Dynamical source term estimation in a multi-compartment building under time-varying airflow

In the case of an airborne contaminant release, it is critical to know the source location and emission rate as soon as possible and take proper actions to ensure the safety of people. In practice, the source emission rates and airflows in buildings are always time-varying. However, the existing algorithms are not effective for estimating the dynamical source term under time-varying airflow. This study proposed a synthetic inverse model to determine the source location and dynamical emission rates under time-varying airflow in multi-compartment buildings. For this model, a transient Markov chain was built on the basis of the airflow; Tikhonov regularization and Bayesian inference were used to estimate the source emission rate and backward location, respectively. We further investigated the effects of the sensor location selection and regularization parameter selection method (L-Curve, generalized cross-validation (GCV), quasi-optimality (Quasiopt)). The results show that the Markov chain combined with Regularization and Bayesian inference model (MCRB) was feasible for estimating the periodic source term under time-varying airflow, and the relative errors were generally smaller than 20%. The number of time nodes below the threshold (relative error is 20%) of the GCV method accounted for 75.1% of the total number of nodes, the ratio of L-Curve method was 66.8%, and that of Quasiopt method was 57.4%, so the GCV regularization method was preferable to determine the regularization parameter. This study found a new and fresh perspective for source term estimation under time-varying airflow in buildings.