Eindhoven University of Technology MASTER Modelling of the ZF-intarder for HIL simulations Pesgens, M.F.M. Award date: 2002 Disclaimer This document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Student theses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the document as presented in the repository. The required complexity or quality of research of student theses may vary by program, and the required minimum study period may vary in duration. General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Download date: 25. Aug. 2018
Transcript
Eindhoven University of Technology
MASTER
Modelling of the ZF-intarder for HIL simulations
Pesgens, M.F.M.
Award date:2002
DisclaimerThis document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Studenttheses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the documentas presented in the repository. The required complexity or quality of research of student theses may vary by program, and the requiredminimum study period may vary in duration.
General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.
Download date: 25. Aug. 2018
TUle technische unlversiteit eindhoven
faculteit werktuigbouwkundeDI1F
Modelling of the ZF-Intarder forHIL simulations
Master's thesis ofM.F.M. Pesgens
DCT report 2002.22DAF report 51211/02-033
Committee:Supervisor - Prof. dr. ir. M. SteinbuchMember - Dr. ir. F.E. VeldpausMember - Jr. P.W.J.M. NuijMember - Dr. ir. L.M.T. SomersAdvisor - Dr. ir. R.G.M. HuismanAdvisor - Dipl.-Ing. W. Hubler
Eindhoven University ofTechnologyFaculty ofMechanical EngineeringControl Systems Technology Group
iVlodell ing c~f tl~.ZF-Intardel.~I~:~!llI, simulations
Contents
CONTENTS 2
PREFACE 3
SyMBOLS 4
SUMMARY 8
INTRODUCTION 9
2 RETARDERS 11
2.1 RETARDERS IN GENERAL 112.2 POSITION IN THE DRIVETRAIN 112.3 WORKING PRINCIPLES 122.4 THEZF-INTARDER 14
3 RETARDER MODEL 17
3.1 OVERVIEW OF THE SUB-MODELS 173.2 MEASUREMENTS 173.3 HYDRAULIC VALVE MODELLING 173.4 SUB-MODELS 183.5 MODEL IDENTIFICATION 193.6 MODEL VALIDATION 19
4 DAF COOLING SYSTEM 21
5 COOLANT HEAT TRANSFER MODEL 22
5.1 RETARDER HEAT EXCHANGER 225.2 ENGINE 235.3 RADIATOR 255.4 FAN AND FAN CLUTCH 265.5 COOLANT PUMP AND THERMOSTAT 275.6 COOLANT DUCTS 275.7 TEMPERATURE SENSORS 275.8 MODEL VALIDATION 27
6 HIL VALIDATION 29
6.1 UPHILL AND DOWNHILL DRIVING 296.2 RETARDER SPEED CONTROL .30
7 CONCLUSIONS 34
REFERENCES 35
APPENDICES'
• Due to confidential infonnation the appendices have been placed in a separate volume
Michie]
Preface
of the ZF-Intardcr for IIIL simulations
This report is my master's thesis for my study at the Mechanical Engineering faculty oftheEindhoven Technical University, and describes my work at DAF Trucks in Eindhoven inorder to get a functioning Hardware-In-the-Loop truck retarder implementation.I would like to everyone at the Product Development department ofDAF Trucks for thepleasant and informal ways of interaction. It is very nice that you can drop in with a shortquestion anytime, and that trainees are treated as full-blown DAF employees. Especially theTechnical Analysis group, who immediately involved me in their group activities. Further Iwould wish to thank my coaches at both DAF and the university (Rudolf Huisman, FransVeldpaus, Bart Veldpaus, Peter Kup, Pieter Nuij, Bram Veenhuizen) for their support andconstructive criticism. Especially Rudolf Huisman for his support throughout this work andwhose views that helped me to regain the total perspective of the problems at hand. AlsoFrans Veldpaus for his constructive criticism and remarks on my reports, which seemedslightly excessive at first, but finally proved vital for my academical writing skills. Thanks goto Rene Nevels and Rege Wilbers who made sure that the cooling system measurementsbecame available within my limited timerange, and to Kees van den Wildenberg for hissupport at the HIL frame. I also very much appreciated the good collaboration of the people atZahnradfabrik Friedrichshafen, who provided the indispensable measurements for my retardermodel. Finally I would like to thank my parents for providing me with food, love and lodgingduring this period.
vehicle pneumatic system air pressuretimeretarder engagement timevehicle speeddriver's demand vehicle speedthermostat positionvalve positionvalve orifice flow-through areaduct flow-through areainjected fuel per strokevalve discharge coefficientengine lumped heat capacitybypass coolant flow factoroil cooler coolant flow factorpump coolant flow factorradiator coolant flow factorretarder coolant flow factorretarder turbine output pressure parameterretarder turbine output pressure parameterretarder turbine output pressure parameterretarder turbine output pressure parameterradiator lumped heat capacityretarder coolant lumped heat capacityproportional valve duty cycleretarder turbine outer diameterelectrical force on valve stemwheels traction forcefuel combustion energyengine heat transfer coefficientretarder booster valve signalretarder proportional valve currentretarder proportional valve current controller setpointretarder booster gainengine coolant power distributionretarder oil pump flow gainretarder oil pump pressure gainretarder proportional valve gainretarder turbine gainretarder turbine gainretarder volume control valve gainretarder volume control valve flow gain into turbine circuitretarder volume control valve gain out of turbine circuitdriveshaft torqueengine torqueretarder torqueretarder controller predicted retarder actual torquedriver's demand retarder torqueretarder torque at driveshaftmaximal driveshaft retarder torquemeasured retarder driveshaft torquemodel predicted retarder driveshaft torque
5
rVlichiel of the IF-Intarder l~)r HIL simulations
Mr,int
Peootant
Peng,btoek
Peng,e
Peng.l1alf.1
Pexhallst
Pjuet
Pjue/.max
Pin/ere
Pintere.he
Piso
Pp,max
Pr
Prad,ea
Prad,e
Prest
P ret.c
Pret,he
P r.max
Rdyn
Roc
Rp
Rrad
Rrad,h
R ret
Rro
TTamb
Teoo/ant
Tduct,in
Tduet, out
Teng,bloek
Teng,c
T eng,in
Teng,oil
Teng,out
Toc,in
Toc,out
Toi
Too
Trad
Trad,in
Trad, out
Tret,in
Tre~out
Tro
Tth
Tth,hyst
Tth,in
Tth,min
Tth,max
retarder controller predicted retarder intended torqueengine power to coolantengine power to engine blockengine net coolant flow powerengine power through combustion chamber wallsfuel power to exhaustfuel powermaximal fuel powerfuel power to intercoolerintercooler heat exchanger powerfuel power to driveshaftmaximal retarder proportional valve dissipated powerretarder powerradiator power to ambient airradiator net coolant flow powerfuel power to unidentified recipientretarder net coolant flow powerretarder heat exchanger powerretarder maximum powerdynamic wheel radiusoil cooler coolant flow resistanceproportional valve coil resistanceradiator coolant flow resistanceradiator coolant flow resistance with engaged cabin heaterretarder coolant flow resistanceretarder coolant temperature sensor resistancetotal timeambient air temperaturecoolant temperatureduct inlet coolant temperatureduct outlet coolant temperatureengine block temperatureengine coolant temperatureengine inlet coolant temperatureengine oil temperatureengine outlet coolant temperatureoil cooler coolant inlet temperatureoil cooler coolant outlet temperatureretarder heat exchanger oil inlet temperatureretarder heat exchanger oil outlet temperatureradiator temperatureradiator inlet coolant temperatureradiator outlet coolant temperatureretarder inlet coolant temperatureretarder outlet coolant temperatureretarder outlet coolant temperaturethermostat temperaturethermostat hysteresis temperaturethermostat inlet coolant temperaturethermostat opening temperaturethermostat maximum opening temperature
6
Michiel of the IF-Intarder !()[ 1111. simulations
Ue
UdSpaee
Uecu
Up
Vb
Vb,max
Veng,e
Vrad,e
VI
VI,max
'leoo/anl
'lintere
AWds
Wr
WrO
Td,b
Td,eng
Tdjan
Tjan
Td,rad
Td,sw
Tduet
Teng,b/oek
Teng,walls
Tintere
Tpr
Tret,m
"Csensor
Tth
PPeL1Poe
L1pret
L1Trad,ea
L1Trad,eaO
<Pb
<Pb,in
<Pb,out
<Pbypass
<Pduet
<Peng
<Poe
<Pp
<Ppump
<Prad
<Pret
<Pv
(/>v,in
(/>v,out
(/>valve
