Home Artificial Intelligence Physics-Informed Neural Networks: An Application-Centric Guide

Physics-Informed Neural Networks: An Application-Centric Guide

Physics-Informed Neural Networks: An Application-Centric Guide

A comprehensive overview of PINN’s real-world success stories

Towards Data Science
Image generated by DALL-E.

With regards to applying machine learning to physical system modeling, it’s an increasing number of common to see practitioners moving away from a pure data-driven strategy, and beginning to embrace a hybrid mindset, where wealthy prior physical knowledge (e.g., governing differential equations) is used along with the information to enhance the model training.

Under this background, physics-informed neural networks (PINNs) have emerged as a flexible concept and led to many success stories in effectively solving real-world challenges.

As a practitioner who’s desirous to adopt PINNs, I’m keen on learning each the most recent developments in training algorithms, in addition to the novel use cases of PINNs for real-world applications. Nonetheless, a pain point I often see is that, although there are abundant research papers/blogs summarizing effective PINN algorithms, overviews of novel use cases of PINNs can rarely be found. One obvious reason is that, unlike the training algorithms that are domain-agnostic, reports of PINN use cases are scattered in various engineering domains and never readily accessible for a practitioner who is frequently an authority in a single specific domain. As a consequence, I often found myself reinventing the wheel as my ways of using PINNs have already been well addressed by practitioners in one other field.

It is precisely my journey and experiences which have sparked the concept of writing this blog: here, I strive to interrupt the data barrier across different engineering domains and distill the recurring functional usage patterns of PINNs. I hope that this review will inform practitioners from different domains about what’s possible with PINNs and encourage latest ideas for interdisciplinary innovation.

Toward that end, I even have extensively reviewed PINN research papers previously three years and got here up with the next 5 primary usage categories:

  • Predictive modeling and simulations
  • Optimization
  • Data-driven insights
  • Data-driven enhancement
  • Monitoring, diagnostic, and health assessment


Please enter your comment!
Please enter your name here