Reducing Cost & Improving Reliability
​
The Problem
​
All machines are prone to wear and downtime, costing operators millions of pounds in unscheduled and reactive maintenance, with particularly acute impact in safety critical, highly-utilised assets like jet engines.
​
Realworld and simulation sensor data can provide insights into asset condition - What if it could be used to predict when maintenance was needed?
The Approach
​
We developed a series of machine learning models to predict turbofan Remaining Useful Life (RUL), using a synthetic dataset made available from the NASA Prognostics Center of Excellence.
After cleaning and analysing the dataset, we developed a variety of machine learning models to predict remaining cycles to failure from real-time data generated by each engine.
Models where evaluated against each other using a cross-validation technique.
The Solution
​
The project produced a variety of predictive models for remaining useful life, with the best performing models capable of estimating remaining useful life to within 30-40 remaining cycles.
​
You can read more about the project, and explore the data, source code, and models, on our GitHub page.
