【摘要】 针对现有聚类方法对机器人故障数据聚类时对初始点选取依赖性大、收敛速度慢且精度低等问题，提出了一种FFDEZOA算法来对KMC聚类算法进行优化。ZOA算法具有寻优能力较强，收敛速度快，且在聚类时对初始点选取依赖性小，但其有几率会陷入到局部最优解。首先针对ZOA算法的缺点，提出了自由觅食策略、非线性收敛因子及斑马进化策略等来对其进行改进，能够有效提高算法搜索范围，从而避免局部最优；进而结合FFDEZOA和KMC算法的互补迭代，既加快了算法的搜索速度，也提升了精度。在多个公开数据集上的实验表明，FFDEZOA-KMC在精确度和归一化互信息的指标上均优于ZOA-KMC、AO-KMC、KMC和MFO-KMC,具有更好的收敛性能和聚类效果。最后依据各故障特征的主成分不同，利用FFDEZOA-KMC对故障数据进行了聚类，可在多种工况下对机器人进行针对性的保养和维护。更多还原

【Abstract】 A FFDEZOA algorithm is proposed to optimize the KMC clustering algorithm in response to the problems of high dependence on initial point selection, slow convergence speed, and low accuracy in existing clustering methods for robot fault data clustering.The ZOA algorithm has strong optimization ability, fast convergence speed, and little dependence on initial point selection during clustering, but it has a chance of falling into local optimal solutions.Firstly, in response to the shortcomings of the ZOA algorithm, free foraging strategy, nonlinear convergence factor, and zebra evolution strategy were proposed to improve it, which can effectively increase the search range of the algorithm and avoid local optima; furthermore, combining the complementary iteration of FFDEZOA and KMC algorithms not only accelerates the search speed of the algorithm, but also improves accuracy.Experiments on multiple public datasets have shown that FFDEZOA-KMC outperforms ZOA-KMC,AO-KMC,KMC,and MFO-KMC in terms of accuracy and normalized mutual information, with better convergence performance and clustering performance.Finally, based on the different principal components of each fault feature, FFDEZOA-KMC was used to cluster the fault data, which can provide targeted maintenance and upkeep for robots under various working conditions.更多还原