DPAI: A Data-driven simulation-assisted-Physics learned AI model for transient ultrasonic wave propagation.

Area of Science:

• Computational physics
• Artificial intelligence in engineering
• Wave propagation modeling

Background:

• Simulating transient ultrasonic wave propagation is crucial for various engineering applications.
• Traditional methods like finite element (FE) modeling can be computationally intensive.
• Developing efficient and accurate simulation techniques is an ongoing challenge.

Purpose of the Study:

• To propose a novel deep neural network model, the Data-driven-simulation-assisted-Physics learned AI (DPAI), for simulating 2D transient ultrasonic wave propagation.
• To leverage a data-driven approach combined with physics principles for enhanced simulation accuracy.
• To demonstrate the model's effectiveness across diverse elastodynamic scenarios.

Main Methods:

• Implementation of a modified convolutional long short-term memory (ConvLSTM) network with an encoder-decoder structure.
• Training the DPAI model using simulation-assisted finite element datasets with various excitation sources.
• Utilizing a data-driven approach to learn the underlying physics of elastic wave propagation.

Main Results:

• The DPAI model successfully simulates transient ultrasonic wave propagation in a 2D domain.
• Demonstrated effectiveness in modeling scenarios with multiple point sources and varying excitation parameters.
• The DPAI model shows comparable accuracy to FE modeling with significantly reduced computational time.

Conclusions:

• The proposed DPAI model offers an efficient and accurate method for simulating ultrasonic wave propagation.
• This AI-driven approach provides a promising alternative to conventional numerical methods for complex wave phenomena.
• The DPAI model has broad applicability in elastodynamic physics and related fields.