Probabilistic μ-analysis for system performances assessment

Abstract

Hinfinity/μ methods are commonly used in Airbus Defence and Space for the design and validation of control solutions. Formulated in a worst-case paradigm, these methods necessarily lead to overly conservative solutions since sized on the extreme cases. However, the acceptance for relaxed control performances requires mastering the risk associated to the detected unlikely events calling for probabilistic performances metrics in the validation process. A probabilistic μ-analysis method is presented in this paper to exhaustively explore the uncertain parametric domain while evaluating the cumulative probability density function of the performance index. Recent μ-analysis tools implemented in the ONERA’s SMAC toolbox are coupled with a dichotomic search algorithm in order to delimit the safe parametric domain while incrementing the probability of success of criteria. The proposed algorithm is applied to a didactic second order system to demonstrate the performances of the method.