Current Projects

Hybrid Deep Learning–FEA Technique for Fracture Prediction

An EPSRC-funded project combining deep learning with finite element analysis to predict fracture in advanced materials. This project builds on the open-source FEA, Akantu, for phase field damage simulations and develops a data+physics-driven simulation framework for fracture evolution. Collaborations with EPFL and the University of Florida.

Phase-field damage field (from finite element simulation) vs. graph neural network (GNN) prediction.

Training of LSTM model (top-left), Stress history prediction (top-right), Correlation plot (bottom-left) and computational time savings (bottom right).

Data-Driven Crystal Plasticity Modeling for Finite Element Simulations

A school-funded project combines deep learning with finite element analysis to develop constitutive models for crystal plasticity simulations. A time-consuming material solver at integration points in a finite element simulation is replaced by a data-driven deep learning model (based on an LSTM architecture).

Multifunctional Composites for Electromagnetic Absorption

This research focuses on developing novel designs of hierarchical multiscale and multiphase composites for combined load-bearing and electromagnetic absorption. An integrated experimental-numerical approach is adopted to realise analysis and design tools for optimisation of such multifunctional material/structural systems.

AI-Driven Damage Detection in Composite Structures

Using deep learning for real-time health monitoring, damage identification, and quantification in aerospace composite structures. (Supported by IMechE Astridge fund and George Daniels Scholarship).

Workflow in feature-engineered sensor data-driven damage detection

Compressor Blade Design Using Generative AI Tools

CERES Industrial consortium-funded project to develop intelligent design techniques for compressor systems using generative AI. The figure shows generative deep learning-derived rotor profiles. An in-house generated dataset of varied profiles was used to train the generative adversarial network (GAN) network while enforcing geometric constraints.

Active Control of Helicoptor Rotor Blades using Deep Learning and Magneto-electro-elastic (MEE) Composites

Funded by the Royal Society- Newton Fellowship, this project focuses on utilising multifunctional materials as actuators for active control of flexible structures through digital twin and deep learning frameworks.

A sandwich composite plate with MEE (functional) facesheets for active control of the vibrations

Transnational Education: Strategy Roadmap for Sustainable Development and Skill Enhancement for Industry 5.0

Funded by the British Council, the project aims to develop a strategic roadmap for curriculum innovation driven by future industry requirements centralized on digitalization and AI in aerospace and mechanical engineering.

Google Sites

Report abuse