رویکرد فرایند تحلیل شبکه ای (ANP) برای برنامه ریزی ترکیب محصول در سازنده نیمه هادی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|6012||2005||22 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : International Journal of Production Economics, Volume 96, Issue 1, 18 April 2005, Pages 15–36
This paper proposes an application of the analytic network process (ANP) for the selection of product mix for efficient manufacturing in a semiconductor fabricator. In order to evaluate different product mixes, a hierarchical network model based on various factors and the interactions of factors is presented. By incorporating experts’ opinion, a priority index can be calculated for each product mix studied, and a performance ranking of product mixes can be generated. The results provide guidance to a fab regarding strategies for accepting orders to maximize the manufacturing efficiency in considering the aspects of product, equipment efficiency and finance. The model can be easily understood and followed by administrators to determine the most efficient product mix for a fab.
Global competitiveness has become the biggest concern of semiconductor industry. How to increase the overall profit and the return on investment of a company is therefore very essential. The purpose of this paper is aimed at the strategic planning level and attempts to present an effective approach for product mix evaluation that allows for the consideration of various factors and important interactions among factors. The product mix selected can best represent a near-optimal utilization to the factory resources and a highest possible profit attained, and it can be a reference for production planning and order acceptance. Wafer fabs involve the most complex manufacturing system in the manufacturing world. Its manufacturing process is of high complexity, with several hundreds of processing steps on a single wafer and a flow time of usually more than 1 month. Different product mix only complicates the already-complex system. Depending on the types of products, the process plan of a product can range from very identical to being extremely distinctive, and the requirement of setups may also be different. The greater the difference, the more diverse the loading demand and batch difficulty on the factory. The actual throughput and cycle time under a given product mix thus depend on how badly the fab is bottlenecked, whether the bottleneck is shifted, and how many machine setups are needed because of product type conversions. Many metrics can be applied for evaluating factory productivity. Leachman and Hodges (1996) evaluated semiconductor wafer fabrication plants around the world to quantify manufacturing performance and to establish comparative benchmarks in manufacturing technology, factory operations, organization, and management. The major technical metrics they used to measure manufacturing performance are cycle time per wafer layer, line yield, die yield, stepper productivity, direct labor productivity, total labor productivity and on-time delivery. Although their study provided a comprehensive performance evaluation and identified those practices that underlie top performance, there was no attempt to correlate the interactions of the metrics. There are at least three aspects that are necessary for measuring the overall effectiveness of a factory: production, utilization of assets, and costs (SEMI, 2002). SEMI provided a guideline for definition and calculation of overall factory efficiency (OFE) and other associated factory-level productivity metrics. The document focused on evaluating production; however, utilization of assets and costs were outside of its scope. Organizing available data and providing a singular metric to compare performances is not an easy task. Chung et al. (2002), however, adopted a good nonlinear programming method called Data Envelopment Analysis (DEA) to deal with multiple inputs and outputs. Without pre-assigning weights, DEA can be used to measure multiple inputs and outputs for product mixes in a semiconductor fabricator, and an efficiency score for producing each product mix relative to other mixes can be obtained. The major advantage of DEA, without pre-assigning weights to any performance measure, can also be its drawback. Managers often have their own opinion on what performance measures are more important than others. In that case, analytic hierarchy process (AHP) and/or analytic network process (ANP) can be a good alternative in evaluating production performance under different product mixes. While AHP has been a popular research and application tool for multi-attribute decision-making, the ANP technique so far has had only a few applications in literature. A matrix manipulation approach, developed by Saaty and Takizawa (1986), is applied to solve a network, which is very similar to a hierarchy but has dependence among criteria and dependence among alternatives with respect to each criterion. Lee and Kim (2000) used the above-cited ANP approach within a zero-one goal-programming (ZOGP) model to suggest an information system project selection methodology, which can reflect interdependencies among evaluation criteria and candidate projects. Karsak et al. (2002) dealt with product planning in quality function deployment by also using a combined ANP and goal programming approach. Chung et al. (2004) adopted Saaty's matrix manipulation concept and suggested a simplified ANP approach to analyze multiple process inputs and outputs, and with experts’ opinion on their priority of importance, to obtain optimal product mixes for