LSTM-ANN-GA A HYBRID DEEP LEARNING MODEL FOR PREDICTIVE MAINTENANCE OF INDUSTRIAL EQUIPEMENT
Predictive maintenance is essential for ensuring the reliability of industrial equipment and minimizing maintenance costs. However, current predictive algorithms sometimes reach their limits in terms of accuracy, necessitating continuous improvement. The fusion of multiple algorithms can potentially...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
University of Kragujevac
2025-06-01
|
| Series: | Proceedings on Engineering Sciences |
| Subjects: | |
| Online Access: | https://pesjournal.net/journal/v7-n2/14.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Predictive maintenance is essential for ensuring the reliability of industrial equipment and minimizing maintenance costs. However, current predictive algorithms sometimes reach their limits in terms of accuracy, necessitating continuous improvement. The fusion of multiple algorithms can potentially enhance model performance. This research proposes a hybrid deep learning model to detect the degradation of industrial equipment and predict their future health status. Unlike traditional predictive approaches based on existing data for fault detection, which rely on a single deep model and often struggle to maintain satisfactory generalization performance across various forecasting scenarios, these approaches encounter difficulties in effectively initializing and optimizing reset parameters, impeding performance and accuracy. The proposed hybrid model incorporates two deep learning architectures: long short-term memory (LSTM) and artificial neural networks (ANN), with a genetic algorithm (GA) applied as an optimization method to simultaneously optimize the parameters of the model structure. The efficiency of the LSTM-ANN-GA hybrid model is evaluated on a real-life dataset to predict engine-bearing failures. The results show that this model significantly outperforms traditional predictive maintenance methods, achieving high prediction accuracy. |
|---|---|
| ISSN: | 2620-2832 2683-4111 |