Markov-chain-based probabilistic approach to optimize sensor network against deliberately released pollutants in buildings with ventilation systems

Abstract If hazardous gaseous pollutants are deliberately released in public buildings, it is necessary to detect them and warn promptly indoor occupants to evacuate. In this study, we propose a Markov-chain based probabilistic approach to optimize an indoor sensor network against a deliberately released contaminant and demonstrate it for a simple multi-zone building and a real experimental cabin. The probabilistic approach is mathematically an ergodic method, which contains two objective functions: to minimize expected time to detection and to maximize the number of successful detections. The Markov chain method uses the multi-zone model or CFD simulation to calculate the transition probability matrix. If such a matrix is determined, simulating indoor pollutant distributions takes almost no time since the Markov chain model does not require iterations. The Markov chain method combines with the probabilistic approach to design the sensor placements for the detection of pollutants. The first case is a simple building with a recirculating air system using the multi-zone-based Markov chain method as the simulation tool. Utilizing the probabilistic method, we determine the optimal sensor placements with different sensor number. The relationship between the sensor network performance and airflow rates of the ventilation system is revealed. Next we conduct a concentration-measured case in an experimental cabin connected with an all-air system and in this case uses CFD-based Markov chain method as the simulation tool. The impact of sensor properties (including sensitivity, the miss alarm rate, sample interval time and response time) on the sensor network performance is discussed for this case.