The Auxiliary Power Unit (APU) engine is an essential part of an aircraft. When the main engines stop working, the APU intervenes to supply the aircraft with electricity and pressure. It is therefore essential that it is always in an extremely good condition and that possible defects are detected early.
Challenge
How can we transform the data streams coming from the APU engine in a way and use them to detect defects early.
Solution
Predictive maintenance for aircraft engines
Hemisphere has developed a deep learning model for EPCOR that looks 280 flights ahead to detect errors. The model had an accuracy of 80% for certain types of defects. Resolving the defect early gives EPCOR a cost saving of 60%.
I was really impressed with the way Hemisphere handled everything. From the first call to everything in between they were very helpful and knowledgeable
Niels van Hofwegen - Program Manager bij EPCOR (Air France-KLM) Data
APU sensor flight reports from thousands of flights from 80 unique aircraft
Technology
Keras DNN LSTM
Scope
Prototype, Test and Implementation
Planning
Realized in 5 months