基于多特征集融合与多变量支持向量回归的回转支承剩余寿命评估

Residual life assessment of slewing bearing based on multivariate eigenvalues fusion and support vector regression

针对回转支承剩余寿命难以评估的问题,提出一种基于温度、扭矩、振动信号时域内多个特征值融合和多变量支持向量回归(MSVR)的剩余寿命评估新方法。该方法通过主成分分析(PCA)求得温度、扭矩、振动信号性能衰退指标量化回转支承性能衰退规律,以此作为输入量构建多变量支持向量回归回转支承剩余寿命评估模型。MSVR克服了结构简单、信息匮乏等缺点,实现变量之间冗余信息的消除和样本数据潜在信息的最大挖掘,采用回转支承全寿命实验数据对评估模型进行检验,结果表明MSVR可获得准确的评估结果。

A residual life assessment for slewing bearing based on characteristic indexes in time domain including temperature,torque and vibration information fusion and multivariable support vector regression was proposed. The principle analysis（ PCA） was used to obtain recession indicators（ temperature,torque and vibration signal） and quantize the degradation pattern of slewing bearing performance. With the three recession indicators being used as input data,a residual life assessment model for slewing bearing based on multivariable support vector regression was established. Overcoming the shortcomings of simple structure and information scarce,the proposed method was able to obtain the potential information in sample data and to eliminate redundant information contained between characteristic values,and was applied in lab slewing bearing data,results showed that multivariable support vector regression（ MSVR）could obtain accurate assessment results.