We address the parallel machine total weighted tardiness scheduling problem with release dates. We describe dominance rules and filtering methods for this problem. Most of them are adaptations of dominance rules based on solution methods for the single-machine problem. We show how it is possible to deduce whether or not certain jobs can be processed by a particular machine in a particular context and we describe techniques that use this information to improve the dominance rules. On the basis of these techniques we describe an enumeration procedure and we provide experimental results to determine the effectiveness of the dominance rules.Keywords: scheduling, parallel machines, total weighted tardiness, dominance rules, possible machines. * corresponding author 1
IntroductionIn this paper we consider the situation where a set of n jobs J = {1, . . . , n} has to be processed on a set of m identical parallel machines M = {1, . . . , m} and where the objective is to minimize the total weighted tardiness. Associated with each job i is a release date r i , a processing time p i , a due date d i and a weight w i . A job cannot start before its release date and preemption is not allowed. Machines cannot execute more than one job simultaneously. The tardiness of job i is defined as, where C i is the completion time of job i. The problem is to find a feasible schedule with minimum total weighted tardiness. This problem, denoted as P m|r i | w i T i , is strongly NP-Hard [25].The single machine problem has been intensively addressed over the last two decades. For the 1|r i | T i problem, Chu and Portmann [17] and Chu [15] introduced dominance rules, heuristics, a lower bound and branch-and-bound procedures. Baptiste et al.[7] also proposed a lower bound and a branch-andbound method using efficient dominance rules. For the 1|r i | w i T i problem, Akturk and Ozdemir [2] proposed dominance rules used in a branch-andbound algorithm [1,2]. Finally, Jouglet et al.[22] studied these two criteria and generalized and improved the above dominance rules which were used in a branch-and-bound method. More recently Jouglet et al.[23] described dominance-based heuristic methods for these problems.As regards parallel machine problems, several exact methods have been described. Yalaoui and Chu [32,33] described dominance rules, lower bounds and branch-and-bound methods for the general P m|r i | C i problem. More recently Nessah, Yalaoui and Chu [28] described dominance rules and a branch-and-bound method for the weighted case P m|r i | w i C i . Less attention has been given to the tardiness criteria. Nevertheless, Azizoglu and Kirca [3], Yalaoui et Chu [32] and Shim and Kim [30] solved the problem with equal release dates P m|| T i using a branch-and-bound algorithm and dominance rules. Moreover, Liaw et al.[26] also proposed a branch-and-bound algorithm using dominance properties for the weighted case considering unrelated parallel machines. Finally, Baptiste [6] showed that the special case P m|p i = p, r i | T i of the total ...