基于PCA-EDT-CNN的柴油发动机故障诊断方法

FAULT DIAGNOSIS METHOD OF DIESEL ENGINE BASED ON PCA-EDT-CNN

针对柴油发动机故障诊断中振动信号呈非平稳、非线性,直接将原始信号输入卷积神经网络(Convolutional Neural Network, CNN)进行故障诊断效果差的问题,提出一种基于PCA-EDT-CNN的新方法。首先,利用主成分分析(Principal Component Analysis, PCA)对传感器采集的原始数据进行自适应降维,并构建符合条件的主成分特征向量矩阵(Principal Component Eigenvector Matrix, PCEM);其次,对PCEM进行欧氏距离变换(Euclidean Distance Transformation, EDT),计算各行间欧氏距离并构建欧氏距离矩阵(Euclidean Distance Matrix, EDM);最后,将PCEM和EDM分别展平为一维向量并合成一个一维样本序列,输入到一维卷积神经网络(One-Dimensional Convolutional Neural Network, 1DCNN)对模型进行训练与诊断。搭建柴油发动机预置故障试验台,验证了该方法的有效性,且通过与传统方法对比,结果表明,该方法对柴油发动机不同故障状态诊断准确率高,具有实际工程应用价值。

Aiming at the problem that the vibration signal in diesel engine fault diagnosis is non-stationary and nonlinear, and the original signal is directly input into the Convolutional Neural Network(CNN) for fault diagnosis with poor effect, a new method based on PCA-EDT-CNN is proposed. Firstly, use Principal Component Analysis(PCA) to adaptively reduce the original data collected by the sensor, and construct a qualified Principal Component Eigenvector Matrix(PCEM);secondly, perform Euclidean Distance Transformation(EDT) on PCEM, calculate the Euclidean distance between each row and construct the Euclidean Distance Matrix(EDM);finally, flatten PCEM and EDM into one-dimensional vectors and synthesize a one-dimensional sample sequence, input into One-Dimensional Convolutional Neural Network(1 DCNN) to train and diagnosis the model. A diesel engine preset failure test bench was built to verify the effectiveness of the method, and through comparison with traditional methods, the results show that the method has high accuracy in diagnosing different fault states of diesel engines and has practical engineering application value.