Mahmoodi et al. [79] examined the effects of a new combi- nation scheduling rule on the performance of  a random FMS and compared it with three existing  rules,  which  were SPT, SIx, and SPT/TOT. The new rule was CR-SPT which calculated the critical ratio, which was due date minus the current time, pided by the estimated remaining processing time. They attempted to use this rule to compensate for SPT’s short- comings considering part due  dates. The FMS was written  in the SIMAN simulation language. The system consisted of three machine types, eight machines, three load/unload stations, 16 AGVs, an infinite central WIP buffer, a transfer station, and a computer-controlled network. Performance measures were aver- age flow-time, average percentage tardy, and average tardiness. Results showed that CR-SPT did not perform the best when compared with the other rules. It was the poorest rule with respect  to  average flow-time.

Jayamohan and Rajendran [80] analysed two different approaches for scheduling flexible flow shops. The two approaches were:

1. Use of different dispatching rules at different stages of the flow shop.

2. Use of the same dispatching rules at all the stages  of  the flow shop.

The model under evaluation had three work centres, and each work centre consisted of two machines. Performance measures were mean flow-time, maximum flow-time, variance of flow- time, mean tardiness, maximum tardiness, variance of tardiness, and percentage of tardy jobs. Scheduling rules used at   different

stages were SPT, ATC, RR, MOD, COVERT,  PT  (process time) WINQ    (work    in    next    queue) SL    (slack), PT WINQ AT (arrival time), PT WINQ SL AT, SPT then EDD then SPT, and SPT then SPT then EDD. Jayamohan and Rajendran concluded that the use of various simple rules at different stages of manufacturing systems might not always be good. They advocated the use of single dispatching rules, especially those incorporating the both process time and due date, at every stage of the    system.

Subramaniam et al. [81] proposed three machine selection rules, namely LAC, LAP, and LACP, in a dynamic  job shop. The proposed rules were applied together with four common dispatching rules, which were RAN, FIFO,  EDD,  and  SPT. The effectiveness of the new rules was evaluated through a simulation study. The system under consideration consisted of three machines. Mean operational cost and mean tardiness were employed as the performance measures. Results showed that LAC and LAP rules performed well for the mean  cost  and mean tardiness, respectively. The authors concluded  that  the use of the machine selection rules significantly improved the scheduling performance of dispatching   rules.

3. Multi-Criteria Scheduling Approaches

3.1 Introduction

Because of rapid change in demand, FMSs are working with different orders, each of which aims at some criteria. Therefore, operating an FMS is, in fact, a multiple criteria activity. Some authors employed these criteria in their modelling. For example, Lee and Jung [82] developed a formulation for part  selection and allocation using goal programming (the basic concept of goal programming involves incorporating all goals into a single model). Their model considered the goal of meeting production requirements, balancing of machine utilisation, and minimis- ation of throughput time of parts. This kind of goal program- ming could be used by decision-makers to satisfy their goals and their prioritisation. Two shortcomings of this kind of modelling are that information on the dynamic working of an FMS cannot be provided and the  effect  of  the  waiting  times on the system performance cannot be taken into account. The greatest disadvantage of this sort of programming is that it is computationally  costly and is expensive  in  practice.