S. Saurí and E. Martín 1
Import Storage Yard Layout in Vehicle Terminals 1
2
S. Saurí and E. Martín. 3
4
S. Saurí (corresponding author), CENIT – Center for Innovation in Transport, 5
BarcelonaTech-UPC, Jordi Girona 29, 2A, 08034 Barcelona (Spain), 6
Tel: +34 934.137.667, Fax: +34 934.137.675, [email protected]. 7
E. Martín, CENIT – Center for Innovation in Transport, BarcelonaTech-UPC, 8
Jordi Girona 29, 2A, 08034 Barcelona (Spain), Tel: +34 934.137.667, 9
Fax: +34 934.137.675, [email protected]. 10
Submission date: July 30th, 2011. 11
5743 words (abstract and references included), 5 figures, 2 tables. 12
13
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 2
ABSTRACT 1
This paper addresses the problem faced by terminal operators in organizing the import storage 2
yard of vehicle terminals. An analytical model is developed to quantify rehandling moves and 3
storage density for each storage layout, characterized by a shape parameter. Afterwards the 4
total cost function of the yard is defined. The target of the paper is to set up the storage layout 5
which minimizes total cost and maximizes terminals’ capacity. A numerical case, based on a 6
real terminal, is tested. Results show the optimal layout is quite sensitive to labor and 7
opportunity costs. Nonetheless, block lines with more than two import vehicles are 8
economically feasible. 9
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 3
1 I0TRODUCTIO0 1
Automobile industry and its logistic operations in the auto supply chain have included port 2
terminals as strategic nodes, hoping for better economies of scale. At the same time, activities 3
such as component, module assembly or other added value activities are being included in 4
vehicle terminals; hence port terminals are a key part of the global supply chain of vehicles. 5
On the other hand, the driving strives for efficiency in a highly competitive 6
environment has resulted in creating a net of dedicated port terminals and hubs. Consequently 7
port terminal operators have strengthened their efforts to optimally manage their resources 8
and storage areas in order to increase terminal’s productivity. 9
Storage yard of port terminals is perceived as one of the main areas; hence their layout 10
should take into account the infrastructural decisions with options to be taken by operational 11
planning of vehicles (1). As it is known, import and export processes differ in the way 12
vehicles are taken into and retrieved from a storage yard. Vehicles for import are stored as a 13
batch but their departure is a subject to stochastically arriving customers’ delivery orders, 14
whereas export vehicles are consolidated according to their final destination in order to 15
warrant an effective loading process in a batch. 16
Therefore, the layout of storage yards provided for processing batches is distinguished 17
from the layout suitable for stochastic retrieval. Another difference between import and 18
export vehicles is a dwell time in the terminal associated to each one. Import vehicles remain 19
in the terminal longer than export vehicles, and the dwell time in the terminal is defined 20
according to different delivering strategies (2). 21
This paper will focus on inbound logistics operations and more particularly on those 22
which take place in the storage yard. Moreover, we will give special attention to rehandling 23
moves and their effect on the operational cost. Rehandle is a consequence of storing import 24
vehicles in lines which length is longer than 2 vehicles because its departing patterns are 25
random and unpredictable, as it is demonstrated in (1). 26
Afterwards, we will propose and analyze economically different storage block layouts 27
in order to maximize the utilization of available space and to minimize operational cost, that 28
is, try to find the trade-off between storage utilization and operational cost due to rehandling 29
moves in economic terms. 30
This article is structured as follows: first literature is analyzed by reviewing the major 31
contributions in the design of port configuration and rehandling moves. Chapter 3 introduces 32
the problem and Chapter 4 defines and develops the methodology used in this article. The 33
analyzed case study and respective results are presented in Chapter 5 that will allow us to set 34
some conclusions (Chapter 6). 35
36
2 LITERATURE REVIEW 37
Optimization of vehicle terminal operations is a field within transport research which has 38
recently become a subject to explore. Generally, the vast majority of studies have focused on 39
improvement and optimization of processes developed in container terminals. 40
In the terminal vehicles are moved by drivers organized in gangs, so terminal 41
operations are very cost-intensive. Moreover, if rehandling moves were required to retrieve 42
vehicles from a line, the operational cost would be highly increased. In order to minimize the 43
cost, storage strategies, decision models for manpower planning or deployment scheduling 44
have been analyzed by few authors, although rehandling moves have been studied mainly by 45
De Castilho and Daganzo (3), Kim (4), Huynh (5) and Saurí and Martín (6) just for container 46
terminals. 47
