基于时频域多指标融合和图模型的滚动轴承早期退化监测方法

Early degradation monitoring method of rolling bearing based on time-frequency multi-index fusion and graph model

在对滚动轴承的全生命周期健康状态进行监测,采用从振动信号中提取监测指标时,存在噪声干扰、检测准确度低的问题,为此,提出了一种基于时频域多指标融合和图模型的滚动轴承早期退化点检测新方法。首先,进行了时频域的指标提取,对采集的轴承原始振动信号进行了分段处理,从每段信号中提取了多个时频域指标,进行了基于图模型的指标优化,即对提取的每一个指标进行了图建模,得到了一系列图模型;接着进行了图模型相似度计算,得到了优化后的指标;然后,进行了多指标融合,对优化后的多指标采用逐点均值及逐点最值相结合的方法进行了融合,得到了综合的监测指标;最后,进行了异常决策,采用假设检验对轴承早期退化进行了监测,确定了退化点,并在滚动轴承全生命周期退化数据集上进行了实验,对上述综合指标的退化点检测性能进行了检验,同时在该数据集上将该方法与其他4种方法进行了对比实验。研究结果表明:针对每一组实验数据,采用该方法均能成功检测到退化点,同时在对比实验中,采用该方法取得了1.27的最高平均排名;实验结果证明了该方法的有效性和先进性,表明该方法在滚动轴承早期退化监测中具有良好的实际应用潜力。

In order to monitor the whole life cycle health status of rolling bearings,there were some problems such as noise interference and low detection accuracy,when extracting monitoring indicators from vibration signals.A new method for detecting early degradation points of rolling bearings based on time-frequency domain multi-indicator fusion and graph model was proposed.Firstly,multiple time-frequency domain indexes were extracted,the collected bearing original vibration signal was processed by windowing and segmentation,and multiple time-frequency domain indexes were extracted from each signal segment.After extracting the indexes,the graph model was used to model and optimize each index,and a series of graph models were obtained.Then,the similarity of graph model was calculated to get the optimized index.The optimized multi-indexes were fused by the combination of point-to-point mean and point-to-point maximum to obtain a comprehensive monitoring index.Finally,the abnormal decision was made,the early degradation of the bearing was monitored by using the hypothesis test,the degradation point was determined,and the experiment were carried out on the whole life cycle degradation data set of rolling bearing to test the degradation point detection performance of the above comprehensive indexes.At the same time,comparative experiments were carried out with four methods on this data set.The experiment results show that the degradation points can be successfully detected for each group of experimental data,and the highest average ranking of 1.27 is obtained in the comparative experiment.The experimental results show that the method is effective and advanced,and shows that this method has good practical application potential in early monitoring of rolling bearing.