Context
Engines of the same category are produced in large quantities. They all have the same structure and theoretical lifespan, however, during use, each motor is used in different contexts and its actual remaining lifespan is different from the others.
Need
Prediction of the remaining life of turbofan engines based on training data from 218 units (NASA Open Data) to provide pre-failure warning, reduce estimation errors and optimize maintenance planning .
Solution
The model we developed, based on progressive learning from successive equipment failures, made it possible to monitor a large number of similar motors and provided a reliable prediction of the remaining useful life over time. learning from the data collected after each flight cycle.
Results
Once 30% of the fleet reaches its end-of-life point, the prediction of the end-of-life date for the remaining engines is quite accurate and the prediction error decreases as more data arrives
