Near real-time approach to statistical flight test

Abstract

A near real-time approach to statistical flight test is proposed that minimizes the number of test condition executions required to adequately describe system performance relative to a specified performance goal or specification. Four different statistical intervals, all based on the Student-/ distribution and its generalization, are summarized and applied to aircraft flight test scenarios. A graphical technique is developed with applicability to near real-time mission operations. Finally, probability statements are associated with sample set statistics and applied to performance goals or specification compliance. ERFORMANCE flight test of aircraft systems, and some system components, is inherently statistical. Data from multiple executions of a given test condition are required to adequately characterize system performance. Examples of system performance that fall into this type of flight test are terrain following at a specified altitude and air-to-air radar target acquisition; system component evaluation would in- clude radar altimeter and laser ranging sensor testing. Two natural questions arise from this type of flight test: 1) What is the minimum number of test condition execu- tions required to adequately describe system performance? 2) Given a specified number of test condition executions and the sample set statistics, what is the probability (or con- fidence) level that the data collected represents the true sys- tem performance? This article provides one approach to answering these ques- tions based on statistical intervals. Furthermore, the approach given has a near real-time application for flight test mission control. The mathematical background of statistical intervals is first presented, followed by the development of a graphical technique applicable to near real-time operations. Finally, a simple example is given to demonstrate this technique.