ICACT20240208 Question.1
Questioner: 202212666@mail.sdu.edu.cn 2024-02-06 오전 3:17:04 |
ICACT20240208 Answer.1
Answer by Auhor jjhu@nttu.edu.tw 2024-02-06 오전 3:17:04
Chrome  Click!! |
| The text mentions the use of the PINN method to simulate blood flow and calculate multiple test cases, can you indicate the specific characteristics or scenarios of these test cases and whether the selection of these cases was made with a particular intent? |
The aim of this paper and future work are to develop and use a one-dimensional fluid-dynamical model with the PINNs method to predict blood flow and pressure in the systemic arteries of a human. |
ICACT20240208 Question.2
Questioner: jllovido@bicol-u.edu.ph 2024-02-08 오전 12:19:36 |
ICACT20240208 Answer.2
Answer by Auhor jjhu@nttu.edu.tw 2024-02-08 오전 12:19:36
Chrome Click!! |
| Are there any potential limitations or areas for further improvement identified with the proposed PINN method? |
We believe that the potential limitations and areas for further improvement with physics-informed neural networks are:
Generalization: PINNs can sometimes overfit to specific training data and fail to generalize to new scenarios.
High computational cost: Training PINNs typically requires many significant computing resources. Further work on optimization methods could help address this.
Robustness: It's not clear yet how sensitive PINNs are to noise or uncertainty in the data and governing equations.
Complex physics: Current PINNs have focused mainly on relatively simple physical systems described by partial differential equations. Extending to more complex physics with coupled multi-physics systems remains an open challenge. |
