An improved inference technique for case-based reasoning and its application to failure modes and effects analysis

• Main Authors: Yuh-Wei Jang, 張育維
• Other Authors: Chi-Bin Cheng
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/55050837320286069337

碩士 === 朝陽科技大學 === 工業工程與管理系碩士班 === 89 === Abstract Failure modes and effects analysis (FMEA) is a preventive method for quality improvement. Due to the prevalence of QS9000 system, the needs of implementing FMEA is growing. Implementation of FMEA is labor-intensive and time-consuming, and it requires domain knowledge of experts from many disciplines. The high complexity of modern products make the implementation even more difficult. The purpose of this study is to build an assistant system for the implementation of FMEA so as to alleviate such difficulties. Considering the similarity existing among products of a certain company, case based reasoning (CBR) is chosen to build this assistant system. The steps are first constructing a case base by capturing characteristic attributes of products, and then infer the most similar old product with the new product from the case base. The inference process of CBR is carried out by a similarity measurement. In this study, we propose a new and improved approach from the old similarity measure methods. In this new approach, we correct the over-estimating or under-estimating tendency of some old similarity measure methods with an adjustment technique. Moreover, the importance of different characteristic attributes in the case base is emphasized, and a weighted technique, which mimics human thinking, conducted on characteristic attributes is suggested. To justify the performance of our approach, an example of air nailers is taken from a company in Taiwan. A case base is constructed upon many old products of this company, and a user interface written in Visual Basic is also provided. Comparing with the new product, a most similar old product is picked from the case base by our inference approach. Meanwhile, an expert is also asked to pick out the most similar one among the old products based on his own experience. The result by our approach and by the expert is consistent. However, when the amount of old products is huge, our approach has the advantage of saving time significantly.