vehicle contact voltagedSpace voltagecontroller ecu voltageproportional valve voltageretarder booster oil volumeretarder maximal booster oil volumeengine coolant volumeradiator coolant volumeretarder turbine circuit oil volumeretarder turbine circuit maximal oil volumefuel power to coolant efficiencyfuel power to intercooler efficiencyretarder shape factordriveshaft speedretarder speedretarder viscous transition speedbooster delayengine coolent flow delayfan 'stick' delayfan time constantradiator coolant flow delayretarder switching valves time delayduct coolant flow delayengine block time constantengine combustion chamber walls time constantintercooler heat exchanger time constantproportional valve time constantretarder heat exchanger metal time costanttemperature sensor time constantthermostat time constantturbine mixture densitycoolant densityoil cooler pressure differenceretarder pressure differenceradiator temperature difference coolant to ambient airradiator linearization temperature differenceretarder booster oil flowoil flow into retarder boosteroil flow out of retarder boosterbypass coolant flowduct coolant flowengine coolant flowoil cooler coolant flowretarder oil pump flowcoolant pump coole flowradiator coolant flowretarder coolant flowretarder volume control valve oil flowretarder volume control valve oil flow into turbine circuitretarder volume control valve oil flow into turbine circuitvalve oil flow
7
Michiel Pesgens
Summary
DAF Trucks use Hardware-In-the-Loop (HIL) simulations to test the functionality of the truckcontrollers. The HIL test system consists of the controllers in combination with a vehiclemodel, simulated in real-time. HIL tests reduce the number of necessary test drives. A HILvehicle model should reproduce road test phenomena well, but the model's computationalload on the real-time simulator should not be excessive.Until recently, the truck retarder controller was not yet implemented on the HIL test system.Retarders are high-torque downhill endurance brakes, sparing the conventional brakes. TheZF-Intarder as used by DAF is a hydrodynamic retarder, which generates a braking torque onthe driveshaft by means of fluid inertia effects. The braking power is transferred to the enginecooling system and then passed on to the ambient air.In this study, the retarder has been modelled in combination with the engine cooling system,since this system strongly effects the retarder performance due to a high coolant temperatureprotection. The retarder model is mainly based on the oil flow dynamics. A lumped heatcapacity approach has been adopted to model the engine cooling system. Both models havebeen developed and validated using vehicle measurements.After the signal connections to the retarder controller were realized, HIL tests wereperformed, simulating the total vehicle model (including the retarder and cooling systemmodels) together with the vehicle controllers (including the retarder ECD). Phenomenaencountered during vehicle tests like retarder torque cutback due to high coolant temperaturesare clearly visible. Continued vehicle speed oscillations occurred during retarder downhillconstant speed control. This phenomenon has been shown to occur in vehicle tests also. UsingHIL measurements, it has been shown that these oscillations are caused by insufficient closedloop stability margins, resulting in limit-cycle behaviour.Recommendations for further investigations have been given. The HIL coolant heat transfermodel can also be used for concept analyses of future changes to the DAF cooling system.
8
1 Introduction
OAF Trucks develops and manufactures trucks in the medium heavy and heavy marketsegments. These trucks are equipped with an increasing amount ofcontrol systems with moreand more mutual interaction. OAF mainly develops controller specifications whereas theactual control systems are developed by a supplier. When a new controller becomes available,the specifications must be verified by OAF. Often, controller settings must be tuned. For all ofthese tests, OAF has developed a Hardware-In-the-Loop (HIL) test system, as reported in[Frid, 1999], [Huisman, 1999] and [van den Akker, 2001]. This is just one ofa range of testmethods used in controller development. Figure 1.1 shows a block diagram ofa Software-Inthe-Loop simulation. Here models ofboth the controller and the vehicle are simulated oflline.Rapid controller prototyping (Figure 1.2) is a means of testing a prototype controllersimulated in real-time using the actual vehicle hardware.
To test the actual controller hardware, Hardware-In-the-Loop tests (Figure 1.3) can beperformed simulating a vehicle model in real-time. Finally, in actual vehicle tests both thevehicle and controller hardware is used (Figure 1.4). In general, SIL and HIL tests areperformed to reduce the number of vehicle tests. SIL and RCP tests are performed to testcontroller prototypes. HIL and vehicle tests are mainly used to validate controller hardware.
Real control system Real control system
sensors
Circum stances
£.A---+.-.:!t.,.'lWJ.l~~
Figure 1.3: Hardware-In-the-Loop (NIL) Figure 1.4: Vehicle test
OAF Trucks performs SIL control system simulations mainly to derive controllerspecifications. OAF use a HIL simulation test system to evaluate the functionality of, andinteraction between, the various Electronic Controller Units (ECUs) of their trucks. Thisreduces the amount ofcostly, time consuming and possibly dangerous vehicle tests.The OAF HIL test system is depicted in Figure 1.5. The PC together with a dSpace real-timesimulation environment (running the vehicle model) is depicted on the left. On the right, thetest frame, equipped with many electrical and electronic systems of the vehicle (including
ECUs) is visible. In the middle, the signals communicated back and forth between the testframe and the vehicle model are also visible. The controls (pedals, levers) needed foroperation are equal to those in a real truck, and are located on the test frame.
Brake pedal postion ["All
Brake si!J1S1 (0/1)
Brake presslJ"e fl'lJ'1t axle (desired)
Brake pressure rear axle rig,t (actual)
Brake pressure rear axle left (actual)~
Brake pressure trailer cortrol valve (desired)~
Engine torque (actLaO
Exhaust brake active (011)
Engine brake active (011)
U",M
Boost pressure
Engine speed
Gearbox output shaft speed
Brake pressure front axle (actual) •Brake pressure trailer control valve (actual)•Wheel speed front axle
Wheel speed rear axle
Outch operv'dosed
Figure 1.5: DAF HIL test system
Until recently, the retarder (downhill endurance brake) controller was not implemented. Torealize this, the vehicle model needed to be expanded with a retarder model. Furthermore, amodel of the vehicle engine cooling system was needed, as it effects retarder performance.Extra actuator and sensor signals had to be transferred between the controller and the model.This report describes the work in order to get a functioning HIL retarder implementation. Thedemands made on the HIL models, are as follows. In principle, the stability margins on theHIL system should not differ much from those for the real vehicle. This requires a sufficientlyaccurate model because modelling errors introduced should not result in unrealistic systembehaviour, compared with an actual vehicle test. Further, important phenomena encounteredin road tests should be reproducable on the HIL system. The model computational load shouldbe such that it can be simulated in real-time.Firstly, a survey on retarders is presented, featuring the hydrodynamic type. The ZF-Intarderwill be introduced and the development ofa retarder model is reported. Then the DAF enginecooling system will be described, followed by a report of the coolant system heat transfermodel development. Both models have been validated during HIL tests, including a stabilityanalysis of the retarder constant speed control loop using HIL measurements, demonstratingthe usefulness ofHIL simulations in both identification and analysis ofdynamicalphenomena. Finally, conclusions and final remarks are given.
