Collaboration with a PhD researcher in structural engineering at Nottingham Trent University, UK. Project investigated the optimum design of three-dimensional steel frames in terms of real-world cost function (weight, embodied carbon). Imposed behavioral constraints were based on EC3 specifications.
Since the problem is hard to formulate in YALMIP, metaheuristics were considered. Fully functional optimization algorithms interfaced:
- Genetic Algorithm (Matlab built-in)
- Particle Swarm Optimization (Matlab built-in)
- Harmony Search Algorithm (custom augmentationy)
- Multi-Objective GA (Matlab built-in)
- Multi-Objective PSO (custom)
Elements of speeding up the code (precomputing, inlining, memory allocations) and metaprogramming - dynamic creation in buildGroup. Possible extension of program editing it's code and compiling it - for hopefully faster execution instead of evaluating the anonymous function.