基于VMD-WOA-LSSVM的汽轮机轴承振动趋势预测

Vibration Trend Prediction of Steam Turbine Bearings Based on VMD⁃WOA⁃LSSVM

在火电厂中,汽轮机故障通常会导致轴承的异常振动,因此预测轴承的振动趋势能够为汽轮机故障提供借鉴依据,降低故障发生概率。针对振动数据具有随机性和波动性的问题,首先利用变分模态分解方法,将振动序列分解成一系列子模态,以降低振动序列的非平稳性。将分解出的分量作为最小二乘支持向量机(Least Squares Support Veetor Machine Classifiers,LSSVM)预测模型的输入,利用鲸鱼算法(Whale Optimization Algorthm,WOA)算法的寻优特性对LSSVM中的参数进行优化,从而建立超前一步预测模型,最后将各个子模态的预测结果相叠加,得到预测振动数据。为评估该模型的预测性能,以江苏某电厂的轴承振动实测数据为例进行仿真实验。结果表明,所提出的预测模型优于其他多种典型预测模型,表现出较好的预测性能。

In thermal power plants,the abnormal vibration of bearings usually leads to steam turbine failure.Therefore,predicting the vibration trend of bearings can provide a reference for steam turbine failure and reduce the probability of failure.The vibration data have the characteristics of randomness and fluctuation.Firstly,the vibration sequence was decomposed into a series of sub modes by using the variational mode decomposition method to reduce the non⁃stationary of the vibration sequence.The decomposed components were used as the input of LSSVM prediction model,and the parameters in LSSVM were optimized by using the optimization characteristics of WOA algorithm,so as to establish a one⁃step ahead prediction model.Lastly,the prediction results of each sub mode were superimposed to obtain the predicted vibration data.In order to evaluate the prediction performance of the model,the measured bearing vibration data of a power plant in Jiangsu Province were taken as an example for simulation verification.The results show that the proposed prediction model is better than many other typical prediction models and shows better prediction performance.