Table 7.2 Fill rates with different retailer reorder points for a distribution system with echelon stock policy
R_ROP 4 6 8 10 12 16 R1 0.882 0.949 0.973 0.992 0.999 1.000 R2 0.909 0.962 0.983 0.993 0.992 0.991 R3 0.786 0.837 0.875 0.965 0.960 0.968
Table 7.3 Costs with different warehouse reorder points for a distribution system with installation stock policy vs. echelon stock policy
W1_ReOrderPoints 4 8 10 12 15 20 25 30 35 Cost with ISP Cost with ESP 61007.54 57195.32 63533.72 59662.00 64695.10 59235.67 65705.00 61503.42 64477.48 61486.65 68236.58 65778.14 68524.75 71714.20 69215.68 74603.34 76408.40 76256.62
Table 7.4 Fill rates with different warehouse reorder points for a distribution system with installation stock policy vs. echelon stock policy
W1_ReOrderPoints 4 8 10 12 15 20 25 30 35 ISP-R1 0.946 0.959 0.962 0.971 0.964 0.973 0.978 0.973 0.973 ESP-R1 0.941 0.949 0.951 0.963 0.949 0.973 0.984 0.981 0.974 ISP-R2 0.975 0.975 0.97 0.973 0.971 0.985 0.981 0.984 0.977 ESP-R2 0.977 0.972 0.973 0.984 0.965 0.977 0.984 0.976 0.978 ISP-R3 0.859 0.854 0.849 0.865 0.856 0.865 0.875 0.889 0.892 ESP-R3 0.841 0.833 0.843 0.874 0.871 0.863 0.868 0.877 0.882
130
Echelon Stock ReOrder Point Policy
1.2 1
R1 R2 R3
Fill Rates
0.8 0.6 0.4 0.2 0 0 2 4 6 8 10 12 14 16 18
Retailers' ReOrder Points
Echelon Stock ReOrder Point Policy
100000 80000 Total cost 60000 40000 20000 0 0 5 10 15 20 25 30 35 40 Warehouse ReOrder Point
Echelon Stock ReOrder Point Policy
1 0.98 0.96 0.94 0.92 0.9 0.88 0.86 0.84 0.82 0 5 10 15 20 25 30 35
RE1FillRate RE2FillRate RE3FillRate
Fill rates
40
Warehouse ReOrder Point
Figure 7.6 Costs and fill rates for a multi-echelon distribution system with echelon stock reorder point policy
131
From Figure 7.6, we see that the optimal reorder point value for warehouse is around 12 and the optimal reorder point value for retailers is around 10.
Installtion Stock Policy vs. Echelon Stock Policy
90000 80000 70000 Total Cost 60000 50000 40000 30000 20000 10000 0 0 5 10 15 20 25 Warehouse ReOrder Point 30 35 40
ISP ESP
Figure 7.7 Costs with different warehouse reorder points for a distribution system with installation stock policy vs. echelon stock policy
Installtion Stock Policy vs. Echelon Stock Policy
1 0.98 0.96
ISP-R1 ESP-R1 ISP-R2 ESP-R2 ISP-R3 ESP-R3
Fill Rates
0.94 0.92 0.9 0.88 0.86 0.84 0.82 0 5 10 15 20 25 30 35 40
Warehouse ReOrder Point
Figure 7.8 Fill rates with different warehouse reorder points for a distribution system with installation stock policy vs. echelon stock policy
132
By comparing the cost and fill-rate performance between installation stock policy and echelon stock policy, we have Figures 7.7 and 7.8. Figure 7.7 shows that the echelon stock policy is better (less cost) than installation stock policy when the warehouse reorder point is below 22 units. From Figure 7.8, it can be seen that the fill-rate performance curves of echelon stock policy and installation stock policy are overlapping. This indicates that either echelon stock policy or installation stock policy may be advantageous. 7.7.2 Output Analysis of A Multi-Echelon Distribution System with Lateral Transshipment In multi-echelon research, the most common assumption is that shipments among retailers are not allowed. One reason is that it will be very difficult, if not impossible, to consider this situation. However, through simulation, the impact of lateral transshipments can be investigated. The following issues will occur in the simulation study: (1) How to evaluate the benefits of adding this flexibility? (2) In emergencies, will the service be better? and (3) How about the cost? Some factors to be considered in the study include: (1) replenishment lead time from warehouse, (2) demand process, and (3) number of retailers in pooling group. Generally, there are two different transshipment policies: (1) random policy and (2) risk balancing policy. The idea behind random policy is as follows: when retailer k faces a shortage and retailers i, j have available on-hand inventory, the source retailer is chosen randomly and transships as much as needed to completely eliminate the shortage. If its inventory is not sufficient, then the other retailer also sends the quantity required to eliminate the remaining shortage. The rationale behind the risk balancing policy is that the determination of transshipment quantities should also take into account the risk of stock-out in at least the following period. Since the risk should be balanced between the two senders or the two receivers, the transshipped quantities must be those that equalize the probability of a stock-out in the following period.
133
The previous research shows that the random policy is simple to implement, but the performance is not good. On the other hand, the system performance by employing riskbalancing policy is good. However, it is hard to implement in practice. In this section, a new transshipment policy called “alternate policy” is presented. The principle of alternate policy is as follows: when retailer k faces a shortage and retailers i, j have available on-hand inventory, both retailers i and j, alternately, transship as much as needed to eliminate the shortage. For this policy, the inventory levels of senders should not fall below their safety stocks. In the following simulation study (three retailers), two cases are identified: Case I: two senders and one receiver, and Case II: one sender and two receivers. The simulation models for a multi-echelon distribution system with different transshipment policies are shown in the following Figures.
