This repo represents the work that I did in my final year project. This project was on utilising Multi-Objective Bayesian Optimisation for a mechanical design. This overview will undisclose any detailed information of the mechanical component, it serves as a guideline for the underlying algorithms used for optimisation.
The whole idea of this project was to test and optimise a mechanical component for crash instances. Usually, this is done through parametric study which sequentially loops over every combination of parameters to find an optimal point that gives a max/min output response. In this project, we used a sampling algorithm that enables us to predict the relationship of output response vs parameter combination. This serves as a new way to craft a Design of Experiment. This project was done by incorporating a CAD (ntopology), an FEA (abaqus), and Python
To read more on Multi-Objective Bayesian Optimisation architecture, click on link below. This is one of the libraries that was used.
| filename | Description |
|---|---|
| MOBO_ANN.py | algorithm used to run Multi-Objective Bayesian Optimisation |
| abaqus | all scripts used to run simulation on CAD files generated on ntopology |
| ntopology | scripts used to generate CAD models based on inital samples, and new parameters to study |