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 4
A multi-agent system for integrated storage allocation and deployment scheduling was 1
developed by Fischer and Gehring (7) to minimize and balance the number of drivers 2
required in the planning period. Similarly, Mattfeld and Kopfer (8) set out a decision planning 3
and scheduling system intended to support automated terminal operations in Bremerhaven. 4
Both studies demonstrate that heuristic tools are suitable to support the current storage 5
allocation problem and deployment scheduling. 6
Regarding the literature concerned with organization of the storage area of vehicle 7
terminals we should mention the contribution of Mattfeld (1) who stated the efficiency of the 8
operational processes and terminal’s productivity depend on the organization of the storage 9
area. 10
There have been many studies on the optimization of the design parameters of storage 11
systems. Berry (9) analyzed the layout of conventional warehouses, Bozer and White (10) 12
suggested an algorithm for designing picking systems, Lee and Kim (11) proposed two 13
methods for optimizing block size and Iranpour and Tung (12), among others, stated that 14
parking areas should be designed according to the storage capacity. They proposed a new 15
design that maximized its capacity and reduced parking moves to the minimum. The 16
implementation of this design has had great success, not only in storage yards but in any 17
parking area. 18
Additionally, Cassady and Kobza (13) focused their studies on finding the optimal 19
parking slot in order to reduce the access time and the distance to the parking lot, considering 20
two strategies. Other authors who provided a quantitative approach to the design of parking 21
areas which can be adapted to car terminals are Bingle et al., (14). They reduced the turning 22
radius to reduce incidents and vehicle damages. 23
In short, this paper will combine the design process of a storage block, limited by the 24
terminal capacity, with the restrictions of inbound logistic processes of vehicles. That is, we 25
will consider the stochastic departure process of vehicles and therefore, the required rehandles 26
to retrieve a vehicle from a line with more than 2 vehicles. All studies regarding import 27
vehicles suggest lines formed by only two vehicles, in order to avoid rehandling moves. 28
However, we will try to demonstrate that in particular conditions it will be economically 29
feasible to design block storage with more than 2 vehicles so the capacity of the terminal will 30
be increased without large investments. 31
32
3 PROBLEM DESCRIPTIO0 33
Let the shape parameter (β) be the reference ratio that will be used to compare the different 34
block layouts which dimensions are Lx and Ly, length and width respectively. The capacity of 35
each block is n slots and the main slot dimensions will be w and l (see Figure 1). 36
37
β = L�L� (1)
38
where: 39 �� = � · (2) �� = � · (3)
where: 40
X: Number of slots (vehicle) per line along the abscissa axis (‘x’) 41
Y: Number of slots (vehicle) per row along the ordinate axis (‘y’) 42
43
44
45
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 5
1
FIGURE 1 Generic block layout. Relationship between Lx and Ly and storage area (Sb+v,i) 2
3
In the layout shown in Figure 1, where vehicles are stored in lines, only the first and 4
the last vehicle can be accessed directly. Since we are dealing with import vehicles, rehandles 5
will be needed for retrievals, due to stochastic departing process. Therefore, as the number of 6
vehicles per line (Y) increase, bigger will be the probability of rehandles, but lower will be the 7
storage space per vehicle. Hence, there exists a trade-off between both criteria (rehandles 8
versus storage space per vehicle). 9
Finally, we will develop an economic analysis where the total cost function of the 10
storage yard will be introduced. Thus, the target will be to minimize total cost. 11
12
4 METHODOLOGY 13
4.1 Storage density 14
Terminal storage areas are organized in blocks, where vehicles are temporarily stored, and in 15
internal paths which help to organize traffic within the terminal. Therefore, the total area 16
available for storage is divided into parking areas and internal paths. 17
Each layout will have paths arranged in different ways; therefore the distribution of 18
paths and parking slots will depend on each layout. For this reason, storage density will be 19
evaluated and the unit area per vehicle will be assessed (m2/vehicle), that is: 20
21 �� = ��,�∁� (5)
where: 22
23 ��,�: Storage yard’s overall surface (m2) for layout i (characterized by β). 24 ∁�: Overall capacity for block layout i (vehicles). 25
26
Internal path width between blocks will be ax and ay meters (regarding the associated 27
axis). Also, slot area is (w · l ) square meters (Figure 1). 28
Therefore, the analytical expressions that quantify the shaded area in Figure 1, 29
applying to storage blocks (S�,�) and internal paths (S�,�), are: 30
31 ��,� = (�� · ��) (6) ��,� = (�� + ��) · �� + ���� (7)
32
Lx+axL y+ay
Sb+v,i Sb,i
ay
ax
Sv,i
SEASIDESEASID
E
STORAGE
YARD
BLOCK
I0TER0AL PATHS
XY
ROW
Slot
Lx
Ly l
w
ROW
rehandled vehicles
