sensitivity analysis of the effective centrifugal pump parameters using the efast method

In the present study, the effective parameters of centrifugal pumps are investigated using the efast sensitivity analysis (sa) method. the sa is performed using gmdh type artificial neural networks (ann) which are based on validated numerical data of flow field in centrifugal pumps. there are four design variables namely: leading edge angle of blades on hub section (β1 hub), leading edge angle of blades on shroud section (β1 shroud), trailing edge angle of blades (β2),  and the stagger angle of blades on mid span (γ mid) and there are two objective functions namely: efficiency (h) and the required npsh of impeller. the results show that among design variables, β2 has the highest effect on variations of h (46%) and npsh (45%). except β2, β1 hub and γ mid has the highest effect on npsh (33%) and h (28%) respectively. the effects of all of the design variables on objective functions are shown in the results.