The aim of this paper is to test the impact of design and control strategies on flexible system of integrated manufacturing. A computer simulation model is developed to evaluate the effects of aforementioned strategies on the make-span time, which is taken as the system performance measure. This system does not have dedicated buffer but instead has central buffer. The impact of delay time for storage and retrieval of parts from the central buffer on the performance of the system is also studied. The result of the simulation shows that there is definite range of pallets for each level of routing flexibility at which the systems performs satisfactorily. It is also observed that different combinations of dispatching and sequencing rules give different levels of performance. Also there is some impact of delay time on the system performance.
Keywords: flexible manufacturing system, make-span, routing flexibility, simulation,
1. Introduction
In the present era, quality, innovation and variety are the primary features of product success. To achieve these requirements, manufacturing companies need to be flexible and responsive to changes in order to produce variety of products in a short time span. Among all the existing manufacturing system, manufacturers are adopting flexible manufacturing system (FMS). FMS is the system, which is equipped with the several computer-controlled machines, having the facility of automatic changing of tools and parts. The machines are interconnected by automatic guided vehicles, pallets and several storage buffers. The process of implementing an FMS is costly as it requires heavy capital investment in machinery and equipment. Hence the design and control of FMS requires an intensive work on planning an efficient and effective system.
Within the study of design strategies, the diverse types of flexibilities are studied. Browne et.al.1984 have comprehended eight types of manufacturing flexibilities, which are known as: machine flexibility, process flexibility, routing flexibility, operation flexibility, product flexibility, volume flexibility, expansion flexibility and production flexibility. Among all these, the routing flexibility is one possible manifestation of manufacturing flexibility at the shop floor.
The controlling action in any manufacturing system is having increasing importance. As far as the role of control strategies is concerned it manifests it self in the form of sequencing and dispatching decision of the parts in the system. Buffer is also an important component of any manufacturing system. In general the advantage of high buffer capacity causing high utilization is offset by the disadvantage of additional floor space and inventory as stated by Saad and Byrne [1]. While low buffer capacities reduce the floor space and higher inventory requirement, they are often not preferred because of reduced utilization and increased delay.
This paper presents a simulation study with the combination of various design and control strategies. In the design strategy, the impact of routing flexibility, location of buffer and number of pallets in the system has been taken into consideration. The sequencing and dispatching rules have been selected as the control strategies. In this study, SPT and MBPT are considered as the sequencing rule whereas the dispatching rules are NR (release the part to the idle machines), MINQ (minimum number in queue) and MWTQ (minimum waiting time in queue). The make span time is considered as the performance measure for the FMS. The different combinations of various strategies are studied with the help of ARENA simulator.
The remainder of the present paper has been organized in the following manner: Section 2 deals with literature review, followed by section 3 which delineates the description of the problem under study. Section 4 reveals the obtained results with the different manufacturing strategies. Finally, the conclusions are reported in the section 5.
2. Literature Review
A literature review was carried out to identify the previous research efforts and direction related to this work. The literature review addressed two broad domains of FMS i.e., design and control strategies.
The design strategies consist of routing flexibility, buffer location and number of pallets in the system. The control strategies consist of sequencing and dispatching rules. The routing flexibility measures the ability to perform operations by more than one machining center to handle machine breakdowns. Exploiting the routing flexibility in the discrete part manufacturing systems is studied by many researchers. Matsui et al. [2] studies the performance of flexible manufacturing systems with finite local buffers and fixed or dynamic routing rules. They showed that for a fixed routing model the system throughput in the case of finite local buffers is greater than in the case of infinite local buffers. Buyurgan et al [3] studied a hypothetical FMS having a central buffering strategy had been adopted where machining centers have no input or output buffers. Parts enter the system from a central input buffer and exit from a central output buffer. A limited buffer capacity in both input and output places of workstations gives the possibility for the system to experience a deadlock situation where no material can move as stated by Caumond et al., [4]. Ali and Wadhwa [5] studied the impact of routing flexibility on the make-span performance of a flexible manufacturing system. They considered three levels of routing flexibility, i.e., no routing flexibility, partial routing flexibility, and full routing flexibility. They observed that partial routing flexibility based system performed better that full flexibility system. Ali and Wadhwa [6] exploited the routing flexibility in the discrete part manufacturing systems involving variety production towards enhancing the make-span performance by integrating different entities flowing in the system. Matta et al. [7], states that it is not easy to estimate the effect that an increase or a decrease of the number of pallets can have on the saturation of machines.
