Physics Informed Neural Networks – Exploring the Future of Computational Electromagnetics

Computational electromagnetics (CEM) is a field of research devoted to simulating the complicated electromagnetic systems foundational in our technology, security, and exploration. Historically, problems have been solved by breaking them into many small pieces that can be solved by a computer; an approach that makes it difficult to solve very complicated problems. Today, a new approach to solving CEM problems has emerged in the form of artificial intelligence, namely, Physics Informed Neural Networks (PINNs). These techniques leverage the hardware and techniques supporting advanced machine learning to “learn” the solution to CEM problems.

PINNs are so new that their capabilities are not entirely known. We seek to learn the state-of-the-art in PINNs, and will attempt to apply them to a relatively simple problem, but important, problem of electromagnetic target dataset creation. CEMWorks, the partner organization, will benefit from creating a foundational understanding of PINNs upon which to build future innovation in their software.

Faculty Supervisor:

Ian Jeffrey

Student:

Partner:

CEMWorks Inc

Discipline:

Engineering

Sector:

Information and cultural industries; Professional, scientific and technical services

University:

University of Manitoba

Program: