应用卷积神经网络的锂电池极片涂布缺陷分类

鲁永帅; 唐英杰; 马鑫然; 刘爽

引用本文:

【打印本页】【下载PDF全文】【查看/发表评论】【EndNote】【RefMan】【BibTex】

←前一篇|后一篇→

过刊浏览高级检索

本文已被：浏览 1817次下载 979次	码上扫一扫！
分享到：微信更多字体:加大+\|默认\|缩小-
应用卷积神经网络的锂电池极片涂布缺陷分类
鲁永帅,唐英杰,马鑫然,刘爽
北京印刷学院信息工程学院，北京 102600

摘要:

目的针对锂电池极片涂布缺陷种类多，传统方法分类检测精度不高，以及人工依赖性强等问题，提出一种基于卷积神经网络的锂电池极片涂布缺陷自动分类算法。方法首先对网络结构以及模型参数进行优化，接着在网络中加入跳跃连接结构，将空洞卷积提取到的多尺度特征与高层特征进行融合以获取更多缺陷特征，并采用LeakyReLU(Leaky Rectified Linear Unit)激活函数保留图像中的负值特征信息，最后通过构建的数据集训练模型，实现锂电池极片涂布缺陷的准确分类。结果实验结果表明，当前方法识别准确率能够达到99.34%，平均检测时间为51 ms。结论改进后的方法能够准确分类出锂电池极片18种涂布缺陷，满足工业生产中实时分类检测的要求。

关键词: 卷积神经网络锂电池涂布缺陷检测图像分类

DOI：10.19554/j.cnki.1001-3563.2022.09.031

分类号:TP391

基金项目:北京市自然科学基金项目-北京市教委科技计划重点项目(KZ202010015021)

Defect Classification of Lithium Battery Pole Piece Coating Using Convolutional Neural Network

LU Yong-shuai, TANG Ying-jie, MA Xin-ran, LIU Shuang

(School of Information Engineering, Beijing Institute of Graphic Communication, Beijing 102600, China)

Abstract:

In order to solve the problems of many kinds of coating defects of the lithium battery pole piece, the low accuracy of the traditional classification and detection methods, and strong artificial dependence, an automatic classification algorithm for coating defects of lithium battery pole piece based on the convolutional neural network was proposed. First, the network structure and model parameters are optimized, then the jump connection structure is added to the network, the multi-scale features extracted by dilated convolution are fused with the high-level features to obtain more defect features, and the LeakyReLU (Leaky Rectified Linear Unit, LeakyReLU) activation function is adopted to retain the negative feature information in the image. Finally, through the constructed data set training model, the accurate classification of the coating defects of the lithium battery pole piece is realized. Experimental results show that the recognition accuracy of the current method can reach 99.34%, and the average detection time is 51 ms. The improved method can accurately classify 18 kinds of coating defects of lithium battery pole pieces, which can meet the requirements of real-time classification and detection in industrial production.

Key words: convolutional neural network lithium battery coating defect detection image classification