Michiel o1't11e ZF-lntarder for TIlL simulations
2 Retarders
2.1 Retarders in generalFriction brakes can generate high torques while braking a fully laden vehicle. The thermalcapacity of these brakes should be very large, because the dissipated power can be very high.This doesn't pose problems for short braking actions, since then the thermal capacity is largeenough to store the energy, and to slowly pass on the energy to the surrounding air over amuch longer period. Problems arise when a long period of braking is needed, especially ondownhill descents. The heat capacity of the friction brakes is not sufficient and overheatingproblems occur, causing brake fade and excessive brake-lining wear [Packer, 1974]. This iswhere retarders come into the picture. Retarders are long-duration, high torque brakingdevices. Engine/exhaust retarders utilize the decompression of the engine to generate an extrabraking torque. An electromagnetic retarder works on the principle of eddy-currents generatedby two opposing sets of electromagnets, whereas the hydrodynamic retarder is based on thegeneration of fluid momentum changes, causing a braking torque.The first hydraulic dynamometer was invented by Froude in 1877 as a device for measuringthe torque of a marine engine. A similar device can be used to generate torques for braking avehicle. Vehicle hydrodynamic retarders mostly feature several constant torque settings,which can be chosen from by the driver, usually by means of a hand lever. More recently,downhill constant speed controlling features have become available, with similar operation ascruise control.
2.2 Position in the drivetrainA retarder can be placed either at the primary (engine) side or at the secondary (driveshaft)side of the gearbox. Primary retarders feature vehicle speed and retarder torque dependent onengine speed and the gear engaged [Packer, 1974]. Unfortunately, engine stall is possibleduring gear shifting with the retarder engaged. In the case of a hydrodynamic retarder, there isa better match between retarder power output and radiator cooling capacity, as the waterpump ofthe vehicle cooling system operates proportionally to engine speed [Packer, 1974].Further, small retarder sizes are possible, due to the gearbox ratio [Forster, 1974]. Mosthydrodynamic retarders are secondary retarders, which feature a torque independent of thegear engaged [Forster, 1974] (Figure 2.1). Limited braking is possible with de-coupled
engine or in case of gearbox failure. Since the hydrodynamic retarder needs high coolantflows, the engine speeds must be kept high, since most engine coolant pumps are coupled tothe crankshaft. Unfortunately, large sizes are needed when used without a step-up gear.Setups using the torque converter of an automatic transmission also exist. Sharinghardware/fluids results in reducing costs and weight, but also in reduced freedom of design,mostly at the cost of a less than optimal retarder.When a retarder is mentioned in the following text, a hydrodynamic retarder is meant.
11
l'Vlichiel of the IF-lntarder for 1111, simulations
2.3 Working principlesA hydrodynamic retarder consists a rotor coupled to the drivetrain, and a stator mounted onthe chassis of the vehicle. The housing of the retarder is filled with a working liquid (usuallyoil). The rotor generates a centrifugal force, causing a radial outward flow, and accelerates theliquid between the blades tangentially. When the liquid reaches the blades of the stator, theflow is diverted, causing a momentum change in the fluid. This generates a braking torque onthe rotor. which decelerates the vehicle.The operation of a hydrodynamic retarder has similarities with a torque converter as used inautomatic transmissions, but a retarder has only two elements. Figure 2.2 shows a picture of aretarder turbine with one toroid (flow chamber). Some retarders use opposing flow cavities(two stators together with one central double rotor), to counteract axial forces on the rotorshaft. The toroid cross-section is usually circular, as in Froude's machine [Packer, 1974], orellyptic.
Stator
Gap
Blade angle a.
Stator
Vortex
Rotor
Figure 2.2 Single toroid retarder turbine
The governing equation for the retarder torque My in stationary situations is [Forster, 1974]:
(2.1)
With W ythe rotor speed, Do the stator/rotor characteristic (outer) diameter, p the liquid densityand A. a shape factor that takes into account all other effects. As can be seen from (2.1), thediameter of the rotor is a very important design parameter. The quadratic dependency on W y istypical for fluid inertia effects. The density p of the medium between rotor and stator stronglydepends on the liquid to air ratio in the retarder circuit. The retarder braking torque is usuallycontrolled by changing the amount of liquid in the retarder circuit and thus the density. Oftenan additional step-up gear with fixed transmission ratio iy is added to the retarder, to increasethe retarder torque at the driveshaft My,ds with a factor i/. As a result, higher torques can begenerated in a smaller package at lower driveshaft speeds. A smaller package needs less oil,which reduces filling response times. Some retarders are fitted with a booster cylinder toobtain faster oil fillings [Forster, 1974]. Apart from the step-up gear ratio in the reardifferential gear ratio idiffand rear wheel radius Rdyn (and the gearbox ratio for primaryretarders) are also very important parameters determining retarder performance (Figure 2.1).
The rotor and stator blades guide the vortex flow. The flow approximately spirals around acircle between the stator and rotor (Figure 2.3, Figure 2.4).
Figure 2.5: Retarder torque asfunction ofbladeangle for a hypothetical retarder {Packer, 1974J
10 80 9040 10 0010 ZO 30
A high mass flow is desirable to obtain high torques,so the liquid for use in a retarder should ideally havea high density and low viscosity. As a result, not theviscous effects (~OJr) dominate the retarder torquegeneration, but the fluid inertia effects (~w/).
Furthermore, a high specific heat is desirable, sothat the retarder liquid temperatures do notbecome excessive. Water would be a well-suitedmedium. Early vehicle retarders used the cooling liquid (which mainly consists of water) ofthe engine as working liquid. Then no additional heat exchanger is required. The drawback isthat the engine cooling liquid is already close to boiling temperature, and the addition ofretarder power to already warm cooling water quickly results into highly undesirable boiling.Most hydraulic vehicle retarders nowadays use oil as working medium. Despite thedrawbacks oflower density, higher viscosity and lower specific heat, oil has the advantagethat it can cope with higher temperatures than water, has excellent lubricating properties andcorrosion resistance, and can be shared with the oil in the gearbox [Packer, 1974].When the retarder is not engaged, a retarder torque is still being generated because of airbetween the stator and rotor. Several devices are used to reduce empty losses by suppressingthe vortex flow, for example by engaging air-swirl suppression slides between rotor andstator.The energy dissipated in the retarder must be transferred to the ambient air. Some retardersuse liquid-to-air heat exchangers, but most hydrodynamic retarders use an oil-to-coolant heatexchanger, adding power to the engine cooling system. Since the retarder maximum poweroften exceeds the maximum cooling capacity of the engine cooling system, the retarder torquewill possibly have to be reduced to prevent coolant boiling.The retarder causes a pressure difference similar to a centrifugal pump. This can be used totransfer the working liquid to the heat exchanger and back (Figure 2.6).
The rotor and stator blades can be set at an angle a tothe tangential and radial plane (Figure 2.2). There isan optimum angle slightly less than a=45° [NarayanRao, 1968], which produces maximum velocitychanges (and thus changes in the fluid momentum),resulting in maximum retarder torque (Figure 2.5).
The pressure on the discharge side (Pro) of the retarder is approximately proportional toretarder torque [Packer, 1974]. The pressure difference pro - pr; across the torus is proportionalto the retarder torque [Packer, 1974],[Pittius, 1990] (Figure 2.6). These relations, relatingtorque to pressure (difference), are often used for retarder control since pressures are easier tomeasure than torques.
2.4 The ZF-IntarderZahnradfabrik Friedrichshafen (ZF) manufacture the Intarder, a hydraulic reTARDER whichis physically INtegrated with a ZF gearbox. The Intarder can be used with a large range of ZFgearboxes. An Intarder in combination with a ZF AS-Tronic gearbox is depicted in Figure 2.7[ZF,1998].
