Sequential indexation of flight data

With the accessibility of more and more aeronautic data opens the possibility to analyze an engine state on line. Today we just begin to implement datalakes able to store measurements recorded by flying aircrafts. This information is linked to the engines configuration and we even slowly acquire details about each embedded part such as production lot, supplier and even derogations and corresponding quotations as well as unexpected events and scheduled maintenance operations. The first data to be collected were stabilized snapshots: instant values of all embedded sensors during stable flight phases. From those values we were already able to detect wear tendencies as specific trends. Using smart combinations of measurements taken from the physic knowledge of the engine behavior we were also able to define an engine state vector, compare the engines together and even search for similar evolution patterns. However this previous analysis known as part as the Prognostic and Health-Management (PHM) process is not fast enough to detect sudden changes and anticipate urgent actions. It is one of the reasons we now collect the whole flight data stored by the aircraft numeric flight recorder and analyze transient phases. Our project is to approach the problem from a statistic point of view. Our idea is to detect the transient phases and categorize similar behaviors. Then it will be possible to detect instantly a weak signal as an unusual pattern.

[1]  Charles Bouveyron,et al.  Robust clustering of high-dimensional data , 2012, ESANN.

[2]  Xavier Flandrois,et al.  Expertise Transfer and Automatic Failure Classification for the Engine Start Capability System , 2009 .

[3]  R. Tibshirani.  Regression Shrinkage and Selection via the Lasso , 1996 .

[4]  Charles Bouveyron,et al.  Anomaly Detection Based on Confidence Intervals Using SOM with an Application to Health Monitoring , 2015, WSOM.

[5]  Jerome Lacaille,et al.  Visual mining and statistics for a turbofan engine fleet , 2011, 2011 Aerospace Conference.

[6]  Madalina Olteanu,et al.  Design Aircraft Engine Bivariate Data Phases using Change-Point Detection Method and Self-Organizing Maps , 2017 .

[7]  Madalina Olteanu,et al.  Comparison of three algorithms for parametric change-point detection , 2016, ESANN.

[8]  Donald J. Berndt,et al.  Using Dynamic Time Warping to Find Patterns in Time Series , 1994, KDD Workshop.

[9]  P. Fearnhead,et al.  Optimal detection of changepoints with a linear computational cost , 2011, 1101.1438.

[10]  Jerome Lacaille,et al.  Standardized failure signature for a turbofan engine , 2009, 2009 IEEE Aerospace conference.

[11]  Madalina Olteanu,et al.  SOM Bound to Realize Euclidean and Relational Outputs , 2015 .

[12]  Michel Verleysen,et al.  Aircraft Engine Fleet Monitoring Using Self-Organizing Maps and Edit Distance , 2011, WSOM.

[13]  Madalina Olteanu,et al.  Unequal time series clustering applied on flight data , 2017 .

[14]  Jérôme Lacaille,et al.  Online Normalization Algorithm for Engine Turbofan Monitoring , 2014 .

[15]  杰罗姆·亨利·诺尔·拉卡耶.  Monitoring of an aircraft engine for anticipating maintenance operations , 2012 .

[16]  Charles Bouveyron,et al.  Model-based clustering of high-dimensional data streams with online mixture of probabilistic PCA , 2013, Advances in Data Analysis and Classification.