Related Experiment Videos
Physics and artificial intelligence collaborative evaluation-prediction framework for rapid underwater acoustic
Jiawei Jiang1,2, Jianxin Wu2,3, Peng Qian4
1Institute of Marine Science and Technology, Shandong University, Qingdao 266237, China.
The Journal of the Acoustical Society of America
|August 11, 2026
Summary
This study introduces a novel physics-AI framework for reliable underwater acoustic field prediction. It enhances accuracy and reduces computation time by intelligently routing tasks between AI and physical models.
Area of Science:
- Ocean acoustics
- Artificial intelligence applications
- Computational physics
Background:
- Artificial intelligence (AI) shows promise for low-cost acoustic field prediction.
- Existing AI methods lack reliability and confidence evaluation in diverse marine environments.
- This limits practical engineering applications of AI in underwater acoustics.
Purpose of the Study:
- To develop a robust framework for reliable AI-driven acoustic field prediction.
- To enhance the practical applicability of AI in underwater acoustic engineering.
- To improve prediction accuracy and reduce computational load.
Main Methods:
- A physics-AI collaborative evaluation-prediction framework was proposed.
- An AI prediction quality evaluation module was developed for confidence assessment and task routing.
- Principal component analysis was used for dimensionality reduction of acoustic fields.
Main Results:
- The evaluation model achieved 94.38% precision in identifying acceptable samples.
- Acceptable predictions had an average RMSE of 2.614 dB, while unacceptable ones had 4.508 dB.
- Overall computation time was reduced by over 50%.
Conclusions:
- The proposed framework ensures accuracy by routing unacceptable predictions to physical models.
- The system effectively balances prediction speed and reliability for underwater acoustic analysis.
- This approach offers a viable paradigm for practical AI implementation in underwater acoustics with minimal fine-tuning.