The Intarder, a secondary retarder, is coupled to the driveshaft with a step-up gear. Valves andhydraulic flow channels are implemented within the Intarder casing. The gearbox oil is usedas the retarder working fluid, and both the oil pump and heat exchanger are shared with thegearbox. An opened up view of the Intarder showing several components is depicted in Figure2.8 [ZF, 1998]. A schematic diagram showing the main retarder components is depicted inFigure 2.10. The Intarder and its operation are explained in Appendix 1.A block diagram of the retarder control (during HIL simulation) is presented in Figure 2.9. Ifa constant torque Mr.dd is wanted by the driver, the driver switches the hand lever to a nonzeroposition (33%,66% or 100%), the EST42 retarder controller sends currents Ib and Ip to theretarder (feedforward), resulting in a relatively constant torque Mr.ds, controlled by hydraulic
14
valves varying the retarder oil filling. The EST42 monitors the signals represented by redlines in Figure 2.9 and only limits Ip if any of the torque limiting criteria is exceeded (e.g.maximal power Pr.max ). If the driver enters a setpoint speed Vveh.dd to the EST42, the downhillretarder speed control is activated (blue lines in Figure 2.9). I b is feed-forwarded to theretarder, and Ip is used to control the vehicle speed Vveh. The torque limiting signals are stillbeing monitored and Ip is limited if necessary.The red blocks in Figure 2.9 represent models for vehicle systems, that were not yetimplemented in the vehicle model on the DAF HIL test system at the start of this researchproject. The models have been developed during the course ofthis work, the results of whichare described in the following chapters.
Vveh
Tachograph
abssignal
---.,aveh I
III
---------- 1
-----,Vveh I
IIIIIIII
------------------,
r----I Mr•diJI ,.. __~
: I Vveh.ddI II II II L _
Il _
- - - - - Driver/vehicle interaction-- Constant torque demand- Torque limiting signals-- Speed control loop_ Systems to be modelled
Figure 2.9: Block diagram ofHIL retarder control simulation
Al proportional valve D3 booster cylinderA2 volume control valve EI heat exchangerA34 switching valves E2 coolant temp. sensorB turbine Gl retarder shaftCI oil pump G2 driveshaftC2 pressure limiting valve G3 step-up gearDI booster valve HI transmission
The ZF-Intarder model consist of several sub-models representing combinations of retardercomponents. The model has mainly been based on physical relations since these enhance thepredictive capabilities of the model. However some empirical relations turned out to beunavoidable. In this chapter, the modelling approach for the retarder model is described. Amore extensive report on the retarder model development is given in Appendix 3.
3. 1 Overview of the sub-modelsA block diagram of the proposed total retarder model is depicted in Figure 3.1, showing theinterconnections between the various sub-models. The inputs of the total model are thecurrent Ip commanding the proportional valve, the current h commanding the booster, the airpressure Psys acting on the booster when the retarder is engaged, and the driveshaft speed Wds.
The output variable is Mr,ds, the retarder generated braking torque acting on the driveshaft.
The variables interconnecting the submodels are the retarder speed W n booster output flowqyb,out, volume control valve flow qyv, pump flow qyP' steering pressure Pst, retarder outputpressure pro and pump output pressure PP' Considering the volume control valve with theturbine circuit as a closed-loop control system, the volume control valve can be seen as afeedback controller. The booster can be regarded as a feedforward. Looking at themeasurements in Appendix 2, it is seen that the booster produces a faster response (by fasteroil filling of the turbine circuit), but introduces overshoot.
3.2 MeasurementsFollowing a request for information about the stationary and dynamic Intarder behaviour, ZFhas performed 8 measurements of Mr,ds, Ip , pst, Pro, PP and nels (driveshaft speed) during stepwise engagement and disengagement ofconstant torque settings. The results of these 8measurements together with an analysis have been given in Appendix 2.
3.3 Hydraulic valve modellingThe following equation for the turbulent flow $va[ve through a single orifice valve (see Figure3.2) has been used throughout the retarder model [Streeter, 1961]:
<I> =C.A .!2·(POU/-P iJvalve d 0 ~ P (3.1)
With discharge coefficient Cd, the orifice flow-through area Ao, pressure at valve outlet Pout,
pressure at valve inlet Pin, and liquid (oil) density p. Equation (3.1) has been derived from
17
Bernoulli's law for a streamline. Multiple orifice valves (such asthree-way valves) can be modelled, using the characteristics ofseveral single orifices [Streeter, 1961]. It is assumed that the oil isincompressible, and that the valve flow-through area Au is a linearfunction of valve position Xvalve . The valve position dynamics havebeen neglected since these are much faster than the valve flowdynamics.
Michie! !','<':O('11'
3.4 Sub-models
of the ZF-lntardel' for FilL simulations
Figure 3.2: Single orifice valveschematical sketch
Figure 3.3: Oil pump with pressure limitingvalve schematical sketch
Xvalve
Stationary, the proportional valve (Figure 3.4) linearlytranslates current Ip to steering pressure pst. Theelectrical force on the valve stem F] is nearly linearwith the current Ip [Lausch, 1990]. Looking atmeasurements of steering pressure pst and current Ip ,
the proportional valve dynamic behaviour can bemodelled as a first order system.
The switching valves (operated by pressure Pst) have the task ofclosing the retarder circuit, making torque generation possible.They have been modelled (based on measurements) as a constant.:..F~]~==~=::time delay on retarder torque generation. In Figure 3.1, theswitching valves have been integrated with the 'Proportionalvalve' block.
The oil pump pressure PP (Figure 3.3) is determined by thepwnp and its pressure-limiting valve. Based on measurements,the combination of pump and pressure limiting valve isasswned to maintain a maximwn pwnp oil pressure PP p, cPdepending on retarder speed COr' The pwnp oil flow cPp p P'P"'~,*,_""'&M@"'MMllW'wx"@""",,-"Th
is limited by the maximum pump flow (againdepending on retarder speed cor). The gearboxpressure Pg is assumed equal to ambient pressurePamb=O.
pst Xvalve
Figure 3.4: Proportional valveschematical sketch
The volwne of oil in the booster follows from flow cPb to or fromthe booster. The booster flows are modelled as described inparagraph 3.3, with constant flow-through areaAo. The oilflow filling the booster is limited by the maximwn pwnp flow.The oil pressure at booster output Pb has been asswned equal to 0(ambient pressure), since no measurements were available. The flowexiting the booster is subject to a delay, representing the time it takesthe oil to reach the retarder turbine.
The volwne control valve (Figure 3.5) controls the oil volwne of theretarder turbine circuit (and thus retarder torque) depending onpressures pst and Pro. Stationary measurements show a linearrelationship between Pst and Pro (which, in turn, is approximatelylinear to the retarder torque). The oil flow cPv has been modelled asdescribed in paragraph 3.3, limited by the maximwn pwnp flow Figure 3.5: Volume control valve
cPp . schematical sketch
18
Michie! of the ZF-Intarder for llIL simulations
The turbine circuit can be filled with oil from both the booster and the volume control valve.It can only be drained through the volume control valve. The volume of oil in the turbinecircuit follows from the flows <Pv and <Pb to and from the retarder circuit. It is limited by themaximal oil volume the circuit can hold. The retarder output torque Mr could be calculatedusing equation (2.1). With the assumption that the air density is zero, the density p is linear tothe oil volume in the retarder circuit.When OJr is small the fluid speeds in the toroid increase with increasing OJr . As a result, fluidinertia effects dominate the retarder torque, so then Mr depends linearly on OJ/. But for veryhigh retarder speeds OJr, the fluid speeds between stator and rotor do not significantly increaseany more with increasing (Or, due to viscous fluid behaviour. Viscous shear forces in the planebetween stator and rotor then dominate the retarder torque, and as a result Mr depends linearlyon OJr. The transition from inertia to viscous effects influences the retarder behaviour in itsworking range, so equation (2.1) has been modified with an exponential transition (withrespect to OJr) from fluid inertia effects to viscous effects.The retarder turbine output pressure Pro depends on both the retarder torque M r and theretarder speed OJr. As OJr increases, pro increases. This implies that in stationary situations, theretarder torque slightly decreases for a given pro (or Ip ) as OJr increases.
The retarder heat exchanger will be modelled in the engine cooling heat transfer model (seechapter 5).
