Time Series Analysis Methods for Improved Flight Test Data Analysis

Daniel LIMIA PEREZ, Capgemini Engineering
Aristeidis ANTONAKIS, Capgemini Engineering

Abstract

Flight test data analysis has historically depended upon the application of classical, model-based system identification techniques, coupled with a qualitative, ‘engineering judgment element’ to extract conclusions on the outcome of tests. Nowadays, despite the recent advances in the data science field, their applications to flight test analysis remain limited, despite the important benefits they would potentially introduce.

With a view to exploring these benefits, this study presents a case study on the implementation of modern predictive models for time series analysis in the Flight Test sector. In this context, an analysis of the characteristics of different time series predictive models is performed (ARX, ARMAX, ARIMAX) followed by test cases on a series of flight test analysis tasks using real flight data. Based on these example applications, the study explores, identifies and demonstrates possible ways in which the use of these analysis techniques can improve the quality of existent processes by providing stronger and quantitative conclusions on test results. The latter include rapid system identification for time response analysis, identification of non-linearities in system response, anomaly or disturbance detection, quantitative analysis of configuration effects on aircraft
dynamic response using flight test data.

At a larger scale, the results of this study suggest that the use of time series analysis methods for flight test data analysis can contribute to improved efficiency of the flight test process allowing for more extended use of available test data, combined with better analysis quality and the possibility to automate a significant amount of the analysis tasks.