Probability-based inverse characterization of the instantaneous pollutant source within a ventilation system

Once a biochemical pollutant is deliberately released into a ventilation system, the source information including the releasing time and location need to be determined promptly and accurately. Successful inversion algorithms to identify airborne contaminant source within enclosed spaces were deeply developed by previous studies. Such mathematical algorithms inversely simulate airflow and concentration field with numerous intricate inverse matrixes and spend plenty of time in the simulation process. However, tracking airborne pollutant sources within a ventilation system has a higher requirement on computation time due to the rapid spread of contaminants in high-speed airflow, which imposes a great challenge on model abstraction and method selection. This paper mainly focuses on a specific source identification scenario: characterizing an instantaneous pollutant source within a ventilation system by employing a probability-based inverse model. The mathematical model and the solving process of both forward propagation and backward identification of the source are investigated and proposed. To verify the feasibility of the forward model and to validate the applicability of the proposed inverse modeling, a concentration-measured experiment was conducted in a real-built ventilation system. The measured concentrations are used as model inputs to calculate the unconditional and the conditional backward time probability density function (PDF). Then, the impact of sensor errors, sensor number and the change of the operating status of the ventilation system on the results of source identification are discussed. Finally, the basis and limitations of this work are extensively commented.