differences between the scheduling rules diminished signifi- cantly at high flexibility   levels.

Kannan and Ghosh [28] described three new truncation procedures, along with two existing ones, which truncated jobs based on their critical ratio, operation slack, and change  in queue rank. Their simulation model consisted of ten machines. In addition to the above five truncation procedures, four dis- patching rules were used to dispatch the truncated jobs in the priority queue. Dispatching rules applied were SPT, FCFS, MSLK, and MODD. The performance measures of this study were mean flow-time, mean tardiness, standard deviations of flow-time, and tardiness. The authors concluded that the per- formance of the SPT rule could be improved using truncation procedures and the selection of the truncated rule was depen- dent  on the selected performance  measure.

Linn and Xie [72] investigated the influence of job sequen- cing rules on delivery performance of an automated storage/retrieval system base on simulation modelling.  They also examined the interaction of the  sequencing  rules  with other control variables such as production load level, product mix, and delivery due time  estimate.  The  simulation  model was written in GPSS/H and FORTARN and the  sequencing rules  were FCFS,  SDDT-F, and SDDT-L.

Gyampah [73] evaluated the part type  selection procedures for different tool allocation approaches. Three tool allocation approaches, three production scheduling rules, and three levels of part mix were assessed using a  simulation  model  of  an FMS coded in SLAM II. Three  tool  allocation  approaches were batching, flexible tooling, and resident tooling. The part type selection rules were LNT, SNT, and EDD. Performance measures in this study consisted of mean flow-time, mean tardiness, percentage of orders tardy, machine utilisation, and robot utilisation.

Sarper [34] examined two criteria against four dispatching rules under three system utilisation levels and five due-date assignment levels. Criteria used in the paper were MAL and MXAL, and dispatching rules were MDD, SPT, EDD, and FIFO. The results showed that the  MDD  rule  was  the  best one to minimise MAL. When MAXL was considered as a criterion, the Sarper did not  conclude  which  rule  dominated the others. In effect, there was no single rule that led to the best result for all system utilisation  levels  and  for  all  due- date tightness.

Kim and Bobrowski [74] used simulation to investigate the job-shop scheduling problems and  tested  the scheduling rules. A simulation model, with only one decision point, was developed using SLAM II which consisted of nine machines. Scheduling rules were CR, SIMSET, JCR, and SPT. The performance measures were set-up related measures, due-date related measures, flow-time related measures, shop utilisation, and cost measures. The model was run for two levels of due- date tightness. They did not conclude from simulation results which rule was the best    overall.

Tang et al. [36] proposed a framework for  a two-phase model of planning and scheduling. The planning phase was involved with part type and tool assignment in which a linear integer programming was used. In the scheduling phase, decisions were made at six decision points and various dis- patching rules  were evaluated  using a  simulation model.    The

model consisted of ten part types, 12 tool types, and five machine centres. Two AGVs were assigned for transporting parts among machines and between load/unload stations and machines. There were six decision points in the model includ- ing AGVs selection by parts, routeings selection, parts selection by machines at input buffers, next operation selection by parts at output buffer, parts selection by AGVs, and AGV destination selection. Decision rules used at each decision point mentioned above  were  as follows: