Vibration-PINNs

This repository presents a series of analysis on the performance of Physics-Informed Neural Networks in vibrational systems. The limitation of PINNs in learning such systems is discussed and the potential solutions are investigated. Specifically, the following topics are covered:

• Free Vibration of a damped translational System (mass-spring-damper)
• Effect of system parameters' magnitude on the performance of PINNs (to be added)
• PINNs shortfalls in learning long temporal domains and the potential solutions (to be added)

Codes are developed using JAX library (https://github.com/google/jax)

Training PINNs on damped vibration system with t_end = 1 s

Training PINNs on damped vibration system with t_end = 2 s

Training PINNs on damped vibration system with t_end = 3 s

Files

This repository contains the following files:

• README.md: The file you are currently reading.
• vib_pinn.ipynb: The code for training PINN on mass-spring-damper system.

Licence

This repository is released under the mit licence, which can be found in the LICENSE file.

Related resources

• https://github.com/benmoseley/harmonic-oscillator-pinn

If you find any issues or errors in the tutorial materials or have suggestions for improvements, please feel free to create an issue or pull request.