[59],  [60], [64],

[70],  [72], [74]

their work.

Hoffmann and Scudder [47] investigated the influence of dispatching rules which directly involved costs, on the time- oriented and cost-oriented measures using a simulation model with one decision point. The simulation model used in their study was written in FORTRAN using GASP-IV simulation language. The model consisted of nine machines that were categorised into three classes of processing cost. Eight different priority rules were examined in this study, including four time- oriented rules and four cost-oriented rules. Time-oriented rules consisted of SPT, EDD, MNSTUP, and CR; and cost-oriented rules consisted of MXPROF that selected the most profitable job in the queue, DOLSHP that selected the job having the highest selling price, IPDOL that tried to minimise in-process cost and selected a job with the highest current value, and VALADD that selected a job with the greatest value added in previous operations. Six performance measures were employed to assess the performance of the proposed rules. These are categorised into two groups: three time-measures and three cost-measures. Mean flow-time, average tardiness, and average lateness were time-measures, whereas average work-in-process dollars, average profit in all queues, and average dollars  of value added for work waiting to be processed, were cost- measures.  They found  that  SPT  minimised average  flow-time

and mean lateness, CR minimised mean lateness. IPDOL and VALADD minimised mean dollars in process, and MXPROF and DOLSHP dominated the other rules on the measure of average profit in the queue. They did not conclude which  rule is, overall, the best one with respect to all performance meas- ures. Scudder and  Hoffmann  [48]  employed  the  same model to explore the effectiveness of more composite time/cost priority-rules on both time and  cost  performance  measures. The model contained nine workstations, and only one decision point was considered. The due-date of jobs was assigned using TWK policy and thirteen priority scheduling rules were tested in this study. Two of them, SPT and CR, were pure time- oriented rules and another two rules, MXPROF and VALADD, were pure cost-oriented rules. Two rules, which were SPTTRN and MXPRFTRN, were truncated SPT and MXPROF at a maximum of 75 h waiting in the  current  queue, respectively. The remaining rules, MXPCRT, VLADCRT, VLADRAT, CRRATP,  CRRATV,  PRF/OPT,  and  PRF/TOPT,  were com-

posite, utilising time and cost information  in  various  ways. Ties were broken in all cases using the SPT rule. Two utilis- ation levels were tested (80% and 90%), each with five differ- ent cases:

1. SPT  compared  to SPTTRN.

2. MXPROF compared to MXPRFTRN and   MXPCRT.

3. VALADD  compared  to VLADCRT.

4. VLADRATCRRAT compared to CRRATP and   CRRATV.

5. PRF/OPT  compared  to PRF/TOPT.

The only ordinary rule  was  SPT  and  the authors  did  not try to compare the above rules with some other ordinary due-date- based rules such as EDD or SLRO. The same performance measures as in the previous study were used, and the authors concluded that VLADRAT was the best rule over all utilisation levels, but did not mention how they reached this conclusion. Dar-El and Sarin [49]  used  a  digital  simulation consisting of two decision points to study a real FMS with six machines. The effect of two dispatching rules and one heuristic algorithm on machine utilisation and minimum job tardiness were com- pared. Alternative routeings and part type due dates were investigated. Due-date information was used only for   releasing