required vehicle
axis of
symmetry
LI0E SLOT
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 6
Finally, the total block surface (�� �,�) will be: 1
2 �� �,� = (�� + ��)·( �� + ��) (8)
3
The storage density per vehicle can be divided into two components: storage area 4
(��,�) and area traffic path (��,�) area, that is: 5
6 �� = !",#∁# = $·!%&',#∁# = ( · )!%,#∁# + !',#∁# * = ( · (��,� + ��,�) (9)
7
Where the parameter k determines the total number of blocks in the storage yard. 8
9
It is worth mentioning that, in equation (9) the first component (��,�) is the same for 10
all layouts, based on the assumption that all blocks have the same capacity and, thus, the 11
assigned area per vehicle will be equal to (� · �) m2. The second component (��,�), is 12
particular for each block layout and therefore the criteria will denote the differences between 13
them. 14
15
4.2 Calculation of rehandles 16
Rehandling is another important criterion to assess the suitability of different layouts. 17
In particular, the objective is to quantify the expected number of rehandles in a row 18
under two different situations: homogeneous rows (each vehicle in a row has the same 19
departure probability or even a gang of 8-10 vehicles with similar properties which will leave 20
the terminal in the same truck) and heterogeneous rows (rows are exclusively made up by 21
vehicles with different departure probability). 22
23
4.2.1 Assumptions and notations 24
The model developed is based on the following assumptions and notations. 25
Assumptions: 26
• All vehicles unloaded from the same ship have the same departure probability. 27
• Ship’s interarrival time is constant (6,15) 28
• The number of vehicles unloaded per ship is the same in all cases (15) 29
• Vehicle’s dwell time at the storage yard is similar to the Weibull distribution 30
function (this hypothesis derives from empirical data from vehicle terminals in 31
Port of Barcelona). 32
• Storage blocks have the same capacity. 33
Additional notations: 34
• h: Number of vehicles in half of a row, obtained according to expression (12) 35
• +,,-: Storage block with Y vehicles per row and X=n/Y vehicles alongside axis 36
‘x’. 37
• . : Index that identifies each block layout (+,,-). 38
• ∆T: Ship interarrival time 39
40
41
42
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 7
4.2.2 Dwell time at the storage yard 1
Vehicle dwell time in the yard (t), or the time vehicles stay in the terminal, is typically 2
considered to be a non-negative continuous random variable, which probability distribution 3
function F(t) might be the exponential (simplest distribution for non-negative random 4
variables) or any other candidate distributions like Weibull, gamma or log-logistic parametric 5
distribution. For this paper, we are choosing the Weibull distribution function (α, β), which 6
includes the exponential one (for α=1) 7
The survivor function, S(t), is defined as the function representing the probability that 8
a vehicle has not left the terminal at the precise time (t),this is: 9
10 �(/) = 0(1 ≥ /) = 1 − 5(/) (10)
11
Let h(t)be the hazard function and represents the instantaneous rate at which the event 12
of interest (a vehicle leaving the terminal) will occur at time t, given that it has not occurred 13
up to this time. That is: 14
15 ℎ(/) = lim9:→< 0(=/ ≤ 1 ≤ / + ?/|1 ≥ /)?/ = A(/)�(/) (11)
16
The probability of the vehicle leaving at time interval (t,t+dt] is approximately h(t)∆t, 17
for infinitesimal values of ∆t (16). 18
For the particular case of an exponential distribution, the instantaneous rate or hazard 19
function exhibits no duration dependence (hazard rate is constant), which entails that the 20
probability of leaving at the following time interval does not depend on the time the vehicle 21
has been parked in the yard. 22
23
4.2.3 Calculation of rehandles for a homogeneous row 24
This section focuses on the calculation of the rehandling moves of a row made up of same-25
ship vehicles (therefore, these vehicles have the same probability of leaving the terminal at 26
time t). The objective is to obtain the expected number of vehicles in the row needed to be 27
handled in order to retrieve a specific vehicle. 28
Any vehicle stored in the first or last positions of the row will have zero rehandling 29
moves associated, while a vehicle located on the second or penultimate position of the row 30
will entail (h-1) rehandling moves, where as h is half the number of vehicles in a row, that is: 31
32
ℎ = B 2 .A .D EFEG + 12 .A .D H?? = (12)
33
As vehicles may be retrieved from both ends of the row, a horizontal symmetry within 34
the rows may be found. In the case that the number of vehicles per row is odd, the central 35
vehicle will be assigned to one of the extremities and then the number of vehicles to be 36
retrieved to access to it will be exactly the same. 37
We will quantify the number of rehandling moves per row at time t by calculating the 38
probability of each possible number of expected rehandles (per row and for each time t), i.e: 39
IJK(/)L = M K · 0N(/)OPQNR< (13)
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 8
where R is the number of possible rehandling moves associated to each vehicle (function 1
