Investigation of airfoil self noise with trailing edge serrations using computational methods

In most of the engineering applications, noise is considered as an important factor. Especially in defense industry, it is necessary for the veichle (ship, submarine, plane etc.) to have decreased sound characteristics in order to decrease the chance of getting identified-located. Therefore, the recent number of researches about flow generated noise and vibration has increased drastically. However, the non linear characteristics of the equations used especially in flow generated noise prediction complicates the computations and requires tremendous amounts of computational resources. Aeroacoustics field investigates the flow generated noise. The very first work considering flow generated noise is done by Strouhal about investigating the tones created by flow aroud cylinder. The first work about Computational Aeroacoustics is made by Lighthill. Following CAA studies are based on Lighthill's work. In this thesis, the Ffowcs Williams and Hawkings method which is also based on Lighthill's work is used for CAA calculations. Idea of the serrations were first suggested by Arndt and Nagel. They investigated the leading edge serrations on airfoils to reduce the flow generated noise. It was observed that while these serrations reduced the noise levels, the aerodynamic characteristics of the airfoil was affected and lift coefficients were reduced. Howe was the first to suggest trailing edge serrations which did not reduce these aerodynamic characteristics. The serration parameters used in this work are based on Howe's calculations. In this thesis, the effects of trailing edge serrations on the flow generated noise of an airfoil is investigated. Flow generated noise around NACA 0012 airfoil is calculated with numerical and analytical methods and effects of the serrations that has beed added to the trailing edge are observed. The pressure fluctiations on the airfoil surface are obtained via solving the Navier-Stokes equations using computational methods. Discretisation of the equations are made with Finite Volume Methods. Also the details of the Finite Volume Method, SIMPLE algorithm used in discretisation of pressure and velocity and the turbulence model SST k-ω are mentioned in section two. The acoustic calculations are made via Ffowcs Williams and Hawkings (FW-H) equations. In this work, the validity of the CFD (Computational Fluid Dynamics) method is inspected at first. For this reason, the aerodynamic experiments of NACA 0012 by NASA are reffered. The results of the experiments that has been done with Reynolds number of 6x106 and Mach number of 0.15 are compared with the CFD results of this work at the same conditions. The results that has been obtained on 0, 10 and 15 degrees of angles of attack and checked the lift coefficients of each case. Maximum difference with the expermental lift coefficients was about %5.4 so the validity of the CFD method was ensured. Moreover, the pressure coefficients obtained from experiments are also compared with CFD results and it was almost an exact match. A mesh independency study was also carried out with creating three different mehes with varying cell counts systematically. The mesh with negligible difference in lift coefficient from its predecessor is chosen to conduct the rest of the analyses. The dimensionless y+ value of 1 was chosen for all of the cases. After ensuring the validity of the CFD method, the validity of the CAA (Computational Aeroacoustics) method was investigated. The NASA acoustic experiments on NACA 0012 profile were also used in this section. Time step is chosen as 1x10-5 to be able to resolve the frequencies of 50 kHz. Solution time is chosen as 0.02 seconds so that the frequencies as low as 1 kHz can be resolved. At 7.3 degrees of angle of attack, 0.208 Mach number and 1.13x106 Reynolds numbers, it was observed that the peak of the SPL-frequency graph was only 13 Hz different from the experiments. Also, the peak of the SPL magnitude was only 7 dB off from the experimental data. Thus the validity of the CAA method was ensured. Furthermore, a case with LES (Large Eddy Simulation) method was carried out to see the difference from URANS (Unsteady Reynolds-Averaged Navier-Stokes) method and it was observed that the LES method was more succesful on calculating the high frequencies but less succesful on predicting the peak frequency of the SPL values. The reason LES is able to calculate more accurately on higher frequencies is that it can resolve eddies as small as the mesh size thus the effect of those eddies are calculated and added to the pressure fluctuations. However, the mesh size requirements were not met in our case and accuracy of the LES results was unreliable. After ensuring the validity of the CFD and CAA methods, the serrations were applied. The serration geometries that has been used in this work was linked to three parametrs which are wave length (λ), serration height (h) and the ratio of λ/h. Seven different serration geometries were created for the analyses. The boundary and initial conditions were same as the CAA validation case. Five different virtual microphones were placed around the airfoil with varying locations. After the analyses, the results were obtained. It was determined that λ and h parameters did not create a trend in reducing the peak SPL and OASPL (Overall Sound Pressure Level) but the λ/h ratio did create a trend in this matter with some exceptions. The case named S7 with λ/h=2.56 (which is highest) has reduced the peak SPL and OASPL values about 4 dB from the reference case with no serrations on microphones M1, M2 and M3. However, the michrophones M4 and M5 did not show the same results since they were placed on the wake field of the profile in computational domain meaning to accuratley calculate the fluctiations on those locations were impossible for the calculation times used in this analyses. When the frequency spectra of the microphones are examined, it was observed that the peak SPL and OASPL values of the cases were compatible meaning the peak SPL of the spectra dominates the spectrum with high energies in low frequencies. When the spectra is examined with more detail, an additional peak at 2700 Hz to the peak at 1300 Hz was observed with SPL value of about 40 to 50 dB. It was observed that on the S7 case with λ/h=2.56, this secondary peak was diminished and absorbed. To examine the differences on noise levels in a more detailed manner, the TLS (Turbulence Length Scale) and vorticity distributions of the cases were investigated. The exceptions of the λ/h trend are examined at first. When the TLS distrubution of the cases with same λ/h=1.28 values (S2 and S4) were checked, it was determined that the 6 dB difference in noise levels was occured due to TLS levels being higher on the case with higher SPL value. Also the vorticity distrubutions of the cases S6 (λ/h=0.32) and S7 (λ/h=2.56) were investigated and it was observed that the case with lower noise level (S7) had about vorticity of 170,000 around the serrations and the case with higher noise levels had 200,000. It was concluded that the vorticity and TLS are in sync regarding the noise levels. Another comparison was between two cases with same λ=1.6 mm and different h. The case with h=2.5 mm (S4) had higher noise levels compared to the case with h=1.25 mm (S7). When vorticity and TLS distrubutions of these cases examined, it was obvious that the case with higher noise levels (S4) had higher TLS and vorticity.