Sensitivity analysis of the probability-based inverse modeling method for indoor contaminant tracking

Abstract

Accurate and prompt identification of contaminant sources ensures that the contaminant sources can be quickly removed and contaminated spaces can be isolated and cleaned. The adjoint probability method shows great potential to identify indoor pollutant sources with limited pollutant concentration data from sensors. Application of the method to the reality with unideal conditions such as transient velocity and inaccurate measurement of contaminant concentration requires a sensitivity analysis of the method to these critical parameters. The study finds that with up to 90% of random errors in indoor air flow velocity, the inverse algorithm is still able to produce acceptable predictions, as long as the flow pattern remains the same. In a reasonable yet wide range of contaminant concentration accuracy ([0.01, 100] of the sensor accuracy), the measurement error will not influence the capability of the inverse algorithm to predict the correct source location. This paper further proposes an approach to prescribing the required but presumed contaminant mass range so that the algorithm is able to properly predict the source location.