The Application of Modern Design of Experiment Techniques to Improve Development Flight Tests

Floris Bremmers, Netherlands Aerospace Centre NLR, the Netherlands

Abstract

In this paper development flight test experiments used to determine aircraft performance characteristics – FAR25B – are reviewed and improved using modern design of experiment techniques. In the conventional tests, local models – the drag polar, lift curve and elevator trim curve – are modelled per aircraft configuration individually.

During this thesis these tests are combined and subjected to multiple linear regression analyses to obtain a single global model. Using a theoretical approach it is demonstrated that the global model requires a lower amount test points compared to the local models. Furthermore, with the global model validated, an optimal input design is determined using several techniques: A full factorial design is used since it already equalled the minimum amount of test points required. This is because nearly all interactions among the independent variables are presents in the regression model.

For the same reason the consideration of a D-Optimal design is irrelevant. In addition, the input design matrix is improved by applying leveraging and replication techniques. This results in a global model less prone to random error effects. The methods applied are expandable to other flight test programmes of which a remaining but solvable challenge is the polynomial model structure validation.