Abstract: The selection of process parameters is a difficult problem in current power battery welding industry.To improve the efficiency of power battery welding and to solve the difficulty in selecting process parameters for meeting multiple objectives,a solution combining kernel ridge regression and Multi-Objective Particle Swarm Optimization (MOPSO) was adopted.In this scheme,the welding lower limit corresponding to process parameters was set,and the kernel ridge regression model was used for simulation based on Gauss kernel function.A set of process parameters was represented as a particle of MOPSO,and by means of a regression model,an optimal solution set for the specified welding objective was effectively obtained through three operational steps consisting of the evolution and variation of the population,the selection and optimization of the guide and the maintenance of solution set.In addition,K-nearest neighbor algorithm was referenced for designing an evaluation criterion to measure the reliability of each solution and further screen better solutions,which ensured that the selected process parameters own higher fault tolerance.The proposed method had solved the difficult problem in selecting process parameters faced by current power battery welding industry,which would enhance the efficiency and effectiveness of power battery welding.

Key words: power battery, laser welding, process parameters, kernel ridge regression, multi-objective particle swarm optimization

摘要： 工艺参数选择是动力电池焊接行业面临的困难,为提升动力电池焊接效率并满足多项目标,采用核岭回归与多目标粒子群优化算法相结合的方法辅助优化工艺参数选择。构造了工艺参数对应的焊接下限,继而利用基于高斯核函数的核岭回归模型进行拟合;多目标粒子群的每个粒子代表一组工艺参数,通过群体进化与变异、引导者选取与优化、解集维护3种操作,并结合回归模型,有效获取了指定焊接目标下的最优解集。该方法还借鉴K近邻算法思想设计评价标准,以度量每个解的可靠性,进一步筛选更优质的解,保证所选工艺参数有更高的容错性。所提方法解决了电池焊接行业目前面临的问题,具有极其重要的应用价值。

关键词: 动力电池, 激光焊接, 工艺参数, 核岭回归, 多目标粒子群优化