of the vehicle position), taking into account the horizontal symmetry between the parking 2
cells, and PR(t) the probability that R rehandling moves occur at t. 3
The next step will be to define the probability that R rehandling moves in a vehicle 4
row (0N(/), K ∈ J1, ℎ − 1L) occur. For example, let’s consider a hypothetical case of a 6-5
vehicle row, i.e. h = 3, and we want to calculate the probability of R rehandling moves, 6
where, in this case K = T1,2U. 7
To determine the probability of R=2 rehandling moves it is first necessary to define all 8
possible situations where there are two associated rehandling moves. In this case, it will only 9
happen when the vehicle located in the middle position of the row is retrieved at t and the two 10
vehicles parked in front of it remain at the yard. 11
Therefore, the probability is determined by the compliance of the events "departure of 12
the vehicle located in the middle position of the row" and the "two remaining vehicles will 13
stay at the terminal." Analytically, this is the result of the probability that a vehicle leaves the 14
terminal at time t (f(t)) and the probability that the two remaining vehicles will stay at the 15
terminal, S(t), as defined in equation (10). 16
To get the probability of R = 1 rehandling moves in a 6-vehicles row for a time t 17
several conditions must occur at the same time. The first of them is that the vehicle parked in 18
the middle position of the row departs from the terminal (f(t)) while the remaining two 19
vehicle stay (S(t)). Another condition: one of the three vehicles that formerly made up half of 20
the row has to be retrieved (F(t)) and, finally, one more vehicle (having another parked in 21
front of it) departs at time t. The occurrence of these three events will define the probability of 22
R=1. Here, the density function, f(t), represents the probability of departure a vehicle at a 23
given time, assuming that it has been stored at the yard until that time. That is: 24
25
0NROPQ (/) = A(/) V �W(/) = A(/)(1 − 5W)(OPQ)OPQWRQ
(14a)
0NROPX(/) = A(/) V �WOPQWRQ (/) + )ℎ1* A(/)5(/) V �W(/)OPX
WRQ (14b) ⋮
0NRQ(/) = A(/) V �W(/) + )ℎ1* A(/)5(/) V �W(/)OPXWRQ + ⋯OPQ
WRQ+ ) ℎℎ − 3* A(/)5(OP\)(/) V �W(/) +XWRQ + ) ℎℎ − 2* A(/)5(OPX)(/) V �W(/)Q
WRQ
(14c)
Once the expected number of rehandling moves in a row of vehicles at t is defined, by 26
combining expressions (13) and (14a, 14b and 14c), the analytical expression that determines 27
the expected number of rehandling moves for a row consisting of n vehicles from the same 28
ship (with equal probability of departing the terminal) and for a period of time, J/<, ∞), is 29
obtained: 30 IJKL = ^ IJK(/)L?/_:` (15)
31
32
33
34
35
36
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 9
4.2.4 Calculation of rehandles for a heterogeneous row 1
The objective is to calculate the number of rehandling moves that occur in a row 2
accommodating vehicles from different ships and, therefore, with different probabilities of 3
leaving the terminal. 4
In order to do that, the volume of vehicles still parked in a row for a given time t i.e., 5
the expected number of vehicles at the block, E[8(t)] will be first determined. 6
The number of remaining vehicles in each row will depend on the vehicle departing 7
rate. Given that the departure rate approaches a Weibull distribution, characterized by 8
parameters α and λ, we can determine a vehicle’s departure probability and then obtain the 9
average number of vehicles remaining in each block of the terminal. 10
Assuming that there were initially Y vehicles from the same vessel in each row, the 11
probability that there are z vehicles, a ∈ J0, L, in a row for a given time instant t, will be: 12
0- (/) = V �c(/) = Vd1 − 5c(/)e-cRQ
-cRQ
(16a)
0-PQ(/) = ) 1* 5(/) V �c-PQcRQ (/) = f !( − 1)!h 5(/) V(1 − 5c
-PQcRQ (/))
(16b)
⋮
0Q(/) = ) − 1* �(/) V 5c-PQcRQ (/) = f !( − 1)!h d1 − 5(/)e V 5c(/)-PQ
cRQ
(16c)
0<(/) = V 5c(/)-cRQ
(16d)
Since a row is made up exclusively of vehicles coming from the same vessel, the 13
distribution function 5c(/) will be the same for all vehicles in the row. Therefore, for 14
simplicity5c(/) = 5(/) ∀a, so the above expressions can be summarized as follows: 15
16 0c(/) = f − ah �(/)c5(/)-Pc (17)
17
From equation (17), the expected number of vehicles per row at time t, E [8 (t)], will be 18
given by: 19
IJj(/)L = M a · 0c(/)-cR< (18)
20
Once the methodology for calculating the probability that there will be z vehicles in a 21
row, akJ0, L, and thus the expected number of remaining vehicles in the rows making up the 22
storage blocks is defined, the expected number of rehandling moves in a row which is made 23
up by vehicles with different departure probabilities will be quantified. 24
Since vehicle arrival times at the terminal are different and taking into account that the 25
reference time corresponds to the arrival time of the vehicles unloaded from the first vessel, it 26
will be necessary to evaluate the occupation of each row for each ship arrival time. 27
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 10
Let´s suppose a yard with vehicles from the same ship A. If we consider a specific 1
