A Gaussian process emulator approach for rapid contaminant characterization with an integrated multizone-CFD model

摘要

This paper explores a Gaussian process emulator based approach for rapidBayesian inference of contaminant source location and characteristics in anindoor environment. In the pre-event detection stage, the proposed approachrepresents transient contaminant fate and transport as a random function withmultivariate Gaussian process prior. Hyper-parameters of the Gaussian processprior are inferred using a set of contaminant fate and transport simulationruns obtained at predefined source locations and characteristics. This paperuses an integrated multizone-CFD model to simulate contaminant fate andtransport. Mean of the Gaussian process, conditional on the inferredhyper-parameters, is used as an computationally efficient statistical emulatorof the multizone-CFD simulator. In the post event-detection stage, the Bayesianframework is used to infer the source location and characteristics using thecontaminant concentration data obtained through a sensor network. The Gaussianprocess emulator of the contaminant fate and transport is used for Markov ChainMonte Carlo sampling to efficiently explore the posterior distribution ofsource location and characteristics. Efficacy of the proposed method isdemonstrated for a hypothetical contaminant release through multiple sources ina single storey seven room building. The method is found to infer location andcharacteristics of the multiple sources accurately. The posterior distributionobtained using the proposed method is found to agree closely with the posteriordistribution obtained by directly coupling the multizone-CFD simulator with theMarkov Chain Monte Carlo sampling.