ET3107 Individual Project and Design III Project Dissertation Brief | City St George's

ET3107 Project Dissertation Brief

Project Overview: Progression from Previous Work to Current Dissertation

1. Work Completed In Previous Year

The previous project focused on analysing structural damage in aerospace composite structures using computational methods. The primary objective was to understand how damage influences the dynamic behaviour of a structure, particularly through changes in modal characteristics.

Key aspects of the previous project included:

• Development of a computational model of a composite aerofoil structure.
• Use of modal analysis to determine natural frequencies and mode shapes.
• Simulation of structural damage by reducing stiffness in selected regions.
• Investigation of how damage severity and location affect modal frequencies.
• Identification of patterns in frequency shifts as indicators of structural damage.- Supporting aerodynamic analysis using CFD to assess performance impact.

The outcome of this work was a strong understanding of the relationship between structural properties, damage, and modal response. This formed the technical foundation for the current project.

2. Proposed Work For Current Year

The current project builds on the previous work by transitioning from analysis to prediction. Instead of repeatedly running simulations to obtain modal characteristics, the aim is to develop a machine learning model that can predict these results instantly.

Overall Aim

To develop a data-driven framework capable of predicting modal frequencies of a structure based on selected material and geometric parameters.

Detailed Objectives

• Define a consistent structural model with fixed geometry and boundary conditions.
• Select key input parameters influencing modal behaviour: Young’s modulus (E), density (ρ), and thickness (t).
• Generate a dataset by running multiple simulations (approximately 80–150 cases).
• Extract the first three natural frequencies from each simulation.
• Organise the data into a structured dataset for machine learning use.
• Train regression-based models (e.g. Linear Regression and Random Forest).
• Validate the models using unseen test data.
• Compare predicted modal frequencies with simulation results.
• Evaluate model accuracy using metrics such as MAE, RMSE, and R².
• Analyse the effectiveness and limitations of the machine learning approach.

Expected Outcome

The project is expected to demonstrate that machine learning models can accurately predict modal frequencies within a defined parameter range, thereby reducing the need for repeated computational simulations and enabling faster engineering analysis.

Key Contribution

This work represents a progression from physics-based simulation to data-driven prediction, highlighting the integration of traditional engineering analysis with modern machine learning techniques.