关键词: 商业智能, 偏最小二乘回归, 主成分分析, 汽车冲压生产

Abstract: To reach the goal of advanced quality control of automobile stamping production,aiming at the problem that production quality relied heavily on manual experience and production data relationship was hard to establish,the Partial Least Square regression (PLS) method was used to constructed a mapping relationship,and a stamping Business Intelligence (BI) analysis model was established to further explore the “process-quality” relationship and determine the key factors affecting the stamping production quality.The multiple groups of independent variables were classified and dimensionality reduced with Principal Component Analysis(PCA) by collecting stamping process data and parts quality data .Four groups of principal components were extracted as comprehensive indicators to interpret the source data,and nine key factors were determined based on the standard regression coefficients in the regression model.Upon examination,model determination coefficient was equal to 0.8,which could prove a practical and reliable prediction result.In practical application,the model could effectively establish one-to-many or even many-to-many mapping relationship of different parts,which improved the consistency and stability of parts' quality and achieved advanced control of production quality.

Key words: business intelligence, partial least square, principal component analysis, automobile stamping production