Cfd-analysis of Fan Aeroacoustics - Comparative Studies

The need of aeroacoustically highefficient fans requires more detailed investigations of three dimensional effects and separated flow analysis. Comprehensive studies and code validation for fans have shown, that further development especially for the influence of sweep -is necessary. Therefore, two different Navier-Stokes Codes, a lifting surface code and a classical method for 3-D cascade flow were used to compute a ducted swept radiator fan. The results of the method are checked against each other and compared with experiments to firstly validate the codes in aerodynamic and secondly deliver data for acoustic calculations. Here the aerodynamic pressure and boundary displacement thickness are taken for the acoustic field computations at the fan face using the acoustic analogy method. The far-field is computed by the use of Rayleighs’s formular. Different flow rates were modeled for a 7-bladed fan geometry. The computational results of static pressure and efficiency are compared with experimental data. Indicated flow separation of both, a forward and an aft swept fan have been analysed. The effect of turbulence modeling and different grid size has been investigated. The linear methods used for the aeroacoustic fan design are validated for local data. In order to check the prediction capability of the DLR acoustic methods, acoustic spectra of periodic and stochastic noise parts have been calculated at different angles around the fan face and compared with measured data. INTRODUCTION In the past intensive research work has been carried out by the German industry and the universities to reduce lowspeed fan noise of air conditioning systems and engine cooling fans due to an acoustically optimized geometry without any loss in performance. Many wind tunnel tests and numerical simulations have been conducted to improve the physical understanding of rotor aeroacoustics and to validate the aerodynamic and acoustic codes for the prediction of fan noise. From the point of industrial view a lot of additional boundary conditions have to be taken into account for fan design as a small size, no stator, low production costs, suitable for recycling, no air losses and a white noise sound with no stressing tones. The aeroacoustic computation of a complete cooling system which consists of one or more parallel or in line working fans is very complicated. Many additional sound sources are introduced by the cooler, ducts behind or parallel to the fan, sharp wedges, changes of the cross section, gaps and slots, irregular shaped in and outlets, which are sometimes covered with a grid. Here the numerical simulation can essential be reduced, when each fan is considered isolated with an disturbed onset flow. Isolated ducted fans can be computed with the classical 2-D cascade flow method or with modern 3-D singularity or field methods. In the subsonic range the non-linear effects can be neglected and therefore the computational results of all the methods mentioned above should be approximately the same. Because of great expenses in grid generation and computer time it seems to be reasonable to use linear methods as the DLR lifting surface code or the radial equilibrium method RE for cascade flow for the numerical optimization. On the other side, only a field method as the Navier Stokes code FLOWer from DLR or TASCflow used by Siemens Electric Ltd. can compute separated flow. Therefore, for a more special analysis field methods have to be used after the design. In this way design time, expensive fan models and wind tunnel tests can be saved. Great experience exists at DLR in the design of aeroacoustically optimal working propulsion systems using supersomputers. Panel and field methods were applied to compute propellers and rotors. The methods were developed in the early eighties and validated within the European GARTEUR activities (GARTEUR = Group of Aeronautical Research and Technology in Europe) [1]. During the last decade DLR made a technology transfer of rotor aerodynamics and acoustics from aviation to the car industry. Here the DLR lifting surface code LBS for rotors was extended to isolated ducted lowspeed fans. Additionally, another code for stochastic noise from randomly fluctuating forces was developed. Both codes were coupled with an optimization code to design aeroacoustically fans [2,3]. During the last 5 years a lot of investigations were made in the field of high speed fans respectively turbomachines, where the acoustic field is calculated from a hybrid computational aerodynamic-aeroacoustic technique (CAA) [4, 5, 6, 7, 8, 9]. With the input of near-field aerodynamics, calculated from a Navier-Stokes CFD Code and a modal pressure analysis according to the theory of Tyler and Sofrin [10], firstly the propagating sound field around the fan is computed by using a Galerkin finite element procedure. Then secondly, at an interface relatively far away from the fan, the decaying pressure modes are input for a velocity potential formulation using a finite difference or higher order finite volume scheme. Outside the fan the essential acoustic signals are captured by a Kirchhoff integral to compute the acoustic far-field. Each acoustic mode has to be calculated separatly. Duct lining can be treated. For the treatment of acoustically absorbing liners in ducted fans also boundary element methods have been developed, which seem to be appropriate as a design tool [11,12]. Ducts of low speed radiator fans are usually too short for the implementation of lining materials. Radiator fans should also have light weight and low production costs. Therefore, in general the radiator fan noise can be only reduced at the source. For this reason a simpler analysis can be employed to optimize the fan geometry. The fan of this study is isolated and has a rotating duct. Therefore, guide vanes and tip clearence effects [13,14] don’t have to be considered. However, blade parameters like twist, chord, sweep or asymmetric blade staggering and blade number have to be optimized by the methods mentioned above [2,3]. Furtherly, the noise at the source can be reduced by avoiding flow separation. Here finally a CFD analysis can help to accomplish the improved design. An essential condition to fullfil the tasks of such an optimized aeroacoustic design are excellent validated codes. Although validated codes for fan aeroacoustics are not the state of art at the time. Therefore, the goal of this study is to 1. validate the Navier Stokes Code FLOWer for lowspeed fans. Comparisons are carried out with computational results of the code TASCflow of Siemens Electric Ltd. and their experimental data. An additional analysis of an aft swept fan should show that the FLOWer code is capable to predict all the special aerodynamic features of a forward and aft swept lowspeed fan. A further goal of this study is to 2. validate the lifting surface code LBS for local aerodynamic data as pressure, boundary layer thickness, lift and drag distributions which are required for the computation of acoustics. The local aerodynamic data can be obtained from the Navier Stokes analysis. For the comparison experiment/theory in acoustics experimental data from Siemens Electric Ltd. is available. DESCRIPTION OF TESTS Experimental measurements of the aerodynamic and acoustic performance for the 390 mm axial flow fan have been conducted at Siemens Elektric Ltd in London. The fan performance was measured in a test tube of 1 x 1 m cross section where the fan sucks air from the tube which can be continuously throttled down to control the airflow (see Fig. 1(a)). When the test commences, the cone is moved inward to close the nozzle outlet. The fan test program calculates equal test points from fully open to close. The cone is then moved to the calculated points and measurements are recorded. For acoustics the testing has been conducted in a semi anechoic test chamber of 6.3 x 5.9x 3.5 m in size with the test rig shown in Fig. 1(b). The chamber has provision to test vehicles for noise and vibration. The floor of the chamber partly consists of a concrete hard surface, providing reflection to sound fields. The background noise level is 25 dB(A). The test rig box is made of wood and surface treated with absorbent material. (The Institute of Noise Control Engineers of USA has recommended specifications for such a box). The fan module can be tested either in a free blow or with a resistive system installed. As test conditions two operating speeds of 2800 and 3500 rpm’s and three radial distances of .5, 1.0 and 1.5 m were selected. For each speed and radius the measuring locations were spaced over a range of 90°. The test results have been recorded as Overall Sound Pressure, Narrow Band and 1/3 octave band noise level. The radial distance is measured from the origin located at the upstream face of the motor shaft. Microphones were positioned stepwise over the meshed ground of the serni-anechoic chamber. test-tube for mess flow and static pressure