Quantitatively Identify Unsteady Gas Pollutant Releases In Indoor Environment By Inverse CFD Modeling

Rapid and precise determination of release strength and contaminant source location is quite crucial for offering emergency protection once there is accidental release of harmful agents. The current inverse modeling is able to pinpoint the pollutant source location but cannot quantitatively identify the complicated unsteady release rates. This thesis presents the inverse fundamentals to determine the unsteady release process of a gaseous pollutant source based on the provided unsteady concentration information at the measurement spot, by assuming the release is from a known fixed location in a steady flow field.The contaminant concentration distribution in indoor environment is governed by the airflow, contaminant source location as well as source release strength. Suppose flow filed is steady and a source releases at a fixed spot, the concentration distribution and the unsteady gas release strength is governed by the linear gaseous pollutant transport equation. To solve unsteady concentration distribution based on known source release strength, the governing equation is well-posed. However, the inverse solution of release strength based on known unsteady concentration is ill-posed. This thesis adopts the well-known Tikhonov Regularization to improve the solution stability by modifying the ill-posed problem into a well-posed problem. The Tikhonov regularization adds a regularized term to the optimizing objective function and imposes a bound to solution. The optimal solution can be obtained by matching with the provided unsteady concentration using the least-squares optimization strategy. To accelerate the solving procedure, the cause-effect relation between concentration and release strength is described by the so-called pollutant transport matrix that is composed of concentration response factors. Within a linear system, the concentration response from an arbitrary release can be expressed as the convolution integral between the unsteady release and the concentration response factor from a unit impulse release. Thus, the transport matrix can be expressed as the series of concentration response factors. For demonstrating the inversed model, this investigation has tested its application in a two-dimensional square cavity. The gaseous CO2 source in constant release rate is placed at the air supply of the square cavity and unsteady concentration evolution is monitored at the outlet. The unsteady release strength is then identified and compared with the realistic release process to validate the inverse modeling. The inverse model is futher applied to identify some more complicated release scenarios like the sinusoidal release from different positions in the square cavity and a three-dimensional office room, respectively. The tracer gas is released from a human repiratory tract intermittently following the sinusoidal mode. In addition, different regularization parameters are tested to evaluate their impacts to the inverse solution.It finds that the inverse model is able to provide results in reasonable agreement with the actual release rate in experiment. Some differences between the experimental data and the indentified results may be due to the concentration measurement errors or response delay by test instruments. The whole model tests indicate the developed inverse models work well either for complex release scenarios or in three-dimensional space. When implementing the inverse identification solution, various regularization parameters can be selected if they fall within a suitable range. However, if the selected regularization parameter is somewhat large, the identified release rate is dominated by the regularization term. In such situation, the identificated release rate versus time is smooth but with a few oscillations in low frequency and large amplitudes. If the selected regularization parameter is too large, some identified solutions may diverge. On the contrary, if small regularization parameters are selected, the solution is dominated by the provided measurement concentration errors. Consequently, a good match between the identified solution and the actual release process can be obtained but with high-fregquency oscillations in small amplitudes.