一种基于二维卷积神经网络的舵机故障检测方法

A Steering Gear Fault Detection Method Based on Two-Dimensional Convolutional Neural Network

针对卷积神经网络对一维舵机数据特征提取不充分,本文提出将一维数据升级为二维数据,采用二维卷积神经网络对舵机故障进行智能检测。首先将一维数据首尾对称排列组成矩阵形式的二维数据,拓宽感受野,增加数据量,打破了空间局限性,避免了数据特征提取不充分,使特征提取具有全局性;其次构建了局部特征学习模块,该模块包含一个卷积层,一个Batch Normalization(BN)层,和一个ReLU激活函数,用于学习数据相关性。最后利用该模型实现对舵机数据的处理,从而实现舵机的智能故障检测。实验结果表明,该模型的准确度高达99.53%,效果优于其他的常用模型,证明了二维卷积神经网络应用于舵机故障检测的可行性。

In view of the insufficient feature extraction of one-dimensional steering gear data by convolutional neural network,it is proposed to upgrade one-dimensional data to two-dimensional data and adopt two-dimensional convolutional neural network for intelligent detection of steering gear faults in this paper.First of all,the one-dimensional data is upgraded to the two-dimensional data in the form of a matrix through the method implementation of stitching at the beginning and end which broadens the sensing field,increases the amount of data,breaks the space limitation,avoids the incomplete of data feature extraction,and makes feature extraction global.Then,the local feature learning module is established,which contains a convolutional layer,a batch normalization(BN)layer and a ReLU activation function for learning data correlation.Finally,the model is used to realize the detection and classification of the servo data,which realizes the rudder intelligent fault detection.Experimental results show that the accuracy of proposed method can reach 99.53%,which is better than other commonly used models and proves the feasibility of the model.