time, t*, and if we analyze the terminal occupation at this same time (Figure 2), we will be 2
able to determine the final configuration of the rows in each cell assuming that the vehicles 3
unloaded from the vessel just docked at the terminal, B, are incorporated to them (figure 3: 4
block configuration at t>t *). The standard configuration of the rows will be type (z,Y-z) 5
where z, a ∈ J0, L, are the vehicles already stored in the cell (from ship A) and Y–z is the 6
number of new unloaded vehicles (from ship B). 7
8
9
FIGURE 2 Possible rows and slots configuration at t <t*(when vehicles belong to ship A) 10
and t > t* (when vehicles belong to ships A and B). 11
From the scheme shown in Figure 2, the analytical expression quantifying the 12
expected number of rehandles at time t in a row with z remaining vehicles stored at the 13
terminal, and (Y-z) vehicles just unloaded to the terminal will be finally: 14
IJK(/)L = M 0c(/∗ − /<) · Ic,-PcJK(/)L-cR<
(19)
Finally, by integrating the expected number of rehandling moves for the time span 15
between [/∗, ∞], the number of rehandling moves resulting from mixing vehicles from two 16
different ships (A and B) is obtained: 17
IJKL = ^ IJK(/)L?/ _:∗
(20)
18
19
Possible block configuration at time t=t* Possible block configuration at time t>t*
Vehicles from ship A Vehicles from ships A and B
*)]([)2,4( tRE*)]([)0,4( tRE*)( ttPz = *)]([),( tRE zYz −
empty slot
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 11
5 STORAGE YARD LAYOUT 1
The assessment of each block layout will be carried out by analyzing the rehandling moves 2
and storage density. Both criteria will be related to the shape parameter (β), as defined in 3
equation (1). 4
5
5.1 Block layout and initial data 6
The block layouts suggested (+,,-) are summarized in Table 1, based on the following input 7
parameters: number of vehicles per block (n) and number of vehicles per row (Y), ∈ (1, G)). 8
Assuming that the slot dimensions are 2.5 meters wide (�) and 5.0 meters long (�) and 9
the internal paths width of the yard is 6.0 meters (�� = ��, is assumed to solve the case 10
study), the dimensions of the block need to be defined. Bearing in mind that n=300 vehicles, 11
the settings detailed in Table 1 will be analyzed. 12
13
5.2 Results 14
The results are organized into two different sections: the first one analyzes the main criterion: 15
storage density and rehandling moves; and the second section presents a numerical case, 16
based on a vehicle terminal, where the optimum block layout will be defined. 17
18
5.2.1 Storage Density and rehandling moves 19
Table 1 shows the results from the application of expressions as defined in Section 4.1 and 20
4.2. As regards the storage density (��), and particularly its components (blocks and internal 21
paths), it should be noted that blocks storage density (��,�) is equal for all block layouts: 12.50 22
(m2/vehicle). 23
On the other hand, Table 1 also shows the expected number of rehandles per block 24
considering two different scenarios: homogeneous and heterogeneous row. In both cases, it 25
has been assumed that vehicle dwell time follows a Weibull distribution which parameters are 26
α=1.5 and λ=0.04 as the average dwell time is assumed to be 10 days. 27
In the second scenario (heterogeneous row) we have considered a row made by 28
vehicles from the first ship and vehicles from a ship which have arrived at the terminal 3 days 29
later. It should be stand out; vehicles arriving earlier are more likely to leave the terminal 30
before those arriving later. 31
32
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 12
TABLE 1 Storage Density (m2/vehicle) and Expected Rehandles per Layout 1
Block
Configuration X Y ββββ
mn (m
2/vehicle)
mo,n (m
2/vehicle)
pJqLrstuv pJqLrstuv ∆pJqLrstuv (%)
+Qx<,X 150 2 37.50 20,30 7,80 0.00 0.00 0.00
+yx,z 75 4 9.40 16,80 4,30 37.50 45.00 20.00
+{<,x 60 5 6.00 16,10 3,60 55.80 67.20 20.40
+x<,{ 50 6 4.20 15,70 3,20 75.00 90.50 20.70
+\<,Q< 30 10 1.50 15,10 2,60 150.00 180.40 20.30
2
Let there be a block with 8 vehicles per row, all from the first ship. At the arrival time 3
of the second ship (∆T = 3 days) we will note that, on average, 5-6 vehicles from the first 4
vessel remain in that row and therefore there will be 2-3 empty slots. As can be appreciated, 5
the number of rehandles (∆IJKL�|`}$) will be increased by 21% over the ones that would 6
have been required for a homogeneous 6-vehicles row with the same departing probability. 7
For the specific case in study, the increase in the percentages will be around 20% for all 8
analyzed configurations, as it is shown in Table 1. 9
10
From the above results it is observed that: 11
12
• The storage density decreases as the shape parameter (β) decreases, that is, the more 13
compacted the block geometry is, the smaller the area needed for internal circulation 14
paths will be. 15
• Data shows that given two blocks with the same storage capacity (number of slots), 16
the required total area can be 34% higher in one case than in the other (comparison 17
between configurations +Qx<,X and +\<,Q<). Thus the application of configuration 18 +\<,Q< would achieve a better utilization of the storage area. 19
• The expected number of rehandles per row increases more than proportionately (to 20
