Streamlining Flight Test with the Design and Analysis of Experiments

Abstract

Traditionale ight-testmethodologytypicallyresultsinlargetestmatricesusingtheapproachofchangingsettings of one input variable at a time, keeping all others at some nominal level. Although this method is functional, another method exists that can dramatically reducetheamount of testing required, increasethe quality of the data analysis, and provide additional insight into system performance. We propose using the principles of statistically designed experiments; and we provide a process for planning e ight test using these principles. The proposed process complements traditional testing methods and places a greater emphasis on structuring test design to increase the precision and interpretability of the data using statistical analysis. Another enhancement to the test processis recommendingsmalland sequentialtesting sothatinformation learned from earliertestsisincorporated in subsequent phases and unnecessary tests can be avoided. The proposed methodology has three phases: building the strategic e ight-test plan, developing detailed test objectives, and planning e ight testing and analysis. This paper discusses the activities conducted in each phase and illustrates the main points with an example from a recent experience with the CV-22 operational test and evaluation. Although our proposal is directed to e ight test, it can be easily implemented in most test environments.