermal & Photovoltaic Systems,” Volume 6 Building America Best Practices Series. 159 pp., NREL Report No. TP-550-41085. [15] Kurtz, S., Lewandowski, A., Hayden, H., 2004, “Recent Progress and future potential for concentrating photovoltaic power systems: preprint.9,” pp. NREL Report No. CP-520-36330. [16] http://www.wbdg.org/resources/bipv.php . Retrieved September 1, 2011. [17]http://www.eere.energy.gov/basics/buildings/water_heaters_solar.html Retrieved September 1, 2011. [18] IEA., 2008, “Process Heat Collectors, State of the Art Within Task 33/IV,” edited by Weiss, W., and Rommel, M., Report by Solar Heating and Cooling Executive Committee of the International Energy Agency (IEA), published by AEE INTEC, Gleisdorf, Feldgasse 19, Austria, 2008. pp. 1-55. [19] DOE NREL (2006). Parabolic Trough Solar Thermal Electric Power Plants. Publication FS-550-40211; DOE/GO-102006-2339. [20] Sargent and Lundy LLC. (2003). Assessment of Parabolic Trough and Power Tower Solar Technology Cost and Performance Forecasts, NREL Report SR-550-34440. [21] Rayegan, R., and Tao, Y. X., 2011, “A Procedure to Select Working Fluids for Solar Organic Rankine Cycles (ORCs),” Renewable Energy, Vol. 36 (2), pp. 659-670.




FUZZY-LOGIC HEALTH MONITORING ROBOTS FOR THE ELDERLY

Wuqayan Alwuqayan, Sabri Tosunoglu Department of Mechanical and Material Engineering

Florida International University Miami, Florida, U.S.A

ABSTRACT

This research paper ventures through a cutting-edge

technology behind personal service robots that are aimed

towards the elderly population. The technology uses sensors

that measure several characteristics about the habits of elderly

people and provide vital feedback on their health. The sensors

are specially designed to fulfill their objectives which may

range from face recognition and/or detecting people’s motion

all the way to measuring their pulse and body temperature. The

technology will enable the service robots to create personal

health profiles of their owners. Through pattern recognition the

robots then can detect any abnormality or irregularity in their

owners’ health or habits and act accordingly to prevent

potential health mishaps.

Key element about this new breed of robots is their

controller system which allows them to mimic human

reasoning in handling complex nonlinear systems. There are

number of methods to conduct such kind of controllers.

However when dealing with imprecise input signal in real-time

operations, fuzzy-logic controller proven to be the best suit for

the job. A detailed research work in fuzzy-logic controller is

compiled by a number of researchers and there have been many

books and articles on the subject as reported in [1-4]. For

simulation purposes a fuzzy logic controller was constructed

using Fuzzy Logic Toolbox provided by MATLAB software, a

performance test of the system was carried out, and the results

were analyzed. Simulation and analysis provide designers with

information on how the fuzzy-logic controller can affect the

output speed and stability of a system with multivariable inputs.

Thus, designers can proceed with a larger and more complex

system. The objective behind this technology is to improve life

quality of the elderly.

INTRODUCTION

The population of people age 65 and older is increasing

rapidly. According to the 2010 U.S. census 13% of the

population is over 65, The U.S. Census Bureau projects that

number will increase to 19.7% in 2030. Current living

conditions for the majority of elderly people are already

alarmingly unsatisfactory. According to US Department of

Health and Human Services, nearly 9% of non-institutionalized

persons 70 years of age and over were unable to perform one or

more activities of daily living such as bathing, dressing, using

the toilet, and getting in and out of bed [5]. At the same time,

the health-care sector is undergoing a dramatic cost increase,

according to US Department of Health and Human Services

current nursing home costs range between $30,000 and $60,000

a year which has been doubled over the last decade. Therefore,

an alternative ways of providing quality care for elderly people

have to be considered. Such ways must not only lower the cost,

but they must also increase the comfort of living while

approaching the elderly with a level of privacy and dignity that

they deserve.

One approach would be a tailor made service robot that

will open the gate for new generation of artificial intelligence

robots. These new service robots are powered with Fuzzy-

Logic Controller (FLC) system that gives them the ability to

imitate humans in decision making when dealing with

multivariable inputs. Integrated with specially designed high

tech sensors, these robots are capable of creating health profiles

of their owners through pattern recognition thus sensing any

potential risk ahead of time. Furthermore they can detect and

report sudden break downs. Their fuzzy-logic controller is

proven to be far more superior to the commonly used

Proportional-Integral-Derivative (PID) controller [6]. FLC has

more robust against disturbances and faster transient response

with less overshoot and oscillation than the PID controller.

Robotic technology is going through major revolution and now

would be the time to take advantage of such revolution and find

an optimum solution for a potentially growing problem.