square) to the number of vehicles per row. Similarly the labor costs will rise. 21
• Finally, it can be seen that the expected number of rehandles increases with the 22
decrease of the storage density. Hence there exists a trade-off between both criteria. 23
24
5.2.2 8umerical case 25
Once both parameters are analyzed, the next step is to analyze a hypothetical terminal with a 26
storage area of 50Ha and a block capacity of 300 vehicles (i.e: automobile terminals in Port of 27
Tacoma (WA) or Port of Zeebrugge (Belgium)). In this case, we will try to find the block 28
layout which minimizes total cost. 29
Table 2 shows the organization of the storage yard: storage density, number of blocks, 30
and the number of vehicles that can be stored in the terminal (in terms of capacity). 31
32
33
34
35
36
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 13
TABLE 2 Storage Density and Overall Capacity for each Layout (ST=50Ha) 1
Block
Configuration β k
~n (vehicles)
mr o,n (m
2/vehicle)
�r,n(%) �o,n (%) +Qx<,X 37.5 82 24,600 20.3 62% 38%
+yx,z 9.4 99 29,700 16.8 75% 25%
+{<,x 6.0 103 30,900 16.1 78% 22%
+x<,{ 4.2 106 31,800 15.7 80% 20%
+\<,Q< 1.5 110 33,000 15.1 83% 17%
2
The table above shows the difference in the use of the allocated space for vehicle 3
storage in a terminal. Considering a block layout of two vehicles per row, 62% of the total 4
area will be useful for storage; while if the blocks are formed by rows of 6 vehicles, surface 5
profitability will be increased by 18% (40Ha used for storage purposes and only 10Ha for 6
internal circulation paths). Additionally, the terminal capacity also depends on the shape 7
parameter (β). Hence, 34% more capacity can be achieved if rows with 10 vehicles 8
(+(� Q<)⁄ ,Q<) are used instead of 2-vehicles rows (+(� X)⁄ ,X). 9
These results validate the suggestion stated by Mattfeld (1), who proposed a storage 10
strategy for stochastic output (import) of two vehicles per row in order to avoid rehandling 11
moves although this will require more space for paths in order to ensure direct access to these 12
vehicles. 13
Following the study case, the expected number of rehandles that would be required a 14
block of 300 vehicles from the same vessel and with a dwell time of 10 days are shown in 15
Figure 3, where the results from the Table 2 are also included. 16
17
18
FIGURE 3 Capacity and expected rehandling moves regarding shape parameter (ββββ) 19
Analyzing the expected number of rehandles per block layout, it can be observed that 20
no additional vehicle move is required when there are 2 vehicles per row (β=37.5). However, 21
if configurations of 4 vehicles per row are applied, moves increase by 12.5% compared to the 22
0
5000
10000
15000
20000
25000
30000
35000
0 5 10 15 20 25 30 35 40
Capacidad (veh)-Remociones
Parámetro de forma (β)
Total expected number
of rehandles
Storage yard capacity
(vehicles)
β=1
Shape parameter (β)
Capacity(veh.)-Rehandles
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 14
previous configuration. Moreover, if storage blocks of 10 vehicles per row are designed (β = 1
1.5), vehicle moves per block will increase by 50 %, i.e, one on three moves within the 2
terminal will be for rehandling purposes. From Figure 3 it can be stated that, when β 3
increases, storage yard capacity decreases linearly (as an approximation) and the total 4
expected number of rehandles drops also but with factor 2. 5
6
5.3 Cost analysis 7
To carry out the economic analysis annual operating costs for a terminal with storage area of 8
50 Ha will be quantified. The considered costs correspond with concession fees (composed of 9
a fixed part based on the surface occupation and a variable part depending on the number of 10
vehicles handled by the terminal), the cost of additional manpower to perform rehandling 11
moves and the opportunity cost of not being able to answer all possible demand (will be 12
calculated as the price charged by the terminal for the pass of each vehicle and taking into 13
account the difference in configuration capacity of the storage area with more capacity, β = 14
1.5, depending on the configuration considered in each case). 15
Another cost to consider is related to the value of the time required for transfers from 16
the vessel to the storage area. Different block layouts require placements in different distance 17
in relation to pier line, meaning a decrease in labor productivity when the shape parameter (β) 18
increases. However, this difference in productivity translated into costs will be disregarded 19
due to its low importance compared to other cost components. 20
First of all, it is necessary to define the potential demand of the terminal. To this end, 21
various levels of using parking slots of the terminal will be established, valued in block 22
rotation a year (�) (number of times slots are filled and empty a year) 23
Since the cost associated with vehicles rehandling is a stochastic variable, the total 24
cost of the yard is expressed in terms of expected total cost: 25
26 IJ��(�)L = dIJK�|`}$L��N + �� �,��� + G���e( + ∆���0 (21)
Where: 27
