Research on Variation in Process Quality Based on Clustering Method

In order to identify the variation sources of quality in the process of manufacturing, this paper proposed a clustering method to solve the problem. For the validity of clustering, Pearson´s correlation coefficient and was introduced as a similarity measure. A new measure criterion based on the significance of differences between the variance and mean was proposed. In this paper, quality affecting factors in each process were disposed wholly as a combination. According to the measures, clustering was performed between and within each group of quality data formed by each combination. By comparing the differences of intraclusters and interclusters, factors causing quality differences and quality variations can be obtained, and the accuracy can be confirmed through the prediction of quality according to clustering results.