134
RE11BkOrd BL16 BL15 RE11BO RE11BkOdr Proc
Case I: Only R3 has insufficient inventory Senders: R1 and R2 Receiver: R3
BL17
W1RE11 RO RS14
B1
RE11Consd
RS13
RE11ReOrd
RS12
ChkRE11 P1Inv
RsW 15
W1T W1T W1T RE11P P1Deliver 1 EOQTranW 2 RdySend 3 TransP1To 1 RE11P1I PR11 C1OrdProce OC14 P1C1Ord W1RE11 1RE11 yRE11 RE11 nv ss s
P1
C1
11
5
BP1
1
T1
RS
P1C1 P1C1 P1C1 BatchTransP T2 RdySen T3 T4 C1Receive P1C1Trans 1C1 dP1C1 sP1
W1
BR1
1
13
LT
RS1 5
PR
BR
OC13
BL 11
12
C1P1Odrs Sg11 C1Order Arrive sP1 OC OC 11 12 C1Odrs Control
11
RP sW1 R 15 1
R11TransR 13
C1BkOrdPr BL13 C1BackO BL12 C1Backlog ocess drs
5 W2 Rs
W1R
E11
RE12BkOrd BL26 BL25 RE12BO RE12BkOdr Proc
BL27
W1RE12 RS24 RS23 RE12Consd RE12ReOrd RO
W1B 2
W1Reor RSW4 der
W1Consd
RSW3
RSW2 ChkW1P W1ReOrd 1Inv
RS22
PR
RS2
RSW5
P1Inv
P1InvT2 P1Delive ryW1
P1W1 P1W1T P1W1 1 EOQTran T2 RdySend T3 TransP1T P1W1 WareHou P1W1 P1W1 oW1 se1
W
E 1R
12
RE12 P1Delive W1T4 EOQTran W1T5 RdySend W1T6 TransP1T P1 RE12P1I PR21 C2OrdProce OC24 P1C2Ord OC23 W1RE12 W1RE12 oRE12 nv ss s ryRE12 BL PR 21 22
P1
1
C2
Rs W 21
ChkRE12 P1Inv
P1C2 P1C2 P1C2 BatchTransP T2 RdySend T3 T4 C2Receiv P1C2Trans 1C2 P1C2 esP1
25
5
T1
BP2
RS 21
BR
C2P1Odrs Sg21 C2Orde Arrive rsP1
21
BR16
O
C2
OC
2
21
BL37
W1RE13 RS34 RE13Consd RO
3
RS3 5
LT232
RE13 P1Delive W1T7 EOQTran W1T8 RdySend W1T9 TransP1T P1 RE13P1I PR31 C3OrdProce OC34 P1C3Ord ryRE13 W1RE13 W1RE13 oRE13 nv ss s
P1
C3
31
1
T1
BP3
Figure 7.9 Simulation model for a multi-echelon distribution system with installation stock reorder point policy and random transshipment policy (Case I)
5 W3 Rs
35 PR 31 RsW
W1 RE 13
LT132
LT231
C2BkOrdPr BL23 C2BackO BL22 C2Backlog ocess drs
C2Odrs Control
R12TransR 13
RE13BkOrd BL36 BL35 RE13BO RE13BkOdr Proc P1C3T P1C3 P1C3T 2 RdySen T3 4 C3Receiv P1C3Trans dP1C3 esP1
RS33
RS32 ChkRE13 RE13ReOrd P1Inv
BatchTransP 1C3
135
W1B
RS
BR
OC33
BL 31
C3P1Odrs Arrive
Sg31 C3Order sP1
OC 31
31
OC
PR 32
32
C3BkOrdPr BL33 C3BackO BL32 C3Backlog ocess drs
C3Odrs Control
Case I: Only R3 has insufficient inventory Senders: R1 and R2 Receiver: R3
RE11BkOrd BL16 BL15 RE11BO RE11BkOdr Proc
BL17
W1RE11 RO RS14
B1
RE11Consd
RS13
RE11ReOrd
RS12
ChkRE11 P1Inv
15
W1T W1T W1T RE11P P1Deliver 1 EOQTranW 2 RdySend 3 TransP1To 1 RE11P1I PR11 C1OrdProce OC14 P1C1Ord W1RE11 1RE11 yRE11 RE11 nv ss s
P1
C1
11
T1
RS
P1C1 P1C1 P1C1 T4 C1Receive BatchTransP T2 RdySen T3 P1C1Trans 1C1 dP1C1 sP1
W1
5
BP1
1
BR1
RS1 5
BR
OC13
BL 11
C1P1Odrs Sg11 C1Order Arrive sP1
11
RsW
LT
PR
12
OC
13
12
OC
1
11
RP sW1 R 15 1
R11TransR 13
C1BkOrdPr BL13 C1BackO BL12 C1Backlog ocess drs
C1Odrs Control
W1R
E11
RE12BkOrd BL26 BL25 RE12BO RE12BkOdr Proc
5
R
2 sW
BL27
W1RE12 RS24 RE12Consd RO
W1B 2
W1Reor der
RSW4
W1Consd
RSW3
W1ReOrd
RSW2 ChkW1P 1Inv
25
RS23
RE12ReOrd
TS2
RS22
5
PR
RS2
TS1
RSW5
P1Inv
P1W1 P1W1T P1W1 P1InvT2 P1Delive 1 EOQTran T2 RdySend T3 TransP1T P1W1 WareHou P1W1 ryW1 P1W1 oW1 se1
W
1
RE
12
RE12 P1Delive W1T4 EOQTran W1T5 RdySend W1T6 TransP1T P1 RE12P1I PR21 C2OrdProce OC24 P1C2Ord OC23 W1RE12 W1RE12 oRE12 nv ss s ryRE12 BL PR 21 22
P1
1
C2
W
21
T1
Rs
ChkRE12 P1Inv
P1C2 P1C2 P1C2 BatchTransP T2 RdySend T3 T4 C2Receiv P1C2Trans 1C2 P1C2 esP1
BP2
RS 21
BR
C2P1Odrs Sg21 C2Orde Arrive rsP1
21
O
C2
O
2
C2
BR16
1
BL37
W1RE13 RS34 RE13Consd RO
3
RS3 5
LT232
RE13 P1Delive W1T7 EOQTran W1T8 RdySend W1T9 TransP1T P1 RE13P1I PR31 C3OrdProce OC34 P1C3Ord ryRE13 W1RE13 W1RE13 oRE13 nv ss s
P1
C3
RS
31
T1
BP3
1
Figure 7.10 Simulation model for a multi-echelon distribution system with installation stock reorder point policy and alternate transshipment policy (Case I)
Rs W3 5
TransCont rol2
C2BkOrdPr BL23 C2BackO BL22 C2Backlog ocess drs
C2Odrs Control