Buitenhek et al. [8] have stated that the control mechanism in FMS specifies which part to process next at a machine upon the competition of the current operation. Chan et al. [9] studies the effect of dispatching and routing decisions on the performance of flexible manufacturing systems. They studied the impact of buffer capacities on the performance of the system by applying three routing policies. The performance measures considered is make-span, average machine utilization, flow time and average delay at the local input buffers. Abou-Ali and Shouman, [10] studied the effect of dynamic and static dispatching strategies on dynamically planned and unplanned flexible manufacturing system. Ali [11] has studied real time scheduling method for a system with continuous job arrival pattern. He has concluded that maintaining a balance of workload on each machine reduces the WIP inventory and that on an average SPT rule performed better than other sequencing rules. Matsui et al. [12] evaluates the performance of flexible manufacturing systems with finite local buffers and fixed or dynamic routing rules, and addresses the optimal design or system configuration problem of maximizing the system throughput. Das and Canel [13] studied the problem of scheduling batches of parts in a flexible manufacturing system (FMS) and developed a model could be used to minimize the total production time (make-span) i.e. minimize production time, minimization of inter-batch setup times becomes an important task. Altinkilic [14] has presented a use of simulation to improve shop floor performance. The performance of the existing system is evaluated by using ARENA. Saygin et al. [15] considered real-time manipulation of alternative routings in flexible manufacturing systems through simulation study. Similarly Shingoli et al. [16] studies the impact of manufacturing flexibility on flexible manufacturing system with help of simulation technique. Yan and Zhong [17] use Petri nets technique to resolve the problem of deadlocks in automated flexible manufacturing systems.
The above literature highlights, that the performance of FMS which is highly dependent upon the design and control decisions. However more work needs to be done in this direction. An attempt has been made in this paper to consider the impact of design and control strategies on the performance of flexible system of integrated manufacturing a sub domain of flexible manufacturing system.
3. Problem Description:
In the present flexible system of integrated manufacturing, a system configuration consists of 6 flexible machines (M1, M2, M3, M4, M5 and M6). All these machines are capable of producing any part types. Figure 1 depicts the layout of the system. It is provided with central buffer (CB) with finite capacity. However the capacity of CB is modeled as variable. The sequencing decision point (SD) and a dispatching decision point (DD) are attached to central buffer.
M1
M4
M5
M6
M2
M3
CB
SD
DD
Figure 1: Sample Flexible Manufacturing Systems
In all 6 part types (P1, P2, P3, P4, P5 and P6), are manufactured in this system. The number of operations required for processing the part type has been taken as 4 to 6. The total operations are 30. The detail of number of operations to be performed on each of the part type are as follows: Part 1=4 operations; Part 2=5 operations; Part 3=6 operations; Part 4=4 operations; Part 5=5 operations and Part P6=6 operations respectively. The parts can be routed through the different machines, for processing depending on the level of routing flexibility available in the system. The performance measure is considered as make-span. Make-span for 600 parts (100 of each six part types) is collected to evaluate the performance of flexible system of integrated manufacturing. First, different parts arrive in the system. These part types are held up at the loading area till it is released whenever the potential machine becomes available. Once the parts are released from the loading area, they are sent to respective machining stations for operation. If the machine is not idle, the parts are sent to the central buffer. Table 1 allocation of machines with processing time in brackets at routing flexibility level 1.
Table 1: Routing for sequence of operation and processing time for routing flexibility (RF=1)
Parts
O1
O2
O3
O4
O5
O6
P1
M1 (40)
M2 (40)
M3 (24)
M5 (24)
M4 (52)
M1 (52)
M6 (33)
M3 (33)
#
#
P2
M4 (17)
M3 (17)
M2 (39)
M1 (39)
M3 (41)
M4 (41)
M5 (38)
M2 (38)
M1 (26)
M5 (26)
#
P3
M5 (72)
M6 (72)
M1 (49)
M4 (49)
M3 (37)
M5 (37)
M2 (94)
M1 (94)
M4 (39)
M6 (39)
M6 (70)
M1 (70)
P4
M2 (55)
M1 (55)
M5 (92)
M3 (92)
M6 (38)
M2 (38)
M3 (92)
M6 (92)
#
#
P5
M6 (67)
M5 (67)
M4 (65)
M6 (65)
M2 (40)
M3 (40)
M5 (94)
M4 (94)
M1 (15)
M2 (15)
#
P6
M3 (64)
M4 (64)
M5 (52)
M2 (52)
M4 (66)
M6 (66)
M1 (36)
M5 (36)
M6 (63)
M3 (63)
M2 (50)
M4 (50)
The simulation model has been developed in the ARENA-11 simulation package. ARENA package was selected for modeling as it provides a good graphical interface and also the animation utilities. In ARENA package, there is no available feature to explicitly model the flexibility features. An extensive effort has been carried out to achieve this feature also. Now the modeling of different strategies is described which will highlight the underlying logic of the model and the associated assumptions.
