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

Abstract

This paper explores a Gaussian process emulator based approach for rapid Bayesian inference of contaminant source location and characteristics in an indoor environment. In the pre-event detection stage, the proposed approach represents transient contaminant fate and transport as a random function with multivariate Gaussian process prior. Hyper-parameters of the Gaussian process prior are inferred using a set of contaminant fate and transport simulation runs obtained at predefined source locations and characteristics. This paper uses an integrated multizone-CFD model to simulate contaminant fate and transport. Mean of the Gaussian process, conditional on the inferred hyper-parameters, is used as an computationally efficient statistical emulator of the multizone-CFD simulator. In the post event-detection stage, the Bayesian framework is used to infer the source location and characteristics using the contaminant concentration data obtained through a sensor network. The Gaussian process emulator of the contaminant fate and transport is used for Markov Chain Monte Carlo sampling to efficiently explore the posterior distribution of source location and characteristics. Efficacy of the proposed method is demonstrated for a hypothetical contaminant release through multiple sources in a single storey seven room building. The method is found to infer location and characteristics of the multiple sources accurately. The posterior distribution obtained using the proposed method is found to agree closely with the posterior distribution obtained by directly coupling the multizone-CFD simulator with the Markov Chain Monte Carlo sampling.

Employing statistical model emulation as a surrogate for CFD

“…Goethals et al (2012) investigated based on surrogate modeling the sensitivity of night cooling performance to room and system design. Finally, Tagade et al (2013) conducted multizone-CFD simulations to train a Gaussian process emulator, and used the emulator to rapidly localize and characterize multiple sources after contaminant detection by sensors.…”

An improved particle swarm optimization method for locating time-varying indoor particle sources

“…The fixed-sensor network method can apply "forward" or "backward" models. A forward model stores all potential release scenarios that are pre-simulated as a database, and then a potential indoor source can be identified by matching the pre-simulated and measured concentrations from an efficient search algorithm such as the Bayesian probability algorithm [5][6][7][8][9] or optimization algorithm [10][11][12][13][14]. By https://doi.org/10.1016/j.buildenv.2018.…”

Experimental study on a comprehensive particle swarm optimization method for locating contaminant sources in dynamic indoor environments with mechanical ventilation

“…Stationary sensor network methods require the installation of one or more sensors in the indoor space in advance, after which the source location must be identified from sensor readings by using forward or backward models. Most of the available studies on stationary sensor network methods have focused on steady-state indoor environments [8,[10][11][12][13][14][15][16][17][18][19][20][21] (a detailed review is provided in [22,23] ), and only very few attempts have been made to locate sources in dynamic indoor environments. Wang et al [10] proposed an adjoint probability-based method for identifying the contaminant source location in dynamic indoor environments.…”