Comparative studies on air borne noise and flow induced noise of a double suction centrifugal pump

Abstract

The Flow Induced Noise (FIN) of a centrifugal pump directly affects the Air Borne Noise (ABN) of the same. But, the information regarding focus on ABN of a pump in relation to its FIN is very scarce. This paper presents the comparative study performed on the ABN and FIN of a double suction centrifugal pump. Experiments were conducted on a double suction pump of specific speed (ns) 19 to measure ABN. Free field noise measurements for ABN, aligning with the industrial standards were conducted using FFT analyzer for 3 speeds such as 990 rpm, 1200 rpm, and 1450 rpm. For each speed, the normalized flow ratio (QQn) covered were 0.4, 0.6, 0.8, 1 and 1.2 for measurements. For FIN prediction, transient numerical simulations were conducted for the same pump at 990 & 1450 rpm. Detached Eddy Simulation (DES) and Fast Fourier Transform (FFT) were used to understand the noise generation at the blade passing frequency (BPF) and the variation of discharge flow recirculation in impeller outlet at different speeds. Moreover, for FIN, a generalized quadratic equation with parabolic form is derived for a speed of 1450 rpm. Also, the derived equation at 1450 rpm has been used to develop a relation to predict the FIN at other speeds. Further, the derived FIN relations are used to predict the ABN at different speeds by expressing ABN in terms of FIN for particular speed. Overall, the developed FIN function is useful in predicting the FIN in dB at BPF of a double suction pump for impeller Reynolds number and specific speeds in ranges 2.7∗106≤Re≤8.2∗106 and 19≤ns≤52 respectively with error ranging from +7.5 % to −7.5 %. Empirical correlation is also developed to predict the FIN with error ranging from +6 % to −6%, where FIN was the function of structural (geometrical) parameters and hydraulic (fluid and flow) parameters of the pump. Also, the methodology is used to predict ABN at BPF from the calculated FIN for that particular speed with error ranging from −2.5 % to +2.5 %.