semiconductor production. Sarkis (2002) presented a systemic ANP model to evaluate environmental practices and programs in analyzing various projects, technological or business decision alternatives. Momoh and Zhu (1998) proposed an application of AHP and ANP to enhance the selection of generating power units for appropriate price allocation in a competitive power industry. Meade and Sarkis (1999) suggested a decision methodology that applied ANP to evaluate alternatives (e.g. projects) and to help organizations become more agile, with a specific objective of improving the manufacturing business processes. Meade and Presley (2002) used ANP to support the selection of projects in a research and development (R&D) environment. Suwignjo et al. (2000) and Bititci et al. (2001) constructed an innovative framework and supporting system to let organizations incorporate and map performance measures in a hierarchical way. The quantitative model for performance measurement system (QMPMS) relies on AHP to quantify both tangible and intangible factors for performance. Bititci et al. further applied the QMPMS for manufacturing strategy evaluation and management in a dynamic environment. Sarkis (2002) revisited the above works and applied ANP to the QMPMS process. Through the utilization of the supermatrix approach, the combined effects of factors on organizational performance measures can be quantified, and the dynamic nature of strategic decisions can be evaluated. Saaty suggested the use of AHP to solve the problem of independence on criteria and alternatives and the use of ANP to solve the problem of dependence among criteria and/or alternatives (Saaty, 1996). The metrics for measuring manufacturing performance, such as production throughput, cycle time, equipment utilization and WIP are highly interrelated. While some metrics are positively dependent, others may be negatively dependent. As a result, ANP is adopted in this research for determining the production performance of various product mixes. Semiconductor companies as well as other industries need good problem solving methods that can be used in real practice. While ANP provides a good quantitative and qualitative tool to assist administrators, they may feel threatened by its complexity if they have no experience or a good understanding of the method. This paper can provide a good prototype for the users to conceptualize the process and to follow the procedure in the determination of a suitable product mix in manufacturing. This paper is organized as follows. Section 2 discusses the process parameters selected for evaluation, and a multi-attribute selection framework represented as an ANP model is presented. Section 3 describes the system environment for simulation. Section 4 applies ANP to the evaluation of the efficiency under different product mixes. Some conclusion remarks are made in the last section, while the appendix briefly reviews the decision making tool ANP.
نتیجه گیری انگلیسی
The ANP is presented in this paper as a valuable method to support the selection of product mix that is efficient for a wafer fab to manufacture. The relative prices of products affect the total revenue trend under different product mixes, and the ANP can be applied to determine the efficiency of product mix when product prices change. Because profitability is the key to success for an enterprise, relative prices of products and the total revenues generated from different product mixes can often determine the desirable product mix. This is shown by the trend of the ranking of alternatives under most cases studied in this paper. However, there are times when the ranking of alternatives is not exactly the same as the ranking of total revenue of alternatives. This means that factors other than total revenue do have impact on the efficiency evaluation of product mixes. In this paper, the approach adopts ANP to deal with interrelated factors and provides users with procedures to be followed for the determination of a suitable product mix for wafer fabrication. When product mix is predictable or when all products use each facility equally, an overall production forecast is sufficient to determine equipment requirements. However, when various products follow distinct routes and have different loadings on machines, any deviation from the product mix target can make the workstation utilizations time-varying over the planning horizon. The time-varying utilizations will in turn cause time-varying cycle times. How product mix should be set in a longer term will be our future research direction. In addition, due to the intensive competition in semiconductor industry, a variety of products are usually required to be manufactured in order to satisfy customer demand. Multiple priority orders, such as hot lots, rush lots and normal lots, are often encountered in fabs. As a result, the manufacturing environment can be even more complex. Green production is especially important for semiconductor manufacturing and will determine the sustainability of a company in the long term. Semiconductor fabs use a lot of hazardous chemicals in their manufacturing process. Different product mixes require different chemicals and gases in the process, and the residual process gases and reactive by-products generally contain toxic and corrosive gases that may be harmful to the environment. As a result, the determination of a good product mix may not only include product, equipment efficiency and finance factors, environmental issues may also need to be taken into consideration. This can also be our future focus of research.