physics-informed neural network for solving elastodynamics (elasticity) problem

This repo includes the implementation of physics-informed neural networks in paper:

Chengping Rao, Hao Sun and Yang Liu. Physics informed deep learning for computational elastodynamics without labeled data.

• PlateHoleQuarter: Training script and dataset for plate with a hole (stress concentration) problem in Sec 3.1;
• ElasticWaveInfinite: Training script and dataset for elastic wave propagation in infinite domain in Sec 3.2;
• ElasticWaveConfined: Training script and dataset for elastic wave propagation in confined (4 edges fixed) domain in Sec 3.2.
• ElasticWaveSemiInfinite: Training script and dataset for elastic wave propagation in semi-infinite (top is traction-free) domain in Sec 3.2.

Defected plate under cyclic load (top: PINN; bottom: FEM.)

Elastic wave propagation in infinite (unbounded) domain (top: PINN; bottom: FEM.)

Elastic wave propagation in confined domain (top: PINN; bottom: FEM.)

Elastic wave propagation in semi-infinite (half-bounded) domain (top: PINN; bottom: FEM.)

• These implementations were developed and tested on the GPU version of TensorFlow 1.10.0.