Source term estimation of a time-varying source around a building based on Bayesian inference and unsteady adjoint equations

Abstract

In actual pollutant dispersion accidents, the location of the source is typically concealed and the intensity of the source varies with time. It is important to accurately estimate source parameters based on limited sensor data. However, previous studies were based on the assumption of stabilized sources and concentration fields, and ignored the process of sensor concentration changes over time, which affects the accuracy of the estimation. Therefore, this study applied a source term estimation (STE) method which combines the Bayesian inference method with unsteady adjoint equations to a time-varying source around building. The influences of the release forms, locations, and heights of the source were analyzed from the flow field and transient stage perspectives. We found that the estimation of the time-varying source performed worse than that of the constant source assumed in existing studies. The uncertainty of the estimated results increased with the complexity of the release forms of the source. In particular, the estimation of the location and strength of the period source had a wider probability distribution, higher uncertainty, and was more susceptible to changes in source location and height. The results showed that for time-varying sources, the estimated results fluctuated strongly over time with the pre-developmental and stabilization phases, and it was critical to estimate the source term based on sensor data at various time points.