Hydro and vibroacoustical solution of underwater vehicles using numerical methods and digital sonar design

In this thesis, using numerical solvers, the hydroacoustical and vibroacoustical calculations are performed for a generic underwater vehicle and its propulsion system. Sound pressure level, sound directivity, acoustical spectrum and acoustical patterns are investigated with the aid of a finite volume method and hybrid acoustical analogies such as Lighthill, Curle and Ffowcs Williams Hawkings (FWH) method. In order to validate these calculation tecniques, the results are compared with test cases in which pressure data was obtained using pressure transducers and microphones. Due to the cost and time requirements, Computational Fluid Dynamics (CFD) codes are exploited in that the analyses are performed to obtain the accurate solutions as much as possible. Although experimental tests are trustworthy for many cases, they also consist uncertainties and, thus, a special afford was given in certain applications to keep test errors in a margin. In order to tune the CFD tools for the analysis of new designs, the solver settings and preferences in software should be validated against experimental results obtained in thefacilities which are arranged for almost similar applications. So as to determine proper solver type, discretization schema, mesh size and turbulence model, the validations are performed in two step, namely hydro/aerodynamical calculations and acoustical solutions. A single propeller VP1304 on an inclined shaft was simulated, which is subjected to an inhomogenous flow at different rotation rates and inflow velocities. The thrust, torque and efficiency curves have been obtained from the literature. Regarding to these testcases, Large Eddy Simulation (LES) has been implemented to solve turbulence quantities and filter large eddies in equations. The propeller rotation is simulated using sliding mesh technique in which the mesh is dynamically rotating at a specified time step. Spatial discretization with tetrahedral unstructured mesh is generated on fine cells at three dimensional flow domain. In the testcase of numerical study, propeller is located at center of the computational domain with boundaries set as the velocity inlet, pressure outlet, symmetry and no slip wall conditions. Optimum mesh size is determined on the basis of both computing time and spatial resolutions resulted for different mesh cases. The rotation of propeller is achieved by using moving reference frame in which flow is circulated around axis instead of rigid body rotation. Discretization scheme was second order. Thrust and torque results in experiment of testcases are taken into account and compared with those of numerical outputs. They are matching well in the region of on- and off-design conditions with a maximum error percentage of 8. Thus, it is validated that propeller attached on an aft body of an underwater vehicle can be simulated with high accuracy in an inhomogenous flow. A numerical calculation on a raingutter is performed in order to compare pressure data obtained in the numerical calculations with the data from the experimental case. Boundary layers are generated fine enough that y+ value is lower than three. The same parameters and solver settings in the propeller calculation are implemented in this case. The nondimensionalized pressure data obtained at different sensor around raingutter shows that the results are matching sufficiently good with the test except for one pressure sensor. The discrepancy on that sensor is found around fifteen percent. It has been shown that numerical simulation can be accurately obtained for the flow around underwater vehicle, which has large pressure gradients and pressure stagnation point. Besides, the pressure data obtained by the farfield sensor are used as an acoustical testcase for the validation of acoustical calculations. The turbulent scales of hydrodynamics and the large wavelengths of sound propagation are well different from each other. Therefore, the calculation of sound propagation is not trivial in engineering application where Direct Numerical Simulation (DNS) is required while obtaining the pressure at each point in the flow. Hence, DNS is not applicable to most industrial flows because of necessity of high computing power. Instead, the flow solutions with a high mesh and low time step can be exploited with an acoustical analogy formulation in a hybrid hydro/vibroacoustical study. In order to attain accurate results in acoustical solutions, validations are performed using available experimental data on sound propagation at the nearfield and farfield. The mesh cell size, boundary conditions, hybrid acoustic solver are investigated through some testcases, as for example flow around cylinder, farfield sound propagation through raingutter, pressure pulses resulted from single propeller VP1304. In those cases, monopole, dipole and quadrupole source terms are attained from CFD calculations to couple them with acoustical solutions. The results of the