FUZZY THEORY

Fuzzy-logic control system can replace many of today’s

complicated control systems. It simplifies the control design

which can save time and money. The main feature of fuzzy

logic is that it is able to deal with imprecise linguistic

information. Compared with traditional logic, fuzzy-logic is

more suitable for analyzing complex nonlinear systems, such as

the ones that involve humans. Fuzzy algorithms can be

implemented with software on standard hardware such as


microprocessors. However, for complex and real-time control

systems hard-wired fuzzy algorithms are required (Fuzzy

hardware); one type of such hardware is the fuzzy-logic

controller. The fuzzy-logic controller is a rule based system (see

Fig 01) where fuzzy control rules and membership functions

are stored in the knowledge base. Control operations then can

be divided into three parts. First part is fuzzification, which

converts crisp input data into linguistic values (fuzzy values).

The second part is fuzzy inference which executes the fuzzy

approximate reasoning. Third and final part is defuzzification

which yields a none-fuzzy crisp data action as the output. In

depth analysis on FLC construction is reported in [7,8].

Figure 01 Fuzzy Logic-Based Control

PROPOSED METHOD

Utilize fuzzy-logic controller to design a state-of-the-

art personal service robot. The model involves specially

developed sensors that enable face recognition, motion

detection, and monitoring vital characteristics such as pulse and

respiration [9]. These high tech motion sensors can be installed

throughout the owner’s home; i.e. doorways, mattress, showers,

chairs, etc. The sensors send their signals to the robot’s control

panel. The collected data are then used to create a personal

heath profile via fuzzy-logic controller embedded within the

service robot which enables it to detect potential medical alerts.

Furthermore, by incorporating the gathered data the robot will

be able to distinguish between similar sounds that belong to

different sources; i.e. owner’s cry for help vs. a loud sound

from a TV set, which will reduce the number of false alarms. In

order to construct a fuzzy logic controller inputs and outputs

variables must be defined. The membership functions can then

be created via set of rules and equations. Figure 02

demonstrates a general design scheme of a fuzzy logic

controller.

CONTROL SYSTEM CONSTRUCTION

In order to simulate the robot’s performance, a fuzzy-

logic controller had to be constructed. For simplicity, a system

of three inputs and two outputs was chosen (Figure 03). The

input variables were defined as Pulse Rate, Body Temperature

and Room Temperature, each of which was assigned five levels;

Alarmingly Low, Low, Normal, High, and Alarmingly High.

The corresponding values for each level were picked according

to nursing practice [10]. Output variables were defined as

Robot’s Mode and Temperature Control. The Robot’s Mode has

four levels; Normal, Standby, Alert, and Emergency. For each

input variable a set of membership functions were constructed

(Figure 04). Table 01 and 02 contain the values of input and

output variables respectively.

Table 01 Input Membership Functions variables

Table 02 Output Membership Functions variables Robot's Mode Factor

Normal 0-0.3

Standby 0.31-0.5

Alert 0.51-0.75

Emergency 0.76-1.00

Figure 02 fuzzy-logic controller general design scheme [7]


Figure 03 Control System Outline

The inputs and outputs membership functions were

defined, a set of membership functions rules were created using

the IF-THEN rule base. As mentioned, one of the important

features of fuzzy-logic is the ability to efficiently handle

multiple-state logic as demonstrated below:

1. IF pulse is Alarmingly Low OR body temp is Alarmingly

Low THEN it is an Emergency.

2. IF pulse is Alarmingly High OR body temp is Alarmingly

High THEN it is an Emergency.

3. IF pulse is Normal AND body temp is Normal THEN it is

Normal.

4. IF pulse is Normal AND body temp is High THEN it is an

Alert.

5. IF pulse is Normal AND body temp is Low THEN it is an

Alert.

6. IF room temp is Low AND body temp is Low THEN it is a

Standby.

7. IF room temp is High AND body temp is Low THEN it is

an Alert.

8. IF room temp is Low AND pulse is Alarmingly Low OR

body temp is Alarmingly Low THEN it is an Emergency.

Figure 04 Input Variables Membership Functions

Membership functions rules are very essential for any

fuzzy logic controller. They provide the FLC with more options

to choose from for each possible scenario. However, having too

much set of membership functions rules can affect the

performance of the controller which may yield some

inconsistence output results. It can also slow the system

response speed which is a very vital aspect of the controller.

Therefore an optimum set of membership functions rules must

be sought after.

When the fuzzy logic controller is constructed, the

controller guides the robot to perform tasks and make decisions

based on the specified membership rules to incorporate a preset

commands. The controller enables the robot to mimic human-

like reasoning when handling none-linear multivariable inputs.