3.5 Model identificationMost parameters have been obtained from least squares fits on the measured data, presentedin Appendix 3. The major part of the retarder dynamical properties is determined by 5parameters throughout the sub-models. Therefore the complete retarder model has beensimulated, using measurements of the retarder model input variables as model inputs. The 5unknown model parameters have been be estimated minimizing the following least squaresfunction err: .
1 T ( )2err =- . M t - M t dtT f r,ds,meas ( ) r,ds,model ( )o
(3.2)
With measured retarder driveshaft torque Mr,ds,meas, model predicted retarder driveshaft torqueMr,ds,model, time t and the simulated time interval T. This expression is nonlinear due tononlinearities in the model equations. The MATLAB fminsearch minimalisation routine hasbeen used. It uses the Simplex (Nelder-Mead) method, can cope with nonlinearities, for a fewparameters, and is relatively fast and reliable since it adapts the search direction to the shapeof the function to be minimized [van den Bosch & van der Klauw, 1994]. The modelparameters have been determined by a minimization for every single measurement, followedby averaging of the results. The intermediate and final results have been reported in Appendix4. A summary of the final retarder model is also presented in Appendix 4.
3.6 Model validationUnfortunately, there is no set of independent measurements available to validate the retardermodel. Instead, the set of measurements presented in Appendix 2 is re-used to evaluate thequality of the model prediction (more validation results have been reported in Appendix 4).Figure 3.6 shows the model prediction for a situation with a low torque demand at lowretarder speeds. The simulation result corresponds very well to the measurement.
19
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
4000
2500
3000
----, - - -- -r -- --,- - --- r- ---,- -- - - T - - - --II I I I I I Measurement III I I I I ,- Model I----'-----r----'-----r----'-- T II I I I I t II I I I I I I
----'-----r----'-----r----~-----T-----I
I I I I I I II I I I I I I
----'-----r----~-----~----~-----T-----II I I I I I II I I I I I I
----1-----r----~-----~----1-----T-----1
I I I I I I II I I I I I I
----4-----~----~-----~----~-----~-----1I I I I I I II I I I I I I
----~-----~----~-----r----~-----+-----II 1I I 1
500 -----1- --1----- .... -----1------1----- -----1I I I I I I II I I I I 1 11---'" --- -I- - - - - .... - - - - -I- - - - - -1- - - - - -1--I I I I I I II I I I I I I
-500 - - - - -1_ - - - - ~ - - - - --1- - _ - - ~ - - - - -1- - - - _.a.. - - - __ II I t I I I II I I I I I I
_ 2IlOO
!~1500~
':Ii:1000
-10000~-~~--.L---~---:---:':'O~--:':12:-----',4
limo lsi
Figure 3.6: Model validation nds =500 [rpm]. 20% torque demand
Figure 3.7 depicts the simulation results at relatively high retarder speeds and maximal torquedemands.
- - - - .... - - ---1----- ~--- --1- -- - - -- - --I 1 I I 1(\I I 1 1 I~ \
.....__-- - - - -I- - - - - -1- _ - - -I- - - - - _1 -I-,\.loo_-1I 1 I I I
1 I I I t 1-500 - - - - ..J - - - - - l- - - - - -1 __ - - - l- - - - - -1- ... _
I I 1 1 I II 1 1 1 1 I
~1500{
:i- 1000
-10000L---'------'------"------""e-----"o----,L
2---'14
limo Is]
Figure 3.7: Model validation nds =1500 [rpm], 100% torque demand
Again the measurement is predicted well. The differences in peak timing (at t=3 [s]) arecaused by the assumption that booster output pressure Pb equals 0 (see also Appendix 3). Itnow appears that for situations with high retarder torques and high retarder speeds (resultingin a high retarder output pressure Pro), booster output pressure Pb is actually higher than zero.It counteracts pressure Psys, resulting in a lower booster output flow and hence the boosterbecomes empty slightly delayed. Therefore, the retarder torque measurement shows lessovershoot than the model prediction and a delayed point of maximum torque.Summarizing, a higher dynamical model accuracy can be obtained by further investigating thebehaviour of the booster output pressure Pb. Neglecting the valve stem position dynamics hasnot led to large errors. Although the fluid inertia effects dominate the retarder torque, viscouseffects could not be neglected when describing the retarder torque generation.
20
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
4 DAF cooling system
T. . , Teng,oul irel,ln i I
i._ , _-'
The OAF cooling system (Figure 4.1) has been modelled, since it effects the cutback ofretarder perfonnance due to a retarder controller protection against high coolant temperatures.
K3 Fan drive and clutchKiLl ThermostatK2 MI Intercooler heat
exhangerNI Engine oil cooler
Figure 4.1 DAF cooling system layout
The coolant is propelled by the coolant pump, and enters the engine block. Here a small partof the flow enters the oil cooler and returns towards the pump. The main coolant flowexchanges heat with the engine and exits towards the retarder (and gearbox) heat exchanger. Itexchanges heat with the retarder/gearbox oil, and exits towards the thermostat. The coolanttemperature detennines the thermostat position. Depending on the thermostat position, part ofthe coolant flow enters the bypass and returns to the pump. The other part passes through theradiator, is cooled due to heat exchange with the ambient air and then returns to the pump.The coolant from the radiator and the bypass flow are mixed in the pump, and enter the. .engme agam.The engine controller (UPEC) monitors the coolant temperature at the engine outlet Teng,ouh
and controls the fan. The fan is located behind the radiator and can raise the air-flow throughthe radiator and as a result the radiator heat exchange.The retarder controller (EST42) monitors the temperature at retarder outlet Tret,ouh and adjuststhe retarder braking power (by lowering the proportional valve current and as a result thegenerated torque) if this temperature becomes too high. The cabin heater, compressor heatexchanger and ventilation lines have been neglected here, since they have a relatively smallinfluence in the total coolant system.
._-- --------_....-.----
21
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
5 Coolant heat transfer model
In this chapter, a description of the coolant heat transfer model is given. A more extensivereport of the model development is presented in Appendix 6. The starting point for modeldevelopment was the model described in [Manders, 2001]. Validation with measurementsshowed that this model did not predict vehicle measurements well. Therefore a new modelhas been developed. After the fIrst new model concepts, a number of vehicle tests wereformulated that ought to provide the necessary data to identify all unknown model parameters.A summary of the measurements (test descriptions and measured variables) is presented inAppendix 5. For each component, a sub-model has been developed separately (Figure 5.1).All sub-models interconnected result in the total coolant heat exchange model, which hasfinally been validated.
T''''''G Retanlcr ,upply h"" Engine
Pjuel
;:~L..-__""
~~
Figure 5.1: Heat transfer model layout
Some general assumptions made are:
• Constant material properties (independent of temperature/time).
• No power losses to the surroundings except in the radiator.
• Where possible, a lumped heat capacity approach has been used.
5.1 Retarder heat exchangerThe total retarder power P r = Mr' (J)r is dissipated in the oil in the retarder turbine. (Figure5.2) and passed on to the coolant through the thin heat exchanger walls separating the oil fromthe coolant. A very schematical representation of the retarder heat exchanger model isdepicted in Figure 5.4.
