基于SVR的滚动轴承剩余使用寿命预测方法研究

发布时间：2018-05-04 10:06

本文选题：滚动轴承 + 剩余使用寿命　；参考：《哈尔滨理工大学》2017年硕士论文

【摘要】：滚动轴承是旋转机械设备中重要的基础零件之一,一旦出现故障可能会引发安全事故造成重大经济损失和恶劣的社会影响。滚动轴承的剩余使用寿命(RUL)是其受损程度的综合反映,因此,准确预测滚动轴承的RUL对旋转机械的视情维修意义重大。论文建立2种模型对滚动轴承的RUL进行预测。(1)研究滚动轴承的特征提取以及特征约简方法。提出将包络解调与数学形态学结合,提取数学形态学特征。同时,基于时域、频域特征指标与数学形态学分形维数构造特征矩阵,并提出多评价标准有效性分析结合核主成分分析(MCEA-KPCA)的特征约简方法。(2)研究滚动轴承运行状态可靠度的预测方法。为了保证滚动轴承运行状态可靠度的预测精度同时增加预测步长,提出一种数学形态学分形维数结合改进果蝇优化算法-支持向量回归(IFOA-SVR)的滚动轴承可靠度预测方法。该方法首先将数学形态学分形维数作为轴承性能退化状态特征;其次,利用IFOA结合SVR建立预测模型,对性能退化特征进行预测,同时利用极大似然估计结合IFOA建立威布尔比例故障率模型(WPHM),进而得到可靠度模型;最后,将预测结果嵌入到可靠度模型中,从而预测出轴承运行状态的可靠度,即RUL。(3)研究滚动轴承当前使用寿命与全寿命的比值p的预测方法。针对训练和测试是同工况、同型号,但轴承个体不同的情况,提出MCEA-KPCA结合组合的支持向量回归(CSVR)预测模型的RUL预测方法。该方法可自适应确定每个评价标准的权重和各SVR模型的权重,对与训练不同的轴承进行测试,将约简后特征输入到CSVR预测模型中,预测轴承的p值,实验结果表明,所提方法可以准确地对滚动轴承进行RUL预测。
[Abstract]:Rolling bearing is one of the important basic parts in rotating machinery and equipment. Once failure occurs, it may lead to serious economic loss and bad social impact. The residual service life of rolling bearing (RUL) is a comprehensive reflection of its damage degree. Therefore, it is of great significance to accurately predict the RUL of rolling bearing for the maintenance of rotating machinery. In this paper, two models are established to predict the RUL of rolling bearing.) the feature extraction and feature reduction method of rolling bearing are studied. Combining envelope demodulation with mathematical morphology, the feature of mathematical morphology is extracted. At the same time, the feature matrix is constructed based on time domain, frequency domain characteristic index and mathematical morphology fractal dimension. The method of feature reduction of MCEA-KPCA (MCEA-KPCA) is proposed to predict the reliability of rolling bearings. In order to ensure the prediction accuracy of rolling bearing reliability and increase the prediction step, a new method for predicting rolling bearing reliability is proposed, which combines mathematical morphology fractal dimension with improved Drosophila optimization algorithm-support vector regression (SVR) IFOA-SVR. Firstly, the fractal dimension of mathematical morphology is regarded as the characteristic of bearing performance degradation state; secondly, the prediction model is established by using IFOA and SVR to predict the performance degradation characteristics. At the same time, the Weibull proportional failure rate model is established by using maximum likelihood estimation combined with IFOA, and then the reliability model is obtained. Finally, the prediction results are embedded into the reliability model to predict the reliability of the bearing running state. The prediction method of the ratio p of the current service life to the full life of the rolling bearing is studied. In view of the fact that the training and testing are in the same working condition and the same type, but the bearing individual is different, a RUL forecasting method based on MCEA-KPCA combined with support vector regression model is proposed. This method can adaptively determine the weight of each evaluation criterion and each SVR model, test the bearings which are different from the training, and input the reduced features into the CSVR prediction model to predict the p value of the bearings. The proposed method can accurately predict the RUL of rolling bearings.
【学位授予单位】：哈尔滨理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.33

【相似文献】