28 �: Rotation of slots per year (number of times each slot is in use annually) 29 G: Block capacity (number of vehicles) 30 �N: Labor cost associated with rehandling moves (€/rehandle). 31 ��: Area cost (€/m2) corresponding to the fixed part of the concession fee. 32 ��: Variable cost corresponding to the concession fee and based on the number of 33
vehicles handled by the terminal a year (€/vehicle). 34 ∆��: Capacity reduction of the layout i in respect of those considered to have 35
maximum capacity (βmin). For instance, taking the results obtained in Table 2 it has 36 ∆��RQ.x = 0 and ∆��R\y.x = 8400 vehicles. 37
P: Opportunity cost per vehicle as it goes through the terminal (€/vehicle) (price 38
charged by the terminal) 39
40
Finally, the objective to be achieved is to determine the optimal block layout that 41
minimizes the operational cost of the terminal ( IJ��(�)L). 42
As an example (Figure 4) we proceed to determine the optimal shape parameter (β*). 43
Here we take into account the following parameters obtained from the terminals of vehicles 44
operating in the Port of Barcelona: �� = 7€/�X, �� = 1€/vehicle, �N = 5€/movement, P= 45
10 € and for the next demand (rotation): � = 20 �G? 80 uses per year. 46
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 15
1
FIGURE 4 Analysis cost and optimal block layout for each scenario 2
3
In Figure 4 it can be observed that when increasing the rotation (ρ), the labor costs 4
increase, in respect of rehandling, also increases the variable part of the concession fee when 5
raising productivity and ultimately the opportunity cost. In the considered cases, for practical 6
purposes, the following parameters can be accepted as optimal values: β=4.2, β=6.0 y β=9.4. 7
Also, since ρ, reduction of labor cost with β is lower as β increases (increases proportionally 8
to the square of number of vehicles per row), while the increase of opportunity cost is linear 9
with β. 10
Moreover, since the cost components that have more sensitivity in the total count are 11
the labor cost and the opportunity cost associated with each layout, we proceed to analyze the 12
adequacy of each layout for different combinations of the following unit costs, �N and P. For 13
this analysis the following parameters will be assumed: ��= 5€/m2, ��=1€/vehicle and 14 � = 20 �G? 80, taking into account the explicit characteristics in Table 2 (see Figure 5). 15
- €
2.000.000,00 €
4.000.000,00 €
6.000.000,00 €
8.000.000,00 €
10.000.000,00 €
12.000.000,00 €
14.000.000,00 €
0 5 10 15 20 25 30 35 40
- €
1.000.000,00 €
2.000.000,00 €
3.000.000,00 €
4.000.000,00 €
5.000.000,00 €
6.000.000,00 €
7.000.000,00 €
0 5 10 15 20 25 30 35 40
β*=6.0
E[CT(β*)]
E[CT(β*)]
β*=6.0
ρ=20E[C
T( ββ ββ
)](€)
ρ=80
E[C
T( ββ ββ
)](€)
OpportunityCostLabor cost
(rehandles)
Labor cost
(rehandles)
OpportunityCost
Total Cost
Total Cost
Port Concession (fixed part)
Port Concession (fixed part)
Port Concession (variable part)
Port Concession (fixed part)
Shape parameter(ββββ)
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 16
1
FIGURE 5 Optimum shape-parameter and minimum cost associated 2
The results shown in Figure 5 lead to the following considerations: 3
4
• When the unit cost of opportunity (P) is low, independent of the demand of the 5
terminal, the optimal block layout is to have two vehicles per row (β = 37.5). In this 6
case, operating costs of rehandling are null and opportunity costs associated with the 7
capacity difference is negligible due to its low unit cost. 8
• As the unit cost of opportunity (P) increases, layouts with more than two vehicles per 9
row become more favorable. This trend is observed in Figure 5(a), when the 10
rehandling cost increases (�N), the parameter decreases until it reaches its minimum 11
value (β = 1.5) Therefore, when both unit costs are raised, it is advisable to design 12
blocks that maximize the use of the storage area (to maximize the capacity of the 13
terminal). In the present case, there has been 34% of capacity increase (with β = 1.5 in 14
regard to β = 37.5) 15
• Finally, when terminal demand increases, and therefore there is greater rotation of 16
slots of the storage area, the use of blocks of storage is still recommended to ensure 17
the maximum terminal capacity (β=1.5) The highest revenues of the terminal by being 18
able to move more vehicles outweigh handling costs (which increase with ρ) 19
20
The results of this section suggest that it is not necessary to heavily invest to the 21
increase of the storage area in order to increase the terminal capacity, whether when storing in 22
extension or vertically. Depending on the potential demand, average staying time of vehicles 23
and operational costs, it is possible to design and configure the storage area characterized by 24
the parameter β, so that the productivity of the terminal increases. 25
In certain situations it will be optimal to choose block layouts involving an additional 26
cost for rehandles (β <37.5), but the increase cost mentioned is compensated by revenues 27
derived from the increase of the storage productivity. 28
However, in many cases to save the compliance of quality and safety requirements set 29
by vehicle manufacturers to minimize the risk of vehicle damage, there are currently two 30