Retarder heat ....""Trer,in .~
exchanger ~ ,Retarderturbine
coolant
Figure 5.2: Retarder turbine and heat exchanger
-------------------------
"
Figure 5.3: Retarder heat exchanger in detail
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
Pret,he
<I> ,T .ret ret In
~Pr 1-.~ -r-.s-+-l I-.~
rel,m
<I> ,Tret ret,OUt
Figure 5.4: Retarder heat exchanger model
~95
110r--,--,-----r--,----,--,-----r----,---,I II II I I I I I
1ai - --+-----'......~-=---+ -- - -1-- - -1-- ---I- --- .... ---I I I I I II I I I I II I I I I I
100 ---+-- ---+---~----~---t---~---I I I I I tI I I I II I I I ~:;...o'"_
---T- --,---,----,--I I I I I I I I
..a I A. I I I I I I- I ~,I I I I I,:::I ]:~I\ :I:::r::r::;:::;:::
I I' I I I II I I I I II I I I I I I I
8Oc:~'V'~-i----i---~---i----:-- - T....outllimUilt8d)I I I I I - Tr8l.out (mesSJred)
This model with a frrst order system with time constant iret,m describes measurements well(Appendix 6). Pret,he is the heat transfer from the oil to the coolant in the heat exchanger, (/>ret
is the retarder coolant flow, c,et,c is the coolant heat capacity and Tret,in and Tret,out are the inletand outlet coolant temperatures, respectively.To estimate the model parameters, the RMS value of the measured and simulated Tret,out hasbeen minimized, using the MATLAB fininsearch routine. For each parameter, the mean valuehas been taken of the optimisation results of two measurements.The model has been validated usingindependent measurements (that havenot been used to develop the model).The model input variables have beenused from measurements. Figure 5.5shows the measurement together withthe model prediction of output variableTret,out. The measurement featuresretarder braking (downhill) followedby a period of engine operation(uphill). The simulation result predictsthe measurement well. Thetemperature offset is probably causedby a sensor offset in the measurement.
Figure 5.5: Simulation and measurement results duringmaximum retarder and engine operation
5.2 EngineOf all the fuel power Pjuel supplied to an engine, only a small part P iso is obtained as usefulpower in the form of torque at the crankshaft. The rest of the fuel energy is mainly lost asheat, part of which must be drained away through the engine's cooling system (Pcoolant in theheat balance [Aldenzee, 1999] shown in Figure 5.6). Unfortunately, the engine power balanceduring part load engine operation differs significantly from the power balance during fullload. Therefore, the engine model has been made dependent on the engine load.
---_.._._-----
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
plue/ =100%
- Pjso
_ Pexhaust
o pcoo/ant
o Pinterc
o Prest
cylinder head
engine block
.L-U-.....L-U..__..IIIII
I I1- ,
cylinder head wall
combustion chamber
piston
Figure 5.6: Power balance during fulilDad Figure 5.7: Engine power flows
Figure 5.7 schematically shows the layout of the upper part of an engine (only a part of onecylinder has been drawn). The coolant enters the engine block (with temperature Teng,in), andflows upward through slots positioned around the cylinder. Then it passes through thecylinder head (containing the air inlet and exhaust gas outlet ports) and exits the engine (withtemperature Teng,out). The engine metal components (block and head) are not uniform intemperature, since parts close to the combustion chamber (the heat source) have much highertemperatures than the bulk of the engine components. That is why the metal thermodynamicshave been split into two groups (Keng in Figure 5.8). In Figure 5.8, 'feng,walls is the time constantrepresenting the combustion chamber walls (and other relatively hot parts close to the coolantducts), Ceng,block is the engine lumped heat capacity, with associated block temperatureTeng,block. Ceng,c is the coolant lumped heat capacity, and Teng,c is the coolant outputtemperature. It is subjected to a delay 'fd,eng representing the time it takes the coolant to passthe engine coolant channels and hence depends on the coolant flow <Peng•
Ceng.block
Teng.block
Teng,c <I> ,Teng eng ,out
Figure 5.8: Engine model
The arrow colors in Figure 5.8 correspond to those used in Figure 5.7. The engine parametershave been estimated in a manner, similar to those of the retarder heat exchanger model(Appendix 6).
24
Michiel Pesgens Modelling of the ZF-Intardcr for HIL simulations
300 400 400200 200 300limo[s)
100 10050
I I I I I II I I I I I
~ ---+---~~---~---+---~----~-I I I I I I
90 1 : : :__ 1__ - :_JI I II I II I I
8B ---1"----1----I II I
~86 ----:----~--1!. I I
I-~ ---i--- I
II
82 ---+--II
94,.-----,-----.-----.----,-----,--.----,---..,.---,The validation is also similar (Figure5.9). The initial part of the simulationpredicts the measurement well.However, after t = 330 [s], thetemperature difference between Teng,in
and Teng,out is not predicted well,because the engine braking power hasnot been taken into account in theengine model. To enhance the accuracyof the engine model this braking powershould be included in the future.
Figure 5.9: Simulation and measurement results duringconstant speed retarder operation
5.3 RadiatorThe radiator exchanges power between the coolant and the ambient air. The air-flow throughthe radiator is determined by the fan speed and the vehicle (wind) speed. The intercooler heatexchanger is positioned in front ofthe radiator and effects the radiator's performance.The temperature of the radiator metal is taken equal to coolant temperature, since the maintemperature gradient will occur between the ambient air and radiator metal. The radiator hasbeen modelled as one buffer, which can extract heat from the cooling system and pass it on tothe surrounding air. A representation of the model is depicted in Figure 5.10. Here, Crad is the
<I> ,Trad rad,in
~• Prod,co
<I> ,Trad rad ,out
Figure 5. J0: Radiator model
total radiator's lumped heat capacity and C/Jrad is the coolant volume flow through the radiator.A time delay id,rad, dependent on the coolant flow through the radiator, represents the timeneeded for a coolant particle to pass through the radiator channels.Using the DAF V99 stationary cooling system simulation program, an interpolation table hasbeen obtained for the calculation of cooling power Prad,ca. Prad,ca depends linearly on thetemperature difference between coolant and ambient air. Since the intercooler heat exchangeris positioned in front ofthe radiator (Figure 4.1), Prad,ca is diminished by the intercooler powerfrom the engine power balance Pinterc. A fIrst order system with time constant Tinterc has beenapplied to the intercooler power, representing the intercooler heat exchange.The variables determining the radiator's dynamical behaviour have been fitted in a similarmanner similar to the retarder heat exchanger and engine (Appendix 6),
Michiel Pesgens Modelling of the ZF-Intardcr for HIL simulations
~ ---+- ~----~---+---~----~---~---~---, t I I I I I II r I I I I I I
~ ---+ - ~----~---+---~----~---~---~---I I I I I I I I
I I I I I _ Trad.oul (simulated)
~ - - - - ~ - - - ~ - - - ~ - - - - ~ - - Trad,out (measJred)L.....o - I I I I I - T
rad.in
(meeSJred)
Figure 5.11 shows the validation resultsfor an experiment with retarder braking,followed by uphill engine operation.Initially, the thennostat is closed. Theinitially flow (/>rad (a model input) has arelatively high error, since it has beenreconstructed from pressuremeasurements (Appendix 5). Hence theinitial simulation errors. The followingpart (with larger radiator flow (/>rad) ispredicted much more accurate.
The fan considered here is a fan, driven by the engine crankshaft by means of a fan drive(with ratio ifan) and an electronically controlled hydraulic fan clutch. The maximal fan speednfan.max is limited by the engine speed n eng• The fan clutch is controlled externally controlledby the UPEC (engine controller), so the fan speed can be manipulated more or lessindependently of the engine speed. has very complex dynamic behaviour, and also varies forindividual fan clutches. When a fan deceleration is demanded, the fan clutch needs to purgeitself of oil, which can take a considerable amount of time. During this time, the fan clutchslip ratio remains fixed, and as a result, the fan speed "sticks" to the engine speed.The dynamic behaviour of the fan itself fan has been modelled as a first order system. The fanclutch however, has only been modelled with a very simple model, with a fixed delay on fandecelerations. In reality, the fan "sticking" delay varies significantly. As this simple fan clutchmodel reproduces the most important phenomenon, it is sufficient for HIL-simulationsreproducing road test phenomena. The fan controller has been modelled as in [Nevels, 2001](SIL implementation), but the actual UPEC actuator signal can be used during HILsimulations when the necessary actuator and sensor signals are transferred between the UPECfan controller and the model.The evaluate the quality of the fan (and associatedcomponents) model, together with the fancontroller and UPEC sensor model, a simulationhas been perfonned, with the model inputs takenfrom measurements. The measured and simulatedmodel output nfan has been compared. The resulthas been depicted in Figure 5.12. There seem tobe significant discrepancies betweenmeasurement and simulation results of the fanspeed nfan. To further reduce the errors, a morecomplicated fan model is necessary. Furthermore,the UPEC sensor time constant Tsensor should beestimated more accurately, since the time the fan Figure 5.12: Fan nwdel validationstarts to accelerate is not equal for measurement andmodel prediction.
