Colloquium aankondiging

Faculteit Engineering Technology

Afdeling Engineering Fluid Dynamics
Master opleiding Mechanical Engineering

In het kader van zijn/haar doctoraalopdracht zal

Marijn Wiecherink

een voordracht houden getiteld:

Development of an optimization tool for the efficiency improvement of centrifugal pump impellers

ZaalNH 115


Centrifugal pumps are encountered in a wide range of a applications. They are used for the transport of a broad range of liquid media. An important performance characteristic of centrifugal pumps is the efficiency. In recent years, SPX made significant efforts in improving the efficiency of pump impellers. However, the current redesign process of doing so is time consuming, since it is based largely on trial and error.

The objective of this research is the development of an algebraically based optimization tool, in order to reduce the trial and error phase in the current design process. This optimization is based on the so-called Two-Elements-In-Series model, that is used to describe the impeller's internal hydraulics. Complemented with volute and additional loss models, a pump's head and efficiency curves can be predicted.

Before being implemented in an optimization tool, the accuracy of these models was assessed. The required accuracy was established to be 5% in order to be considered useful in optimization. The validation of the models was performed on three pumps. The accuracy of the model proved to be highest for the pump with lowest specific speed; the average error between the predictions and test results was found to be 4%. For the two pumps with highest specific speeds, the average error between model predictions and test result was around 9%.

The research has focused on impeller optimization. Starting from an existing impeller, the main goal of the optimization tool is to find promising directions in the impeller design space in terms of efficiency improvement. As an optimization algorithm, a genetic algorithm was used. This algorithm is available as a software extension in Excel.

Multiple optimization runs were performed of which two resulting geometries were selected, whose predicted efficiency improvements are 1.8 and 3%, respectively. Numerical flow analyses (CFD) were used to validate the given results. Here, a system of RANS equations, extended with the turbulent flow model of Spalart-Allmaras, was solved numerically for the new impellers.

Of the two generated geometries, the predicted improvement of only one impeller is confirmed. This is expected to be caused by phenomena such as incidence and recirculation losses, of which the descriptions in the Two-Elements-In-Series model are not yet fully developed. More work should be done to integrate these aspects in the model. However, the optimization tool as established in this research has the potential to save significant amounts of time due to its potential of generating promising geometries, thus reducing the amount of trial and error in the impeller redesign process.