Based on the fact that the explicit relation expression between product multi-quality characteristics and the dimensional tolerance of component is not educed directly with geometry principle and the product-level quality loss cost is difficult to establish, a quality loss cost model with neural network method and analytic hierarchy process (AHP) was processed. In order to decrease the impact of subjective randomness, the weights are determined with neural network method to product historical data process. Taking rework costs and scrap costs into account during the course of manufacturing mechanical parts, the piecewise quality loss function was used to improve Taguchi quality loss function, which is very close to engineering practice. The corresponding quality loss costs were deduced with the different distribution of part dimension. The results show that the quality loss cost model provides an effective method for the implicit relation model.%针对产品多质量特征与零部件的尺寸之间不能直接利用几何原理得出显式关系表达式,很难建立产品级的质量损失模型的现状,提出一种基于神经网络法和层次分析法的质量损失模型.为了尽量减少主观随意性对权重确定的影响,采用神经网络法处理产品历史数据确定权重.考虑零件加工过程中的返修成本和报废成本,采用分段质量损失函数对田口质量损失函数进行改进,使其更加贴近工程实际,并根据零件尺寸服从不同的分布,导出相应的零件质量损失成本.研究结果表明:该质量损失模型为隐式关系建模提供了有效解决方法.
展开▼