The levels of routing flexibility are explained as: RF=0, means that there is exactly one machine for an operation on a given part i.e. there is 0 (zero) alternatives. RF=1, implies that there are two possible machines for processing the same operation i.e. there is exactly 1 more alternative machine (other than the machine which is available at RF=0) for any operation on any part. RF=2, implies that there are in all three possible machines for processing the same operation i.e. there are exactly 2 more machines available for processing the same operation (other than the machine which is available at RF=0). Similarly RF=3 and RF=4 imply 3 and 4 alternative machines are available respectively for any part/operation.
The dispatching decision is referred as the selection of the machine for processing the next operation on the part out of available alternative machines. The machine is selected on the basis of the dispatching rule enforced. The two sequencing rules and three dispatching rules have been used in the control strategies are explained here.
Sequencing Rules: SPT: Select the part that has Shortest Processing Time on the machine; MBPT: Select the part which has Maximum Balance Processing Time left for completing the total processing.
Dispatching Rules: MINQ: Select the next machine for processing the next operation on the part which has Minimum number of parts in the input buffer; MWTQ: Select the next machine for processing the next operation on the part which has Minimum sum of the processing times (on that machine) of all the parts waiting in the input buffer i.e. Minimum Waiting Time in Queue; NR: Release the part to the idle machines.
Verification and validation are two technique employed during the development of simulation modeling of system. In the present study, the verification of the simulation model has been carried out with the help of built in feature of ARENA package and by continuous monitoring of the values of some important variables. The facility of graphical animation of system during simulation run provides great support for verification of the model. The other facilities provided in ARENA, which help in verification, are SIMAN coding, which is in generated parallel, when the model is run. This helps for quick references. The second aspect of simulation model is validation. The validation of present simulation model is done according to the approach suggested by Fryer [18] that hypothetical simulation model, should perform in accordance with the assumption and logic. This is achieved by tracing step by step of the simulation run to validate, that the model is performing in accordance with assumption and under lying logic.
4. Simulation Results
In the following sections, we analyze and discuss the results obtained by performing simulation experiments keeping in view the above mentioned issues. We consider the following issues:
Impact of MST on RF at given NP with different DR/SR combination of rules.
Impact of DT for Storage and Retrieval of parts from CB.
Impact of NP and RF with different combinations of DR/SR
In this section, we present the results of the simulation experiments conducted under different combinations of dispatching and sequencing rules (i.e., control strategies). The simulation results indicate that by keeping the capacity of the centralized buffer fixed, different levels of routing flexibility and given number of pallets indicates different results for alternative control strategies. The result of NR/MBPT combination of sequencing and dispatching rules are shown in Figures 2.
Figure 2: Impact of MST on NP at different levels of RF with NR/MBPT
It is seen from Figure 2, that for all the levels of routing flexibility there is decrease in make-span when the number of pallets is increased from 6 to 12. When number of pallets is further increased from 12 to 60, the make-span remains almost constant at given level of routing flexibility. The systems blocks when the number of pallets is increased beyond 60 at RF=0, 1, 2 and 3. However at RF=4, and 5, the system blocks at 66 pallets. This shows, that with increase in flexibility levels, more pallets are allowed in the system. However the increase in number of pallets, does not improve system performance. When routing flexibility is increased to 1 we observe substantial decrease in the make-span at all the levels of pallets. We can observe from Figure 2, that there is fixed range of pallets within which the system performs satisfactorily i.e., 60 pallets for RF=0, 1, 2, and 3 and it is 66 for RF=4 and 5.
Table 2, below shows the summary of impact of various factors on the system performance. For example at RF=1, PR =6 to 60, NPMB = 24, NPB > 60 and reduction in MST is 20.44%, when number of pallets is increased from 6 to 24. Table 2 shows the comparison of performance of the system with NR/MBPT and NR/SPT as the combination of sequencing and dispatching rules.
