基于数据挖掘的汽车生产线螺栓打紧异常识别

Recognition of Bolt Tightening Anomaly in Automobile Production Line Based on Data Mining

针对汽车装配螺栓打紧过程,提出了一种可以准确识别打紧异常的模型。首先对原始数据进行处理,包括去除噪声和集中主体数据,并根据过程扭矩提取特征。对异常样本过少问题,提出了适合的SMOTE改进算法,通过对异常样本过采样平衡了数据集,挖掘并利用更多的样本信息。最后基于加权随机森林算法构建分类器,完成整个识别模型的建立。通过真实数据验证,模型可准确识别出打紧异常螺栓。

Aiming at the tightening process of automobile assembly bolts,a model that can accurately identify tightening abnormalities is proposed.First,the original data is processed,including noise removal and main body data collection,and features are extracted according to the process torque.For the problem of too few abnormal samples,a suitable improved SMOTE algorithm is proposed,which balances the data set by over-sampling abnormal samples,and mines and utilizes more sample information.Finally,a classifier is constructed based on the weighted random forest algorithm to complete the establishment of the entire recognition model.Through real data verification,the model can accurately identify abnormally tightened bolts.