Using big-data and surface fitting to improve aircraft safety through the study of relationships and anomalies

Wooder, D.; Purvis, A.; McWilliam, R.P.

Abstract

The aim of this paper is to assess the utility of a Big-Data approach to fault detection for ‘systems of systems’, utilising the derivation of empirical relationships identified through surface fitting. So-called Big-Data Integrated Vehicle Health Management systems do currently exist, but tend to analyse the health of vehicle systems based on the behaviour of individual sensors and readings. This paper proposes that it is possible to consider vehicle systems with a ‘macro’ approach and identify relationships between key variables which may not be initially apparent. Used in this paper is the open source flight simulation software FlightGear which has previously been assessed for the development of fault detection systems with positive results. The relationships found can be combined into a model of expected results against which real-time data is tested. Surface fitting and the assessment of ‘goodness of fit’ is used to identify these relationships. It is proposed that this technique need not be limited to fault detection in vehicle systems but is also applicable to other vital systems which require redundancy and constant health analysis. This paper concludes that this method is a viable approach and that relationships can be successfully identified for fault detection purposes.

Citation

Wooder, D., Purvis, A., & McWilliam, R. (2017, December). Using big-data and surface fitting to improve aircraft safety through the study of relationships and anomalies. Presented at TESCONF-2016, Cranfield

• Presentation Conference Type: Conference Paper (published)
• Conference Name: TESCONF-2016
• Acceptance Date: Oct 18, 2016
• Online Publication Date: Mar 2, 2017
• Publication Date: Mar 2, 2017
• Deposit Date: Oct 26, 2016
• Publicly Available Date: Oct 27, 2016
• Volume: 59
• Pages: 172-177
• Series Title: PROC-CIRP18
• Series ISSN: 2212-8271
• DOI: https://doi.org/10.1016/j.procir.2016.10.126
• Public URL: https://durham-repository.worktribe.com/output/1149476