Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 58325 (05)
- University: Sharif University of Technology
- Department: Electrical Engineering
- Advisor(s): Hashemi, Matin; Pourmohammad Namvar, Mehrzad
- Abstract:
- Predictive maintenance plays a crucial role in enhancing productivity and reducing operational costs in industries. This study aims to predict failures in an eight-stage compressor, which experiences impeller damage and breakage over time. One of the main challenges was defining the problem based on the available dataset in a way that enables early failure detection. A review of different approaches revealed that, due to the sudden nature of compressor failures, conventional methods based on estimating the remaining useful life (RUL) are not highly effective. Therefore, anomaly detection was selected as the primary approach. Initially, an Autoencoder neural network—one of the most widely used networks for anomaly detection—was employed. However, given the sequential nature of the data, it was combined with a Long Short-Term Memory (LSTM) network for improved performance. Evaluation results demonstrated that the proposed method can accurately detect failures before they occur and provide timely warnings. This research contributes to the development of intelligent predictive maintenance systems and can help optimize the performance of industrial equipment. Finally, recommendations for model improvement and future research directions are presented
- Keywords:
- Predictive Maintenance ; Anomaly Detection ; Neural Network ; Autoencoder ; Autoencoder Neural Networks ; Long Short Term Memory (LSTM) ; Time Series ; Industry 4.0
Digital Object List
-
محتواي کتاب
- view
Bookmark