Table 2: Impact of RF and NP on System Performance
Level of Routing Flexibility
PR
NPMB
NPB
Max. % Reduction in MST
0
6-60
36
>60
24.35%
1
6-60
24
>60
20.44%
2
6-60
42
>60
13.11%
3
6-60
54
>60
8.98%
4
6-66
48
>66
4.79%
5
6-66
36
>66
0.21%
Table 3 shows the different combination of dispatching and sequencing rules gives different level of performance. The maximum improvement in the system performance depends on the level of routing flexibility, number of pallets and control strategies. With NR/MBPT, the maximum benefit (24.35 %) is obtained at RF=0, and NPMB =36. With NP/SPT, the maximum benefit (19.45 %) is obtained at RF=1, and NPMB =12. It is seen that merely increasing the number of pallets in the system does not necessarily improve system performance.
Table 3: Comparison of impact between NP, RF and DR/SR on MST
DR/SR
RF
PB
NPMB
NPB
% Reduction in MST
NP/MBPT
0
6-60
36
>60
24.35%
1
6-60
24
>60
20.44%
2
6-60
42
>60
13.11%
3
6-60
54
>60
08.98%
4
6-66
48
>66
04.79%
5
6-66
36
>66
00.21%
NP/SPT
0
6-60
12
>60
7.24%
1
6-60
12
>60
19.45%
2
6-60
24
>60
12.68%
3
6-60
36
>60
08.56%
4
6-66
66
>60
04.45%
5
6-66
36
>66
00.21%
Next we change the dispatching rule from NR to MWTQ and evaluate the performance of the system. It is seen from Figure 3 also, that for all levels of routing flexibility there is decrease in MST when NP is increased from 6 to 12. At RF>=1, with increase in number of pallets beyond 66 the system starts blocking. If we compare this figure with Figure 2 we see that in this case also, it is fruitful to increase routing flexibility from 0 to 1 to get maximum benefit.
Figure 3: Impact of MST on NP at different levels of RF with MWTQ/MBPT
Table 4, below shows the summary of impact of various factors on the system performance. For example at RF=1, PR =6 to 66, NPMB = 60, NPB > 66 and reduction in MST is 26.42%, when number of pallets is increased from 6 to 60.
Table 3: Impact of MST on RF and NP (MWTQ/MBPT)
Level of Routing Flexibility
PB
NPMB
NPB
Max. % Reduction in MST
0
6-60
36
>60
24.35
1
6-66
60
>66
26.42
2
6-66
60
>66
14.15
3
6-66
36
>66
09.59
4
6-66
48
>66
04.69
5
6-66
30
>66
00.21
Table 5 shows the comparison of performance of the system with MWTQ/MBPT and MWTQ/SPT as the combination of sequencing and dispatching rules. It is observed from Table 4 that different combination of dispatching and sequencing rules gives different level of performance. The maximum improvement in the system performance depends on the level of routing flexibility, number of pallets and control strategies. With MWTQ/MBPT, the maximum benefit (26.42 %) is obtained at RF=1, and NPMB =60. With MWTQ/SPT, the maximum benefit (26.02 %) is obtained at RF=1, and NPMB =48. It is seen that merely increasing the number of pallets in the system does not necessarily improve system performance.
