Applicability of several machine learning models in estimation of vortex tube trapping efficiency

Abstract A vortex tube ejector comprises a with slitted crown that lies flush across the entire width of channel bed surface. The and suspended loads are ejected minimal flushing water through slit same efficacy as any other alternative extractor. whirling flow phenomena duct very complex, so ordinary classical models have results contrary to required design guidelines. So, machine learning (ML) artificial neural network (ANN), deep (DNN), gradient boosting (GBM), stacked ensemble (SE), adaptive neuro-fuzzy inference system (ANFIS) used predict trapping efficiency (VTE). input dataset takes size sediment (Sz), concentration (I) sediment, ratio thickness diameter (th/dia), extraction (Extro) while (TE) is taken output. Based on statistical assessments, GBM appears be better than all proposed models. However, ML give comparable performance. models, multivariate linear, nonlinear regression techniques also provide comparatively good results. According sensitivity analyses, most relevant parameter in evaluating VTE.