Nasser Mebarki

Data mining based job dispatching using hybrid simulation-optimization approach for shop scheduling problem

A data mining based approach to discover previously unknown priority dispatching rules for job shop scheduling problem is presented. This approach is based on seeking the knowledge that is assumed to be embedded in the efficient solutions provided by the optimization module built using tabu search. The objective is to discover the scheduling concepts using data mining and hence to obtain a set of rules capable of approximating the efficient solutions for a job shop scheduling problem (JSSP). A data mining based scheduling framework is presented and implemented for a job shop problem with maximum lateness as the scheduling objective.

Best-Case Lower Bounds in a Group Sequence for the Job Shop Problem

Group sequencing is a well-studied scheduling method for the job shop problem. The goal of this method is to have a sequential flexibility during the execution of the schedule and to guarantee a minimal quality corresponding to the worst case. But the best case quality of a group sequence should also be interesting. This article presents new methods to evaluate the best case quality for any regular objective function. More particularly, three new makespan lower bounds are presented. The experiments performed with these lower bounds exhibit very good performances.

A study on the group sequencing method in regards with transportation in an industrial FMS

We are interested in jobs scheduling in flexible manufacturing systems. We try to combine predictive methods which guarantee optimal or good results but are not very robust to perturbations and reactive methods which take into account the real state of the system. For this, we use a method called the group sequencing method which aims at characterizing a set of solutions for the scheduling problem, instead of a single one. This method enables to add sequential flexibility to the operations to process. At the same time, this method guarantees a minimal quality in regards with the performance objectives. In this paper, we are interested in the behavior of this method in regards with both predictive and reactive methods, especially with various degrees of uncertainties. We present the experiments conducted on a real flexible manufacturing system to compare the performances of three scheduling methods in regards with uncertainties on the transportation times: the group sequencing approach, a predictive one and a reactive one. The results show very good performance with the group sequencing method.

Evaluation of a new human-machine decision support system for group scheduling

Abstract The group sequencing method is a proactive reactive scheduling method which aims at bringing flexibility as well as quality. This method is based on a human-machine system allowing to choose in real-time the next operation to perform in a group of permutable operations. In this paper, we propose a new decision support system to improve the cooperation between the human operator and the machine. This new system intends to favor the activity of the operator. To evaluate the pertinence of this new method, an experiment has been conducted. The results show the interest of this new decision support system.

Correlation among tardiness-based measures for scheduling using priority dispatching rules

One of the most commonly used methods to schedule manufacturing systems is to use priority dispatching rules (pdrs). It is frequently desired to distinguish the behaviour of pdrs with regard to tardiness-based performance measures. However, the relation among these performance measures is generally not obvious even for simple scheduling strategies such as pdrs. In this paper, we first focus on the maximum tardiness, which is a very interesting performance measure for the decision-maker on the shop floor. However, because of its singularity, it is not trivial to assess. We conducted a simulation study on a benchmark model of a dynamic job-shop system to evaluate the relative performance of a set of pdrs chosen either because they are used extensively or because they exhibit very good performance. Based upon the distribution of the maximum tardiness of these benchmark pdrs, we identify two sub-sets of pdrs. From this, we conducted experiments on the root-mean-square tardiness, which is used to distinguish a system with a few very tardy jobs from a system with a lot of less tardy jobs. The experiments show a positive correlation between maximum tardiness and root-mean-square tardiness. Because of the fact that the root-mean-square tardiness is an aggregate measure, it is much easier to assess than the maximum tardiness. This provides an opportunity to predict the relative performance of pdrs with regard to the maximum tardiness as well as the width of the tardiness by evaluating the root-mean-square tardiness only.

A procedure to compute a probabilistic bound for the maximum tardiness using stochastic simulation

Abstract Dispatching rules are widely used to dynamically schedule the operations in a shop. Their efficiency depends on the performance criteria of interest. One of the most important objectives to deal with in a manufacturing system is the tardiness which can be measured through several performance measures. This paper proposes an effective procedure to estimate the first two central moments (i.e., the mean and the variance) of the conditional tardiness and from this to compute a probabilistic bound for the maximum tardiness. These estimates are computed from the evaluation of the total tardiness, the number of tardy jobs and the root mean square tardiness obtained through a stochastic simulation. Different evaluations done by simulation show the effectiveness of the bound obtained.

Robust Machine Scheduling Based on Group of Permutable Jobs

This chapter presents the “group of permutable jobs” structure to represent set of solutions to disjunctive scheduling problems . Traditionally, solutions to disjunctive scheduling problems are represented by assigning sequence of jobs to each machine. The group of permutable jobs structure assigns an ordered partition of jobs to each machine, i.e. a group sequence. The permutation of jobs inside a group must be all feasible with respect to the problem constraints. Such a structure provides more flexibility to the end user and, in particular, allows a better reaction to unexpected events. The chapter considers the robust scheduling framework where uncertainty is modeled via a discrete set of scenarios, each scenario specifying the problem parameters values. The chapter reviews the models and algorithms that have been proposed in the literature for evaluating a group sequence with respect to scheduling objectives for a fixed scenario as well as the recoverable robust optimization methods that have been proposed for generating robust group sequence against scenario sets . The methods based on group sequences are compared with standard robust scheduling approaches based on job sequences. Finally, methods for exploiting group sequences in an industrial context are discussed and an experiment for human decision making in a real manufacturing system based on groups is reported.

A Study of Tardiness Based Measures for Benchmark Priority Dispatching Rules used in Dynamic Job Shop

Abstract One of the most important objectives to deal with in a manufacturing system is to minimize tardiness which can be measured through several performance measures. In this paper, a set of benchmark priority dispatching rules is used in a series of simulations under different operating conditions, to study the behavior of tardiness based measures, especially of the maximum tardiness. Two sets of rules have been identified based upon the distribution of the maximum tardiness. Experiments have shown that maximum tardiness and root mean square tardiness of the system are positively correlated. This provides an opportunity to predict the worst-case behavior of the manufacturing system in regards with tardiness as well as the width of the tardiness by evaluating root mean square tardiness.

Investigations on the flexibility of a group sequence and its effect on the best-case schedule

The group sequence is a well-known proactive-reactive scheduling method that brings sequential flexibility to an initial schedule in order to absorb uncertainties. Hence, group sequence favors the cooperation between human and machine. This method guarantees a minimal quality corresponding to the worst-case. This indicator associated with the best-case schedule provides the decision maker two bounds helping him to choose which operation will be executed first. In this paper, we consider the effect of the group sequence flexibility on the best-case schedule. The experiments made on very well-known instances of the job shop problem, using the makespan objective, brought two new results. The first one is that the best-case schedule of a group sequence is not guaranteed to be an optimal solution for the job shop problem, and the second one is that the more flexibility in a group sequence the better is the quality of the best-case.

Perturbation Analysis to set the number of automated guided vehicles in a flexible manufacturing system

By using perturbation analysis on a simulation model of a flexible manufacturing system, this paper proposes a procedure to size this system in order to maximize its productivity. This procedure is based on the identification of the function between flowtime (i.e., the flowtime for each product is computed as the delay between its completion time and its arrival time) and the decision variable, here the number of automated vehicles in the system. A formal evaluation of the performance gradient in each point of the simulation model is then possible.