Investigation of Unsteady Flows in a Centrifugal Fan using High-Speed PIV and Numerical Simulations

B3 - Unsteady Aerodynamics and Flow Simulations

In the design of centrifugal fans the trend goes towards increasing circumferential speed with higher loads. In addition to the centrifugal forces, fluctuating fluid forces become more and more important. A major source of unsteadiness is the rotor stator interaction between the impeller and the volute casing. In this case the excitation frequency is well known in advance and can be considered during the design process. Especially in part load flow instabilities with unknown frequencies like rotating stall may arise, which can lead to severe dynamic loads and structural vibration. Due to the increased availability of computational power, unsteady flow simulations of the entire fan including inlet ducting, impeller and volute casing become feasible on high performance clusters. With a proper prediction of unsteady fluid forces, the designer is able to account for possible flow induced vibrations. In order to reach this aim, adequate numerical modeling is required. This includes especially sufficient grid resolution and accurate turbulence modeling. Hybrid LES RANS modeling is regarded as a promising compromise, allowing the resolution of large scale turbulent motion, while avoiding a full LES resolution of attached turbulent boundary layers. Since these advanced models should not be applied as black box, experimental results are necessary for validation.
In order to provide high quality experimental results for validation of CFD simulations, High-Speed PIV investigations of the flow field in the impeller of a centrifugal fan have been conducted. The impeller and parts of the casing were made from plexiglass for optical access. For this reason the rotational speed is limited to 2500 rpm. The PIV measurements cover several rotational speeds and operating points in the whole operating range of the fan. All three components of the velocity were measured in the center plane of the rotor using stereoscopic PIV.
Numerical simulations were done covering several approaches at selected operating points. Steady state simulations using the standard SST turbulence model were applied near design point. Transient simulations were conducted in design point and at deep part load. Beside the traditional SST model, the Scale Adaptive Simulation (SAS) approach was applied. This model is able to resolve turbulent motion with sufficient grid resolution and time step, with reduced grid dependence in comparison to the more common Detached Eddy Simulation (DES) approach. An extensive comparison of the different approaches with the measurement data will be presented at the conference.