Predictive Maintenance for Oil & Gas

Overview

This project predicts equipment failures in the oil & gas industry using machine learning models (Logistic Regression, Random Forest). It analyzes sensor data (Temperature, Pressure, Vibration, Humidity, Flow Rate) and provides an interactive dashboard for monitoring failure risks.

Features

• Preprocessing & Cleaning: Handles missing values and normalizes sensor data.
• Machine Learning Models: Predict failures using Logistic Regression & Random Forest.
• Failure Risk Visualizations: Heatmaps, sensor trend graphs, and SHAP feature importance.
• Interactive Dashboard: Filter failure trends by equipment type and sensor readings.

Project Structure

Predictive_Maintenance_OilGas/
│-- README.md                 # Project Overview & Instructions
│-- data_preprocessing.py      # Cleans & processes sensor data
│-- ml_model.py                # Trains ML models & evaluates performance
│-- visualization.py           # Generates failure risk graphs
│-- dashboard.py               # Interactive Streamlit dashboard
│-- predictive_maintenance_data.csv  # Sample sensor dataset
│-- requirements.txt           # Dependencies for setup

Installation & Setup

1️⃣ Install Dependencies

Run the following command:

pip install -r requirements.txt

2️⃣ Run Data Preprocessing

python data_preprocessing.py

This cleans and prepares the dataset.

3️⃣ Train the Machine Learning Model

python ml_model.py

This trains Logistic Regression & Random Forest models to predict failures.

4️⃣ Run Data Visualizations

python visualization.py

5️⃣ Launch the Interactive Dashboard

streamlit run dashboard.py

The dashboard provides real-time failure analysis and risk visualization.

Technologies Used

• Python (Pandas, NumPy, Matplotlib, Seaborn, Plotly, Streamlit)
• Machine Learning (Logistic Regression, Random Forest, SHAP)
• Data Processing (Feature Engineering, Normalization, Outlier Detection)
• Database (SQLite for storing structured sensor data)

Author

Charles Eleri

Next Steps

• Enhance model performance using LSTM for time-series predictions.
• Integrate real-time data ingestion from IoT sensors.
• Deploy the dashboard to AWS/GCP/Azure for global monitoring.

---

🔹 GitHub Repo: github.com/charleseleri