Oxygen Transfer in Aerated Flows: Identification and Estimation of Liquid Film Coefficient and Specific Interface Area

The process of oxygen transfer from the atmosphere to water is termed aeration. Aeration improves the quality of water in streams by replacing dissolved oxygen (DO) consumed in the biodegradation of organic matter. The present study focuses on the identification of parameters, specifically liquid film coefficient (KL) and specific interface area (a), that affect the aeration process. Additionally, the study examines the effect of data errors and the selection of an appropriate objective function while estimating these parameters through inverse techniques. A parameter estimation code was developed by coupling a genetic algorithm with the numerical model that simulates oxygen transfer in aerated flows. It was noted that with DO concentration data alone, it is not possible to uniquely estimate both the KL and a simultaneously. The existence of local minima in the KL–a parametric space substantiates this inference. Further, in the case of a parametric model, the accuracy of parameter estimates depends on the selection of the objective function adopted for estimation, given the existence of data errors. This study performed a thorough statistical analysis to investigate the influence of three distinct objective functions on the parameter estimates, considering the presence of data errors. The results revealed that incorporating these objective functions brings a certain degree of bias into the estimated values. Thus, the careful selection of an appropriate objective function is crucial in the context of parameter estimation. Laboratory aerated flow experiments were also conducted, and DO concentrations were obtained at discrete time intervals. The inverse procedure was applied to estimate KL for a known a obtained from image processing. Using an optimized value of KL, simulated and observed values of DO concentration at discrete time intervals were compared. The results showed a satisfactory agreement between the simulated and observed DO values. Further, both parameters were estimated simultaneously using observed DO concentration data, leading to nonunique parameter estimation, and optimal parameter estimates differed significantly from observed parameters.