The Future of Solutions: Ansys SAF for Physics-Informed Thermal Field Prediction with Seamless Integration of ML

A technology demonstration leveraging Ansys Solutions Framework capabilities for rapid web application development, and deployment of Physics-Informed Machine Learning (ML) models is presented. It aims to simulate samples of thermal conductivity profiles and perform finite element solutions using PyAnsys ecosystem to generate input and output image pairs. Deep learning models are then trained to provide capabilities of temperature field prediction from unseen thermal conductivity profiles. As this solution is built using Ansys Solutions Framework (Ansys SAF- GLOW), it offers a user-friendly interface for varying user input and output parameters. It also enables rapid post-processing of predictions and validation against benchmarks. The integration of physics-informed machine learning algorithm to this solution results in a potential use case in leveraging Ansys SAF to develop and deploy ML models at scale as a SaaS. The solution extends to controlling features like input data generation parameters, visualization of output images, tweaking ML hyper-parameters, real-time monitoring of loss-function and learning-rate during training, and optimization of parameters. Overall, this integrated ML solution results into a powerful tool and technology demonstration to streamline digital simulation capabilities of various industries and drive significant technological advancements.