types of solutions: expansion of the terminal storage area (when that option exists and the cost 31
of land is low) or building vertical storage (when no option to expand and / or the land cost is 32
expensive). The latter has been used at various terminals in Europe, for example, Port of 33
Barcelona (Spain) or Southampton (UK), among others. 34
0 2 4 6 8 10 12 14
0
2
4
6
8
10
12
14
CR (€/movement)
P (
€/v
eh
icle
)
€
€
€
€
€
β*=37.5
β*=9.4
β*=4.2
β*=1.5
β*=6.0
ρ=20
0 2 4 6 8 10 12 14
0
2
4
6
8
10
12
14
CR (€/movement)
P (
€/v
eh
icle
)
€
€
€
β*=1.5
β*=37.5
β*=4.2
ρ=80
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 17
Nevertheless, a third option is also plausible, that is through reconfiguration of the 1
storage area. It is, ultimately, choosing a block layout that fits the operating characteristics of 2
the terminal and that is most compatible for the trade-off between rehandling cost and 3
productivity of the yard. 4
5
6 CO0CLUSIO0S 6
This paper addressed the problem faced by terminal operators in organizing the storage yard 7
of port terminals where import vehicles will be stored temporarily. An analytical model was 8
developed to quantify the expected rehandles that result from storing more than two import 9
vehicles per line. Moreover, storage density was considered as a criterion to evaluate storage 10
layouts. 11
It has been demonstrated that there exists a trade-off between the rehandles and the 12
storage density, that is, when the number of vehicles per row increases, the expected number 13
of rehandles moves increases and storage density per vehicle decreases. In order to find this 14
trade-off, the cost analysis was provided. Total cost function includes labor costs, port fees 15
(fixed and variable) and opportunity costs regarding the layout which maximizes terminal 16
capacity. 17
This paper has explored the way to increase the capacity of port terminals which 18
handles a large amount of import vehicles by reorganizing storage yard. This option involves 19
an increase of operational cost and some risk situations regarding quality and security of 20
vehicles. However, large investments will not be required. 21
The results have shown that when labor and opportunity costs are high, the trend is to 22
use a block layout which shape parameter (β) value is close to unity. 23
The model developed here was based on expected values, so total cost is an expected 24
value too. A particularly interesting line of further research could be the inclusion of risk cost 25
analysis, especially regarding labor cost associated to rehandling moves. 26
TRB 2012 Annual Meeting Original paper submittal - not revised by author.
S. Saurí and E. Martín 18
REFERE0CES 1
2
[1] Mattfeld, D.C. The Management of Transshipment Terminals. Decision Support for 3
Terminal Operations in Finished Vehicle Supply Chains. Springer Science and 4
Business Media, 2006, USA. 5
[2] Holweg, J., Miemczyk, J. Delivering the ‘3-day vehicle’-the Strategic Implications for 6
Automotive Logistics Operations. Journal of Purchasing and Supply Management, 9, 7
2, 2003, pp.63-71. 8
[3] De Castilho, B., Daganzo, C.F. Handling Strategies for Import Containers at Marine 9
Terminals. Transportation Research B, 27, 2, 1993, pp.151-166. 10
[4] Kim, K.H. Evaluation of the number of rehandles in container yards. Comput. Ind. 11
Eng. 32 (4), 1997, pp.701–711. 12
[5] Huynh, N. Analysis of Container Dwell Time on Marine Terminal Throughput and 13
Rehandling Productivity. TRB Annual Meeting, 2008. 14
[6] Saurí, S., Martín, E. Space Allocating Strategies for Improving Import Yard 15
Performance at Marine Terminals. Transportation Research E, 2011. In Press. 16
[7] Fischer, T., Gehring, H. Planning Vehicle Transshipment in a Seaport Automobile 17
Terminal Using a Multi-Agent System. European Journal of Operational Research. 18
166, 2005, pp.726-740. 19
[8] Mattfeld, D.C., Kopfer, H. Terminal Operations Management in Vehicle 20
Transshipment. Transportation Research A, 37, 2003, pp.435-452. 21
[9] Berry, J.R. Elements of Warehouse Layout. International Journal of Production 22
Research, 7, 1968, pp.105-221. 23
[10] Bozer, Y.A., White, J.A. Design and performance models for end-of-aisle order 24
picking systems. Management Service, 36, 1990, pp.852-866. 25
[11] Lee, B.K., Kim, K.H. Optimizing the Block Size in Container Yards. Transportation 26
Research E, 46, 2010, pp.120-135. 27
[12] Iranpour, R., Tung, D. Methodology for Optimal Design of a Parking Lot. Journal of 28
Transportation Engineering, 115, 2, 1989, pp.139-160. 29
[13] Cassady, C.R., Kobza, J.E. A Probabilistic Approach to Evaluate Strategies for 30
Selecting a Parking Place. Transportation Science, 32, 1998, pp.30-42. 31
[14] Bingle, R., Meindertsma, D. and Oostendorp, W. Designing the Optimal Placement of 32
Spaces in a Parking Lot. Mathematical Modelling, 9, 10, 1987, pp.765-776. 33
[15] Kim, K.H., Kim, H.B. Segregating Space Allocation Models for Container Inventories 34
in Port Container Terminals. Int. J. Production Economics, 59, 1999, pp.415-423. 35
[16] Kiefer, N. Economic Duration Data and Hazard Functions. J. Econ. Lit. 26(2), 1988, 36
pp.646-679. 37
TRB 2012 Annual Meeting Original paper submittal - not revised by author.