The output responses would be predefined as well. For

instance, an “Emergency” output means a 911 call, while an

“Alert” output means a call to a doctor or relative. On the

other hand a “Cool” output would require the robot to cool

down the air condition unit. By interface with specially

developed sensors, the controller can read the transmitted

signals from the sensors and gather all the data to create a

health profile of the robot’s owner. The controller then can

compare trends and/or detect potential health risks. The robot’s

response will be according to the FLC membership functions

rules (see Figure 05).

Figure 05 Membership rules


RESULTS SIMULATION The assigned inputs and outputs variables were run

through a fuzzy logic controller to simulate the outcome based

on the specified rules. The system showed that it can handle a

wide range of possibility combinations and the ability to

process nonlinear variables. Some set-value modifications were

needed to reach optimum results. A performance test of the

system was conducted and the results were plotted (Figures 06

and 07). The performance plot shows a generally coherent and

smooth surface with no significant spikes or breakouts which is

an indication of the system success.

CONCLUSION With the rapidly growing number of elderly population

and the dramatically cost increase of nursing homes and home-

care services a problem arises. In an effort to control such a

problem a cutting-edge service robot is proposed that can

mimic human control via set of specially developed sensors and

a fuzzy logic controller embedded within the robot. For

simulation purposes a fuzzy logic controller was constructed

and its performance was tested. The results show that the

system can handle multiple inputs and outputs variables. It also

shows that the system can be easily modified to reach optimum

results.

REFERENCES [1] H. J. Zimmermann, 1991, Fuzzy Set Theory and Its

Applications. Boston, MA: Kluwer

[2] Ahmad, M. Ibrahim, 2003, Fuzzy Logic for Embedded

Systems Applications, Elsevier Science, Burlington, MA

[3] Kevin, M. Passino, Stephen, Yurkovich, 1998, Fuzzy

Control, Addison Wesley Longman Inc, Menlo Park, California

[4] H. Bandemer and S. Gottwald, 1995, Fuzzy Sets, Fuzzy

Logic, Fuzzy Methods With Applications. New York: Wiley

[5] Nicholas, Roy, Gregory, Baltus, Dieter, Fox, Francine,

Gemperle, Jinnifer, Goetz, Tad, Hirsch, Dimitris, Margaritis,

Mike, Montemerlo, Joelle, Pineau, Jamie, Schulte, and

Sebastian, Thrun, 1999,”Towards Personal Service Robots for

the Elderly,” Computer science and Robotics Carnegie Mellon

University

[6] Herbert, Eichfield, Thomas, Kiinemund, and Manfred,

Menke, 1996,“A 12b general fuzzy logic controller chip,” IEEE

Transaction on Fuzzy systems, vol 4, no 4

[7] Faizal, A. Samman, and Eniman, Y. Syarnsuddin, 2002

“Programmable Fuzzy Logic Controller Circuit on CPLD

Chip,” IEEE 0-7803-7690-0/02/$17.00 02002

[8] Sheroz Khan, Salami Femi Abdulazeez, Lawal Wahab

Adetunji, AHM Zahirul Alam,Momoh Jimoh E. Salami, Shihab

Ahmed Hameed, Aisha Hasan Abdalla and Mohd Rafiqul

Islam, 2008, “Design and Implementation of an Optimal Fuzzy

Logic Controller Using Genetic Algorithm,” Journal of

Computer Science 4 (10): 799-806

[9] Glascock, Anthony,; Kutzik, David,; “The Impact of

Behavioral Monitoring Technology on the Provision of Health

Care in the Home” Journal of Universal Computer Science

2006, 12:59-72.

[10] McDermott, M., 1996, Central venous pressure. Nursing

Standard 9(35): Cardiol-30

Figure 06 Output Variables Membership Functions

Figure 07 Performance Plot of FLC




DESIGN OF CONTROLLERS FOR A LASER BEAM STABILIZER USING PID AND OBSERVER-BASED STATE FEEDBACK CONTROL

Kwabena A. Konadu, Sun Yi North Carolina Agricultural and Technical State

University Greensboro, North Carolina, USA

ABSTRACT This paper presents the design and implementation of a

Proportional Integral Derivative (PID) controller and an

observer-based state feedback controller to regulate a laser

beam position. In the presence of noise and active disturbance,

the two different control strategies are compared regarding

performance and design procedures. The PID controller

estimates the Root-Mean-Square (RMS) of the signal on the

Position Sensing Device (PSD) and dynamically computes

control commands. However, the PID controller can become

unstable if filters are not selected carefully. Thus, a Low-Pass-

Filter (LPF) and sampling rate should be analyzed and selected

carefully to reject disturbances avoiding destabilization.

Alternatively, the state feedback method stabilizes the laser

beam by continuously observing the state of the laser beam

system (LBS). Since all the state variables in the model of the

LBS are not directly measurable, an observer is designed to

estimate the state of the plant. The estimated state is used to

complete state feedback control. Simulations and experiments

conducted on the LBS are presented to demonstrate the

effectiveness of the two controllers on stability and

performance.

