|
| Location of Abnormal Vibration Source in Helicopter Repair Based on Optical Fiber Sensing Detection |
| |
| View Full Text View/Add Comment Download reader |
| DOI：10.7643/issn.1672-9242.2023.05.018 |
| KeyWord:optical fiber sensing helicopter repair vibration monitoring abnormal vibration source deep learning |
| Author | Institution |
| SUN Xiao-ming |
Nanjing University of Aeronautics and Astronautics, Nanjing , China;Chengdu State-owned Jinjiang Machinery Factory, Chengdu , China |
| SU Shi-wei |
Chengdu State-owned Jinjiang Machinery Factory, Chengdu , China |
| YIN Hua-guang |
Chengdu State-owned Jinjiang Machinery Factory, Chengdu , China |
| WANG Hai-tao |
Nanjing University of Aeronautics and Astronautics, Nanjing , China |
|
| Hits: |
| Download times: |
| Abstract: |
| The work aims to detect the vibration source and analyze the frequency spectrum of the vibration source for the troubleshooting of abnormal helicopter vibration and accurately locate the abnormal vibration source, guide the adjustment of the vibration source and detect the adjustment to improve the efficiency of abnormal helicopter vibration troubleshooting. Combined with the application analysis of the vibration detection characteristics of the optical fiber sensor, a set of abnormal vibration source locating system for helicopter repair based on optical fiber sensing detection was developed to realize the dynamic monitoring of the vibration source of the helicopter. Firstly, the main vibration source of the helicopter was identified, the frequency of the vibration source was analyzed and counted, and the typical vibration faults of the helicopter in test run and test flight were sorted out. Then, the real-time modulation method of optical fiber vibration sensor and the light wave demodulation method subject to signal modulation were studied. Finally, combined with the application demand, the mature distributed optical fiber sensing system available on the market was subject to optimal design and deep learning technology was adopted to redesign the software of the back-end data processing platform. The model accuracy can reach 66.37% after a targeted recognition model library is created and the model library is trained with actual samples. |
| Close |
|
|
|