AIStructDynSolve 1.11

AI-powered framework to solve both forward and inverse problems in structural dynamics

AIStructDynSolve

Author:

• Ke Du(杜轲)
  Email: [e-mail]

AIStructDynSolve is an artificial intelligence (AI) powered framework designed to solve both forward and inverse problems in structural dynamics. It leverages advanced artificial intelligence methods - particularly physics-informed neural networks (PINNs), Physics-Informed Kolmogorov-Arnold Network(PIKANs) and their extensions - to model, predict, and analyze dynamic structural responses under various loading scenarios, such as seismic excitations.

The framework solves the following ODE of MDOF:

M*U_dotdot+C*U_dot+K*U=Pt

• Initial Conditions:
• U(t=0)=InitialU
• U_dot(t=0)=InitialU_dot

The framework aims to:

• Accurately simulate time-dependent structural behavior (forward problems).
• Identify structural parameters or input forces from measured responses (inverse problems).
• Incorporate domain knowledge and physical laws for improved generalization and interpretability.
• Address challenges in multi-frequency, multi-scale dynamics, especially in earthquake engineering applications.

Date:

• 2023/12/26

Update Log

• Version 1.11 - Jul 23, 2025 Added support for trainable_params ("mass" or "stiffness" or "damping"or "pt" or "eq" or "eqX" or "eqY" or "eqZ") for inverse problems.

• Version 1.10 - May 11, 2025 Improved accuracy and stability of PIKANs

• Version 1.9 - Apr 30, 2025 Added support for EQ2D and EQ3D of structural dynamic simulations, enabling users to analyze 2D and 3D earthquake loading scenarios

• Version 1.8 - Apr 12, 2025 Improved accuracy of high-frequency response prediction

• Version 1.7 - Apr 11, 2025 Minor bug fixes

• Version 1.6 - Apr 7, 2025 Minor bug fixes

• Version 1.5 - Jan 13, 2025 Minor bug fixes

• Version 1.4 - Dec 31, 2024 Minor bug fixes

• Version 1.3 - Dec 19, 2024 Minor bug fixes

• Version 1.2 - Dec 18, 2024 Minor bug fixes

• Version 1.1 - Dec 7, 2024 Minor bug fixes