INTRODUCTION Laser beam stabilization experiment is used to correct or

minimize dynamic laser beam pointing errors in application

systems. Representative examples of applications that utilize

laser beams are [1]:

Medical: a variety of surgeries are performed by using

LBSs,

Military: most firearms applications use LBSs as a tool to

enhance the targeting of weapon systems and a target

designator in aircraft,

Industrial and commercial: cutting, preening, welding,

marking of metals and other materials are done by using

LBSs,

Electronics and data communications: LBSs are used for

optical communications over optical fiber and free space as

well as storage of data in optical discs. Also, applications

include nuclear fusion, microscopy, laser cooling, material

processing, photochemistry, etc.

In these applications, errors are introduced by auxiliary

devices such as fans, external light sources from fluorescents,

computers, pumps and any device that introduces high-

frequency signals to measurement. The errors have the effect of

deviating the laser beam from its accurate intended target [2].

Controllers are designed to reduce these errors or deviation that

is introduced on the laser beam. This is achieved by using an

active Fast Steering Mirror (FSM) (known as the Voice Coil

Actuator) to eliminate beam motion by sampling a small

percentage of the beam and using feedback from Position

Sensing Detectors (PSD) to compensate or control these errors

[3]. A controller that stabilizes a laser beam on the PSD in the

presence of noise and active disturbance needs to be designed

and built. The designed controller uses feedback from a

position sensing device to calculate the amount of voltage that

rotates the voice-coil actuator so that the incident laser beam on

its mirror is reflected or directed accurately to the middle of the

position sensor even in the presence of noise and active

disturbance. The stability of the closed-loop system when using

the PID controller and filters will be analyzed carefully because

the PID controller could become unstable if filters are not

selected appropriately [4].

Recent literature on control of laser beams showed that

work has been done on the control of laser beams using the PID

approach [2, 5]. However, not much work has been done on the

control of laser beams using observer-based state feedback [6].

In most laser beam control systems, bi-axial sensors, such as,

optical position sensor (OPS) charge-coupled devices (CCDs)

or quad-photo detectors, are used to determine the coordinated

location of the beam projection on a plane [7], which is fed

back into a two-input/two-output controller [6]. This paper

considers the LBS as a linear time-invariant (LTI) system with

single-input single-output (SISO). The PID design method was

presented in [3], the design method is followed for a PID

controller of different design specifications. An alternative

control scheme is the use of observer-based state feedback. The


response of the system when using the PID is compared with

the state feedback controller.

The rest of the paper is organized as follows. The next

section involves designing the PID controller, followed by

simulations to test the stability of the response. A controller

using an observer-based state feedback is introduced in the

subsequent section. Simulations of controlled systems and

comparison of the two design strategies are then presented,

followed by Conclusion.

DESIGN OF A PID CONTROLLER The purpose is to design a proportional-integral-derivative

controller (Figure 1.) that uses all these three terms to

compensate for any error. This controller will determine the

right amount of voltage that will steer the actuator in a way that

the beam is always reflected directly to the center of the PSD

even in active disturbance.

Figure 1. Block Diagram of a Closed-Loop of a Laser Beam Stabilization System with a PID Controller

Design of RMS Estimator

The RMS estimator (Figure 2.) is used to determine the

actual position of the laser on the PSD. It records the deviation

of the beam from the middle of the PSD (the reference center)

and displays the value digitally. The aim is to estimate signal

values that are close to the ideal or theoretical [3].

The performance of the estimator when tested showed that

the effect of noise and disturbance alters the accuracy of its

values. A second-order low-pass filter is then introduced in the

estimator to reduce these errors.

Figure 2. RMS Estimator Model with Low-Pass Filter

Derivation of the Ideal PID Gains for the Closed-Loop Laser Beam Stabilizer System

The block diagram that describes the operation of the LBS

is shown in Figure 1 in closed-loop. The intended target of the

beam is the middle of the PSD, thus the reference point is zero.

The ideal PID gains without filtering and noise is first derived

by comparing the second-order characteristic equation to the

denominator of the transfer function equation of the system in

Laplace domain. The ideal gains are then used to select desired

filter parameters before the actual practical PID gains are

obtained.

Assumptions made in designing PID 1) There is no actuator saturation and amplifier offset,

2) Vc,amp = Vc,

where Vc(s) is the Laplace Transform of the voice-coil

digital-to-Analog voltage and Vc,max is the maximum

voltage that can be supplied to the voice-coil by the power.

The transfer function (T.F.) of the closed loop disturbance-to-

position of the system, Gx,d, is given as;

(1)

where X(s) is the Laplace Transform of the position measured

by the PSD, D(s) is the Laplace Transform of the disturbance.

