基于混沌理论和CNN-OSVM的水轮机空化状态识别方法

State Recognition Method of Hydraulic Turbine Cavitation Based on Chaos Theory and CNN-OSVM

针对水轮机空化声发射(AE)信号非线性强,导致水轮机空化状态识别准确度不高的问题,建立基于混沌理论和卷积神经网络结合优化支持向量机(CNN-OSVM)的水轮机空化状态识别方法。对不同空化状态下的水轮机空化AE信号进行相空间重构,获得相图作为数据集,通过卷积神经网络提取不同空化状态下的相图特征,输入经网格搜索算法结合K折交叉验证算法全局参数寻优的优化支持向量机分类器完成空化状态识别。结果表明:输入混沌相图数据集的CNN-OSVM模型能够准确识别4种空化状态,平均准确率高达98.8%;同时证实相较于CNN模型、OSVM模型,CNN-OSVM模型对非线性信号分类具有更高的识别准确率和泛化性。

Aiming at the problem of low recognition accuracy of hydraulic turbine cavitation condition,which caused by strong nonlinearity of acoustic emission(AE)signal,a cavitation state recognition method of hydraulic turbine based on chaos theory and convolutional neural network combined with optimized support vector machine(CNN-OSVM)was established.The phase space reconstruction of the AE signal under different hydraulic turbine cavitation conditions was carried out to obtain the phase diagram as a data set.The phase map features of different cavitation states were extracted by convolutional neural network,and the optimization support vector machine classifier was optimized by grid search algorithm and K-fold cross-validation algorithm to identify the cavitation state.Results show that the CNN-OSVM model with the input chaotic phase diagram could identify four cavitation conditions accurately,with an average accuracy of 98.8%.Meanwhile,it was proved that CNN-OSVM model had higher recognition accuracy and generalization than CNN model and OSVM model for nonlinear signal classification.