基于IWAO-SVDD的工业机器人异常检测研究  

Industrial Robot Anomaly Detection Based on IWOA-SVDD

作　　者：何成刚 朱润智 向珍琳 汪晓鑫 刘吉华 HE Chenggang;ZHU Runzhi;XIANG Zhenlin;WANG Xiaoxin;LIU Jihua(School of Rail Transportation,Wuyi University,Jiangmen Guangdong 529020,China;Foshan Institute of Intelligent Equipment Technology,Foshan Guangdong 528234,China)

机构地区：[1]五邑大学轨道交通学院,广东江门529020 [2]佛山智能装备技术研究院,广东佛山528234

出　　处：《机械设计与研究》2024年第4期37-43,共7页Machine Design And Research

基　　金：广东省教育厅特色创新项目(2023KTSCX151);广东省基础与应用基础研究基金项目(2020B1515120010);五邑大学高层次人才科技计划资助项目(AG2018001)。

摘　　要：针对工业机器人运行数据的非平稳性与信号特征提取困难等问题,提出了一种基于集合经验模态分解与离散小波分解、连续均方误差结合的信号去噪方法,再对去噪信号进行时域特征提取,并使用改进鲸鱼优化算法(IWOA)优化支持向量数据描述(SVDD)参数与特征,形成多目标优化异常检测算法。首先对高维度数据进行集合经验模态分解,根据连续均方误差寻找到纯净模态分量与含噪音模态的临界点,使用离散小波对噪音模态去噪进行信号重构后再提取时域特征。然后利用改进鲸鱼优化算法(IWOA)对多模态特征与SVDD核参数进行寻优,进而构建异常检测模型。利用该模型对工业机器人运行中的反馈电流,反馈力矩等信号进行异常检测。结果表明该模型能有效的判断出工业机器人的异常情况,精确度能达到97%-99%,相较其他方法精确度能提升4%-5%。Aiming at the non-stationarity of industrial robot operation data and the difficulty of signal feature extraction,a signal denoising method based on the combination of ensemble empirical mode decomposition and discrete wavelet decomposition is proposed.The denoising signal is extracted in time domain,and the support vector data(SVDD)parameters and features are optimized by the improved whale optimization algorithm(IWOA)to form a multi-objective optimization anomaly detection algorithm.First,the high-dimensional data is subjected to ensemble empirical mode decomposition,and the critical point between the pure modal component and the noise-containing mode is found according to the continuous mean square error,and the discrete wavelet is used to denoise the noise mode to reconstruct the signal and then extract the time domain feature.Then,the IWOA is used to optimize the multimodal features and SVDD kernel parameters.Then an anomaly detection model is constructed to detect abnormalities in signals such as feedback current and feedback torque during the operation of industrial robots.The results show that the model can effectively judge the abnormal situation of industrial robots,and the accuracy can reach 97%-99%,which is 4%-5%higher than other methods.

关 键 词：工业机器人 特征提取 异常检测 IWOA SVDD 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]