LT132
TS 3
LT2 31
35 P R 31 RsW
W1 RE1 3
TransCont rol1
TS4
R12TransR 13
RE13BkOrd BL36 BL35 RE13BO RE13BkOdr Proc P1C3T P1C3 P1C3T 2 RdySen T3 4 C3Receiv P1C3Trans dP1C3 esP1
RS33
RS32 ChkRE13 RE13ReOrd P1Inv
BatchTransP 1C3
136
W1B
BR
OC33
BL 31
C3P1Odrs Arrive
Sg31 C3Order sP1
OC 31
31
OC
PR 32
32
C3BkOrdPr BL33 C3BackO BL32 C3Backlog ocess drs
C3Odrs Control
LT 13 1
Sender: R2 Receivers: R1 and R3
BR1 5
P1Deliver yRE11
RsW 15
W1T W1T W1T RE11P 1 EOQTranW 2 RdySend 3 TransP1To 1 RE11P1I PR11 C1OrdProce OC14 P1C1Ord W1RE11 1RE11 RE11 nv ss s
BP1 1
P1 C1 T1
Case II: Both R1 and R3 have insufficient inventory
RE11BkOrd BL16 BL15 RE11BO RE11BkOdr Proc
BL17
W1RE11 RS14 RO RS13 RS12 ChkRE11 P1Inv P1C1 P1C1 P1C1 T4 C1Receive BatchTransP T2 RdySen T3 P1C1Trans 1C1 dP1C1 sP1
B1 W1
RE11Consd
RE11ReOrd
RS 11
RS1 5
PR 12
11 BR
OC13
BL 11
C1P1Odrs Sg11 C1Order Arrive sP1 OC OC 11 12 C1Odrs Control
RP sW1 R 15 1
R11TransR 13
C1BkOrdPr BL13 C1BackO BL12 C1Backlog ocess drs
5 W2 Rs
W1R E11
RE12BkOrd BL26 BL25 RE12BO RE12BkOdr Proc
BL27
W1RE12 RS24 RS23 RE12Consd RE12ReOrd RO
2 W1B
W1Reor RSW4 der
W1Consd
RSW3
W1ReOrd
RSW2 ChkW1P 1Inv
PR 25
RS22
RSW5
P1Inv
P1InvT2 P1Delive ryW1
P1W1 P1W1T P1W1 1 EOQTran T2 RdySend T3 TransP1T P1W1 WareHou P1W1 P1W1 oW1 se1
2 E1 1R W
RE12 P1Delive W1T4 EOQTran W1T5 RdySend W1T6 TransP1T P1 RE12P1I PR21 C2OrdProce OC24 P1C2Ord OC23 W1RE12 W1RE12 oRE12 nv ss s ryRE12 BL PR 21 22
RS2 5
BP2 1
P1 C2 T1
Rs W 21
ChkRE12 P1Inv
P1C2 P1C2 P1C2 BatchTransP T2 RdySend T3 T4 C2Receiv P1C2Trans 1C2 P1C2 esP1
21 RS
21 BR
C2P1Odrs Sg21 C2Orde Arrive rsP1
BR16
O C2 2
OC 21
BL37
W1RE13 RS34 RS33 RS32 ChkRE13 RE13Consd RE13ReOrd RO P1Inv
RS3 5
LT232
RE13 P1Delive W1T7 EOQTran W1T8 RdySend W1T9 TransP1T P1 RE13P1I PR31 C3OrdProce OC34 P1C3Ord ryRE13 W1RE13 W1RE13 oRE13 nv ss s
BP3 1
P1 C3 T1
Figure 7.11 Simulation model for a multi-echelon distribution system with installation stock reorder point policy and random transshipment policy (Case II)
5 W3 Rs
LT132
LT231
C2BkOrdPr BL23 C2BackO BL22 C2Backlog ocess drs
C2Odrs Control
35 PR 31 RsW
RE13BkOrd BL36 BL35 RE13BO RE13BkOdr Proc
13 RE W1
R12TransR 13
BatchTransP 1C3
P1C3T P1C3 P1C3T 2 RdySen T3 4 C3Receiv P1C3Trans dP1C3 esP1
137
3 W1B
RS 31
31 BR
OC33
BL 31
C3P1Odrs Arrive
Sg31 C3Order sP1
OC OC 31 32
32 PR
C3BkOrdPr BL33 C3BackO BL32 C3Backlog ocess drs
C3Odrs Control
Case II: Both R1 and R3 have insufficient inventory Sender: R2 Receivers: R1 and R3
15
RE11BkOrd BL16 BL15 RE11BO RE11BkOdr Proc
BL17
W1RE11 RO RS14
B1
RE11Consd
RS13
RE11ReOrd
RS12
ChkRE11 P1Inv
W1T W1T W1T RE11P P1Deliver 1 EOQTranW 2 RdySend 3 TransP1To 1 RE11P1I PR11 C1OrdProce OC14 P1C1Ord W1RE11 1RE11 yRE11 RE11 nv ss s
P1
C1
11
T1
RS
P1C1 P1C1 P1C1 T4 C1Receive BatchTransP T2 RdySen T3 P1C1Trans 1C1 dP1C1 sP1
W1
BP1
1
RS1 5
BR
OC13
BL 11
C1P1Odrs Sg11 C1Order Arrive sP1
11
RsW
LT
PR
12
OC
12
12
OC
1
11
RP sW1 R 15 1
R12TransR 11
C1BkOrdPr BL13 C1BackO BL12 C1Backlog ocess drs
C1Odrs Control
BR
W1R
E11
RE12BkOrd BL26 BL25 RE12BO RE12BkOdr Proc
15
Rs
5 W2
LT12
BL27
W1RE12 RS24 RE12Consd RO
W1B 2
W1Reor der
RSW4
W1Consd
RSW3
W1ReOrd
RSW2 ChkW1P 1Inv
25
RS23
RE12ReOrd
TS2
RS22
5
PR
RS2
TS1
RSW5
P1Inv
P1W1 P1W1T P1W1 P1InvT2 P1Delive 1 EOQTran T2 RdySend T3 TransP1T P1W1 WareHou P1W1 ryW1 P1W1 oW1 se1
W
1R
2 E1
RE12 P1Delive W1T4 EOQTran W1T5 RdySend W1T6 TransP1T P1 RE12P1I PR21 C2OrdProce OC24 P1C2Ord OC23 W1RE12 W1RE12 oRE12 nv ss s ryRE12 BL PR 21 22
P1
1
C2
W
21
T1
Rs
ChkRE12 P1Inv
P1C2 P1C2 P1C2 BatchTransP T2 RdySend T3 T4 C2Receiv P1C2Trans 1C2 P1C2 esP1
BP2
2
RS 21
BR
C2P1Odrs Sg21 C2Orde Arrive rsP1
21
O
C2
O
2
C2
BR16
1
BL37
W1RE13 RS34 RE13Consd RO
3
RS3 5
LT232
