基于神经网络的质量控制图异常诊断

针对质量控制图异常模式到异常原因的映射存在模糊不确定性,采用4层前馈BP神经网络,实现从控制图异常模式到异常原因的推理诊断,构建质量控制图异常模式异因推理诊断系统。首先,将控制图异常模式进行归类,构建控制图异常模式集;其次,针对每种异常模式,确定与之对应的异常原因集;最后,针对每一种异常模式,利用其模式数据作为输入,与之对应的异常原因的异常度作为输出,构建与其对应的神经网络,实现质量控制图异常模式到异常原因的推理诊断,并在此基础上将神经网络输出异常度按大小进行排序,缩小异常原因查找范围,提高查找效率。

Diagnostics on Abnormity in Quality Control Chart Based on Neural Network

Aiming at the fuzzy uncertainty in the mapping of anomaly patterns and anomaly causes in the quality control chart,a 4-layer feedforward BP neural network is introduced to achieve ratiocination and diagnosis ranging from anomaly patterns to anomaly causes in the control chart with an exclusive ratiocination diagnostic system for anomaly patterns in the quality control chart is built up.First,the anomaly patterns in the control charts are classified to establish a set of anomaly patterns for the control charts;Secondly,for each anomaly pattern,it is required to define a corresponding set of anomaly causes;Finally,for each type of anomaly pattern,the pattern data is used as the input while abnormal level of corresponding anomaly causes is used as an output to establish corresponding neural network and achieve ratiocination diagnosis of anomaly patterns and anomaly causes in the quality control charts,The anomaly levels output by the neural network will be sorted by size on this base to narrow the range of anomaly causes for searching and improve the searching efficiency.