METHODOLOGY

The methodology follows a structured hybrid workflow integrating CAD modelling, simulation-based optimization, and AI-driven predictive modelling.

1. Parametric CAD Development

A base fixture model was created with fully parametric dimensions to fit different workpiece sizes. Key features include locating holes, clamping pillars, and a vertical bracket, all constrained for flexibility and machining compatibility.

2. Material Selection

Material properties for Aluminium 6061-T6 were selected for optimization due to its excellent stiffness-to-weight ratio, machinability, and fatigue strength. Structural steel and bronze were considered as references.

3. Static & Modal FEA

ANSYS Workbench simulations evaluated deformation, von Mises stress, and natural frequencies. Fixtures were fixed at locating holes to represent CNC mounting. Loads of 1300–1500 N were applied on clamping regions.

4. Topology Optimization

Design and non-design regions were defined. Up to 50% material reduction was targeted. Material was automatically redistributed around primary stress paths to improve mechanical efficiency.

5. Generative AI-Assisted Design

LLMs assisted in evaluating geometry feasibility, proposing rib patterns, voids, and fillets to improve manufacturability and load flow. Five distinct design iterations were produced.

6. ANN Model Training

A feedforward ANN with 10 hidden neurons was trained in MATLAB using FEA-derived datasets.
Inputs: Load, design ID, fixture mass
Outputs: Deformation, stress, stiffness

Training used the Levenberg–Marquardt algorithm with a 70/15/15 split for training/validation/testing.

7. Validation & Design Selection

ANN predictions were compared to FEA results to verify accuracy. Designs were compared based on deformation, stress, natural frequency, manufacturability, and stiffness-to-mass efficiency.