Scholarlink

Background <p>Fault diagnosis of two-stage gearboxes using deep learning techniques has gained significant attention in the last decade. Spur gearboxes are widely used for power transmission in industries which makes the fault diagnosis study of spur gearboxes very crucial.</p> Purpose <p>A significant amount of work has been done to diagnose faults in spur gearboxes at higher loads and high speeds. However, fewer studies are performed to diagnose faults in spur gearboxes running at comparatively low speeds and lower loads. In this paper, fault diagnosis of spur gearbox for faults with different degrees of severity has been performed at relatively low speeds and loading conditions.</p> Methods <p>For this purpose, an experimental test rig is designed to collect vibration signals of the spur gearbox at speeds varying from 180 to 420 rpm and load varying from 0 to 30 kg. The vibration signals are collected for three pinions with different crack lengths of 1, 2, and 3 mm in the tooth root, two pinions with different degrees of eccentricity, and for healthy gear.</p> Results <p>Particle Swarm Optimization (PSO) method is used for DCNN hyperparameter optimization at low speeds and low loads. Persistence Power Spectrum (PPS) image is a histogram in power–frequency space which allows even less dominant frequencies to be visible in the image. The discriminant features of PPS images are explored using the DCNN model optimized using PSO. Apart from PPS, four other types of images including HHT spectrum and CWT spectrum are also analyzed to classify faults at low speed and low loads.</p> Conclusions <p>Among these five methods, the PPS-PSO-DCNN model has shown an accuracy of 97.17% with an image size of 32 pixels. The proposed method is also tested on the publicly available gearbox dataset.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fault Diagnosis Analysis of Two Stages Spur Gearbox by Image Classification Using Deep Convolution Neural Network

  • 期刊论文

  • 发表日期：

    2025年9月6日

查看全文
  • Mahendra Singh Raghav,
  • Shivdayal Patel,
  • Jay Govind Verma

摘要

Background

Fault diagnosis of two-stage gearboxes using deep learning techniques has gained significant attention in the last decade. Spur gearboxes are widely used for power transmission in industries which makes the fault diagnosis study of spur gearboxes very crucial.

Purpose

A significant amount of work has been done to diagnose faults in spur gearboxes at higher loads and high speeds. However, fewer studies are performed to diagnose faults in spur gearboxes running at comparatively low speeds and lower loads. In this paper, fault diagnosis of spur gearbox for faults with different degrees of severity has been performed at relatively low speeds and loading conditions.

Methods

For this purpose, an experimental test rig is designed to collect vibration signals of the spur gearbox at speeds varying from 180 to 420 rpm and load varying from 0 to 30 kg. The vibration signals are collected for three pinions with different crack lengths of 1, 2, and 3 mm in the tooth root, two pinions with different degrees of eccentricity, and for healthy gear.

Results

Particle Swarm Optimization (PSO) method is used for DCNN hyperparameter optimization at low speeds and low loads. Persistence Power Spectrum (PPS) image is a histogram in power–frequency space which allows even less dominant frequencies to be visible in the image. The discriminant features of PPS images are explored using the DCNN model optimized using PSO. Apart from PPS, four other types of images including HHT spectrum and CWT spectrum are also analyzed to classify faults at low speed and low loads.

Conclusions

Among these five methods, the PPS-PSO-DCNN model has shown an accuracy of 97.17% with an image size of 32 pixels. The proposed method is also tested on the publicly available gearbox dataset.