龚浩天,付泽军,徐竹田,等.基于图像机器学习的极片辊压褶皱形貌快速判断方法[J].精密成形工程,2025,17(5):177-187.
GONG Haotian,FU Zejun,XU Zhutian,et al.Method for Fast Morphology Judgment of Electrodes Calendering Defects Based on Image Machine Learning[J].Journal of Netshape Forming Engineering,2025,17(5):177-187. |
| 基于图像机器学习的极片辊压褶皱形貌快速判断方法 |
| Method for Fast Morphology Judgment of Electrodes Calendering Defects Based on Image Machine Learning |
| 投稿时间：2024-10-15 |
| DOI：10.3969/j.issn.1674-6457.2025.05.020 |
| 中文关键词: 锂电池极片 辊压成形 褶皱形貌，缺陷检测 卷积神经网络 |
| 英文关键词: lithium battery electrode calendering wrinkle morphology defect detection convolutional neural network (CNN) |
| 基金项目:国家自然科学基金-区域创新发展联合基金(U23A20629) |
|
| 摘要点击次数: 69 |
| 全文下载次数: 19 |
| 中文摘要: |
| 目的 针对连续辊压制造过程中锂电池极易产生褶皱集合缺陷的问题，研究快速判断测量方法。方法 提出了一种基于图像机器学习的极片褶皱缺陷的快速识别方法。首先，使用激光测量仪获取极片几何缺陷形貌高度形貌，并将其作为标定信息。其次，利用高速工业相机得到不同光源下的极片几何缺陷表面图像，以不同光源图像数据作为输入、以缺陷实际高度形貌作为输出，建立卷积神经网络的预测模型。结果 使用实测形貌高度数据训练后，预测高度点云与实际成形结构形貌的趋势一致，预测高度数据的高度平均偏差在9%以内，计算得到的起皱缺陷高宽比与实际结果误差在6%以内。结论 利用该方法可在激光测量标定后，仅通过高速相机拍摄图像直接在线获得极片褶皱形貌及严重程度，解决了现有产线上无法准确高效得到辊压褶皱缺陷形貌的难题，有望提升锂电池极片连续辊压效率并为质量控制提供指导。 |
| 英文摘要: |
| The work aims to explore a fast judgment and measurement method to deal with the problem that continuous calendering process of lithium battery electrode sheets is prone to generating geometric defects such as wrinkles. A fast recognition method for electrodes calendering defects was proposed based on image machine learning. First, a laser measurement device was used to obtain the height morphology of the electrode’s geometric defects as calibration data, and high-speed industrial camera were employed to capture surface images of the geometric defects under different lighting conditions. With image data from different light sources as input and the actual height of the defects as output, a convolutional neural network prediction model was established. After training with measured surface height data, the predicted height data was consistent with the actual heights. The average deviation in the predicted height data was within 9%, and the error in calculating the wrinkle defect aspect ratio was within 6% compared with the actual results. With this method, after laser measurement calibration, the electrodes wrinkle morphology and severity can be obtained online directly by high-speed camera imaging. This addresses the issue of the current production line being unable to accurately and efficiently acquire the morphology of roll-pressed wrinkles, and it holds promise for improving the efficiency of continuous calendering of lithium battery electrodes and providing guidance for quality control. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
渝公网安备 50010702501719号