Detection of faulty products using data mining

The manufacturing process is complex due to the large number of processes, diverse equipment set and nonlinear process flows. Manufacturers constantly face yield and quality problems as they constantly redesign their processes for the rapid introduction of new products and adoption of new process technologies. Solving product yield and quality problems in a manufacturing process is becoming increasingly difficult. There are various types of failures and their causes have complex multi-factor interrelationships. High innovation speed forced today's manufacturers to find failure causes quickly by examining the historical manufacturing data. Data mining offers tools for quick discovery of relationships, patterns, and knowledge in large databases. This has been applied to many fields such as biological technology, financial analysis, medical information, etc. Application of data mining to manufacturing is relatively limited mainly because of complexity of manufacturing data. Growing self-organizing map (GSOM) algorithm has been proven to be an efficient algorithm to analyze unsupervised DNA data. However, it produced unsatisfactory clustering when used on some manufacturing data. Moreover, there was no benchmark to monitor improvement in clustering. In this study a method has been proposed to evaluate quality of the clusters produced by GSOM and to remove insignificant variables from the dataset. With the proposed modifications, significant improvement in unsupervised clustering was achieved with complex manufacturing data. Results show that the proposed method is able to effectively differentiate good and faulty products.