Wind Turbine Anomaly Vibration analysis using Machine Learning [Edge Impulse]

Title here

With the World moving towards zero CO2 emission Wind Power represent one of the key sources for such clean energy sources. Wind Turbine is the back backbone of the Wind Power Eco-System, so making sure that the wind turbine is fully function for longer time is a goal that must be achieved. With the ongoing advancement in machine learning and pattern recognition, analyzing the wind turbine vibration would be more accurate and robust. Not only the vibration data that would be analyzed but also the Environment data in the which the wind turbine is working.

What you are going to build

This project implements a remote Predictive Maintenance for Wind Turbine that collects information about the wind turbine (like temperature of the turbine , speed of the turbine and measure vibration) this data will be analyzed on the Edge using a pre-train ML module to decide the current state of the wind Turbine.

How Does the solution work

FRDM-MCXN947 will work as the brain of the system, it will be responsible for collecting the data from different sensors (Temperature Sensor, Motor Encoder, and IMU) and will run the ML model on the Edge to detect with anomaly. Also it would be responsible for sending the Sensor data with the state of the turbine for the backend.

Edge Impulse and Machine Learning

Edge impulse is the leading development platform for machine learning on edge devices, with the support of edge impulse we will be able to collect data (Vibration data from IMU) and cleanup this data, then we will be able to train our model from the classified data we had collected. At the end will be able to get our model as C++ library and will be deployed on FRDM-MCXN947 Board.

With the current data set we have reached 80% accuracy which can be increase dramatically to reach accuracy +98% by training the model for longer period.

System Block Diagram