A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load. (1st February 2018)

• Record Type: Journal Article
• Authors: Zhang, Wei; Li, Chuanhao; Peng, Gaoliang; Chen, Yuanhang; Zhang, Zhujun
• Is Part Of: Mechanical systems and signal processing. Volume 100(2018)
• Page Start: 439
• Page End: 453
• Publication Date: 2018-02-01
• Subjects: Intelligent fault diagnosis -- Convolutional neural networks -- Load domain adaptation -- Anti-noise -- End-to-end; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2017.06.022
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A novel and simple end-to-end learning framework with tricks that are easy to implement is proposed. This algorithm performs pretty well under the noisy environment, and works directly on raw noisy signals without any pre-denoisng methods. The model has strong domain adaptation ability, and therefore achieves high accuracy under different working load. The inner mechanism of proposed TICNN model in feature extraction and classification is explored by visualizing the feature maps learned by TICNN. Abstract: In recent years, intelligent fault diagnosis algorithms using machine learning technique have achieved much success. However, due to the fact that in real world industrial applications, the working load is changing all the time and noise from the working environment is inevitable, degradation of the performance of intelligent fault diagnosis methods is very serious. In this paper, a new model based on deep learning is proposed to address the problem. Our contributions of include: First, we proposed an end-to-end method that takes raw temporal signals as inputs and thus doesn't need any time consuming denoising preprocessing. The model can achieve pretty high accuracy under noisy environment. Second, the model does not rely on any domain adaptation algorithm or require information of the target domain. It can achieve high accuracy when working load is changed. To understand the proposed model, we will visualize the learned features, and try to analyze the reasons behind the high performance of the model.