PERFORMANCE EVALUATION OF FAULT DETECTION ALGORITHMS BY MONTE CARLO METHODS

摘要

The problem of estimating the performance of a fault detection system based on an analytical dynamic model in the presence of modelling uncertainties is considered. This performance is expressed in terms of false and missed alarm probabilities over a specified time horizon. To estimate such quantities, Monte Carlo simulations are used. Three sampling techniques, namely the random sampling, the stratified sampling and the Latin hypercube sampling are reviewed and compared on a case study. The superiority of the last two approaches is demonstrated in terms of convergence speed and variance of the estimate.