USE CASE

Transport

Anomaly prediction on an autopilot system

Context

Some anomalies appear from time to time in autopilot systems which cause discomfort to pilots when piloting aircrafts. An autopilot system is monitored by hundreds of sensors.

Need

Customer wished to know first which sensors were the most correlated to the anomalies and second for each anomaly, if they could be predicted during a flight preceding the one where the anomaly might occur.

Solution

DiagFit was first used to find correlations between the 400+ data sensors and the anomalies; all sensors were ranked by correlation importance order. 20 sensors with a correlation higher than 80% were selected to build a model predicting the anomalies before they occur.

Results

Ranked sensors have let our customer finding the root causes of anomalies. Goal was to feed his R&D team with these root causes, to analyse potential changes to be made to system and avoid such anomalies in the future. The predictive model succeeded to predict the occurrences of anomalies during flights preceding flights where anomalies might occur, which was the objective set by customer.

The graphic above displays the data acquired by one sensor during a time period of 1 hour and a half. Red areas show the anomalies that our model had predicted one flight before.