🔌 Predictive Maintenance System for Diesel Generator

A machine learning-based solution for predicting maintenance needs and monitoring diesel generator parameters in real-time.

📊 Project Overview

This project implements a predictive maintenance system for diesel generators using machine learning algorithms to analyze time series data. The system monitors critical parameters such as:

1. Coolant Temperature
2. Oil Pressure
3. Vibration Levels
4. Exhaust Gas Temperature
5. Fuel Consumption Rate
6. Battery Voltage
7. Engine Speed (RPM)
8. Load Current
9. Frequency (Hz)
10. Running Hours

The machine learning model predicts potential issues like the ones listed above, before they cause failures, reducing downtime and maintenance costs.

🎯 Key Features

• Machine learning models for predictive analytics
• Real-time parameter monitoring via MQTT protocol
• Web-based dashboard for parameter visualization
• Comparative analysis of multiple ML algorithms

🧠 Machine Learning Models

The system evaluates multiple algorithms to find the optimal predictive model:

• Linear Regression
• Logistic Regression
• Decision Trees
• Random Forest Regression (achieved 90.56% accuracy)
• Support Vector Machines

🔧 Technical Implementation

Files

• training.py - Code for training the machine learning models
• predicting.py - Implementation of prediction functionality
• data.csv - Historical training dataset
• inputdata.csv - Input data format for predictions
• test.html - Web dashboard for real-time monitoring

Data Flow

1. Data collection from diesel generator sensors
2. Data preprocessing and feature engineering
3. Model training and evaluation
4. Real-time data transmission via MQTT
5. Parameter visualization on web dashboard

📈 Results

• Best Performing Model: Random Forest Regression
• Accuracy: 90.56%
• Key Parameters Monitored:
  • Voltage
  • Current
  • Fuel levels
  • Engine temperature
  • Operational hours

🚀 Getting Started

Prerequisites

Python 3.7+
pandas
scikit-learn
seaborn
paho-mqtt (for MQTT communication)
matplotlib (for visualization)

Installation

1. Clone this repository

git clone https://github.com/yourusername/diesel-generator-predictive-maintenance.git

2. Install required packages

pip install -r requirements.txt

3. Run the training script

python training.py

4. Set up prediction system

python predicting.py

🔮 Future Work

Data Engineering & Big Data Integration

• Real-time Streaming Architecture:

  • Implement Apache Kafka for high-throughput, fault-tolerant data streaming
  • Build real-time analytics with Kafka Streams or Apache Flink

• Workflow Orchestration:

  • Migrate to Apache Airflow for robust pipeline scheduling and monitoring
  • Implement DAGs for complex maintenance prediction workflows

• Big Data Processing:

  • Scale to distributed computing with Apache Spark for handling fleet-wide generator data
  • Implement batch processing with Hadoop ecosystem for historical analysis

• Data Warehousing & Storage:

  • Implement Snowflake data warehouse for flexible scaling and analytics
  • Utilize AWS S3 for cost-effective long-term storage of sensor data

• Cloud Infrastructure:

  • Migrate to AWS cloud infrastructure (EC2, Lambda, SageMaker)
  • Implement containerization with Docker and Kubernetes for deployments

• Advanced Analytics:

  • Develop a data lake architecture for combining structured and unstructured maintenance data
  • Implement dbt (data build tool) for analytics engineering and transformation

Demo

Dashboard Visualization

Real-time monitoring dashboard showing generator parameters

Model Performance Visualization

Comparison of different ML algorithms with Random Forest achieving 90.56% accuracy

Video Demonstration

Click to watch the demo video of the system in action

📝 Project Information

• Development Period: August 2023 - May 2024
• Institution: SSN College of Engineering
• Contributors: [Sriram Vivek , Sethuram Gautham Rajakumar]

📫 Contact

Feel free to reach out if you have any questions or would like to collaborate!

• Email: [email protected]
• LinkedIn: https://www.linkedin.com/in/sriram-vivek-58a673269/

---

This project was developed as part of academic curriculum at SSN College of Engineering.