RE13 P1Delive W1T7 EOQTran W1T8 RdySend W1T9 TransP1T P1 RE13P1I PR31 C3OrdProce OC34 P1C3Ord ryRE13 W1RE13 W1RE13 oRE13 nv ss s
P1
C3
RS
31
T1
BP3
1
Figure 7.12 Simulation model for a multi-echelon distribution system with installation stock reorder point policy and alternate transshipment policy (Case II)
Rs W3 5
TransCont rol2
TS 3
C2BkOrdPr BL23 C2BackO BL22 C2Backlog ocess drs
C2Odrs Control
LT23 1
35 P R 31 RsW
W1 RE1 3
TransCont rol1
TS4
R12TransR 13
RE13BkOrd BL36 BL35 RE13BO RE13BkOdr Proc P1C3T P1C3 P1C3T 2 RdySen T3 4 C3Receiv P1C3Trans dP1C3 esP1
RS33
RE13ReOrd
RS32 ChkRE13 P1Inv
BatchTransP 1C3
138
W1B
BR
OC33
BL 31
C3P1Odrs Arrive
Sg31 C3Order sP1
OC 31
31
OC
PR 32
32
C3BkOrdPr BL33 C3BackO BL32 C3Backlog ocess drs
C3Odrs Control
From the corresponding simulation models, the following performance results can be obtained:
Table 7.5 Costs with different warehouse reorder points for a distribution system with different transshipment policies (Case I)
W1_ReOrderPoints Random Policy Alternate Policy Regular (no transshipment) 4 228294.74 175966.38 237651.38 8 245647.16 148343.34 238801.87 10 230534.43 157234.59 238384.84 12 224913.27 157993.04 242837.31 15 224560.48 141397.59 245474.13 20 273261.01 146140.38 245076.86 25 264802.84 158731.68 252274.62 30 259250.28 157222.57 255144.90 35 269341.03 162011.65 259312.16 38 275891.04 175962.96 259903.95 40 274629.24 165883.67 262485.88 43 278206.78 189995.96 268108.86 45 277806.15 174809.24 270517.50
Correspondingly, we have the plots as shown in Figure 7.13. Random Transshipment vs. Alternate Transshipment Policies
300000 250000
Random Alternate Regular
Total Cost
200000 150000 100000 50000 0 0 10 20 30 40 50
Warehouse ReOrder Points
Figure 7.13 Costs with different warehouse reorder points for a distribution system with different transshipment policies (Case I)
139
The fill-rate performance is given in the following Table.
Table 7.6 Fill rates with different warehouse reorder points for a distribution system with random transshipment policy vs. alternate transshipment policy (Case I)
W1_ReOrderPoints 4 8 10 12 15 20 25 30 35 38 40 43 45 RA-R1 0.752 0.759 0.799 0.776 0.779 0.787 0.828 0.793 0.822 0.817 0.807 0.809 0.817 Alt-R1 0.941 0.939 0.930 0.937 0.912 0.949 0.953 0.926 0.937 0.950 0.955 0.941 0.953 RA-R2 0.862 0.865 0.885 0.878 0.896 0.882 0.889 0.895 0.898 0.903 0.897 0.887 0.913 Alt-R2 0.963 0.958 0.946 0.951 0.959 0.948 0.995 0.991 0.990 0.987 0.984 0.992 0.991 RA-R3 0.990 0.992 0.991 0.991 0.991 0.992 0.993 0.994 0.993 0.993 0.993 0.993 0.993 Alt-R3 0.929 0.929 0.930 0.930 0.940 0.945 0.946 0.946 0.947 0.944 0.945 0.947 0.946
The corresponding fill-rate performance plots are shown in Figure 7.14.
Random Transshipment vs. Alternate Transshipment Policies (Case I)
1.2 RA-R1 Alt-R1 RA-R2 Alt-R2 RA-R3 Alt-R3
1
0.8 Fill Rates
0.6
0.4
0.2
0 0 5 10 15 20 25 30 Warehouse ReOrder Point 35 40 45 50
Figure 7.14 Fill rates with different warehouse reorder points for a distribution system with random transshipment policy vs. alternate transshipment policy (Case I)
140
From above Tables and Figures, we have the following observations: (1) Alternate policy is more cost-effective than regular situation (no transshipment) and random policy. (2) Under random policy, the fill rate of retailer 3 is much better. However, retailer 1 becomes much worse and the fill rates of three retailers are still unbalanced. (3) Under alternate policy, the fill rates of retailers 1 and 2 are still good and the fill rates of three retailers become balanced. Similarly, for case II, we have the following simulation results (see Tables 7.7, 7.8 and Figures 7.15, 7.16, 7.17, 7.18).
