平均値置換寄与による異常原因因子の特定

抄録

In a Lithium Ion Battery process, it is desired to find a causal factor that reduces the stability of the process. In this paper, we propose the mean replacement contribution. Each representative value is replaced with an average value one by one, and the difference between those Mahalanobis distances is used as a contribution. The proposed method was compared with T² contribution, Q contribution, reconstruction contribution (RBC), and random forest importance using the sensor signal of the coating device. As a result, the mean replacement contribution was able to identify the causal factor most accurately based on the knowledge of the engineer. Normally, an engineer needs a lot of man-hours to perform analysis, so that the proposed method can reduce the time.