I am a lecturer in the School of Physics, Engineering and Computer Science (SPECS). I hold a PhD in Computer Science from the University of Nottingham and a master's degree in Mechanical Engineering Design from the Indian Institute of Technology Madras. My interests lie at the intersection of mechanical engineering and computer science with a focus on artificial intelligence (metaheuristics and machine learning), structural topology optimisation, and acoustics of porous materials. 

Keywords:

Metaheuristics, Acoustics, Topology optimisation, Structural Dynamics, Artificial intelligence

Summary

Structural Topology Optimisation methods are the current state-of-the-art for designing shapes of lightweight and performant mechanical parts and structures. Given a fixed amount of material, these methods can identify near-optimal shapes or layouts that maximise the structural strength. The resulting intricate optimal shapes can be fabricated using additive manufacturing techniques. As an example, using topology-optimised lightweight parts in Airbus A380 has resulted in a 1000 kg weight reduction per aircraft. With an enormous potential weight and cost savings, there is an immense benefit to improving these techniques and their applicability to various domains. My interests are focused on making topology optimisation methods more efficient in terms of computational resources requirement using artificial intelligence methods and bridging the gaps in terms of the manufacturability of optimised shapes.

Metaheuristic techniques such as genetic algorithms and covariance matrix adaptation evolution strategy have contributed to tackling challenging optimisation problems that cannot be solved in practicable times using exact methods. Using a combination of gradient-free heuristic and conventional gradient methods like solid isotropic material with penalisation (SIMP) has resulted in better Pareto-optimal shapes by avoiding getting stuck at local optimal solutions within the same computational budget (https://doi.org/10.1121/10.0019455). Future interests include better-understanding the fitness landscapes in different problem domains and devising domain-specific strategies.

A particular application area of my interest is the use of topology optimisation in passive noise control. By introducing cavities of air pockets within sound-absorbing foams, one can create favourable resonances that improve sound absorption in low frequencies, which are particularly hard to attenuate.

Module support

Semester A (Oct-Dec)

4ENT1163 Applied Design (CDIO) 

7ENT1145 Data Analytics and Artificial Intelligence 

Semester B (Jan-April)

4ENT1165 Programming for Engineers

5ENT1125 Industrial Mechanics (CDIO)

6ENT1159 Acoustics

Teaching Philosophy: Nurture > Nature

Given enough time and proper support, anyone can become a genius at any topic. The Polgar sisters' experiment is a testament to this. 

Game perspective

Conventional learning methods such as books, lectures and tutorials cannot be replaced in providing formal learning. However, with the ever-growing forest of human knowledge, these learning paradigms would limit one's ability to master a wide range of topics in a given lifetime. In this regard, unconventional learning methods such as physical demonstrations, interactive demos or games can reduce the learning time and impart a more profound intuition.

One of my specialisms is developing interactive web demos to help students understand concepts quickly and deeply. Examples of my work include:

1. Genetic Algorithm tutorial: A simple interactive web app where one can understand the concept behind genetic algorithms in 5 minutes or less.

2. Damping ratio demo: An interactive web app that helps gain an intuition for the concept of damping ratios in vibrating spring-mass systems. 

3. Co-variance Matrix Adaptation Evolution Strategy: A demo of the CMA-ES optimisation strategy.