Aircraft Engine Fault Diagnosis Model Based on 1DCNN-BiLSTM with CBAM
Jiaju Wu1,2, Linggang Kong1, Shijia Kang1
1Institute of Computer Application China Academy of Engineering Physics, Mianyang 621999, China.
Sensors (Basel, Switzerland)
|February 10, 2024
Summary
This study introduces an advanced aircraft engine fault diagnosis model using 1DCNN-BiLSTM with CBAM. The model accurately identifies fault modes directly from raw data, enhancing operational reliability and Remaining Useful Life (RUL) prediction.
Area of Science:
- Aerospace Engineering
- Mechanical Engineering
- Artificial Intelligence
Background:
- Aircraft engine operational status and fault modes evolve over time.
- Degradation trends necessitate improved fault diagnosis methods.
- Existing methods may require complex feature extraction.
Purpose of the Study:
- To propose an effective aircraft engine fault diagnosis model.
- To enable direct diagnosis from raw monitoring data.
- To improve the accuracy and reliability of fault detection.
Main Methods:
- A novel model combining 1D Convolutional Neural Network (1DCNN), Bidirectional Long Short-Term Memory (BiLSTM), and CBAM (Channel and Spatial Attention Mechanism).
- 1DCNN for local feature extraction, CBAM for feature weighting, and BiLSTM for temporal sequence analysis.
- Direct application to raw sensor data, eliminating manual feature engineering.
Main Results:
- The proposed 1DCNN-BiLSTM with CBAM model achieved higher classification accuracy on the NASA CMAPSS dataset.
- Successfully categorized faultless, High-Pressure Compressor (HPC) fault, and HPC & Fan mixed fault conditions.
- Demonstrated superior performance compared to other existing models.
Conclusions:
- The developed model offers a robust and accurate solution for aircraft engine fault diagnosis.
- Its ability to process raw data directly simplifies implementation.
- Significant practical implications for enhancing aircraft engine operational reliability and Remaining Useful Life (RUL) prediction.
