Abstract: As a way of energy saving by recovery of residual pressure, centrifugal pump as turbine (PAT) has been widely used in many fields. As PAT is gradually developed for high power, flow-induced noise becomes one of the most important issues that cause negative effect on reliability. In order to improve both hydraulic and acoustic performances of PAT, an optimization method combining sensitivity analysis and response surface was established. Firstly, through comparison of impeller parameter impact on hydraulic and noise performances, the geometric parameters with great influence on acoustic were filtered based on sensitivity analysis. Further more, with the efficiency and A-weighted overall sound pressure level (OASPL) as target, the multiple regression models connecting variables and multi-objective functions were constructed. Optimization for high efficiency and low OASPL of PAT was carried out with the focused research on the interaction among key parameters. During the investigations, a synchronous acquisition of hydraulic parameters and noise signals were realized on the basis of INV3020C data acquisition system and hydraulic test system in an open test loop. The liquid was pressurized through booster pump, and then impacted the turbine's impeller to make it rotate. The dynamometer consumed and measured the turbine's energy. The operating condition was adjusted by regulating the converter's frequency to change the booster pump's capacity. The flow-induced noise signals were collected using hydrophone at a sampling frequency of 25,600Hz. The signals were amplified and recorded by INV3020C data acquisition system, and Fast Fourier Transform was used to compute the spectra with the Hanning window for reducing the spectrum leakage. The hydraulic performance (such as head, shaft power and hydraulic efficiency) was calculated. The unsteady numerical simulation was performed to obtain noise-generating fluid forces using k-ε turbulence model. After that, a time series of pressure fluctuations at fluid-wall interface was obtained. Then, the boundary element method (BEM) was applied to study flow-borne noise caused by impeller and casing dipole sources in interior acoustic. Validated by experiments, the flow-borne interior noise prediction by k-ε turbulence model combined with BEM acting by casing dipole source was feasible. The results showed that the factors including impeller inlet diameter, blade outlet width, blade number as well as blade wrap angle had a great influence on OASPL in interior noise. The interaction between inlet diameter and outlet width was significant, while it was not obvious between inlet diameter and blade number, inlet diameter and wrap angle. The impact of blade number on OASPL was more significant than outlet width. The best match of outlet width and wrap angle was near the diagonal. The impact of blade number on OASPL was more significant than wrap angle. After variance analysis on regression equation, the response surface model of hydraulic efficiency and OASPL could reflect the correlation between parameter and response value. According to the experiments, after optimization using this proposed method, one third octave OASPL of the PAT was reduced by 4.25dBAwith efficiency increase of 1.98%. All these showed that the response surface method can improve acoustic performance without losing the original hydraulic performance.