26
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
Ring
Piston
Wax
I I
T th.min Tth.max
Tth.hYSII I
--------------------~~----
ot
5.5 Coolant pump and thermostatThe coolant pump determines the total volume flow 1Jpump, linear with engine speed. Theradiator flow 1Jrad and the bypass flow 1Jbypass are mixed in the coolant pump and determinethe engine inlet coolant temperature Teng,in· The thermostat position can be regarded as afluctuating resistance in the coolant duct system, and as a result it effects all coolant flows.The thermostat is a conventional wax thermostat. The thermostat divides the retarder coolantflow 1Jret in a flow component 1Jrad through the radiator, and a flow component through thebypass 1Jbypass (Figure 5.14).The thermostat's dynamical behaviour has beendetermined from measurements (Appendix 5), andis modelled as a fIrst order system. The stationarycharacteristics have been modelled as shown inFigure 5.13.
The influence of the thermostat position and engine speed on coolant flows has beendetermined using the measurements described in Appendix 5.
5.6 Coolant ductsThe relatively long coolant hoses and ducts (retarder supply duct, retarder drain duct, indirectcoolant return duct) have been modelled free of power losses, but with a delay time dependingon coolant flow, representing the time it takes a fluid particle to pass through the duct.
5.7 Temperature sensorsThe temperature sensors for the UPEC and the retarder temperature sensor have beenmodelled as a fIrst order system with time constant l'senson representing the heat exchange withthe coolant.
5.8 Model validationA summary of all coolant heat transfer model parameters is given is Appendix 7.To evaluate the model quality, the model will be simulated using measured model inputs of anindependent measurement that has not been used for model parameter estimation. The modeloutput temperatures have been compared with their measured counterparts (Appendix 7contains more validation results). The initial temperatures have been matched. The measuredfan speed has been used as a model input. Figure 5.15 shows a long retarder braking action inretarder constant speed control mode set at 90 [km/h].
27
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
I I I I I I - Trl1t,in- r - - - - T - - - -.., - - - - -,- - - - -,- - - - - r - T'eI.out
........ ~ ........ __l_ ............ J................ _: ................ _:.................... ~ - Tth~I I I I I I Trad,in
I I I I I I - Trld,out
70
~....60
50
40
30
3> -
100
Figure 5.15: Temperature comparison, downhill with retarder constant speed control
The initial thermostat opening time is different for measurement and simulation. This couldbe caused by a thermostat in the measurement that opens at a lower temperature. The initiallyhigher temperatures are also caused by the delayed opening ofthe thermostat, causing theamount of cool radiator coolant to mix with the bulk of the coolant at a later time. Thefollowing model time responses do correspond well to the measurements. The overalltemperature level is slightly higher, since no power losses to the ambient air have been takeninto account (except the radiator). As a result the radiator model needs a higher temperaturedifference of the coolant versus the ambient air temperature to dispose of this extra power.
28
Michiel Pesgcns Modelling of the ZF-Intarder for HIL simulations
6 HIL validation
After the individual validation of the retarder and heat transfer models they have beenimplemented on the Hardware-In-the-Loop test system. The necessary signal connections andsignal conditioning operations have been described in Appendix 8.
6.1 Uphill and downhill drivingA HIL simulation has been performed to validate the functionality of the models with theretarder controller and all other vehicle systems. The input was a sinusoidal road slopevarying with time (Figure 6.1), representing a mountainous route. Only the accelerator pedal(with gear shifting) and maximal (100%) manual retarder braking have been used, with a fullyladen vehicle (40 tons total mass).
Figure 6. I: Uphill/downhill driving H1L test with retarder and engine operation
The retarder power cutback due to high retarder outlet coolant temperature (Tret,out) is clearlyvisible. The proportional valve current Ip is lowered, and the retarder torque Mr,ds drops. Theretarder braking is not sufficient any more, and the vehicle speed Vveh keeps increasing (duringnormal road driving, the driver would operate the brake pedal to prevent this). The initialtemperatures are higher at the start of the second retarder braking action, so the time thatmaximal retarder torque is available is shorter than for the fIrst retarder braking. Figure 6.3shows a similar road test, with a downhill full retarder torque demand, followed by uphillengine operation with shifting. The main differences between measurement and simulationare caused by differences in external conditions, like the road slope.
--------------------29
Michiel Pesgens Modelling of the ZF-Intarder for HIL simulations
12Jr---r---,---,---,---,..---.,..------,
100
£90,..
110 ----J-----L----~-----t-----:-----+----I I I I
---~-----r-----I-----T----I I
="110_..........\ ..... - ,- - - - - r - - 1- - - - - r - - --I I I I
Figure 6.2: Downhillfull retarder braking and uphill engine operation during vehicle test
6.2 Retarder speed controlDuring HIL tests of the retarder constant speed control function (Bremsomatic) in downhilldriving, continued oscillations in the vehicle speed Vveh occurred (Figure 6.3) with allsimulated vehicle configurations and speeds.
Figure 6.3: Bremsomatic vehicle speed response during Figure 6.4: Bremsomatic vehicle speed response duringHIL test vehicle test
During real life vehicle tests, this phenomenon is also seen in similar long descents with areasonable constant road slope (Figure 6.4). Moreover, Zahnradfabrik Friedrichshafen (ZF)states that they experienced similar behaviour in their SIL simulations and that the stationary(oscillating) behaviour is sufficiently comfortable and not noticeable for the driver. TheBremsomatic speed control loop is depicted in Figure 6.5. The Bremsomatic controller in theEST42 is a PI-type controller. Details of this controller, however, are unknown to DAF, andZF is not willing to give more information.
Vvehsp ... EST42 DC~ Ip Mrds VVeh...... Bremsomatic PWM ... Retarder model ' .. Vehicle model... controller...,
demodulation... ... ...
... ,VVeh
,,,,,Tachograph ~
, DfF .......I
....,,,.-- -- -- -------------------------------------------------Figure 6.5: Bremsomatic control loop
The controller loop is nonlinear, since the retarder model and some parts of the vehicle modelare nonlinear, especially with respect to the retarder speed. Furthermore, the vehicle dynamicbehaviour can vary substantially with speed, load, gear etc. That is why a qualitative approachto explain the origin of these speed oscillations will be pursued. For small vehicle (andretarder) speed variations, fixed gear, and constant mass load, the system can be assumed tobe linear, so a linear analysis of the closed-loop system will be performed. To investigatewhat could be the cause of the oscillations, Frequency Response Functions have beendetermined from measurements taken on the HIL system. A block diagram of the closed-loopsystem consisting of the EST42 retarder constant speed controller (C) together with the totalHIL vehicle model (including the retarder model) (P), is depicted inFigure 6.6. Signal conditioning operations have been omitted for clarity.
-----------------------------------------dSpace
!V\!\NM1------------------------,,EST42 I,
WIII
V +" u:Ysp y, ..~ C ~ P ...~
I + -~~II
YIIII I,, II IL________________________ _ 1
Figure 6.6: Block system ofsumming node HIL test
A summing node approach has been adopted to determine the open loop Frequency ResponseFunction Hoi. The digital signal Ip has been chosen for implementation of the summing node ischosen. An excitation signal w is injected into the closed loop. Here, a repeating 'sweep'signal with frequencies from 0.02 to 1 [Hz] is chosen. The following signals have beenrenamed: v is the EST42 output after PWM demodulation, U is the retarder model input and ythe vehicle speed (vehicle model output). Since the HIL models are simulated at samplingfrequency Is = 1 kHz, signals U and v need to be downsampled to reduce the amount of data tobe processed (resulting in signals ufand vfi since the total time length Tfor each measurementis 4000 [s]), to obtain enough frequency points with satisfactory accuracy. To preventaliasing, two 8th order digital Butterworth low-pass filters have been used to resample thesignals U and v in two steps (Figure 6.7). Two subsequent filtering and downsampling stepshave proven necessary to avoid digital filtering problems.