Table 7.7 Costs with different warehouse reorder points for a distribution system with different transshipment policies (Case II)
W1_ReOrderPoints Random Policy Alternate Policy Regular (no transshipment) 4 87971.66 63197.78 78284.40 8 95991.76 61787.65 79120.77 10 90102.48 65609.13 77963.17 12 95167.50 61692.92 77983.61 15 98110.95 68949.24 78300.23 20 98875.31 69539.21 84303.58 25 107687.66 79278.66 91755.61 30 107966.09 80897.76 93117.99 35 109615.01 81600.50 94997.71 38 115765.26 83297.71 100163.20 40 116247.55 82424.05 96697.58 43 118738.99 87169.97 99273.72 45 125469.03 91165.99 102650.03
Random Transshipment vs. Alternate Transshipment Policies (Case II)
Random A lternate Regular
140000 120000
Total Cost
100000 80000 60000 40000 20000 0 0 10 20 30 40 50
Warehouse ReOrder Points
Figure 7.15 Costs with different warehouse reorder points for a distribution system with different transshipment policies (Case II)
141
Table 7.8 Fill rates with different warehouse reorder points for a distribution system with random transshipment policy vs. alternate transshipment policy (Case II)
W1_ReOrderPoints 4 8 10 12 15 20 25 30 35 38 40 43 45 RA-R1 0.973 0.977 0.977 0.976 0.977 0.983 0.979 0.982 0.981 0.981 0.981 0.980 0.981 Alt-R1 0.882 0.892 0.881 0.883 0.930 0.895 0.890 0.890 0.902 0.905 0.917 0.908 0.919 RA-R2 0.800 0.782 0.785 0.796 0.800 0.779 0.859 0.871 0.868 0.879 0.861 0.857 0.862 Alt-R2 0.974 0.966 0.969 0.963 0.963 0.964 0.987 0.992 0.991 0.992 0.991 0.988 0.991 RA-R3 0.976 0.976 0.975 0.973 0.975 0.974 0.981 0.981 0.984 0.983 0.982 0.982 0.981
RE1FillRate RE2FillRate RE3FillRate
Alt-R3 0.756 0.724 0.746 0.774 0.677 0.765 0.839 0.827 0.821 0.839 0.822 0.829 0.835
Unbalanced Demand (Case II)
1.2 1
Fill Rates
0.8 0.6 0.4 0.2 0 0 10 20 30 40 50
Warehouse ReOrder Points
Figure 7.16 Fill rates with different warehouse reorder points for a distribution system without transshipment Random Policy (Case II)
1.2 1 Fill Rate 0.8 0.6 0.4 0.2 0 0 10 20 30 Warehouse ReOrder Points 40 50 RE1FillRate RE2FillRate RE3FillRate
Figure 7.17 Fill rates with different warehouse reorder points for a distribution system with random transshipment policy
142
Alternate Policy (Case II)
1.2 1 Fill Rates 0.8 0.6 0.4 0.2 0 0 10 20 30 40
RE1FillRate RE2FillRate RE3FillRate
50
Warehouse ReOrder Points
Figure 7.18 Fill rates with different warehouse reorder points for a distribution system with alternate transshipment policy
From above Tables and Figures, the following observations can be obtained: (1) By adopting random transshipment policy, the fill rate of retailer 1 increases by 10%, the fill rate of retailer 3 increases by 34%. The fill rate of retailer 2 decreases by 16%. (2) By changing transshipment mechanism from random policy to alternate policy, the fill rate of retailer 1 increases by 2%, the fill rate of retailer 3 increases by 15%. The fill rate of retailer 2 decreases by 1%. In summary, the following observations can be obtained from the simulation study for a multi-echelon distribution system: (1) Depending on the structure of the distribution systems, either echelon stock or installation stock policies may be advantageous. (2) For Case I (2 senders and 1 receiver), the alternate transshipment policy improves the service levels and reduces the costs. (3) For Case II (1 sender and 2 receivers), the alternate transshipment policy can provide more balanced service levels and reduce the costs.
143
7.8 Component Commonality in Integrated Supply Chain Networks
7.8.1 Commonality Index (CI) The commonality index is a measure of how well the product design utilizes standardized components and is similar to work done by Collier (1981). A component item is any inventory item (including a raw material) other than an end item that goes into higher-level items. An end item is a finished product or major subassembly subject to a customer order. Different from Collier, two types of commonality indexes are defined in this research. One is called componentlevel (denoted as CIi), which is to provide an indicator on the percentage of a component being used in different products. The other is called product-level (denoted as CIp). There are three variables that will affect the commonality index, which are, number of unique components (denoted as u), number of total components along the product line (denoted as c), and final number of product varieties offered (denoted as n). To get the appropriate product-level CI, all these three variables along with component-level CI should be considered. The basic idea is that, by ranking the different component-level CI values, the average for the differences of CI values is computed. Then, this average difference will be multiplied by a weight, which is the ratio of (c-n) and u. A special case appears when all component-level CI values are same, u u and u ≤ n i i CI p = max{CI i } − min{CI i } i otherwise i × (n − c), u
(
)
(7-14)
u = number of unique components n = final number of product varieties offered c = total number of components along the product line In general, a higher CI is better since it indicates that the different varieties within the product family are being achieved with more common components. Figure 7.19 illustrates the use of the CI measures for five sets of two end products (labeled as A and B). Calculation of the CI is shown below each case. Here, we assume that all demands of products are same, i.e., d1 = d2. 7.8.2 Impact of Component Commonality on Integrated Supply Chain Performance The purpose of the simulation study is to evaluate the performance of “integrated supply chain with component commonality” versus “integrated supply chain without component commonality.” The simulation model for an integrated supply chain network with echelon stock policy and commonality index of 1 is shown in Figure 7.20. This simulation model is a comprehensive model since it contains raw material procurement, manufacturing processes, assembly operations, warehousing, and distribution functions. The corresponding source code and output are given in Appendix B. Three different performance measures are employed in the experiment: order fill rate, delivery time and total cost. The experimental results for fill rate, delivery time, total cost and resource utilization rate are summarized in Table 7.9.
