In this article we look at a practical example of Predictive Maintenance where we are given synthetic data from a fleet of 100 turbofan engines. Sensors on the engines capture the temperature, pressure, flow rate and speed. The objective is to apply machine learning to this data to predict when the engine will fail.
This Live Script shows how to estimate the Remaining Useful Life (RUL) using MATLAB. In this first use case, we are given failure data from our engines with either run to failure information or with unexpected failures. The data has been labelled to indicate a short, medium, or long time until failure will occur. Because information about the failures are known we can make use of a supervised learning approach like a classification algorithm. In production, this system would generate a warning for the operator when there is a medium duration to failure or an alert when a short time to failure is identified. We will take you through:
- The typical Predictive Maintenance workflow for a labelled dataset
- How to deal with Big Data using MATLAB
- The Classification Learner App
- Building and evaluating Machine Learning models using MATLAB
In the next post we will look at how you can estimate RUL using data from a conservative maintenance schedule where failure data is absent.
What Can I Do Next?
- Request a trial.
- Download the Live Script.
- Find out more from the team.
- Visit our Smart Mining & Manufacturing Focus Area page to learn more.