Structured learning-based sinusoidal modelling for gear diagnosis and prognosis

Mengqiu Tao, Wenyi Wang, Zhihong Man, Zhenwei Cao, Hai Le Vu, Jinchuan Zheng, Antonio Cricenti

Research output: Chapter in Book/Report/Conference proceedingConference PaperOther


In this paper, a structured learning based sinusoidal modelling approach is developed for the prognostics of gear tooth cracking process. According to the vibration signal properties, a learning structure is firstly proposed for gear mesh vibration signal modelling. The learning model is deigned to be a representation of the vibration signal, including all harmonic components generated by the normal gear operation, the residual periodic signal affected by the system load and equipment wear, and the residual non-periodic part induced by gear tooth crack and system disturbances. Motivated by neural network learning, the signal model parameters are optimized by the error backpropagation process, using stochastic gradient descent. The proposed structured learning approach is applied to model the gear tooth cracking process using a set of rig test data. The excellent performance of this approach shows that different parts of the vibration signal are decomposed, and the gear tooth cracking development can be successfully monitored.

Original languageEnglish
Title of host publicationProceedings of the 7th International Conference on Fracture Fatigue and Wear, FFW 2018
EditorsMagd Abdel Wahab
PublisherPleiades Publishing Ltd
Number of pages10
ISBN (Print)9789811304101
Publication statusPublished - 1 Jan 2019
EventInternational Conference on Fracture Fatigue and Wear 2018 - Ghent, Belgium
Duration: 9 Jul 201810 Jul 2018
Conference number: 7th

Publication series

NameLecture Notes in Mechanical Engineering
ISSN (Print)2195-4356
ISSN (Electronic)2195-4364


ConferenceInternational Conference on Fracture Fatigue and Wear 2018
Abbreviated titleFFW 2018
Internet address


  • Gear fault diagnosis
  • Sinusoidal modelling
  • Structured learning

Cite this