146
Table 7.9 Simulation results for fill rate, delivery time, and resource utilization rate
CI Replications Delivery Time R1 Fill Rate R2 Fill Rate R3 Fill Rate M1UtilRate M2UtilRate 1 2 3 4 5 … 1 496 497 498 499 500 Mean SD 1 2 3 4 5 … 496 497 498 499 500 Mean SD 1 2 3 4 5 … 0 496 497 498 499 500 Mean SD 4223.63 4250.13 4222.55 4230.74 4240.48 … 4175.87 4207.11 4228.98 4260.51 4249.00 4239.24 146.91 8288.54 8284.00 8290.37 8286.03 8286.22 … 8294.76 8287.80 8290.99 8285.17 8298.49 8294.51 18.14 10211.95 10208.29 10198.74 10206.93 10202.91 … 10212.09 10208.49 10202.85 10203.70 10201.42 10210.21 29.80 0.918 0.882 0.915 0.911 0.881 … 0.918 0.960 0.929 0.934 0.956 0.920 0.107 0.840 0.844 0.850 0.844 0.815 … 0.846 0.838 0.828 0.859 0.803 0.816 0.072 0.758 0.728 0.761 0.762 0.760 … 0.745 0.722 0.747 0.763 0.746 0.740 0.057 0.952 0.963 0.964 0.956 0.952 … 0.939 0.953 0.966 0.960 0.945 0.955 0.044 0.909 0.883 0.908 0.927 0.914 … 0.927 0.913 0.941 0.871 0.923 0.939 0.069 0.908 0.904 0.895 0.907 0.904 … 0.906 0.896 0.902 0.894 0.918 0.907 0.031 0.918 0.940 0.897 0.934 0.942 … 0.971 0.888 0.888 0.888 0.916 0.919 0.105 0.838 0.873 0.834 0.811 0.865 … 0.819 0.834 0.819 0.872 0.848 0.829 0.077 0.775 0.785 0.761 0.767 0.778 … 0.767 0.802 0.802 0.781 0.772 0.786 0.071 0.981 0.976 0.981 0.979 0.977 … 0.991 0.984 0.980 0.973 0.976 0.978 0.031 0.991 0.991 0.991 0.991 0.991 … 0.991 0.991 0.991 0.991 0.991 0.991 0.001 0.686 0.686 0.687 0.686 0.687 … 0.685 0.686 0.686 0.686 0.686 0.686 0.002 0.952 0.947 0.952 0.950 0.949 … 0.962 0.956 0.952 0.944 0.947 0.948 0.030 0.958 0.959 0.959 0.959 0.958 … 0.958 0.958 0.958 0.958 0.957 0.958 0.003 0.778 0.778 0.779 0.779 0.778 … 0.777 0.778 0.779 0.778 0.779 0.778 0.003 M3UtilRate 0.901 0.903 0.867 0.917 0.911 … 0.888 0.889 0.879 0.899 0.916 0.901 0.068 0.839 0.840 0.838 0.840 0.840 … 0.838 0.839 0.838 0.839 0.837 0.838 0.003 1.000 1.000 1.000 1.000 1.000 … 1.000 1.000 1.000 1.000 1.000 1.000 0.000
1/6
147
RE11BkOrd BL16 BL15 RE11BO RE11BkOdr Proc
RsW 15
BL17
P1C1 P1C1 P1C1 BatchTransP T2 RdySen T3 T4 C1Receive P1C1Trans 1C1 dP1C1 sP1
RSW5
P1W P1W1 P1W1 P1 P1Delivery 1T1 EOQTran T2 RdySend T3 TransP1T W1 WareHou W1 P1W1 P1W1 se1 oW1
P1In
Rs W2 5 W1RE11
PR2
RS1
PR
W1Reor RSW4 der
5
11
W1Consd
RSW3
W1ReOrd
RSW2 ChkW1P 1Inv
15 RPR sW1 1
W1RE11 RS14 RS13 RS12 ChkRE11 RS RE11Consd RE11ReOrd 11 C1OrdProce OC14 P1C1Ord OC13 C1P1Odrs Sg11 C1Order RO P1Inv s Arrive sP1 ss O OC C11 BR BL 11 11 12 W1T W1T W1T RE11P C1Odrs P1Deliver 1 EOQTranW 2 RdySend 3 TransP1To 1 RE11P1I PR12 C1BkOrdPr BL13 C1BackO BL12 Control C1Backlog W1RE11 1RE11 yRE11 RE11 ocess drs nv
W1B1
P1C2 P1C2 P1C2 BatchTransP T2 RdySend T3 T4 C2Receiv P1C2Trans 1C2 P1C2 esP1
P1C1T 1
BP11
5
vT2
W
1R
BL27
P1C2 T1
Rs
RE12BkOrd BL26 BL25 RE12BO RE12BkOdr Proc
W 21
Rs W
E1
2
PR
ChkP1Inv
RS42
P1ReOrd
W1 RE
RE12 P1Delive W1T4 EOQTran W1T5 RdySend W1T6 TransP1T P1 RE12P1I PR22 C2BkOrdPr BL23 C2BackO BL22 C2Backl ocess drs og W1RE12 W1RE12 oRE12 nv ryRE12
RS
25
21
P1Inv
W1RE12 RS24 RE12Consd RO
2
RS23
RE12ReOrd
RS22
ChkRE12 RS P1Inv
21 C2OrdProce OC 24 P1C2Ord ss s
BP21
BR21
OC2 3 C2P1Odrs Sg21 C2Orde Arrive rsP1 OC BL O 21 21 C2 2
C2Odrs Control
2
P1Consd
P1T 1
BL37
Aaccumul C31 AmtC3 CO36 PreMfg ateC3 ToMfg Comp3
C4P2Odrs Arrive
OC43
W1RE13 RS34 RE13Consd RO
RS33
RE13ReOrd
RS35
RS32 ChkRE13 RS31 C3OrdProce OC34 P1C3Ord OC33 C3P1Odrs Sg31 C3Order sP1 Arrive ss s P1Inv OC BR BL 31 31 31
BP31
PR
31
P1C3 T1
RsP1 C4Odrs Control O
35
W1B
PR 35
Rs
W3 1
RS43
C4Order sP2
C4
RE13BkOrd BL36 BL35 RE13BO RE13BkOdr Proc
BatchTransP 1C3
P1C3T P1C3 P1C3T 2 RdySen T3 4 C3Receiv P1C3Trans dP1C3 esP1
