Prediction, classification and diagnosis of spur gear conditions using artificial neural network and acoustic emission
The gear system is a critical component in the machinery and predicting the performance of a gear system is an important function. Unpredictable failures of a gear system can cause serious threats to human life, and have large scale economic effects. It is necessary to inspect gear teeth periodicall...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/79590/1/YasirHassanAliPFKM2017.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The gear system is a critical component in the machinery and predicting the performance of a gear system is an important function. Unpredictable failures of a gear system can cause serious threats to human life, and have large scale economic effects. It is necessary to inspect gear teeth periodically to identify crack propagation and, other damages at the earliest. This study has two main objectives. Firstly, the research predicted and classified specific film thickness (λ) of spur gear by Artificial Neural Network (ANN) and Regression models. Parameters such as acoustic emission (AE), temperature and specific film thickness (λ) data were extracted from works of other researchers. The acoustic emission signals and temperature were used as input to ANN and Regression models, while (λ) was the output of the models. Second objective is to use the third generation ANN (Spiking Neural Network) for fault diagnosis and classification of spur gear based on AE signal. For this purpose, a test rig was built with several gear faults. The AE signal was processed through preprocessing, features extraction and selection methods before the developed ANN diagnosis and classification model were built. These processes were meant to improve the accuracy of diagnosis system based on information or features fed into the model. This research investigated the possibility of improving accuracy of spur gear condition monitoring and fault diagnoses by using Feed-Forward Back- Propagation Neural Networks (FFBP), Elman Network (EN), Regression Model and Spiking Neural Network (SNN). The findings showed that use of specific film thickness has resulted in the FFBP network being able to provide 99.9% classification accuracy, while regression and multiple regression models attained 73.3 % and 81.2% classification accuracy respectively. For gear fault diagnosis, the SNN achieved nearly 97% accuracy in its diagnosis. Finally, the methods use in the study have proven to have high accuracy and can be used as tools for prediction, classification and fault diagnosis in spur gear. |
|---|