A comparative study of some priority dispatching rules under different shop loads

Abstract

Priority dispatching rules and shop load affect jobshop performance. This paper evaluates ten different priority dispatching rules with respect to six different performance criteria under light, medium and heavy shop loads. Simulation was used as a tool to determine the rankings of the dispatching rules for a given shop load and performance criteria. A comparative study was conducted to investigate the performance of these rules. Two rules, 'shortest processing time' and 'least work remaining'. performed well under criteria related to the processing time. It was found that the rules that perform well in average and r.m.s. tardiness perform poorly under percentage-of-jobs-late criterion. The experimental results are conveyed with critical comments on the performance of the dispatching rules under different loading conditions of the shop.

A comparison of priority rules for the job shop scheduling problem under different flow time- and tardiness-related objective functions

“…The middle column of Figure 1 shows the comparative priority rule rankings used to solve the traditional as well as the flexible job shop problem for each of the five objective functions. The graphs show that the relative performance of the priority rules to solve the problems under the mean flow time, mean tardiness and proportion of tardy jobs objectives does not differ much from their performance on the classical job shop scheduling problem, which partly confirms the conjecture of Waikar et al (1995). However, the best performing rule for the traditional job shop problem is not always the best one for its flexible variant.…”

“…This block is inspired by the remarks of Dominic et al (2004) who concluded that 9 The results of the comparative study are displayed in Table 2 for each of the five objective functions. It is worth noting the similar comparative studies has been made by other authors, such as Waikar et al (1995), Holthaus and Rajendran (1997), Rajendran and Holthaus (1999), Holthaus and Rajendran (2000), Jayamohan and Rajendran (2000), Jayamohan and Rajendran (2004), Dominic et al (2004) and Chiang and Fu (2007), but with a slightly different focus or limited scope compared to our full comparison on the five objective functions. The results can be briefly summarized per objective function along the following paragraphs.…”

“…Section A.2 list the single priority rules together with their mathematical formulation. For this overview, we based ourselves on the papers of, among others, Panwalkar and Iskander (1977), Blackstone et al (1982), Baker (1984), Russell et al (1987), Vepsalainen and Morton (1987), Anderson and Nyirenda (1990), Waikar et al (1995), Holthaus and Rajendran (1997), Rajendran and Holthaus (1999), Jayamohan and Rajendran (2000), Holthaus and Rajendran (2000), Dominic et al (2004), Mizrak and Bayhan (2006) and Chiang and Fu (2007). Finally, in Section A.3, more information is given on the combination of the single rules.…”

“…For this reason, the priority rule list on the x-axis of Figure 1 is slightly different for the flexible job shop column (column 2) than for the other columns. Indeed, priority rules that cannot be used to solve the (integrated) flexible variant of the 17 Waikar et al (1995) assumed that the performance of priority rules for the traditional job shop problem would not be affected by the addition of alternative machines for each operation, which is the subject of the computational tests of this section. To that purpose, the vdata-set proposed by Hurink et al (1994), containing the most flexible instances, are used.…”

“…In doing so, the priority rules can be used as dispatching rules, which dynamically assigns jobs to the partial schedule upon arrival. In most research papers, arrival patterns are generated from an exponential (Waikar et al, 1995;Holthaus and Rajendran, 1997;Rajendran and Holthaus, 1999;Jayamohan and Rajendran, 2000;Dominic et al, 2004) or a Poisson distribution (Vepsalainen and Morton, 1987;Barman, 1997;Taner et al, 2003), where the distribution parameters are linked to the machine utilization levels. However, in their paper, Waikar et al (1995) argue that the job shop performance is relatively unaffected by the type of arrival time distribution, but instead, the average number of jobs arriving per time unit is important since it affects the shop load level.…”