Deep Gaussian Process-Based Bayesian Inference for Contaminant Source Localization

Abstract

This paper proposes a Bayesian framework for localization of multiple sources in the event of accidental hazardous contaminant release. The framework assimilates sensor measurements of the contaminant concentration with the integrated multizone computational fluid dynamics (multizone-CFD)-based contaminant fate and transport model. To ensure online tractability, we build deep Gaussian process-based emulators approximating multizone-CFD model. To effectively represent the transient response of the multizone-CFD model, the deep Gaussian processes are extended to matrix-variate architecture by adopting Kronecker products to the output covariance for each GP layer. The resultant deep matrix-variate Gaussian process emulators are used to define the likelihood of the Bayesian framework, while Markov chain Monte Carlo approach is used to sample from the posterior distribution. The proposed method is evaluated for single and multiple contaminant sources localization tasks modeled by CONTAM simulator in a single-story building of 30 zones, demonstrating that proposed approach accurately perform inference on locations of contaminant sources. Moreover, the proposed model not only shows outstanding regression performance but speed up training.