摘要
针对潜油电泵特征提取及状态判别问题,提出一种基于冗余第二代小波包变换、多尺度熵和概率神经网络的诊断方法。首先,利用冗余第二代小波包对拾取的信号样本进行处理,得到相应的子带信号分量,继而计算所得子带信号分量的多尺度熵值,并构造能够表征电泵状态的特征向量,最终将特征向量输入到概率神经网络中实现潜油电泵故障的自动识别。实测数据分析结果表明,所述方法能够有效对潜油电泵的工作状态进行识别,具有一定工程应用价值。
Aiming at solving the problems of feature extraction and condition judgment for electrical submersible pumps,this paper proposes a diagnosis method based on redundant second generation wavelet package transformation(RSGWPT) , multiscale entropy and probabilistic neural network (PNN) . Firstly, the acquired signal samples areprocessed using redundant second generation wavelet package, and the corresponding subband signal components can beobtained. Then, the multiscale entropy of each obtained subband signal components is calculated, and the feature vectorswhich could characterize the conditions of the electrical submersible pumps are constructed. Finally, the feature vectorsare input into the probabilistic neural network, and the different fault types of the electrical submersible pumps can beidentified automatically. The analysis results of the measured data show that the proposed method could effectivelyidentify the work condition of the electrical submersible pumps, and has a certain value for engineering application.
出处
《机械工程师》
2018年第1期167-170,172,共5页
Mechanical Engineer