FYP-ML: Machine Learning Part Classification System

This project was developed with the help of AI tools, using them as a guide, assistant, and for significant code generation. All final creative decisions and deployments reflect my own work and intentions.

Overview

This repository contains the code and documentation for the final year project (FYP) titled Machine Learning Part Classification System. The project integrates machine learning techniques with industrial automation to classify parts efficiently, leveraging both software and hardware components.

Project Goals

The main objectives of this project include:

• Integration of Machine Learning with Industrial Automation: Utilizing TensorFlow and Keras to develop a part classification model.
• Real-time Part Classification: Deploying the model in an industrial setting using a Siemens PLC and a Raspberry Pi.
• Data Collection and Preprocessing: Gathering and preprocessing data from various sensors for model training and validation.
• System Implementation: Combining the model with a PLC-controlled system to automate the classification process.

Key Components

1. Machine Learning Model

• Frameworks: TensorFlow, Keras
• Task: Part classification using image data.
• Performance: Optimized for high accuracy in a real-time industrial setting.

2. Industrial Automation Integration

• PLC: Siemens PLC for controlling the automation process.
• Raspberry Pi: For deploying the ML model and handling communication between the PLC and the model.
• Communication Protocols: Use of MQTT and Modbus for communication between devices.

3. Data Pipeline

• Data Collection: Using cameras and sensors integrated into the automation line.
• Preprocessing: Image processing techniques such as normalization, resizing, and augmentation.
• Training: Supervised learning on labeled datasets.

Setup Instructions

Prerequisites

• Python
• TensorFlow and Keras
• Siemens TIA Portal (for PLC programming)
• Raspberry Pi with Raspbian OS
• MQTT Broker (e.g., Mosquitto)

Installation

1. Clone the repository:

   git clone https://github.com/cmac-ire/fyp-ml.git
   cd fyp-ml

2. Install Python dependencies:

   pip install -r requirements.txt

3. Setup Raspberry Pi:

   • Ensure Raspbian OS is installed.
   • Install necessary libraries and setup communication protocols.

4. PLC Programming:

   • Use Siemens TIA Portal to program the PLC according to the provided logic.

5. Model Deployment:

   • Train the model using provided datasets.
   • Deploy the model to the Raspberry Pi for real-time classification.

Usage

1. Data Collection:

   • Ensure the system is connected to the sensors/cameras.
   • Run the data collection script to gather images of parts.

2. Training the Model:

   • Preprocess the collected data.
   • Train the model using the fyp.py script.
   • Save the trained model for deployment.

3. Deploy and Run:

   • Deploy the model on the Raspberry Pi.
   • Start the system and monitor the real-time classification through the PLC interface.

Results

The project achieved an impressive 99% accuracy in part classification.

Future Work

• Enhancements: Implementing advanced ML techniques like deep learning for improved accuracy.
• Expansion: Extending the system to handle different types of parts and materials.
• Optimization: Reducing latency and increasing processing speed for higher throughput.

License

This project is licensed under the MIT License.

Contact

For any queries or collaboration opportunities, please contact Cormac Farrelly at cfarr311y@gmail.com