مطالعه استفاده از مدل ANP برای ارزیابی قوانین توزیع جهت ساخت ویفر

Wafer fabrication is a capital-intensive high-tech sector with highly complex manufacturing processes. Therefore, most of the fabs resort to production dispatching as a means to enhancing production efficiency. The commonly seen methods of production dispatching at the present time are devised to meet single performance indicators. Few methods take into account multiple, or even conflicting performance indicators. Therefore, different production control managers adopt different criteria. Also, as performance indicators change with the variances of production lines and actual demands, it is necessary to clarify the rules of varying dispatching methods and their impacts on all the production performance indicators so that it is possible to explore an architecture for multiple-rule or multiple-target production dispatching in order to meet dynamic performance targets. This paper uses analytical network process (ANP) method to construct a dispatching model based on the characteristics of all the production facilities on-site (such as the utilization of bottleneck machines), in order to explore the relationship among various performance indicators and correlation between performance indicators and the dispatching rules. The aim of this paper is to analyze the production dispatching issues of wafer fabs in an effective and systematic approach, so as to provide an on-site dispatching analysis model that takes into consideration production characteristics and indicator adjustments. This paper finds that the most optimal dispatching method for ANP dispatch model is EDD dispatching method, followed by LS dispatching method. FIFO dispatching method yields the worst performance. The ANP dispatching assess model proposed in this paper can surely serves as an analytical architecture for decision makes to evaluate production dispatching models of multiple production indicators in the future.

The wafer fabrication industry has a highly demanding, complex and yet highly repetitive manufacturing process. There are characteristics such as re-entrance and long production cycles, making it very difficult to arrange production and dispatching schedules. Missing the deadlines or suffering from an overly high inventory of work-in-process is not unusual. For the wafer fabrication industry, the advantages lie in flexible productions and reliable high yields. In order to maintain these two core competencies, it is necessary to constantly improve the manufacturing processes in order to upgrade production performances. There are many disruptions in the production systems, such as equipment breakdowns, urgent orders taking priority, shortage of materials, rework of wafers, and improper operations of operators. Any of these factors may directly or indirectly affect production performances. Therefore, in the face of market competition, all the wafer fabrication plants spare no efforts in the considerations and evaluations of various kinds of control policies, in order to gain insights into the variances caused by production factors (characteristics) to production systems. The purpose is to construct a reliable and stable production environment where the volatility of production performances is immune to impacts due to various scenarios. The impacts of production characteristics aside, the changes of all kinds of production performance indicators are often highly complex. Koriyama, Yazaki, Kato, and Kisakibaru (1995) believes that total throughputs, yields, production cycles, quantities of work-in-process are all issues that production control managers are concerned about in the process of wafer fabrications. However, there may exist correlation among these performance indicators. All the dispatching rules applied in wafer fabrication have different assumptions regarding the environments and performance indicators chosen. There is no single dispatching rule that is superior in every aspect (Chang, Sueyoshi, & Sullivan, 1996). Therefore, managers usually have to measure all the pros and cons and strike a balance in their planning of the systems. Besides, different production control managers adopt different dispatching rules. Therefore, it is an issue that requires across-the-board considerations to incorporate different production performance indicators and perspectives from production control managers into the choice of dispatching rule. It is a critical issue for the wafer fabrication industry amid fierce market competition. In order to solve this problem, this research paper centers its focus on the interdependencies of performance indicators so as to screen out the appropriate combination of production performance indicators and to identify potential factors that may affect production performances. This paper applies multi-principle planning concept (Wein, 1988) and analytic network process (ANP) to examine all the characteristics of the production sites (such as bottleneck equipment utilization rates), in order to explore the relationship among performance indicators and the correlation between performance indicators and dispatching rules. This paper aims to provide a dynamic valuation model that considers a set of production characteristics and performance indicators. This paper presents the whole picture of all the productions issues and the synergies of various dispatching rules via expert questionnaires. With an evaluation model that factors into these performance indicators and production factors that impact performances, production controllers will be able to effectively adjust on-site dispatching methods.

The production indicators of wafer fabs vary in accordance with production demands and changes of production lines. Oftentimes, it is not possible to measure production performances with single indicators. It is necessary to resort multiple indicators to properly express the outcomes of the systems. However, due to the conflicting or interdependent relationships among all the performance indicators, it is usually necessary to choose among them, in order to meet the planned goals of the systems. Due to the high complexity of wafer fabs and the high costs of equipment, dispatching methods are usually an important means for manufacturing sites to enhance production efficiency. However, the use of various dispatching rules great varying impacts to different production performances. Meanwhile, dispatching methods are often confined to the limitations of environments and the performance indicators chosen. As a result, it is possible to have a good performance for one indicator, but poor performance for other indicators. Therefore, the selection of appropriate and efficient dispatching methods is able to enhance performance of the systems. The aim of this paper is not to provide a fixated optimal dispatching method for users to follow; rather, it aims o provide decision makes a procedure to select the appropriate dispatching methods in a swift and convenient manner after the evaluations of multiple production performances. Therefore, this study proposes a dispatching model based on analytical network process method, by considering different production performances and commonly used dispatching methods. This paper constructs an architect for dispatching models suitable for wafer fabs in order to meet production goals of different production performances. Therefore, this paper examines production performance indicators and factors into production characteristic factors that affect production performances, such as product mix, machine down, in order to make the evaluation model as realistic as possible by imitating the actual operations of wafer fabs. This study is sure that this dispatching model is able to effectively solve the dispatching issues and performance indicators for wafer fabs. This model is more feasible than these dispatching methods designed to meet single goals. This paper comes to the following conclusions: 1. This paper applies analytical network process to examine the decision making process, and gathers the judgments of 11 veteran experts to compare all the performance indicators, production characteristic factors that affect product performance and dispatching methods. The purpose is to derive desirability indexes of each option (relative weightings) in order to determine the optimal dispatching methods. The application of ANP method is able to improve the limitations of the traditional approach of AHP method which assumes elements must be mutually independent of each other. The ANP approach allows external interdependency or feedback relationship among elements. This avoids the shortcomings of bias estimates resulting from over simplifications. 2. According to the evaluation results of the ANP dispatching rule evaluation model for wafer fabs, the most optimal dispatching method is EDD, followed by CR and then LS dispatching methods. FIFO dispatching method yields the worst results.