An Intelligent Deep Feature Learning Method With Improved Activation Functions for Machine Fault Diagnosis

Wei You,Xingxing Jiang,Zhongkui Zhu,Changqing Shen,Dong Wang,Liang Chen

doi:10.1109/access.2019.2962734

Abstract

Rotating machinery has been developed with high complexity and precision, and bearings and gears are crucial components in the machinery system. Deep learning has attracted considerable attention from researchers in this area. The convolutional neural network (CNN) is a typical deep learning model that has a strong capability for automatically extracting features from raw data. This capability minimizes dependence on expert knowledge during feature extraction and selection. In CNN, hyperparameters, such as activation functions, can directly influence the performance of the model. In this study, the improved rectified linear units (ReLU)-CNNs are proposed for machinery fault diagnosis. The model's input are raw vibration signals without feature extraction and selection. It is experimentally validated for fault diagnosis using bearing and gearbox datasets. Results show that the proposed method can obtain satisfactory accuracy with enhanced convergence speed. For both datasets, the proposed method gives better diagnosis accrues than the other compared models. The proposed model can take advantages of standard ReLU-CNN, and these advantages can overcome traditional activation functions' vanishing gradient problems. Meanwhile, the improved ReLU-CNN has a new property that makes it perform better than the standard ReLU-CNN.

Highlights

Over the past few years, the great progress which machine learning especially deep learning achieved make these techniques gain increasingly widely applications in many different areas, such as the rotating machinery industry area [1]
In this work, the improved rectified linear units (ReLU)-convolutional neural network (CNN) was proposed for machinery fault diagnosis
Raw vibration signal was directly utilized as input data, and the proposed method can eliminate dependence on manual feature extraction and selection

Summary

INTRODUCTION

Over the past few years, the great progress which machine learning especially deep learning achieved make these techniques gain increasingly widely applications in many different areas, such as the rotating machinery industry area [1]. The main contributions are listed as follows: (1) an improved rectified linear units function based on a CNN model for machine fault diagnosis is proposed; (2) model’s hyperparameters’ selection, especially for an activation function, is fully investigated; (3) the proposed model with satisfactory accuracies with an enhanced convergence speed is experimentally validated using two different machinery datasets. Feedforward propagation is the process of multiplying various input values of a particular neuron by their associated weights, summing the results, and scaling the values between a given range before forwarding this information to the succeeding layer. A. ACTIVATION FUNCTIONS IN IMPROVED RELU-CNNS From the discussion in Section II Part C, ReLU can eliminate negative input values, it may discard some important information contained in these values, which may furtherly generate some problems. LReLU-CNN is a specialist of PReLU when α is equal to 0.01

MODEL BUILDING PROCESS

Findings

CONCLUSION