Scholarlink

<p>This paper investigates an 18-degree-of-freedom planetary gearbox test rig under no-load conditions. The dynamic vibration model includes three planetary gears, a sun gear, a ring gear, and a carrier. The main novelty of this study lies in presenting an integrated methodology for early fault detection and failure prevention in planetary gearboxes, with a particular focus on high backlash conditions. Using the Newton–Lagrange approach and MATLAB/Simulink simulations, both normal and excessive backlash scenarios were analyzed, and a novel analytical formulation was derived from experimental vibration data to enhance sensitivity to tooth-backlash anomalies. Experimentally acquired vibration signals were further analyzed using frequency-domain features and classified with machine learning algorithms, including support vector machine (SVM), random forest (RF), and long short-term memory (LSTM) networks. The RF classifier achieved the highest accuracy (98.12%). The proposed framework, which combines dynamic simulation, signal analysis, and machine learning, offers a comprehensive and effective approach for real-time condition monitoring and the prevention of catastrophic gearbox failures.</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

Backlash Fault Diagnosis in Planetary Gearboxes Using Vibration Analysis and Signal Processing: A Comparative Study of SVM, Random Forest, and LSTM for Catastrophic Failure Prediction

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • Ali Hemati,
  • Alireza Shooshtari

摘要

This paper investigates an 18-degree-of-freedom planetary gearbox test rig under no-load conditions. The dynamic vibration model includes three planetary gears, a sun gear, a ring gear, and a carrier. The main novelty of this study lies in presenting an integrated methodology for early fault detection and failure prevention in planetary gearboxes, with a particular focus on high backlash conditions. Using the Newton–Lagrange approach and MATLAB/Simulink simulations, both normal and excessive backlash scenarios were analyzed, and a novel analytical formulation was derived from experimental vibration data to enhance sensitivity to tooth-backlash anomalies. Experimentally acquired vibration signals were further analyzed using frequency-domain features and classified with machine learning algorithms, including support vector machine (SVM), random forest (RF), and long short-term memory (LSTM) networks. The RF classifier achieved the highest accuracy (98.12%). The proposed framework, which combines dynamic simulation, signal analysis, and machine learning, offers a comprehensive and effective approach for real-time condition monitoring and the prevention of catastrophic gearbox failures.