The LBS utilizes the unity negative feedback in the closed-loop

system as;

(2)

the plant transfer function is;

(3)

the PID controller, C, is given as;

(4)

where is the proportional control gain, is the derivative

control gain and is the integral control gain. Equations (2),

(3) and (4) are substituted into equation (1) to obtain the closed

loop transfer function, , as;

(5)

is the natural frequency, is the damping ratio and K is the

open-loop steady-state gain.

Finding the Ideal PID Gains

For the Ideal PID gains, the denominator of equation (5),

(closed-loop transfer function) is compared with the third-order

characteristic Equation below;


(6)

where is the zero location. Comparing the coefficients in

Equations (6) to the denominator of equation (5) yields the

control gains;

(7)

(8)

(9)

System specification of PID controller 1) The damping ratio of the ideal PID controller is set to 1

.

2) The closed-loop gain should not exceed 0.05;

, the disturbance frequency,

3) The zero location specification is at 0.5

Substituting the closed-loop system specification parameters

into the gain Equations, the ideal closed-loop proportional gain,

, is , the derivative gain, , is 0.0021

, and the integral gain, , is /s

Obtaining natural frequency of system The natural frequency of the LBS is obtained by

substituting the gains and parameters of the PID specification

into the closed-loop transfer function equation to obtain the

frequency response of the system. The natural frequency is

found from the magnitude of the frequency response as;

(10)

Substituting the specified damped frequency, the system gain,

and time constant into equation (10) yields the natural

frequency, 563 rad/s. Hence, the natural

frequency, of the system is obtained as 563 rad/s.

Practical PID Controller Because the LBS is prone to capturing noise, these noise

turns to magnify after taking the derivative of the signal. It is

important to include a filter in the design to remove this noise

from the system. In this design of practical PID controller, a

filter (Low-pass filter) is used to obtain the displacement of the

PSD signal.

Specifications of filter The transfer function of the second-order band-pass filter is

of the form;

(11)

is the cutoff frequency and is the damping ratio of the

filter. The bandwidth is obtained from equation (12)

(12)

The cross-over frequency is the frequency at which the

magnitude of the Transfer Function is 1 or 0 dB. The cross-over

frequency, , is obtained from the magnitude of frequency

response as;

(13)

Table 1. Filter Specifications

Filter parameter Specification

Filter bandwith

Phase margin Cut-off frequency

Selecting filter specifications for LBS

Figure 3. shows a set of filter parameters for different

damping ratios. The parameters are compared with the

specifications to find the set that best meet the given

requirement. The white markers show the parameters for

corresponding damping ratios that are rejected and the black

markers are filter parameters for corresponding damping ratios

that meet the requirements.

Figure 3. Plot of Filter Specifications against Damping

Ratio

From Figure 3., the damping ratio that results in a bandwidth

greater than 6752rad/s and a PM greater than 75degrees is

determined as 0.5 and 0.6 but a damping ratio of 0.5 is selected

0

20

40

60

80

100

120

140

160

180

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

0.3 0.4 0.5 0.6 0.7

Ph

ase

Mar

gin

(⁰)

Fre

qu

en

cy (r

ad/s

)

Damping ratio (ζ)

bandwidth

Cross-over frequency

Phase Margin (PM)


as the choice for designing the filter because its phase margin is

closer to the desired specification.

Controller of Laser Beam Stabilization System The filter with its desired parameters has been selected,

now, the PID controller is built in Simulink and its performance

is tested to determine if the specified gain requirement for the

LBS is met. Figure 4. shows a block diagram of the designed

laser beam controller. The PID gains after applying the low-

pass filter is: = 0.722 V/mm; = 0.002 V.s/mm; = 0.360

V/mm/s.

Figure 4. Block Diagram of the Laser Beam Controller

Experimental Set-up The LBS experiment set up (Figure 5.) consists of four

main components: PC, LBS component, Quanser Personality.

Intelligent Data acquisition board (QPID) and a Peripheral

Component Interconnect (PCI) express board. These are inter-

connected and act as a hardware-in-the-Loop (HIL). The PCI

board is inserted into the CPU and connected to the QPID

terminal board through analog cables. The terminal board is

then connected to the LBS component through analog and

encoder cables before the system is powered. Experiments are

run on this system by generating real-time codes from Simulink

models that runs on a real-time kernel of the processor of the

PC. After designing the appropriate controller, the design is

built and tested in Simulink on the computer.

Fast steering mirror

DC motor for active disturbance

Laser Source PSD

QPIDAmplifier

Laser beam

Figure 5. Schematic Diagram of a Laser Beam

Stabilization Experimental Set-Up

The LBS is subjected to active disturbance by increasing

the disturbance voltage in Simulink which causes the motor to

slide back and forth, thereby displacing the beam from the

middle of the PSD. The controller intended to stabilize the

system is then turned on and the response is tested.