13
Sg41
OC 41
W1B
PO28
BatchC om3
CMO3
TransR3T ReCM3 CM3Inv oCM3
Com3T oMfg
P3
P1
PreMfgC omp21
RdySend CM3
CM3 T2
SetupMfgC omp3
MA7
BO2
CheckC M3Inv RS7 2
RsCM3
P2Assembly P2Assembly1 P1Assembly P1Assembly1 A A A A
CO13
CO43
CO12
CO42
C32
CO
DelivCM3
CM3Con sd
CM
C3Inspe C34 Inspect C3 ctRdy
CO31
CO1
Mach3I dle
MA10
C13
MA
C
2 9C O O62
1
3
Mach1I dle
Aaccumul ateCom1
CA1
B3
3 C3
MA
MA
EOQTran CM3 CM3 T1
6
Aaccumul ateC1
Assemble 1P2 CO82
Assemble 2P2
Assemble 1P1
CO
Assemble 2P1
EOQTran CM1 CM1 T1
CM1Re Ord RS53
MfgComp1
22
CM3Re Ord RS73
MfgComp3
Aaccumul ateCom2 CA2
MA2
AmtC 1ToMf g
MA2 C12
1
RdySend CM1
CM1 T2
RA13 RA 43 1 RsCM RsCM2 CheckC SetupMfgC 1 M1Inv 21
MA4
RS5 2
SetupMfgC omp1
AmtC omp1 ToMfg
MA1
AmtCo mp2T oMfg
CO26
CA4
COB31
RS44
3
OC
32
PreMfgC13
PO27 P2C4Or ds
P1InvR O P1Inv BO
PO21
RE13 P1Delive W1T7 EOQTran W1T8 RdySend W1T9 TransP1T P1 RE13P1I PR32 C3BkOrdPr BL33 C3BackO BL32 C3Backlog ocess drs ryRE13 W1RE13 W1RE13 oRE13 nv
C3Odrs Control
BO3
PO29
C4Recei vesP2
P2T 1
PO7
BO1
PreMfgComp PO8 PreMfgC omp1
C10
P2BO
PO9
MvP2ToC4
P4
MvP1ToP1Inv
PreMfg Comp2
TransR1T ReCM1 CM1Inv oCM1
MA9
CM3 T3
RA71
MA5
RA 33
PO6
P2
CM1 T3
BO4
C30
SetupMfgC omp2
RA23 ReC CM2Inv Rs CM 2 CheckC M2Inv
M2
TransR2T oCM2
C21
CM2 T3
C42
CO23
CO22
CO53
CO52
C52
C22
MA 8
C11
MA
2
MfgComp2
A2 3
RdySend CM2
CM2 T2
11
RS6 2
COB
32 2 CO C P
21
C2 3
CO72
2
CA3
O
C35
Comp3Chk
BC5
RA11
GoodC3
GC31
MoveC3 ToInv
MoveCom 1ToInv
GoodCo mp1
GC
22
MoveCo m2ToInv
GC5
GoodC omp2
G
C2
3
CM
G
O
GC
GC43
GC
2
CM3Sup plier
CM3RO
7
BadC3
BC6 ReMfg C3
C8
O
1
BC1 BadCo BC2 mp1
GC 1
2
ReMfg C1
COB1
4
BatchC1 P2
DelivCM1 Cmp3Inv Cmp1Inv GC4 Cmp2Inv
CM
CM1Con sd
Com1Ins C14 Inspect pectRdy Comp1
BatchCo m2 COB2 1
M
C15
C1Chk
BatchC om1
Mach2I dle
Com2T oMfg
C1ToMf g
CM1Sup plier GC3
CM1RO
Com1T oMfg
ReMfgC2
CM2Re Ord Com2In Inspect C24 spectRd C2 y Comp2 5 Comp2Chk CM2Con sd BC4 BadCo BC3 mp2
EOQTran CM2 CM2 T1
MA4
CP21
RS7 4
RS 63
MA2 4 MA2 5
RS5 4
DelivCM2 RA 21
RS6 4
Figure 7.20 Simulation model for an integrated supply chain network with echelon stock policy and commonality index of 1
GC6
CM2RO
CM2Sup plier
148
For each performance measurement, an analysis of variance (ANOVA) is conducted to compare the performance of “integrated supply chain with different component commonality indexes” and “integrated supply chain without component commonality.” Here, the performance measures include delivery time and fill rates for different retailers. In the ANOVA, the level of confidence is set as α = 0.05. H0: µ1 = µ2 = µ3. H1: At least two of the means are not equal. The ANOVA are conducted as follows: (1) Analysis-of-variance for delivery time
Table 7.10 Analysis-of-variance for delivery time
Anova: Single Factor SUMMARY Groups CI=1 CI=1/6 CI=0
Count 500 500 500
Sum 2114450 4144618.5 5102868.5
Average 4228.9 8289.237 10205.737
Variance 594.0537555 20.70920108 20.28362331
ANOVA Source of Variation Sum of Squares Degrees of Freedom Mean Square Computed f P-value f critical Between Groups 186272964.43 2 93136482.21 439982.602 1.18E-61 3.00 Within Groups 316888.24 1497 211.6821933 Total 186589852.67 1499
Decision: Since P<0.05, or computed f > fcritical, reject H0 and conclude that the average delivery time are not all the same. However, we still don’t know which of the delivery-time means are equal and which are different. We need to perform the further multiple comparison tests. Here, we adopt Tukey’s test (Walpole et al., 1997). This test allows formation of simultaneous 100(1-α)% confidence intervals for all paired comparisons. The method is based on the studentized range distribution.
