TSMSE结合IOOA-BiLSTM的水电机组轴系故障诊断方法OA北大核心CSTPCD
TSMSE combined with IOOA-BILSTM for the fault diagnosis method of hydropower unit shafting
为提高水电机组轴系振动故障诊断的准确率,本文提出了一种新的诊断方法.首先,基于完全自适应噪声集合经验模态分解(CEEMDAN)进行振动信号预处理;其次,基于时移多尺度思想,引入时移多尺度样本熵(TSMSE)模型,以克服传统多尺度样本熵鲁棒性差以及粗粒化不足的问题;最终,将TSMSE提取的故障特征集输入经过改进鱼鹰算法(IOOA)优化的双向长短时记忆网络(BiLSTM),进行故障特征分类.通过对原始信号添加SNR=5 dB噪声并引入两种特征熵与TSMSE对比,分析TSMSE的抗噪性能.仿真实验表明,在给定数据集下TSMSE特征提取能力明显优于另外两种方法.同时,所提故障诊断模型应用在原始信号和含噪信号两种情况下,分别取得了100％以及97.22％的准确率,验证了所提模型的良好性能,为水电机组故障诊断提供新的科学方法.
In order to improve the accuracy of shafting vibration fault diagnosis of hydropower units,a new diag-nostic method is proposed.Firstly,the vibration signal decomposition was carried out based on the CEEMDAN.Secondly,based on the idea of time-shifted and multi-scale,a TSMSE model is proposed to overcome the poor ro-bustness and lack of coarse granulation of traditional MSE.Finally,the fault feature set extracted by TSMSE was input into the BiLSTM optimized by IOOA for fault feature classification.With adding SNR=5 dB noise to the origi-nal signal and introducing two multiscale entropies to compare with TSMSE,the anti-noise performance and robust-ness of TSMSE are analyzed.The results show that the stability and anti-noise performance of TSMSE feature ex-traction are obviously better than the other two in a given data set.At the same time,the accuracy of the proposed fault diagnosis model is 100％and 97.22％respectively in the case of original signal and noisy signal,which veri-fies the good performance of the proposed model and provides a new scientific method for fault diagnosis of hydro-power units.
张兼博;李想;曾云;唐跨纪
昆明理工大学冶金与能源学院,云南昆明 650093昆明理工大学冶金与能源学院,云南昆明 650093||云南省高校水力机械智能测试工程研究中心,云南昆明 650093
动力与电气工程
水电机组特征提取时移多尺度样本熵IOOA-BiLSTM故障诊断
hydroelectric generating setfeature extractionTSMSEIOOA-BILSTMfault diagnosis
《水利学报》 2024 (007)
862-873 / 12
国家自然科学基金项目(52079059)
评论