Real-time Topology Optimization in 3D via Deep Transfer Learning

Introduction

This project provides a graphical user interface (GUI) for Real-time Topology Optimization in 3D via Deep Transfer Learning that performs topology optimization using transfer learning models. The application allows users to select from pre-trained models and apply them to predict structures efficiently. The pre-trained models can be downloaded from here.

Table of Contents

• Introduction
• Features
• Installation
• Usage
• Configuration
• Examples
• Troubleshooting
• Contributors
• License

Features

• User-friendly GUI: Built using PyQt5 for an intuitive user experience.
• Pre-trained Models: Includes multiple pre-trained models for topology optimization.
• Visualization: Utilizes Matplotlib and VTK for 2D and 3D visualizations.
• Interactive Controls: Provides sliders, buttons, and forms for easy interaction and customization.

Installation

Prerequisites

• Python 3.x
• PyQt5
• Keras
• NumPy
• Matplotlib
• SciPy
• VTK

Steps

1. Clone the repository:

   git clone https://github.com/MohammadBeh/TO-TL-GUI.git
   cd TO-TL-GUI

2. Install the required packages:

   pip install -r requirements.txt

3. Ensure the pre-trained models are in the same directory or update the paths in the script.

Usage

1. Run the application:

   python main.py

2. Use the GUI to select the desired pre-trained model and configure the input parameters.

3. Visualize the results using the built-in plotting tools.

Configuration

The paths to the pre-trained models are hardcoded in the script. Ensure the models are available in the specified paths or update the script accordingly.

Examples

To visualize a sample optimization process:

1. Select the TL method.

2. Specify the dimension, domain, resolution, and boundary condition.

3. Click OK.

4. Change the force parameters to generate the structures interactively.

   Note: Changing the parameters (except force parameters) requires clicking on Reset button. After that you can repeat the steps 1 to 4.

   Note 2: In the method, you can select TL + SIMP which runs a few iteration of SIMP solver on the predicted structure to obtain high quality and well connected structures (Only works for 2D structures).

Troubleshooting

• Ensure all dependencies are installed and properly configured.
• Verify the paths to the pre-trained models.
• Check the console output for any error messages.

Contributor

• Mohammad Behzadi - MohammadBeh

License

This project is licensed under the MIT License. See the LICENSE file for details.