| 魏豪， 权伟， 何建国， 张玮.基于DTCWPT分频特征和BiLSTM的滚动轴承剩余寿命预测[J].轻工机械,2022,40(6):88-95 |
| 基于DTCWPT分频特征和BiLSTM的滚动轴承剩余寿命预测 |
| Residual Life Prediction of Rolling Bearings Based on DTCWPT Crossover Features and BiLSTM |
| |
| DOI：10.3969/j.issn.1005 2895.2022.06.015 |
| 中文关键词: 滚动轴承 剩余寿命预测 快速谱峭度 双树复小波包 双向长短时记忆网络 |
| 英文关键词:rolling bearing residual life prediction fast spectral kurtosis DTCWPT（Dual Tree Complex Wavelet Packet） BiLSTM（Bi Directional Long Short Term Memory） |
| 基金项目:陕西省科技厅自然科学基础研究计划 面上项目（2022JM 219） |
|
| 摘要点击次数: 237 |
| 全文下载次数: 224 |
| 中文摘要: |
| 针对滚动轴承退化状态不稳定和传统退化指标不能准确描述轴承退化状态从而导致预测精度不高的问题，课题组提出一种基于快速谱峭度与双树复小波包DTCWPT结合双向长短时记忆网络BiLSTM的轴承寿命预测方法。该方法首先使用快速谱峭度计算故障中心频率；然后使用双树复小波包对信号进行分频处理，选取包含故障中心频率的分频带重构信号提取退化特征，并通过时间相关性、鲁棒性进行特征筛选；最后使用BiLSTM进行寿命预测。试验结果表明课题组所提出的预测方法可以准确预测轴承剩余使用寿命，对比LSTM方法进一步验证了课题组所提方法的有效性。 |
| 英文摘要: |
| Aiming at the problem of low prediction caused by the unstable degradation state of rolling bearings and inaccurate traditional degradation index to describe bearing degradation state, a bearing life prediction method was proposed based on fast spectral kurtosis and dual tree complex wavelet packet（DTCWPT），combined with bidirectional long short time memory network（BiLSTM）.Firstly, the fault center frequency was calculated with fast spectrum kurtosis.Then the dual tree complex wavelet packet was used for frequency division processing of the signal, and the frequency division reconstruction signal containing the fault center frequency was selected to extract the degradation features, and the feature screening was carried out through time correlation and robustness. Finally BiLSTM was used for life prediction. The experimental data verify that the prediction method proposed can accurately predict the remaining service life of bearings, and the effectiveness of the proposed method is verified by comparing with LSTM method." |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
