Markov-chain-based inverse modeling to fast localize hazardous gaseous pollutant sources in buildings with ventilation systems

Abstract

Fast locating the sources of hazardous pollutants in indoor environments is extremely important for ensuring indoor air quality and indoor environmental safety. Considering the timely and smooth implementation of the subsequent emergency measures, tracking hazardous sources in buildings with ventilation systems has a higher requirement on computation time. It imposes a great challenge on model abstraction and method selection. In this work, a Markov-chain-based inverse approach was established to locate hazardous pollutant sources in case of emergencies. If the complete prior concentration distribution was obtained, the posteriori source location could be directly identified. If the prior concentration distribution was incomplete, the Bayes Theorem that based on the Markov chain simulations was introduced to determine the source location. To validate the applicability of the proposed inverse approach, both the CFD simulation and concentration-measured experiment were conducted. The simulated data from the Markov chain method and CFD model, as well as the measured data from the tests were used as model inputs respectively to identify the source location. The results showed that both inverse approaches were able to correctly locate the source and could significantly reduce the computing cost and time. The impact of measuring errors, measuring points number and placement was extensively discussed. Such sensitivity analysis could be used to determine the upper limit of the sensor errors and the most economical measuring point number for a specific case. The expansion of the Markov-chain-based Bayes method for multiple sources localization was also speculated.