摘要
文章针对滚动轴承运行辐射声信号,提出一种滚动轴承性能退化预测的方法。轴承辐射声信号通过改进变分模态分解(improved variational mode decomposition, IVMD)得到K层有限带宽固有模态函数(band-limited intrinsic mode functions, BLIMFs);对BLIMFs分量的能量进行相对能量熵分析并作为轴承性能退化特征指标,将退化指标构成的时间序列分解成趋势项和残余项,对残余项进行平稳性检验和白噪声检验,对趋势项和非白噪声的残余项分别通过堆栈长短时记忆(stack long short-term memory, SLSTM)神经网络进行预测,使用自适应矩估计(adaptive moment estimation, ADAM)优化器反向优化网络权值、阈值;采用拉伊达法则确定阈值作为预警线,当预测退化曲线超过阈值报警线时实现报警。实验结果表明:SLSTM模型预测与分解趋势项之间的均方误差为4.149 2×10^(-5),均方根误差为0.003 6,相关系数为0.975 3;SLSTM模型预测与未去除残余项的相对能量熵之间的相关系数为0.776 3,模型的拟合程度较高,轴承性能退化评估良好,预测曲线在轴承早期退化阶段时能够予以报警。
In this paper,a method for predicting the performance degradation of rolling bearings is proposed based on the radiated acoustic signals of rolling bearings.The band-limited intrinsic mode functions(BLIMFs)of K layer were obtained by improved variational mode decomposition(IVMD)of bearing radiated acoustic signals.The relative energy entropy(REE)analysis of the energy of BLIMFs component was carried out to establish the characteristic index of bearing performance degradation.The time series composed of degradation index was decomposed into trend term and residual term,and the stationarity test and white noise test were carried out for the residual term.The trend item and the non-white noise residual item were predicted by the stack long short-term memory(SLSTM)neural network respectively,and a reverse optimization of the network weights and thresholds was made by the adaptive moment estimation(ADAM)optimizer.Pauta criterion was adopted to determine the threshold value as the warning line.Alarm was realized when the predicted degradation curve exceeds the threshold warning line.The experimental results show that the mean square error between SLSTM prediction and decomposition trend term is 4.1492×10^(-5),the root mean square error is 0.0036,and the correlation coefficient is 0.9753.The correlation coefficient between the prediction of the SLSTM model and the REE of the unremoved residual term is 0.7763,which indicates that the model has a high degree of fit and the bearing performance degradation evaluation is good.The prediction curve can give an alarm in the early degradation stage of the bearing.
作者
陈剑
曹昆明
张磊
孙太华
程明
阚东
CHEN Jian;CAO Kunming;ZHANG Lei;SUN Taihua;CHENG Ming;KAN Dong(Institute of Sound and Vibration Research,Hefei University of Technology,Hefei 230009,China;School of Mechanical Engineer-ing,Hefei University of Technology,Hefei 230009,China)
出处
《合肥工业大学学报（自然科学版）》
CAS
北大核心
2022年第8期1009-1015,共7页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金青年科学基金资助项目(11604070)
安徽省科技重大专项资助项目(17030901049)。