31
u 8th order 8th order uf---. Butterworth .... Z.O.H. .... Butterworth r--+ downsampling r+[breakoff = 50 Hz
... r. = 200 Hz...
[breakoff = 10Hz r. = 50 Hz
Figure 6.7: Downsampling ofsimulated signals
-1-1.5-2
1 1 1 1 _ 104' O.l000[AJ1 1 1 1 _ 104' 0.0500 [AJ
1 - L - - - - - i - - - - - .J - - - - - -1- - --=--~-_~_ - 104' 0.0250 (AJ1 I I 1 I _ 104' 00125 [AJ
I -lJnitcitcteI1
1~~~~~~~j ~ l _
III1
1
I 1----r-----r-----
1 1
I II II I1 1
--r------t------1 1I 1
1
I
1
I1
1
1 I-0.5 - t- - - - - - + - - - -
I II I1 1
I 1I I I I
-1 -~-----~-----~------I---I I I II I I I
(6.1)
Figure 6.8 shows the Nyquist plots ofthe open loop transfer functions forfour experiments with differentamplitudes ofthe input signal w, forequal frequency ranges, omitting lowcoherence data. Almost identicalresults have been obtained byreconstructing Hoi from closed loopsensitivity FRF data. Appendix 9shows the Bode diagram, togetherwith the coherence of this Hoi FRFdata.
The closed loop transfer function (from ufto vf) has been detennined with the MATLAB tferoutine using a Hanning window. Here, UJ...s) and VJ...s) are the frequency domain equivalentsofufand vf
The phase margin is the limiting factor with respect to the (linearized) closed loop stability. Itdecreases with decreasing disturbance signal amplitude Iwi. This indicates that the dynamicbehaviour is nonlinear. Furthermore, the magnitude of the open loop transfer functionincreases in amplitude for low frequencies, with decreasing disturbance signal amplitude Iwi.This indicates that an integral gain is increasing with decreasing input amplitude, such that forvery small amplitudes, the closed loop system will become unstable. The time response of theoutput y converges to an oscillatorystate, which is marginally stable forthe linearized system (Nyquist plotthrough the point -1 +o·i).Figure 6.9 shows a phase plot of threeresponses (with different initialconditions), together with thelinearized system stability. For thenonlinear system, the blue lineconstitutes a limit cycle, the setpointVveh,sp is an unstable equilibriumpoint.A situation where similar behaviouroccurs is in the case of a feedbackcontrolled mechanical system withdry friction together with an integralcontroller. Another situation for whichsuch dynamic behaviour could occur is anonlinear integral controller integrating
, .., ... ~
the error's sign with time. For small errors the linearized closed loopgain would inevitably become too high, resulting in an unstable system.
Figure 6.10 shows the input andoutput of the retarder controller forthe measurement of Figure 6.3.Obviously this controller is nonlinear,with constant slope sections in Ip
(green dotted lines), switching at thesetpoint speed, which could be anindication of similar controllerbehaviour as just mentioned. Moreinformation from ZF or furtherresearch is necessary to determinewhether or not the oscillating systembehaviour is really caused by thecontroller. On the HIL system, theEST42 retarder controller could bereplaced by a linear PI-controller. Itwould be very interesting if thiseliminates the limit cycle, as thatwould imply that the EST42controller causes this unwantedbehaviour.
Hardware-In-the-Ioop models for the ZF-Intarder as well as the cooling system have beendeveloped. Both models have been validated with measurements. The retarder dynamicbehaviour can be improved slightly with a model d~scribing the booster output pressure. Theheat transfer model can be further improved by the development of a more accurate fanmodel, by taking into account the engine braking power and by accounting for the powerlosses to the surroundings.The models have been implemented on the Hardware-In-the-Loop test system, and theinteraction with the retarder controller and other vehicle systems has been verified.Phenomena encountered in road tests like retarder torque cutback due to high coolanttemperatures can be reproduced during HIL tests. Oscillations in the vehicle speed occurringwith retarder downhill constant speed control in both HIL tests and vehicle tests have beenanalysed using HIL measurements. The stability margins ofthe closed loop system have beenshown to be insufficient, using HIL measurements. Since these measurements are almostimpossible during vehicle tests, HIL simulations proved to indispensable in stability analysesof the retarder controller. A further analysis of the origin of the retarder speed controloscillations and the development of alternative control strategies to counteract or suppress theoscillating responses could be the subject of further research.The HIL coolant heat transfer model can also be used for concept analyses of possible futurechanges the cooling system. Examples would be the use of different coolant temperatures forthe fan controller or an electronically controlled thermostat, to obtain faster responses to highcoolant temperatures during retarder braking.
34
References
[van den Akker. 2001]H.M.A. van den Akker, "The setup of a hardware-in-the-loop test frame for electronic brakesystems", Afstudeerrapport TU Eindhoven, DAF intemal report 51211/01-089, DCT report2001-41,2001.
[Aldenzee, 2000]A. Aldenzee, "Definitieve warmtebalansen XE-C, 10500 in de 315-280 en 250 kWuitvoering", DAF internal report 51406/00-337, 2000.
[Arici, Johnson and Kulkami, 1999]O. Arici, J.H. Johnson and A.J.Kulkami, "The Vehicle Engine Cooling System Simulation,Part 1- Model Development", SAE Technical paper 1999-01-0240, 1999.
[van den Bosch & van der Klauw, 1994]P.P.J. van den Bosch, A.C. van der Klauw, "Modeling, Identification and Simulation ofDynamical Systems", 1994.
[Forster, 1974]H.-J. Forster, "Retarders Built into Automatic Transmissions", IMech 1974 Conference"Retarders for Commercial Vehicles", C7/74, 3 january 1974.
[Frid, 1999]I.J.G. Frid, "Het opstellen van Hardware-in-the-Loop simulatie ten behoeve van hetparameteriseren van motor ECU's bij DAF Trucks N.V." Afstudeerrapport TU Delft, 1999.
[Packer, 1974]M.B. Packer, "The Development of Hydrokinetic Retarders", IMech 1974 Conference"Retarders for Commercial Vehicles", C1174, 3 january 1974.
[Perset and Jouannet, 1999]D. Perset and B. Jouannet, "Simulation of a Cooling Loop for a Variable Speed Fan System",SAE Technical paper 1999-01-0576, 1999.
[Pittius, 1990]Reinhold Pittius, "Zum Bremsverhalten zweiachsiger Nutzfahrzeuge mit Retardern",Hannover 1990.
[Schoeber, 1990]E.J.1. Schoeber, "Remslijtegebeperking door retarders", Eindhoven Technical University,Voertuigtechniek, report number WOC - VT - 90.01, DAF Trucks report number St9006089,1990.
[Streeter, 1961]Victor L. Streeter, "Handbook of Fluid Dynamics", First edition, 1961.
[Wilbers, 2001]R. Wilbers, "Flow en drukmetingen in het koelsysteem", DAF internal report 51510/01-288,2001.
[ZF,1998]"ZF-Intarder, Der integrierte Retarder fUr ZF-Getriebe mit EST 42, HandbuchfUr Einbau und Inbetriebnahme", ZF Friedrichshafen AG, Publ. nr. 6085 765 004a., Ausgabe03.98.
[ZF,2001]W. Hubler, "Simulationsmodell zum 420 kW ZF-Intarder von H. Eiden", ZF Friedrichshafen,TE-H,2001.
