1. FMS scheduling problems within the context of a general decision-making process.

2. An overview of multi-criteria decision-making approaches and  its feasibility  to  FMS scheduling problems.

3. The literature of FMS scheduling involving multiple objec- tives.

4. The major findings.

Ishii and Talavage [86] demonstrated a mixed dispatching rule (MDR) approach, in contrast to a single dispatching rule (SDR) approach, for assigning different dispatching rules to FMS machines, which considered one decision point only. The developed approach proposed different dispatching rules for different machines. In addition, a search algorithm that selected appropriate mixed dispatching rules using predictions based on discrete-event simulation, was developed for this approach. The effectiveness of the mixed dispatching rule and the efficiency of the search algorithm relative to an exhaustive search were shown using an FMS. It  was  reported  that  it  was  up  to 15.9% better than the  conventional  approach,  and  4% better on average. The best combination of  dispatching  rules  that they found were NINQ, SPT, SLACK, and FIFO. The system performances were based on mean flow-time, mean tardiness, weighted mean flow-time, weighted mean tardiness, and combi- nations of them. The FMS example was modelled using SLAM II.

Yang and Sum [87] selected total cost as a better overall measure of satisfying a set of different  performance measures. A penalty was applied to both early and tardy jobs. A simul- ation model using SLAM II was  employed  to  examine  the cost performance of various dispatching rules.  The  total cost and its cost components were  reported  as  the  costs incurred per simulated year. Dispatching rules were SPT, SWPT, CR, CRTC, WCRTC, and VLADRAT. Dispatching rules were investigated under several environmental factors including machine utilisation (low and high), due-date allowance (loose and tight), tardiness penalty rate (low and high),  and  interest rate and cost rate for holding inventory (low and high). Only one decision point was considered in the simulation model.

Maheshwari and Khator [88] demonstrated that machine loading and control strategies, buffer size, number of pallets, etc., could be evaluated simultaneously. Several issues were stated  as follows:

In which sequence should the parts be launched into the sys- tem?

What  priority should be assigned to the  parts  at the  machin- ing centre?

What priority should be assigned to the parts for transportation? What dispatching rules should be adopted  for  material hand- ling vehicles?

For a machine-loading model, an integer programming  was used to determine part assignment and tool allocation with material handling considerations. Four objectives were used    as

a loading strategy including minimisation of operational cost, minimisation of operational time, balancing of machines, and minimisation of the sum of penalty and operational cost. A discrete-event simulation model was used for the control level. The model was developed using SIMAN and consisted of four workstations, one load/unload station, one  staging  area,  and two AGVs to transport the parts between machines. Part releas- ing rules consisted of LPR, LNV, and STPT RAN. Dispatching rules applied at the input buffer were SPT, SRPT, SPT/TPT, and FIFO. Vehicle dispatching rules were MWIQ, MRV, MRQS, and FIFO. As a result, Maheshwari and Khator con- sidered one loading and three control decision points, and two other parameters considered were buffer size and number of pallets. Two performance measures were used, e.g. makespan and mean flow-time. It was concluded  that  when  makespan was considered, a combination of LPR, MWIQ, SPT/TPT, five buffer spaces, and ten pallets worked best. When considering mean flow-time, a combination of LPR, MWIQ (or FIFO), SPT/TPT, three buffer spaces, and  six  pallets,  outperformed the others. When both performance measures were considered, the combination of LPR, MWIQ, SPT/TPT, four buffer spaces, and eight pallets was best. Machine workload balancing strat- egy outperformed the other three loading strategies for both performance measures. They did not mention what procedure was used to arrive at such  a  conclusion.  No  due-date-based rule  or  performance  measure was  used in  the model.