HISTORY MATCHING WITH SUBSET SIMULATION

Computational cost often hinders the calibration of complex computer models. In this context, history matching (HM) is becoming a widespread strategy, with applications in many disciplines. HM uses statistical approximation, also known as an emulator, to model output, order mitigate computational cost. The process starts observation physical system. It then produces progressively more accurate emulators determine non-implausible domain: subset input space that provides good agreement between output and data, conditional on structure, sources uncertainty, implausibility measure. HM, it essential generate samples from nonimplausible domain, run train emulator until stopping condition met. However, sampling can be very challenging, since domain become orders magnitude smaller than original quickly. This paper proposes solution problem using simulation, rare event technique works efficiently high dimensions. proposed approach demonstrated via robust design examples field aerospace engineering.