System gain of LBS after switching to closed-loop

After building the design, the controller is tested to

determine if the design specification has been met. Figure 6.

shows that the system maintains a desired gain below 0.05 for

all disturbance frequencies when the system is switched from

open-loop to closed-loop. Therefore the gain requirement is met

for this design.

Figure 6. Plot of Closed-Loop System Gain against Disturbance Frequency When Using the PID Controller

System response after switching from open-loop to closed-loop

Figure 7. is a plot of the LBS in open-loop that is switched

to closed-loop after 12.5 seconds. Observation shows that there

is an amount of error that is introduced immediately when the

controller is initially switched from open-loop to PID. This

error however is seen to decrease over time and gradually

approaches steady state at zero.

Figure 7. Plot of Change in Response from Switching

From Open-Loop to Closed-Loop

Stability analysis of controller The Phase Margin (PM) is the amount of phase that

exceeds -180 degree at the cross-over frequency. The more it

exceeds -180 degrees, the more stable the system is. The bode

diagram shows the PM for the transfer function of the loop. The

introduction of filter and sampling has an effect of delay and

shift in the stability of the system. It is, therefore, necessary to

0

0.01

0.02

0.03

0.04

0.05

0.06

8 9 10 11 12 G

ain

=X(m

m) /

D(W

d)

Disturbance Frequency (Hz)

System Gain

Specified System Gain

11.5 12 12.5 13 13.5 14-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

time (s)

Positio

n M

easu

re

d o

n th

e P

SD

x (

mm

)

Open loop

Closed- loop


select a sampling interval such that the stability of the system is

not affected.

Figure 8. Margin Plot for Practical PID Controller

Figure 8. shows a Bode diagram for crossover frequencies

of the loop transfer function of the controller with a filter. For a

signal, it is shown that the phase margin of the practical loop

transfer function has 73.4 degrees. Although the phase margin

is not close to 90 degrees (ideal), the system is still considered

to be stable because it is greater than the desired PM of 60

degrees. The reduction in PM can be accounted for as the effect

of the filter.

Figure 9. Margin Plot for Practical PID Controller after Sampling

Figure 9. shows that after sampling, the PM has reduced to

67.5 degrees meaning the stability of the system has reduced.

Even though there is a reduction in PM, the system is

considered to be stable because the resulting PM is still greater

than 60degrees.

DESIGN OF STATE FEEDBACK CONTROLLER Alternatively, Figure 10. shows a block diagram of the

LBS that utilizes state feedback observer for control. The real

plant of the LBS is modeled in its state space form and

considered as a linear time-invariant (LTI) with single-input

single output (SISO) system.

Figure 10. Block Diagram of the LBS Using a State

Feedback Observer for Control

Assumptions made in designing state feedback controller 1) Not all the state variables, x, of the LBS is available for

measurement.

2) Sensors are not available and they are expensive to obtain

all the physical conditions (initial), , for the LBS

3) There is an amount of error due to estimation.

Because information about the dynamics of the LBS is limited,

the design of an observer that computes an estimate of the

entire state vector with limited information from the output of

the LBS for control is proposed.

Modeling of LBS (Plant) and Observer The dynamics of the plant is obtained by modeling the

LBS in the form;

(14)

(15)

where A and B are system and input matrices respectively,

and are actual state vectors, is the output matrix. [8, 9].The

observer is constructed from the model of the LBS dynamics

as;

(16)

(17)

where is the estimate of the actual state . Since the exact

initial condition, of the actual LBS is not available, the

observer will be used to provide that information. However,

because the observer gives estimated but not exact information

about the LBS, a continuous increase in error may occur if a

poor estimate for is made, this will cause the observer to

give erroneous information about the true state of the LBS.

Error introduced, e, is;

(18)

This divergence in estimated state from the actual LBS state,

can be eliminated and the error made to disappear very fast by

controlling the error with feedback. The difference between the

actual measured LBS outputs and the estimated outputs are

taken and fed back into the observer to continuously correct for

this error as shown in Figure 11.

-200

-150

-100

-50

0

50

Mag

nitu

de (d

B)

101

102

103

104

105

106

107

-270

-225

-180

-135

-90

Phas

e (d

eg)

Bode Diagram

Gm = 15.5 dB (at 5.74e+003 rad/sec) , Pm = 73.4 deg (at 1.02e+003 rad/sec)

Frequency (rad/sec)

-200

-150

-100

-50

0

50

Mag

nitu

de (d

B)

101

102

103

104

105

106

107

-6.912

-4.608

-2.304

0x 10

4

Phas

e (d

eg)

Bode Diagram

Gm = 12.1 dB (at 4.48e+003 rad/sec) , Pm = 67.5 deg (at 1.02e+003 rad/sec)

Frequency (rad/sec)


Figure 11. Block Diagram of the Observer Using Error

Feedback for Accuracy

The difference in measured output and estimated output is;

(19)

Adding the error to the observer gives;

where L, is the observer gain. Re-writing the equation for the

observer gives;

(20)

Deriving the State Space model The plant transfer function for the laser beam is given in

equation (3) where K, is the open-loop steady state gain is

2200mm/(V.s), , the open-loop time constant, is 0.005 s

(21)

Taking the Laplace inverse of equation (21), the equation of

motion of LBS is obtained as;

(22)

From the equation of motion, the System matrix A, and Input

matrix B is derived as

System Matrix

Input Matrix

Output Matrix Control Matrix

Writing the equation of motion in state space form gives

, where , is the state vector and is the input

vector [8, 9].

