Schriber and Stecke [55] provided a simulation study of an FMS that investigated the effect of various modellings. They

demonstrated successfully the validity of the theoretical results and tested the sensitivity of parameters such as the number of AGVs, the number of buffers, and  the  level  of  WIP.  They used SPT as a dispatching rule, and two major performance measures were machine utilisation and production rate. Schriber and Stecke [56] further extended their previous study and provided a comparison of FIFO and SPT scheduling rules, but they did not extend the number of performance measures.

Denzler and Boe [57] used simulation to study a dedicated FMS by using actual data  of  routeing  and  operation  times. The model was comprised of 16 CNC machines and pallet loading for the investigation of part-dispatching rules. The influence of the number of pallets and  different  scheduling rules on machine utilisation was demonstrated, which was the only performance measure used in the model. Results indicated that FMS performance was significantly affected by the choice of  scheduling rules.

Co et al. [58] investigated the influence of queue length on five sequencing rules, which were FCFS, SPT, LWKR, TWK, and NXQL. A computer simulation model was developed to evaluate the performance of the alternative sequencing rules listed above, under various system configurations. The simul- ation model was written in SIMAN and contained n part types and m single server stations. Each part type followed a fixed routeing sequence, which defined the number of operations required to complete each job, the sequence of  machines to visit, and the related processing times. The only performance measure reported was mean flow-time. The authors confirmed that the influence of the sequencing rules on the performance of the system, even in the case of short queue length, should not be ignored. No due-date-based rules and related perform- ance measures were used in the simulation     model.

Chryssolouris et al. [59] compared  the  system performance of a manufacturing system with different dispatching  rules using a simulation model with only one decision point. Four dispatching rules and four performance measures were employed. The performance measures included mean  flow- time, number of task orders completed,  average  WIP,  and mean tardiness. They did not conclude which rule dominated the  others for all performance   measures.

Choi and Malstrom [60] described the use of a simulator to evaluate work scheduling rules in an FMS. An FMS was modelled using actual data. The model consisted of a miniature closed-loop system with eight NC machines, five robots, one washing station, and I/O queues. The combinations  of seven part selection rules with four machine centre selection  rules were investigated. Part selection rules were RANDOM, FSFS, EDD, SPT, SLACK, S/PT, and VALUE. Machine selection rules were RANDOM, FMFS, NINQ, and WINQ. Scheduling rules were evaluated using the following performance measures: actual and relative system effectivity, total and average travel- ling time, actual production output, achievement rate, total and average manufacturing throughput time, total and average wait- ing time, imminent operation work content, and total and average production lateness. The authors concluded that the SLACK/WINQ and SPT/WINQ scheduling rules dominated the other major decision rules for the due-date- and  flow-time- based criteria.

Slomp and Gaalman [61] proposed three scheduling pro- cedures. All procedures executed four functions, but the sequences of execution were different. The first function sched- uled workstations and used the earliest possible moment rule. The second function scheduled transport devices and used the earliest moment rule and, in the case of a tie, the least moving time was used. The third  one  scheduled  operators  and  used the earliest moment rule and, in the case of a tie, the operator who needed the least walking time was chosen. The  last function scheduled operations and employed four dispatching rules consisting of SPT, SPT.TOT, SPT/TOT, and EFTA. The simulation model of a case study was used including two workstations, one loading station, and one unloading station. There was one input buffer with a capacity  of  two  jobs  in front of each workstation. Only two performance measures, makespan per part and mean flow-time, were used  in  the model. In addition to the case study, several simulation tests were carried out to demonstrate the behaviour of the procedures developed. In each simulation, scheduling started with an empty system, but the author did not  mention  how  the  initial  bias (i.e. the transient state) of the results was eliminated. Results showed that the EFTA rule performed worse than SPT/TOT regarding both measures of performance. The main shortcoming is that due dates were completely ignored in this model.