Prediction of aeration efficiency on stepped cascades by using least square support vector machines

It is important to predict aeration efficiency in stepped cascades because they are used in most water treatment applications for re-oxygenation. The flow conditions on stepped cascades have been classified into nappe, transition and skimming flows. Due to the different mechanisms of air entrainment in the nappe, transition and skimming flow conditions, the aeration efficiencies of the three flow conditions differ significantly from each other. In this paper, two intelligent models were created to predict flow condition and aeration efficiency in stepped cascades using critical flow depth, step height and channel slope information. Least square support vector machine (LS-SVM) was used as intelligent tool. The performances of LS-SVM models were evaluated by 3-fold cross validation test method. The correlation between observed and predicted flow condition is 0.99 and the correlation between measured and predicted aeration efficiency is 0.89. The test results indicated that the LS-SVM can be used successfully in predicting flow condition and aeration efficiency in stepped cascades.