Yih-Long Chang

A new heuristic for the n-job, M-machine flow-shop problem

Abstract This paper describes a new improvement heuristic for solving the flow-shop scheduling problem. This new method minimizes gaps between successive operations in solutions generated by other heuristics. We compare the new heuristic with five well-known heuristics and the random case by using simulation with makespan, mean flow time and mean utilization as the performance measures. The simulation results show that the new heuristic improves the best heuristic in the three measures by a significant percentage, while the increase in CPU time is modest.

Minimizing the number of tardy jobs for m parallel machines

The importance of the criterion of number of the tardy jobs has been well recognized in the literature, but there is no efficient heuristic to minimize this objective in the parallel machine case. This paper proposes two fundamental approaches, job-focused and machine-focused, to minimize the number of tardy jobs in the parallel processors. The machine-focused approach heuristic is an important extension of Moores (also called Hodgsons) algorithm to solve the number of tardy jobs problem for the one-machine case. An extensive simulation experiment is conducted to test the effectiveness of the heuristics, and shows that three of the proposed heuristics (two heuristics from the job-focused approach and the machine-focused heuristic) perform very well. We also compare the heuristic solutions with the optimal solution for 800 small problems so as to access the average deviation above the optimum. The simulation results show that the best heuristic is about 1.2% of the optimum.

A simulated annealing approach to scheduling a manufacturing cell

This article proposes a new heuristic based on simulated annealing that schedules part families, as well as jobs within each part family, in a flow-line manufacturing cell. The new scheduling approach is compared to a branch and bound algorithm as well as two other family-based scheduling heuristics for different cell configurations. The results reveal that all the heuristics provide comparable solutions to the optimal procedure for small problems. However, when the problem size increases, the simulated annealing heuristic outperforms the other procedures not only in solution quality but also by requiring substantially less computation time.

A bottleneck-based beam search for job scheduling in a flexible manufacturing system

In this paper, we study job-scheduling methods for flexible manufacturing systems (FMSs). Routeing flexibility is a feature that distinguishes FMS scheduling from a classic general jobshop problem....

Ranking Dispatching Rules by Data Envelopment Analysis in a Job Shop Environment

This paper uses Data Envelopment Analysis (DEA) to measure multiple performance criteria for 42 dispatching rules in a job shop environment. We introduce a DEA application in production scheduling. Seven performance measures are considered in the evaluation. Without pre-assigning weights to any performance measure, DEA evaluates the efficiency of each dispatching rule relative to the other rules. After running a large number of experiments, the results show that two extreme subgroups of dispatching rules perform consistently. The shortest processing time related rules form the top group, while the longest processing time related rules form the bottom group. The due date or slack-related rules perform well in tardiness. However, they are ranked low if all seven criteria are considered together. The results provide guidance to scheduling practitioners in choosing priority dispatching rules when there are multiple objectives.

Heuristics for minimizing mean tardiness form parallel machines

The concept of parallel operations has been widely used in manufacturing and data processing. However, not many efficient methods have been proposed to reduce job tardiness. This article proposes an efficient heuristic to minimize the mean tardiness of a set of tasks with known processing times and due dates for single and m parallel machines. For the single-machine case, the proposed heuristic is compared with the well-known Wilkerson and Irwin algorithm; for the m parallel machine case, it is compared with an extension of the Wilkerson-Irwin algorithm. We also introduce a simple dispatching rule, and it is compared with some existing dispatching rules. The comprehensive simulation results show that the proposed heuristic performs better than the Wilkerson-Irwin algorithm at a significantly reduced computational time.

An integrated MRP and JIT framework

A new production planning and scheduling framework is developed to address the multi-stage, production-inventory system problem by integrating Material Requirements Planning (MRP) and Just-in-Time (JIT) production. The objective is to find detailed shop-floor schedules, which specify the quantity of an operation to be processed, at what time, and by which machine, so as to minimize total production cost. The integrated system gets rid of the major problems existing in MRP and JIT. First of all, the proposed integrated system incorporates both the scheduling and capacity planning aspects, simultaneously. Secondly, the integrated system eliminates the need to specify planned lead time. Thirdly, the integrated system, unlike MRP, provides detailed shop-floor schedules. Lastly, the integrated system does not have to operate in the level schedule case as in JIT, so it can handle a very general production environment.

Multi-criteria sequence-dependent job shop scheduling using genetic algorithms

Real world job shops have to contend with jobs due on different days, material ready times that vary, reentrant workflows and sequence-dependent setup times. The problem is even more complex because businesses often judge solution goodness according to multiple competing criteria. Producing an optimal solution would be time consuming to the point of rendering the result meaningless. Commonly used heuristics such as shortest processing time (SPT) and earliest due date (EDD) can be used to calculate a feasible schedule quickly, but usually do not produce schedules that are close to optimal in these job shop environments. We demonstrate that genetic algorithms (GA) can be used to produce solutions in times comparable to common heuristics but closer to optimal. Changing criteria or their relative weights does not affect the running time, nor does it require programming changes. Therefore, a GA can be easily applied and modified for a variety of production optimization criteria in a job shop environment that includes sequence-dependent setup times.

Cell formation in group technology: A combinatorial search approach

This paper addresses the cell formation problem in group technology (GT). We develop two heuristic methods for generating solutions to the problem. These methods are based on two powerful combinatorial search methods‐ simulated annealing and tabu search. The performance of the heuristics is examined using randomly generated, published and industry data. The results indicate that the simulated annealing based heuristic is the preferred technique in the context of the problem addressed in this paper. Further, we also demonstrate that the simulated annealing based heuristic generates near-optimal solutions to the cell formation model formulated in this paper.

Schedule generation in a dynamic job shop

SUMMARY An extension of Giffler and Thompsons algorithm (1960)is developed to create all active schedules in a dynamic job shop. A partitioning scheme is also developed that works extremely well in reducing the number of active schedules created. This reduced enumeration method is much more efficient than the full generation scheme and, based upon the results of test problems, it appears to be nearly as effective.