A Dynamically Optimized Fluctuation Smoothing Rule for Scheduling Jobs in a Wafer Fabrication Factory

Introduction

Semiconductor manufacturing is a capital-intensive industry. A 200mm wafer fabrication factory can cost over a billion dollars. How to effectively use expensive machinery has been a key element for a semiconductor manufacturer to survive and compete in today’s marketplace. Job scheduling is one of the most important tasks, to achieve this goal. However, this is not an easy task, because a wafer fabrication factory is a very complex manufacturing system, and has the following characteristics: changes in demand, a variety of product types and priorities, un-balanced capacity, job re-entry into the machines, alternative machines with unequal capacity, and shifting bottleneck (Chen et al., 2010b). Nevertheless, it still attracts the attention of many researchers (Gupta Sivakumar, 2006).

In Kim et al. (2001), wafer fabrication factory scheduling problems were divided into three categories: release control, job scheduling in serial processing workstations, and batch scheduling in batch processing workstations. In addition, the scheduling methods can be divided into global approaches and local approaches. Local approaches are usually for specific workstations, while the global approaches can be applied to all workstations in the wafer fabrication factory. On the other hand, the existing methods in this field can be divided into five main categories: dispatching rules, heuristics, data mining-based approaches, agent technologies, and simulation.

This study aims to propose a better dispatching rule. In this regard, some advanced methods have been proposed. For example, Chen (2009a) modified the fluctuation smoothing rule for mean cycle time (FSMCT), and proposed a nonlinear FSMCT (NFSMCT) rule, in which the fluctuation in the estimated remaining cycle time is smoothed, and then its influence is balanced with that of the release time or the mean release rate. Subsequently, the difference among the slack values is enlarged by using the ‘division’ operator instead. After that, Chen (2009c) presented the one-factor tailored nonlinear FSMCT (1f-TNFSMCT) rule and the one-factor tailored nonlinear fluctuation smoothing rule for cycle time variation (1f-TNFSVCT) rule, including an adjustable parameter to customize the rules. Taking into account the two performance measures (average cycle time and cycle time variation) at the same time, Chen and Wang (2009) proposed a bi-criteria nonlinear fluctuation smoothing rule (1f-biNFS). There is also an adjustable factor. To increase the flexibility of customization, Chen et al. (2010a) extended the rules, and proposed a bi-criteria fluctuation smoothing rule with four adjustable factors (4f-biNFS).

The motivations to introduce the new rule are as follows. In the existing fluctuation smoothing rules, the adjustable factors are static. In other words, they will not change over time. Chen (2009b) therefore designed a mechanism to dynamically adjust the values of the factors in Chen and Wang’s bi-criteria nonlinear fluctuation smoothing rule (dynamic 1f-biNFS). However, the adjustment of the factors is based on a pre-defined rule. This treatment is too subjective, but also not takes into account the status of the wafer fabrication factory. In addition, these rules have not been optimized, and there is considerable room for improvement. To tackle these problems and to further improve the performance of job scheduling in a wafer fabrication factory, a dynamically optimized nonlinear fluctuation smoothing rule is proposed in this study, with the following objectives:

• 1.

  It is dynamic. In other words, the content of the new rule can be adjusted dynamically in a convenient way.

• 2.

  It is optimized. In other words, its performance meets some requirements of optimality.

Two performance measures, the average cycle time and cycle time standard deviation, are considered. The proposed methodology has the following innovative features: