基于DK-SVDD的轮毂电机轴承状态识别方法

Research on Condition Recognition Method Based on DK-SVDD for In-wheel Motor Bearing

为了进一步提高电动汽车轮毂电机轴承状态识别技术的高效可靠性,提出一种基于双核支持向量数据描述(double kernel based support vector data description,简称DK-SVDD)的轮毂电机轴承状态识别方法。首先,针对轮毂电机轴承样本数据结构混杂致使SVDD识别率较低问题,通过一定的比例权重将径向基(radial basis function,简称RBF)核函数和高斯差分(difference of Gaussians,简称DOG)核函数结合构建DK核函数;其次,根据最优二叉树原理逐层设计状态识别分类器,并搭建DK-SVDD轮毂电机轴承状态识别模型,同时使用粒子群优化算法对模型参数寻优以提高DK-SVDD的学习能力和泛化能力;最后,基于轮毂电机轴承台架试验数据,验证所提方法的有效性和优越性。结果表明:针对轮毂电机轴承目标状态识别,DK-SVDD方法平均训练时间为0.0655 s,平均状态识别率为97.06%;与采用RBF或DOG核函数相比,DK-SVDD方法在多种工况下可以有效提高状态识别率并降低训练时间。

In order to further improve the efficiency and reliability of electric vehicle in-wheel motor bearing condition recognition technology,a condition recognition method based on double kernel based support vector data description(DK-SVDD)is proposed.Aiming at the lower recognition rate of SVDD caused by the mixed data structure,the DK kernel is constructed by combining the radial basis function(RBF)kernel function and the difference of Gaussians(DOG)kernel function with a certain proportion weight.According to the optimal binary tree principle,the condition recognition classifier is designed layer by layer,and the DK-SVDD in-wheel motor bearing condition recognition model is built.At the same time,the particle swarm optimization algorithm is used to optimize the model parameters to improve the learning ability and generalization ability of DK-SVDD.Based on the bench test data of in-wheel motor bearing,the feasibility of the proposed method is verified.The results show that the average training time of DK-SVDD method is 0.0655 s and the average condition recognition rate is 97.06%.Secondly,compared with RBF or DOG kernel function,DK-SVDD method can effectively improve the condition recognition rate and reduce the training time under various working conditions.According to the above results,the validity and superiority of proposed method based on DK-SVDD are verified.Obtained re-sults can provide reference for the subsequent development of in-wheel motor bearing state identification to im-prove the safety and reliability of electric vehicles.