基于反向传播神经网络的白车身在线监测

A Study on Body in White Online Detection Based on Back Propagation Neutral Network

针对焊装车间的装配工艺过程的复杂性以及白车身在线监测异常模式的多样性,为了更全面更有效地监控白车身的质量状态,根据实际测量数据提出了方差异常模式的仿真函数构造方法,并提取方差异常模式的统计特征和形状特征,采用MATLAB数学软件建立三层反向传播BP神经网络对生产过程中的7种常见异常模式进行识别,并利用生产线实际测量数据进行验证。结果表明,建立的异常模式监测模型可行,并具有较高的准确性,能够有效代替人工识别。

According to the complexity of the welding assembly process and the diversity anomaly pattern of Body in White Online Detection,In order to monitor quality status of BIW more comprehensively and effectively,A simulation function of variance anomaly pattern is constructed according to the online data,extract statistical features and shape features of anomaly patterns and a three-layer Back-Propagation(BP)Neural Network is built with MATLAB to identify seven anomaly pattern in the production process and validate with actual online measurement data.The results show that the monitoring model is feasible and accuracy,which can be applied to replace the manual identification.