The state equation of the LBS can be written as:

(23)

where is the displacement and is the velocity of the beam Checking For System Controllability

The first step to determine if the system is controllable is to

compute the controllability matrix [10, 11]. The controllability

matrix, is derived from Matlab using ctrb (A, B) as

Let be vectors of columns 1 and 2 of matrix

respectively. If are scalar, then

Then are linearly independent of each other, thus

columns 1 and 2 are linearly independent. Since columns 1 and

2 of controllability matrix are linearly independent of each

other, it shows that the rank of controllable matrix is 2. Since

the size of the state vector is 2and the rank of the controllable

matrix is 2, then the system is controllable.

Checking for system observability Observability matrix, is derived from Matlab using obsv

(A, C) as

Since columns 1 and 2 of matrix are linearly independent,

the rank of observability matrix is 2. The system is observable

because the rank of the observable matrix is equal to the size of

the state vector, the system is observable.

Deriving the Desired Pole Locations and Gains Finding the Pole Locations

In order to achieve a desired response, poles are selected so

that the system response to disturbance is dominated by the

dynamic characteristics of the observer and not the control law.

Poles are locations in closed-loop system where desirable

response is achieved when control effort is applied. Placing the

poles are important because the location of the poles

correspond directly to the eigenvalues of the system, which

control the characteristics of the response of the system [12,

13]. If the selected poles are not desirable it will require a larger

effort to control the system making the design expensive [12].

Pole locations are theoretically derived by finding the

eigenvalues of the characteristic equation of the system from

the denominator of the closed-loop response of the LBS. The

equation for the closed loop response is

(24)

From the denominator, we got


Therefore the desired poles are theoretically obtained as

But observation showed that there is a lot of noise in the

response of the system, new desired poles are then selected to

be . The control gain K, is derived

from Matlab using Ackermann formula as

K=acker(A,B,p) (25)

P is the desired pole locations. The control gain, K is [-196

20]. The Estimator Gain, L, is obtained from the Ackermann

formula using Matlab as

L’=acker(A',C', ) ' (26)

where A' and C' are the transpose of system matrix and the

output matrix respectively is desired observer pole location.

For a faster decay of the estimator error, the desired Estimator

Pole location is chosen by a factor of 5 [12].

The observer gain, L, is obtained as

The model of the LBS (plant) and its observer is built in

Simulink (Figure 12.) and simulations are performed to

determine the response of the system when using an observer-

based feedback for controlling the LBS.

Figure 12. Simulink Block Diagram of the Designed State

Feedback for Stabilizing the Laser Beam

Response of the System

Figure 13. shows the response of the LBS in Simulink that

utilizes the designed observer for control. The closed-loop

system maintains a desired response throughout for a sine

input.

Figure 13. Plot of Response of the Designed State Feedback Controller

Figure 14. shows the response of the LBS in Simulink without

a state observer. For a step input, it is observed that the beam is

displaced from the reference point and maintains the distance

throughout whiles the velocity keeps on increasing infinitely.

Figure 15. shows the response of the LBS in Simulink with a

state observer. It is observed that even after displacing the beam

from the reference point with a step input, the beam returns

back to its original position after 2 seconds onto the reference

point whiles the velocity is stabilized after 1.5 seconds.

Figure 14. Plot of the Unstable Response of the LBS

Model without State Observation

Figure 15. Plot of the Stabilized Response of the Lbs

Model with a State Observer

Comparing the Response of the State Feedback Observer and the PID Controller

The Table 2 below shows a summarized comparison of the

two controllers designed for LBS to regulate the position of the

laser beam. One of the main concerns in using PID control is

X Y

Y^

Xdot

X^ dot X^

K*uU

Sine Wave

Scope1

Scope

K*u

L

1

s

Integrator1

1

s

Integrator

K*u

C1

K*u

C

K*u

B

K*u

A1

K*u

A

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

Time (seconds)

Dis

pla

ce

me

nt(

mm

) a

nd

Ve

locit

y(m

m/s

)

displacement

velocity

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

Time (seconds)

Dis

pla

ce

me

nt(

mm

) an

d V

elo

cit

y(m

m/s

)

Displacement

Velocity

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Time (seconds)

Dis

pla

ce

me

nt(

mm

) an

d V

elo

cit

y(m

m/s

)

Velocity

displacement


obtaining the derivative of the output by differentiating

measured response.

