Research On Laser Welding Detection System Of Power Battery Module

With the continuous upgrading of intelligent manufacturing technology for power batteries of new energy vehicles,efficient and high-precision laser welding equipment is playing a key role in industry manufacturing process.However,proficient are concerned about how to evaluate the laser welding quality of power batteries.So far,the welding quality inspection of power batteries are mainly completed by manual or manual participation,which will lead to problems such as low inspection efficiency,high labor cost and low reliability.It can not meet the requirements of modern welding production of power batteries for high quality,high efficiency and high safety.In the high-speed continuous power battery welding production line,timely and effective welding defect detection is very important to improve the core manufacturing process and energy storage quality of battery.In order to realize the real-time detection of power battery laser welding,this thesis builds a welding quality detection system based on machine vision,introduces the deep learning target detection algorithm,and proposes an improved welding defect recognition algorithm to realize the intelligent detection of welding defects.The research content of this thesis can be divided into the following points:Firstly,the main welding defect detection methods of power batteries are investigated and analyzed,and the detection scheme is selected.By comparing the advantages and limitations of different detection methods,combined with the requirements of real-time and accuracy of inspection tasks,a welding defect detection method based on machine vision is finally adopted.Secondly,a welding quality detection system based on machine vision is designed and built.Referring to the requirements of welding detection tasks,the hardware architecture of the system is designed and completed.At the same time,the software interface of the welding detection system is designed and implemented to meet the setting requirements of flexible production.Thirdly,the characteristics of various common defects produced by laser welding are analyzed,and the applicable detection algorithm is selected.According to the characteristics of different defects,considering the requirements of detection speed and accuracy,the machine vision deep learning algorithm is screened,and finally the YOLO algorithm is applied to welding defect detection.Then,according to the welding image data collected at the work site,the data set is constructed and the improved algorithm is proposed.Consider of the hardware configuration and data set characteristics,a deep learning algorithm training environment is established to train the algorithm.According to the training data results,an improved idea of detection algorithm is proposed.In order to improve the detection accuracy and speed of the algorithm,this thesis changed the network structure,multi-level feature network and detection head components of YOLOv4 algorithm.Finally,the trained algorithm is compressed to reduce the requirements of the model for equipment performance and computing power resources on the premise that the accuracy loss is small.The completed detection system is deployed to the actual project for testing the accuracy and speed of the system to identify welding defects.The applicability and effectiveness of the system are verified according to the relevant data.