联合ACO-K-Means与MCS-SVM的设备寿命预测  

Equipment Life Prediction Based on ACO-K-Means and MCS-SVM

作　　者：刘勤明[1] 孙钰栋 陈扬 张坤 LIU Qin-ming;SUN Yu-dong;CHEN Yang;ZHANG Kun(Business School,University of Shanghai for Science&Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学管理学院,上海200093

出　　处：《系统工程》2024年第3期1-10,共10页Systems Engineering

基　　金：国家自然科学基金资助项目(71840003);上海市自然科学基金资助项目(19ZR1435600);教育部人文社会科学研究规划项目(20YJAZH068);上海理工大学科技发展项目(2020KJFZ038)。

摘　　要：针对实际生产中小样本情况下多分类数据缺乏明确的样本标签、样本存在噪声、样本匮乏等问题,本文提出了一种基于改进蚁群优化K-Means(Ant Colony Optimization-K-Means)与多分类自新增SVM(Multi Classification Self-Adding-SVM)的设备健康状态分析与寿命预测模型。首先,基于模糊数据集遵循传统SVM对数据进行第一次分类,得到初次分类结果。随后,通过基于蚁群算法的改进K-Means算法对初次分类后的数据集进行聚类,从而得到更多不同状态下的设备健康状态标签。其次,建立噪声比例系数,并通过引入不均衡比例标准与自适应新增法则优化数据集分布,在忽略噪声影响下丰富匮乏标签样本容量。在此基础上根据聚类出的种类个数引入SVM集,实现数据集的多分类处理。再次,通过拟合设备振动方根均值与剩余寿命变化趋势评估设备未来健康趋势。最后通过算例证明,噪声小样本不均衡数据下,本文提出的ACO-K-Means联合MCS-SVM模型在设备健康状况分类与未来寿命预测方面均有不错的效果。In view of the problems of lack of clear sample labels,noise in samples and lack of samples in the case of small and medium-sized samples in actual production,this paper proposes equipment health status analysis and life prediction model based on improved ant colony optimization K-Means and Multi Classification Self Adding(SVM).First,based on the fuzzy data set,the data are classified for the first time according to the traditional SVM,and the first classification result is obtained.Then,the improved K-Means algorithm based on Ant Colony Algorithm is used to cluster the data set after the initial classification,so as to get more device health status labels in different states.Secondly,the noise scale coefficient is established,and the data set distribution is optimized by introducing the unbalanced scale standard and the adaptive addition rule,so as to enrich the sample size of deficient tags without considering the influence of noise.On this basis,the SVM set is introduced according to the number of clustering categories to realize the multi classification processing of the data set.Thirdly,the future health trend of the equipment is evaluated by fitting the root mean value of vibration and the change trend of residual life.Finally,an example shows that the ACO-K-Means combined with MCS-SVM model proposed in this paper has good results in equipment health classification and future life prediction under the unbalanced data of small noise samples.

关 键 词：状态识别 SVM K-MEANS 剩余寿命预测 不均衡数据 噪声数据 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]