基于改进深度卷积神经网络的轴承故障诊断

轴承为风电机组的重要且故障频发部件，传统基于轴承振动数据的图像转换的卷积神经网络(CNN)的故障诊断技术存在一定局限性。提出了一种基于改进深度卷积神经网络(IDCNN)的直接时间序列特征提取方法，依据采样频率将原始振动数据划分为单个样本，构建诊断模型训练数据集。设计了一种新型的深度卷积神经网络(IDCNN)，自动提取复杂样本数据的故障特征，提高DCNN的鲁棒性和泛化性，并将IDCNN提取的高维故障特征输入到分类器中，从而实现轴承故障的智能诊断。对比实验结果表明本方法有效提升了故障诊断精度。

Based on Improved Deep Convolutional Neural Network for Bearing Fault Diagnosis

Bearing is an important part of wind turbine with frequent faults. The traditional fault diagnosis technology of convolutional neural network (CNN) based on image conversion of bearing vibration data has certain limitations. A direct time series feature extraction method based on improved deep convolutional neural network (IDCNN) is proposed. The original vibration data is divided into a single sample according to the sampling frequency, and the training data set of the diagnosis model is constructed. The robustness and generalization of DCNN are improved, automatic extraction of complex sample data fault characteristics,and the high-dimensional fault features extracted by IDCNN are input into the classifier, so as to realize the intelligent diagnosis of bearing faults. The comparative experimental results show that the proposed method effectively improves the accuracy of fault diagnosis.