Skip to Content
e

Mr Amir Alizadeh
PhD Student

Summary

I am a final-year PhD candidate at the Brunel Centre for Advanced Solidification Technology (BCAST) at Brunel University London. My research falls under the wider Materials Made Smarter Research Centre (MMSRC) programme and is conducted in collaboration with the University of Cambridge. My work sits at the critical intersection of data-driven machine learning and fundamental physics, where I formulate and solve complex multi-physics optimization problems.

Research Focus: Physics-Informed Machine Learning

My primary research investigates the integration of physics-based insights with experimental data to model precipitation hardening (PH) in 6xxx series aluminium alloys. Purely data-driven models often struggle with data scarcity and a lack of physical interpretability, while traditional physics-based approaches require complex, computationally expensive parameter calibration.

To overcome these limitations, I develop frameworks that embed physical laws directly into optimization and machine learning architectures. Key contributions of my current work include:

  • Developing gradient-based optimization frameworks that utilize Adaptive moment estimation (Adam) and smoothed approximations of conditional logic to ensure physically plausible parameter generation.

  • Implementing inverse-problem Physics-Informed Neural Networks (PINNs) for systems governed by partial differential equations (PDEs), enabling robust parameter estimation even under data-scarce conditions.

  • Automating the manual calibration of complex metallurgical mechanisms—like diffusion, nucleation, and coarsening—using backpropagation and dimensionality reduction.

Academic Background Before transitioning to computational metallurgy, I earned my M.Sc. in Telecommunications Engineering (Distinction) from Ferdowsi University and my B.Sc. in Electronics (First Class Honours) from the University of Guilan. This multidisciplinary foundation in signal processing, control theory, and probabilistic optimization deeply informs my current approach to complex, non-linear physical systems.

When I am not writing modular Python implementations or exploring advanced deep learning paradigms, you can usually find me out hiking or running far away from my computer.