摘要

Although various multivariate process monitoring techniques have been developed, they do not diagnose the process for finding the root causes of irregularities during production. There have been recent studies on a new method that involves process-oriented basis representation, which links the process variation to its causes, and thus helps in monitoring and diagnosing a process. However, all the studies done so far focused on its application. In this paper, a method is proposed to build the process-oriented basis for a process irrespective of the number of variables characterizing it. Along with various other statistical techniques, factor analysis and cluster analysis, with customized distance function, are used in developing the method. The built in process-oriented basis is further used for multivariate statistical process control and process capability analysis. Multivariate solder-paste problem from electronics industry is used for illustration.

  • 出版日期2013-12