149
From the analysis-of-variance table, we know that the error mean square is s2= 211.68 (1497 degrees of freedom). The sample means are given by (ascending order): 4239.24, 8294.51, 10210.21
With α = 0.05, the value of q(0.05, 3, 1497) = 3.32. Thus all absolute differences are to be compared to 3.32 211.68 = 2.16 500
As a result, the following represent means found to be significantly different using Tuksy’s procedure: 1 and 2, 2 and 3, 1 and 3.
Therefore, we conclude that the delivery time of integrated supply chain with higher commonality index is significantly (with 95% C.I.) less than that of integrated supply chain with lower commonality index. (2) Analysis-of-variance for retailers’ fill rates
Table 7.11 Analysis-of-variance for retailer 1’s fill rate
Anova: Single Factor SUMMARY Groups CI=1 CI=1/6 CI=0
Count 500 500 500
Sum 460.2 418.35 374.6
Average 0.9204 0.8367 0.7492
Variance 0.00069449 0.00028468 0.00021218
ANOVA Source of Variation Sum of Squares Degrees of Freedom Mean Square Computed f P-value f critical Between Groups 0.146571267 2 0.073285633 184.545201 1.8E-16 3.00 Within Groups 0.59 1497 0.000397115 Total 0.74 1499
150
Decision: Since P<0.05, or computed f > fcritical, reject H0 and conclude that the average fill rate for retailer 1 is not all the same. The Tukey’s test is conducted as follows. From the analysis-of-variance table, we know that the error mean square is s2= 0.000397 (1497 degrees of freedom). The sample means are given by (ascending order): 0.74, 0.816, 0.92
With α = 0.05, the value of q(0.05, 3, 1497) = 3.32. Thus all absolute differences are to be compared to 3.32 0.000397 = 0.00296 500
As a result, the following represent means found to be significantly different using Tuksy’s procedure: 1 and 2, 2 and 3, 1 and 3.
Similarly, for retailer 2, we have:
Table 7.12 Analysis-of-variance for retailer 2’s fill rate
Anova: Single Factor SUMMARY Groups CI=1 CI=1/6 CI=0
Count 500 500 500
Sum 477.5 455.8 451.7
Average 0.955 0.9116 0.9034
Variance 7.4444E-05 0.00044027 5.2267E-05
ANOVA Source of Variation Sum of Squares Degrees of Freedom Mean Square Computed f P-value f critical Between Groups 0.015377867 2 0.007688933 40.6837815 7.1E-09 3.00 Within Groups 0.283 1497 0.000188993 Total 0.298 1499
151
Decision: Since P<0.05, or computed f > fcritical, reject H0 and conclude that the average fill rate for retailer 2 is not all the same. The Tukey’s test is conducted as follows. From the analysis-of-variance table, we know that the error mean square is s2= 0.000189 (1497 degrees of freedom). The sample means are given by (ascending order): 0.907, 0.939, 0.955
With α = 0.05, the value of q(0.05, 3, 1497) = 3.32. Thus all absolute differences are to be compared to 3.32 0.000189 = 0.00204 500
As a result, the following represent means found to be significantly different using Tuksy’s procedure: 1 and 2, 2 and 3, 1 and 3.
For retailer 3, we have:
Table 7.13 Analysis-of-variance for retailer 3’s fill rate
Anova: Single Factor SUMMARY Groups CI=1 CI=1/6 CI=0
Count 500 500 500
Sum 459.1 420.65 389.5
Average 0.9182 0.8413 0.779
Variance 0.00080773 0.00050934 0.00019733
ANOVA Source of Variation Sum of Squares Degrees of Freedom Mean Square Computed f P-value f critical Between Groups 0.097238467 2 0.048619233 96.313147 5.1E-13 3.00 Within Groups 0.756 1497 0.000504804 Total 0.853 1499
152
Decision: Since P<0.05, or computed f > fcritical, reject H0 and conclude that the average fill rate for retailer 3 is not all the same. The Tukey’s test is conducted as follows. From the analysis-of-variance table, we know that the error mean square is s2= 0.0005048 (1497 degrees of freedom). The sample means are given by (ascending order): 0.786, 0.829, 0.919
With α = 0.05, the value of q(0.05, 3, 1497) = 3.32. Thus all absolute differences are to be compared to 3.32 0.0005048 = 0.003336 500
As a result, the following represent means found to be significantly different using Tuksy’s procedure: 1 and 2, 2 and 3, 1 and 3.
From the above analysis, it can be shown that the fill rates of retailers 1, 2 and 3 of the integrated supply chain with higher commonality index are significantly (with 95% C.I.) higher than those of retailers 1, 2 and 3 of the integrated supply chain with lower commonality index, respectively. Therefore, the fill rates of integrated supply chain with higher commonality index are significantly (with 95% C.I.) higher than those of integrated supply chain with lower commonality index. Furthermore, the relative benefits from component commonality increase with the difference of commonality index values for two supply chain commonality configurations. (3) Resource utilization rates By comparing the machines’ utilization rates for the network configurations with different degree of commonality (see Table 7.9), it can be shown that the integrated supply
153
network with higher commonality index will generate more balanced machines’ utilization rates than the one with lower commonality index.
7.9 Simulation Model Verification and Validation
7.9.1 Verification The purpose of simulation model verification is to build the model right. In this research, verification is achieved by following steps: 1. Translation of conceptual model into simulation model (logic flowcharts); 2. Simulation program performs as intended (debugging in modules or subprograms, trace); 3. Graphical representation by STROBOSCOPE; 4. Examine the output for reasonableness under a variety of settings; and 5. Compare the output with the analytic results under simple assumptions. 7.8.2 Validation The goal of simulation validation is to build the right model. The following model validation steps are employed in this research: 1. Examine whether or not the simulation models are accurate representations of the system under study; 2. Examine whether or not the decisions based on the simulation models are consistent with the decisions based on the physical system. For instance, if the variance of warehouse replenishment lead time increases, then the safety stock and reorder point should also increase.
154