the  parts into  the  system,  not for scheduling.

Lin and Lu [50] studied a simulation method, with two decision points, that allowed parts  having  alternative  routes into the system. The impact of various methods  of  parts entering the system and the scheduling rules on four perform- ance measures were investigated. The performance measures were mean flow-time,  mean  waiting  time,  ratio  of maximum to minimum average queue length, and machine utilisation variance. To put parts into the system, two heuristic methods were used, one based on balancing workload and the  other based on balancing workload plus the minimum  number  of tardy parts, in addition to EDD and FIFO.  Scheduling rules used in the model were  WINQ,  SPT, and FIFO. WINQ  led  to a better result for mean flow-time. Although  the  due-date- based dispatching rule was not employed, one due-dated-based performance  measure  was examined in  the model.

Wilhelm and Shin [51] described a study which investigated the  influence  that  alternative  operations  might  have  on    the

performance of FMSs. Four process selection rules (NA, AP, AND, APD) were evaluated, which can be applied to a loop- type FMS with only an infinite common buffer. A SIMSCRIPT model was developed to evaluate the performance of proposed rules. The system contained four machining centres, a load/unload station,  a  set  of  AGVs,  and  three  part  types. The performance measures were makespan, system utilisation, utilisation of inpidual machines, flow-time, maximum spaces required in the common storage, and maximum number of vehicles required. It was concluded that there was no single process selection rule which was superior to the  others. The main shortcoming of this work is that only one decision point was  used  in  the model.

Kimemia and Gershwin [52] compared the LIFO rule with their developed algorithm on the system utilisation and pro- duction rate, using a simulation model that consisted of two workstations. The question that arises here is why the authors selected only the LIFO rule to compare with the algorithm results.

Chang et al. [53] reported a two-step method for scheduling parts using simulation. The suggested procedure was compared with some dispatching rules including SPT, LPT, FCFS, MWKR, and LWKR. The performance measure for this com- parison was mean flow-time. The simulation model consisted of four machines and three parts. One disadvantage of the proposed method was its high computational time which  made it unsuitable  for a real-time  scheduling.

Chan and Pak [23]  studied  a  hypothetical  FMS consisting of four machines and one loading/unloading station. They explored the effect of three heuristic dispatching rules on the cost of tardiness, makespan, and average lead time using  a digital simulation developed in FORTRAN. The influence of rules was tested in both static and dynamic  conditions  for a finite plan horizon. In both conditions the developed heuristics were compared with one due-date-based rule (SLACK) and one processing-time-based rule (SIO). They considered alternative operations for parts, but did not  use  any  operation selection rule in their simulation   model.

Abdin [54] studied a scheduling problem of a job-shop type FMS with machine breakdown and considered three levels of decision making (decision point), that is, selection of machine tool, selection of transport device, and selection of parts from input buffers. An alternative machine was considered  only when the buffer of the original  machine  was  full.  The FMS was modelled by a discrete-event simulation using SLAM II. The model consisted of one loading/unloading station, four multi-purpose CNC machines, and two  carts.  The  SPT  rule was the only dispatching rule used to select a part from the input buffer. The SDS rule determined which transport device to select if some were available. Five performance measures were employed consisting of machine utilisation, WIP, system throughput, mean flow-time, and makespan. The author found that schedules with alternative machine tools were better than schedules without alternative machine tools, and concluded that FMSs without alternative machine tools resemble transfer line systems. No effort had been made to combine the scheduling rules and apply them to the three decision points.