信息不完备情况下多因素工序质量诊断方法

Diagnosis method of multi-cause process quality under incomplete information

摘要:

为解决信息不完备情况下的多因素工序质量诊断问题，在工艺机理分析的基础上，提出基于贝叶斯网络模型构建与推理的问题溯源方法. 在贝叶斯网络结构学习过程中，利用基于评分/搜索的思想对基于工艺的预先假设结构，通过互信息参量排序降低学习复杂度. 针对生产过程中随机因素对诊断准确性的影响问题，结合Leaky Noisy-OR模型引入随机参量节点，对数据需求和推理进行降解优化. 以沟道磨削表面形貌质量问题的诊断为例，给出模型构建与推理程序，并验证了所构建模型及优化方法的可行性和有效性.

Abstract:

Aiming at the problem of multi-cause process quality diagnosis under the circumstance of information losing, a method based on construction and inference of Bayesian network model is proposed. In the learning process of Bayesian network structure, the thought of score/search is adopted for the assumption structure so as to reduce the learning complexity through the mutual information parameters sorting. In view of the influence of random factors on the diagnostic accuracy, the Leaky Noisy-OR model is adopted, which simultaneously degrades the requirement quantities of data and reasoning. In the end, a problem diagnosis for channel grinding is taken as an example to verify the feasibility and effectiveness of the proposed model and optimization method.