Table 2. Comparison of Controllers

State Observer PID

No filter for controllers is

required. One does not need to

obtain the signal for D action in

PID.

A filter is required when taking

the derivative of the signal to be

multiplied with kd.

Since the model is expressed in

matrix-vector form, the

calculation is friendly to software

packages (e.g., MATLAB). Thus,

even if systems have a large

number of state variables, design

processes can be performed in a

relatively simple way.

Design of control requires a

relatively complicated process. It

handles scalar multi-variable

models, and requires designing

extra filters for tuning.

The state variables are obtained

through estimation. There is no

effect of an additional filter.

Difficult to get the dynamics of

the actual signal because the filter

alters the shape of the actual

signal

However, sensors are supposed to add noise to responses in

measuring, differentiating the signals can amplify the noise and

the derivative cannot be used for PID control. Thus, one should

add filters in the control loop, and the filters change the

dynamics of the system. In some cases, the filters induce

instability.

CONCLUSION This paper presented the design of a PID controller that

uses feedback from a position sensing device to rotate the

voice-coil actuator. The incident laser beam on its mirror is

reflected or directed accurately to the middle of the position

sensor even in the presence of noise and active disturbance.

Simulations results and real-time implementation of the

Laser Beam System when using the PID under dynamic

disturbances, demonstrated the necessity for a more effective

and robust design method. The paper also proposed a state

feedback method for controlling the LBS by modeling the LBS

as a linear time-invariant plant and observing the state of the

plant at all conditions, even at high frequencies. Simulations

demonstrate that the proposed state feedback method is more

effective regarding design procedure with satisfactory

responses of the system.

In future, simulations and experiments of real-time

implementation of the LBS using the state feedback observer

under dynamic disturbances needs to be conducted to confirm

the effectiveness and robustness of the method by the Authors.

This will be achieved by replacing the real plant in the

Simulink block diagram of the state feedback controller with

the LBS. The performance of the controller in real-time when a

time-delay is introduced to the system will also be analyzed.

REFERENCES [1] Duarte, F. J., 2009, Tunable Laser Applications, CRC, New

York.

[2] Ying, B., Hanqi, Z., and Roth, Z. S., 2005, "Fuzzy logic

control to suppress noises and coupling effects in a laser

tracking system," Control Systems Technology, IEEE

Transactions on, 13(1), pp. 113-121.

[3] Quanser, 2010, "Laser Beam Stabilization Instructor

Manual," Quanser Speciality Experiment series: LBS

Laboratory Workbook, Quanser, ed.

[4] Tsao, T.-C., Gibson, S., Chiu, K. C., and Chen, S.-J.,

"Adaptive control for focusing of optical drive read/write

heads," Proc. American Control Conference (ACC), 2011, pp.

588-593.

[5] Kawaji, S., and Matsunaga, N., "Design of fuzzy model

following servo control systems," Proc. Fuzzy Systems, 1995.

International Joint Conference of the Fourth IEEE International

Conference on Fuzzy Systems and The Second International

Fuzzy Engineering Symposium., Proceedings of 1995 IEEE

International Conference on, pp. 1763-1768 vol.1764.

[6] Perez-Arancibia, N. O., Gibson, J. S., and Tsu-Chin, T.,

"Laser beam pointing and stabilization by intensity feedback

control," Proc. American Control Conference, 2009. ACC '09.,

pp. 2837-2842.

[7] Hara, K., Maeyama, S., and Gofuku, A., "Navigation path

scanning system for mobile robot by laser beam," Proc. SICE

Annual Conference, 2008, pp. 2817-2821.

[8] Krokavec, D., and Filasová, A., 2007, "Pole assignment in

robust state observer design," Robust and Adaptive Control,

PLUS 2, pp. 75-78.

[9] Luenberger, D. G., 1964, "Observing the State of a Linear

System," Military Electronics, IEEE Transactions on, 8(2), pp.

74-80.

[10] Chen, C.-T., 1999, Linear System Theory and Design,

Oxford University Press, New York.

[11] Haddad, W. M., and Bernstein, D. S., 1992, "Controller

design with regional pole constraints," Automatic Control,

IEEE Transactions on, 37(1), pp. 54-69.

[12] Franklin, G. F., Powell, J. D., and Emami-Naeini, A., 1995,

Feedback Control of Dynamic Systems, Addison-Wesley,

Reading, MA.

[13] Ogata, K., 2002, Modern Control Engineering, Prentice

Hall Inc., Upper Saddle River, New Jersey.


Date post:	15-Apr-2020
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

EVALUATION OF THE EFFECT OF SIMPLIFIED AND PATIENT ... · Understanding blood flow patterns in...

Documents