Anomaly Detection in Oil Pumps Using Isolation Forest

This project demonstrates a machine learning approach for detecting anomalies in oil pump operations using an Isolation Forest model. The project is designed to identify abnormal behavior in synthetic oil pump data, such as temperature spikes, vibration inconsistencies, and pressure fluctuations.

Key Features:

• Data: The dataset used consists of synthetic sensor readings, including variables such as vibration, temperature, pressure, and flow rate. Anomalous data points were artificially introduced to simulate potential failures.
• Model: The Isolation Forest algorithm was employed to detect anomalies, trained exclusively on normal (non-anomalous) data to simulate a real-world scenario where failures are rare.
• Evaluation: The model was evaluated using common metrics such as accuracy, precision, recall, and the F1-score, showing that it effectively detects anomalies with high accuracy.

Results Summary:

• Accuracy: 93%
• Precision: 100% for normal data, 43% for anomalies
• Recall: 93% for normal data, 100% for anomalies

Purpose:

This project provides a step-by-step guide for applying the Isolation Forest model in the context of oil pump maintenance. It aims to help engineers and data scientists identify early signs of failure in industrial equipment, reducing downtime and maintenance costs.

How to Run:

The provided Jupyter Notebook (anomaly_detection_oil_pumps.ipynb) includes all the necessary code and instructions to replicate the experiment. Simply run the cells to generate synthetic data, train the model, and evaluate its performance.