基于尺寸关联和偏最小二乘回归的多工序质量分析与预测

Multistage manufacturing quality analysis and forecasting on dimension association and partial least square regression

针对数字制造下多工序机械加工过程的质量控制问题,根据零件加工特征和基准,构建了基于尺寸链的工序关联矩阵,应用特征描述法存储零件工艺信息。在此基础上提出一种依据加工特征和基准的工序关联检索算法,用于提取关联工序。建立基于偏最小二乘回归方法的多工序质量分析与预测模型,通过偏最小二乘回归方法,提取对工序质量影响强的成分,以解决工艺过程中存在的自变量之间的多重相关问题,提高了多工序质量分析与预测的精度,该模型已应用于轴套零件加工过程中多工序质量的分析与预测。

To meet with the needs of multistage quality control in digital manufacturing, a process association matrix based on the dimension chain was put forward on the basis of analyzing machining features and datum. Part process information was stored by feature based description method. Based on this, a process association searching algo-rithms was designed to extract association between processes. An analysis and forecasting model for multistage man- ufacturing quality was constructed based on partial least square regression. Some principal components which were major influencing factors on multistage manufacturing quality were extracted by partial least square regression. Then, the multi-collinearity among processes could be solved and the forecasting precision of multistage manufacturing quality model could be improved greatly. Finally, a case study on bush machining was given to demonstrate the above-mentioned ideas.