Efficient predictive models using physics-informed machine learning

Overview

This research explores how advanced physics-informed neural network models can guide the development of simplified yet accurate predictive systems across scientific and engineering domains. The work spans machine learning, computational physics, and applied mathematics, addressing the critical challenge of creating efficient models that maintain physical consistency and predictive reliability.

Recent advances in neural operator learning and physics-informed architectures have demonstrated potential for dramatically reducing model complexity while preserving domain-specific knowledge. This research investigates generalisable frameworks for developing simplified predictive models that leverage physics constraints as guiding principles rather than computational burdens.

Research activities

In this project you will:

• design simplified architectures guided by physics-informed models
• develop frameworks for reducing model complexity without sacrificing physical accuracy
• and create novel training strategies that embed domain knowledge into lightweight models
• implement scalable, physics-informed approaches for efficient and accurate predictive modelling.

You will also conduct a literature review and prepare technical writing.

Outcomes

The research aims to develop generalisable frameworks for simplified physics-informed predictive modelling, achieving significant efficiency gains without compromising accuracy. It seeks to establish new paradigms for accessible, scientifically grounded machine learning, enabling wider adoption in resource-constrained settings.

If sufficient novelty is achieved, the work holds strong potential for publication in reputable scientific journals or conferences.

Skills and experience

You will need:

• a strong quantitative and mathematical background
• programming skills in Python and machine learning
• a basic understanding of physics or engineering
• an interest in interdisciplinary research.

It would be preferred if you also have:

• a background in computational science or applied mathematics
• experience in predictive modelling or data analysis
• familiarity with physics-informed methods.

Scholarships

You may be eligible to apply for a research scholarship.

Explore our research scholarships

Keywords

• physics informed machine learning
• model simplification
• predictive modelling optimisation

Contact

Contact the supervisor for more information.