Table 4: Comparison of impact of MST on NP, RF and DR/SR
DR/SR
RF
PR
NPMB
NPB
% Reduction in MST
MWTQ/MBPT
0
6-60
36
>60
24.35
1
6-66
60
>66
26.42
2
6-66
60
>66
14.15
3
6-66
36
>66
09.59
4
6-66
48
>66
04.69
5
6-66
30
>66
00.21
MWTQ/SPT
0
6-60
12
>60
07.24
1
6-66
48
>66
26.01
2
6-66
30
>66
14.13
3
6-60
54
>60
09.48
4
6-66
60
>66
04.51
5
6-66
36
>66
00.37
5
6-66
36
>66
00.26
4.2 Impact of DT for storage and retrieval of parts from CB
In the above sections, the impact of routing flexibility and number of pallets on the system performance was investigated. The study was conducted by assuming zero delay time in storing and retrieving the parts from CB. In this section some delay time for storage and retrieval of parts from the CB is considered. It is important for FMS designers and controllers to have knowledge of the impact of these delays on the performance of the manufacturing systems. This knowledge will help designers and controllers to decide on the level of automation that will be beneficial to them. With delay mode we try to tackle the following issues:
Impact of DT: Performance of RF
It has been reported by the study on FMS with real time capabilities that the increase in flexibility results in a better performance. It is interesting to study whether this also holds true for system having CB having finite capacity and delay being incurred for storage and retrieval of parts from it. This delay time is added in the processing time of the parts whenever it arrives in CB. Further, how these delays affect system performance with increase in levels of routing flexibility is also quite interesting. Figure 4 shows the variation of MST with change in delay time on the performance of system at various levels of routing flexibility. The impact of delay time is being investigated at all the levels of routing flexibility (i.e., from RF=0 to RF=5) with MINQ/SPT as the combination of dispatching and sequencing rules. Figure 4 shows that there is continuous reduction in MST with increase in the levels of routing flexibility when DT>0. It is also seen that when the level of routing flexibility is increased, the variability in MST due to delay time reduces. This is due to the fact that as the levels of routing flexibility are increased there is better distribution of parts among different machines. This leads to less parts being routed to CB, hence reduction in the variability of MST with increased levels of routing flexibility. At RF=5, the impact of delay time is insignificant. This is due to the fact that at this level of routing flexibility fewer parts are moved to CB hence the less impact due to delay time.
At DT=0, with the increase in the levels of RF there is decrease in MST till RF=1. There after there is slight deterioration in the system performance. When RF levels are increased from 0 to 1, and DT is increased, there is decrease in percentage reduction in the MST. This is because when RF level is increased from 0 to 1, there is only one alternative choice of the machines on which the part can be processed. At the given number of pallets in the system and the capacity of CB, parts not finding the potential machine are moved to the CB, incurring DT which increases the processing time of the parts and ultimately the MST. This leads to percentage decrease in MST with increase in DT. However with the increase in RF from 1, and increase in DT there is increase in percentage reduction in MST. This is because with increase in RF levels there is more option to parts to select the potential machines. These results in less probability that the parts are moved to CB, hence less chance of incurring DT thereby improving system performance. From this study we conclude, that practitioners can consider some delay time for storage and retrieval of parts from CB and still get good performance. As we are aware that the maximum reduction in MST occurs when routing flexibility is increased from 0 to 1, the impact of delay time to some extend at this level of routing flexibility is not very much. Hence practitioners can consider delay time and still get almost same amount of benefit when delay is not used. The impact of routing flexibility on the system performance depends on the amount of delay incurred in storing and retrieving the parts from CB.
Figure 4: Impact of MST on RF at different levels of DT (MINQ/SPT)
5. Conclusion
The study in this paper was carried with flexible system of integrated system having centralized buffer. First we analyze the impact of design and control factors on the make-span performance of the system. We see that with system configurations, it can operate satisfactorily with 60 pallets at all the levels of routing flexibility. However simulation results indicate that, increase in routing flexibility and increase in number of pallets is not always beneficial. There is a suitable flexibility and pallet level, beyond which system performance deteriorates, as judged by the make-span measure of performance. It is observed in all the experiments that increasing the number of pallets beyond 60 or 66 lead to the blocking of the system. For different levels of routing flexibility, the system blocks, when the number of pallets is increased beyond 60 or 66, as the case may be. The range of pallets for all the level of routing flexibility, at which the system performs satisfactorily is 6 to 60 or 6 to 66. Hence we conclude from the experiments, that merely increasing the number of pallets in the system does not necessarily improve performance. We have also observed that the maximum reduction in MST occurs when the level of routing flexibility is increased from 0 to 1. There after increase in routing flexibility proves to be insignificant. Also there is some impact of control strategy on the system performance. Also there is impact of delay time for storage and retrieval of parts from the CB. We observe that the system performs satisfactorily even in the presence of delay time at all the level of routing flexibility. There is continuous reduction in MST with increase in routing flexibility at a fixed level of delay time. When routing flexibility is further increased, the variability in MST due to delay time reduces. This is due to the fact, that at higher levels of routing flexibility, fewer parts are moved to CB, hence less impact of delay time. Different control rules perform differently at different level of routing flexibility and delay time. Relative performance of all the control rules is influenced by the delay time that is incurred in storing and retrieving the parts in CB. Hence, it is necessary to decide the best design, and control strategies to get maximum benefit out of the proposed configuration of the system.
