Results for the Underlying Problem
The rules mentioned above and several variants of these rules that became out to be inferior are tested for eight various conditions of the FMS for the production of turbochargers. All these presentation criteria are evaluated for 25 various simulations run and after that averaged. The 25 various simulation runs are produced from five various random basic part type sequences and five various random due date sequences. For a importance analysis of the dissimilarity of two policies, the outcome of all 25 runs was utilized in a Wilcox on test.
Initially, the easy rules were compared. This twisted out that for the three criteria T_{mean}, T_{rms}, and T_{max} and the eight cases identify, the subsequent rules outperformed the others:
(a) ODD best rule along with respect to all criteria under high time pressure;
(b) SL/OPN best rule along with respect to each criterion under low time pressure;
(c) CR best mean and rms tardiness in case 2 under average pressure.
Amongst the combined and split queue rules, the excellent are as:
(a) CR + SPT best rule along with respect to T_{mean} in all the conditions;
(b) SPT - T r = 0;
(c) SPT / Sl u = 0;
The last two rules are the best ones along with respect to T_{ax} in all the conditions and along with respect to T_{rms} under medium and high pressure. For case 2 under low pressure, CR + SPT offer better results. For the underlying system also, the most complicated rules are outperformed via the easier ones.
Clearly, the identified priority rules for job shop scheduling can be divided in two classes depending upon their behavior under high pressure as:
(a) Rules that use SPT for critical jobs; and
(b) Rules that use SL or similar criteria for critical jobs.
The first class of rules attains the best values of T_{mean} at the cost of high values of T_{max}. The second class avoids excessive delays of a minute fraction of the jobs at the cost of a superior value of T_{mean}. Inside these classes, some rules show a comparable performance. In the initially class, CR + SPT for our problem provided the excellent results; in the second best were ODD and SL/OPN. These three rules generated Pareto optimal results along with respect to the three criteria. In the individual machine case, the Pareto-optimal rules were EDD and MOD that in this case are identical along with ODD respective CR + SPT.