Abstract: To accurately evaluate the working state of Rotate Vector (RV) reducer,in view of the problems of RV reducer such as inconspicuous faults,few sample data and difficult diagnosis,the periodic feature of torque transmission of RV reducer during normal operation was proved theoretically.The characteristics of data periodicity was reflected effectively by the periodic evolution characteristics of rotating machinery test signals and the Ensemble Empirical Mode Decomposition (EEMD).An Extreme Learning Machine (ELM) fault diagnosis model by Particle Swarm Optimization (PSO) was proposed based on EEMD,whose performance was verified with the bearing experimental dataset of Xi'an Jiaotong University.On this basis,the data measured the RV reducer test platform was substituted into the proposed model,and finally compared with other models.The comparison results showed that the proposed model could more effectively judge the working state of the RV reducer.

Key words: rotate vector reducer, empirical mode decomposition, particle swarm optimization, extreme learning machine, fault diagnosis

摘要： 为了准确评估旋转矢量(RV)减速器的工作状态,针对RV减速器故障不明显、样本数据少、难诊断的问题,首先从理论上证明RV减速器正常运行时扭矩传递时具有周期性,利用旋转机械测试信号周期演变特征和集成经验模态分解(EEMD)可以有效反映数据周期性的特点,提出一种基于EEMD的粒子群优化(PSO)算法的极限学习机(ELM)故障诊断模型,同时采用西安交通大学轴承实验数据集验证了模型性能。在此基础上,将RV减速机测试平台所测得的数据代入所提模型,最后与其他模型进行对比,结果显示所提模型能够更有效地判断出RV减速机的工作状态。

关键词: 旋转矢量减速器, 集成经验模态分解, 粒子群优化算法, 极限学习机, 故障诊断