Research on multitask fault diagnosis and weight visualization of rotating machinery based on convolutional neural netwo
- PDF / 4,210,444 Bytes
- 14 Pages / 595.276 x 790.866 pts Page_size
- 0 Downloads / 234 Views
(2020) 42:603
TECHNICAL PAPER
Research on multitask fault diagnosis and weight visualization of rotating machinery based on convolutional neural network Fuzhou Feng1 · Chunzhi Wu1 · Junzhen Zhu1 · Shoujun Wu1 · Qingwen Tian1 · Pengcheng Jiang1 Received: 9 December 2019 / Accepted: 5 October 2020 © The Brazilian Society of Mechanical Sciences and Engineering 2020
Abstract Recently, deep learning methods such as convolutional neural networks (CNN) have been widely used in fault diagnosis of rotating machinery. However, most methods are not designed to consider the influence of the working conditions and limited to classifying several fault types. In this paper, we propose two CNN-based multitask models, i.e., sigmoid multitask fault diagnosis model and multisoftmax multitask classification model, which can classify the fault types and the working conditions of the signal simultaneously. All samples are jointly learned at the shared network layers, and different types of learning tasks are completed at different subnetwork layers. Results show that the proposed sigmoid multitask fault diagnosis model achieves the overall classification accuracy of 96.8% when testing the CWRU bearing dataset. The proposed multisoftmax multitask classification model is used to classify different working conditions and fault types of planetary gearbox. With frequency signals as inputs, the accuracy rate is able to reach 96.2%, and the classification accuracy of gear and speed conditions reaches more than 99%. Additionally, the gradient-weighted class activation mapping (Grad-CAM) method is used to visualize the weight vectors of different convolutional layers and locate the signal segment of interest to the model. Keywords Convolutional neural network · Multitasking learning · Fault diagnosis · Grad-CAM
1 Introduction With the fast development of intelligent manufacturing, it is important to protect manufacturing systems from mechanical failures caused by various faults, which may further lead to economic losses and safety issues. With the help of sensors, massive data are collected for fault prediction and health management of mechanical equipment. It brings the field of mechanical equipment failure prediction and health management into the era of “big data” [1]. Recently, the machine equipment fault diagnosis based on machine learning and deep learning has gradually become a hot area. Among different deep learning methods, CNN is one of the most widely used in the field of mechanical equipment Technical Editor: Wallace Moreira Bessa. * Fuzhou Feng [email protected] Chunzhi Wu [email protected] 1
Department of Vehicle Engineering, Academy of Army Armored Forces, Beijing 100072, China
fault diagnosis [2]. When using CNN, some researchers continue the idea from the shallow model to diagnose the fault types. They use the manually extracted features as input and make use of the strong fitting ability of the CNN to improve the intelligent identification ability of the device’s health state. Janssens et al. [3] perfo
Data Loading...
