Abstract: To avoid the waste of remanufactured blank value and economic loss caused by process defects,it is necessary to predict its process reliability and identify the process defect elements after the remanufacturing process of a used product is determined.However,remanufactured blanks are highly variable and have high residual value,which is difficult,poorly generalized and costly to establish a mechanism model affecting process reliability elements.To this end,a data-driven process reliability prediction method for processing elements was proposed.A Bayesian neural network prediction model reflecting the mapping relationship between each processing element and reliability was constructed by using the deviation value between the quality index obtained by processing and the quality requirement as the quantitative index for process reliability evaluation.The proposed prediction method was validated using a remanufactured CNC machine tool as an example.The results showed that the proposed method could identify defective parts and their machining elements and guide the improvement of process quality of remanufactured products.

Key words: process reliability prediction, remanufactured products, Bayesian neural network, processing elements, CNC machine tool

摘要： 为避免由于工艺缺陷造成的再制造毛坯价值浪费和经济损失,废旧产品在其再制造工艺制定后,需要首先对工艺可靠性进行预测,识别工艺缺陷要素。然而,由于再制造毛坯差异性大、剩余价值高,建立影响工艺可靠性要素的机理模型难度大、通用性差、且成本高。为此,提出一种加工要素数据驱动的工艺可靠性预测方法。利用加工获得的质量指标与质量要求之间的偏差值,作为工艺可靠性的定量指标,构建了反映各加工要素与可靠性映射关系的贝叶斯神经网络预测模型。以某再制造数控机床为例,对所提出的预测方法进行了验证。结果表明,该方法能够识别缺陷零件及其加工要素,指导再制造产品工艺过程的改进。

关键词: 工艺可靠性预测, 再制造产品, 贝叶斯神经网络, 加工要素, 数据模型, 数控机床