Calibration of a meanline centrifugal pump model using evolutionary algorithms

Abstract A one-dimensional meanline prediction code suitable for performance prediction of small-to-medium size turbopumps (3–40 kW) is developed, which is able to predict both on and off-design performance of a turbopump given the loss of the diffusion system at the design point. This model is then calibrated against experimental data using an optimization technique. Due to the rapidity of code computations, a single-objective evolutionary algorithm is used for model calibration in order to obtain a wide exploration of the search space. Accuracy of the calibrated model is finally demonstrated by means of a comparison between predicted and actual pump performance.