AI-Based Hydraulic Pump Fault Classification

Inspiration

Hydraulic pumps are widely used in industrial machinery, and unexpected failures can cause downtime and maintenance difficulties. We wanted to explore how sensor data and machine learning could help identify potential hydraulic pump faults at an early stage.

What it does

Our project explores hydraulic pump condition monitoring using vibration and temperature data. The intended system aims to classify pump conditions into three categories: Normal, Abnormal, and Critical. It also explores comparing vibration-only data with combined vibration and temperature data.

How we built it

We developed a prototype hydraulic pump monitoring setup using an MPU6050 accelerometer, a DS18B20 temperature sensor, and an ESP32 microcontroller. The sensors collect vibration and temperature measurements from the pump. The AI/ML model development and training are currently in progress.

Challenges we ran into

One of our key challenges was monitoring pump behaviour through sensor readings and exploring how different operating conditions could be distinguished. Another challenge is determining how vibration and temperature measurements can be combined to support reliable fault classification.

Accomplishments that we're proud of

We are proud of developing a prototype for hydraulic pump monitoring and integrating sensor-based data collection with an ESP32. This provides a foundation for our ongoing work on machine-learning-based fault classification.

What we learned

We learned about integrating sensors with an ESP32, collecting vibration and temperature measurements, and exploring the potential of machine learning for equipment condition monitoring. We also gained a better understanding of how sensor data can support predictive maintenance.

What's next for AI-Based Hydraulic Pump Fault Classification

Our next steps are to prepare and analyse sensor datasets, develop and train the AI/ML model, and evaluate the proposed fault-classification approach under different pump operating conditions. Further testing will help us assess its reliability and potential application in industrial predictive maintenance.