In order to improve the quality inspection efficiency of the automobile seat belt production site,an experimental platform is built according to the"QC/T987-2014 Automotive Seat Belt Retractor Performance Requirements and Test Methods"to collect the sound signals during the working process of the retractor qualified and defective products.Before the convolutional attention module(CBAM)is embedded into the residual network(ResNet-18)residual block,a CBAM-ResNet-18"Before Blocks"model is designed to classify the collected retractor sound signals.Compared with the ResNet-18 model without the attention mechanism,the CBAM-ResNet-18"Within Blocks"model with the attention mechanism after the residual block,the traditional classification model support vector machine and random forest,the designed model performs well in the aspects of confusion matrix,accuracy,precision,recall rate and F1 value in the task of retractor sound signal classification,which is very effective for the qualitaty detection of automobile safety belt retractors based on the sound signal.
关键词
汽车安全带/声音信号/卷收器/质量检测/CBAM-ResNet
Key words
car seat belts/sound signal/retractor/quality inspection/CBAM-ResNet
维普
万方数据