In last few years, there has been growing interest in the application of computational techniques to various areas of civil and environmental engineering. The paper investigates the potential of Gaussian Process, Neural Network and Support Vector Machines based regression approaches to model the oxygen-transfer capacity of multiple oblique jets oxygenation systems. These approaches are used in the prediction of the overall volumetric oxygen transfer coefficient from operational parameters of an oxygenation system having multiple oblique jets, number varying from 1 to 16, plunging at an impact angle of 60°. The results computed from these techniques and obtained from empirical relationship derived on the basis of multivariate linear regression are compared in terms of correlation coefficient, root mean square error and coefficient of determination. The results suggests the utility of these computational techniques in the designing and performance evaluation of plunging jets oxygenation systems; however, support vector machines have predicted overall volumetric oxygen transfer coefficient with highest correlation coefficient (0.985) and coefficient of determination (0.968), and lowest root mean square error (0.002) in comparison to other computational techniques as well as empirical relationship. Further, scattering (within ± 10 percent) is lowest in case of support vector machines approach. A comparison of results suggests that support vector machines and neural network approaches work well and can be successfully used in modeling oxygen transfer from multiple oblique jets oxygenation systems.
Oxygen transfer, multiple oblique plunging jets, support vector machines, Gaussian process, neural networ