hybrid noise prediction method are presented here to show the acoustical properties of a three-dimensional flow around a cylinder. Although investigation of the acoustical spectrum around a cylinder is a simple case, it shows the potential of the method of coupling the flow sources and acoustical formulations. Besides, Direct Method (DM) and FWH analogy are compared with respect to the spectrum predictions. In this test-case, an unsteady LES solver which is coupled to a wave propagation model based on FWH acoustic method and DM is used to predict the far-field spectra of flow over a cylinder. The flow-generated sound sources (monopole, dipole, quadrupole) are calculated to examine the suitability of the LES methodology. The results show strong attenuation of turbulent structures occurring in the far-field. They also indicate that it is not so easy to capture the small scales by using the LES formulation. On the other hand, although the number of samples used in Fourier transformation is not sufficiently high, the harmonics of the pressure waves are well matched, and the amplitudes of sound pressure levels are in good agreement with the peak values in the low end. The only shortcoming in the present computational setup was the mesh resolution in the far-field which is apparently not enough to capture the fine turbulent scales accurately at remote locations. A testcase in the literature about thin backward facing elbow was mounted on a flat plate and sound pressure signal was measured at a distance in the test setup. The pressure was presented at the axial locations with the intervals to observe a consistency with measured flow data. Except for the locations just after the rain-gutter where strong separations and large scales occur, the pressure coefficients are in good agreement. Thus, the acoustic information in time domain can be accepted as accurate to evaluate spectrum range. It is concluded that there is no specific vortex street visible and, thus, the broadband noise is apparent in spectrum range. The results from numerical model do not precisely follow the test data at the mid region, though they overlay each other at the lower and upper spectral bands. In the other testcase, the propeller-induced pressure pulses at the three sensor locations were measured in the cavitation tunnel by Potsdam University. The blades produced a source, which was obtained from the flow calculations over one revolution. First three harmonics were predicted in frequency domain. The significant issue is that the overall BPF (Blade Passing Frequency) level is related to the monopole terms resulting from the movement of blades in the flow domain. Additionally, the dipole terms are contributing from the dynamic loading on the blades. In this study, the first three harmonics of the fluctuations were investigated and compared with respect to those of the experimental data. It is clear that for non-cavitating cases, the calculations seemed to underpredict the harmonics observed in the experiments. The first and second locations were close to the blade tip and, therefore, the amplitudes of the fluctuations are much higher. Except for the first harmonic of the second sensor, the results seem to be matching well with test data. The determining factors on those pulses are mainly the blade passing rates and pressure dominated dipole sources on the moving surfaces. The third harmonic was difficult to distinguish for all receivers in the CFD data. The generic underwater vehicle body with a 5 meters length was supposed to be propelled by the counter rotating propellers (CRP), which rotated with a specific design rpm. The vehicle speed was 15.42 m/s while the CRP with 7-bladed forward and 5-bladed aft propeller was rotating at 600 rpm and 550 rpm, respectively. For that, an unstructured tetrahedral mesh was generated with around 20 million cells with 8 layers attached on the blade surfaces. Pressure based LES calculations were implemented using similar schemes as before. The time required to establish a converged solution at one revolution step was around 2 minutes in real-time which takes 0.6 seconds in simulation time to complete 5 revolutions. The time step-size was selected around 2.77E-04 sec in order to achieve frequency range up to 1800 Hz maximum where the resolution of bandwidth was around 1 Hz. The sound pressure level from pressure transducers located at different locations, sound directivity around vehicle, acoustical pattern is investigated by hydroacoustical calculations. The contributions of different vehicle sound source on the spectrum are observed and classified with respect to their frequency levels. Regarding to the spectrum of the CRP system, the blade passing frequency, which is the dominant characteristic in spectral density of the rotating system, was well represented. The harmonics and tonal noises were also clearly identified in the low band but, since the length scales of the sound and the turbulence are fairly separated, the far-field acoustics at higher frequencies than 1800 Hz could not be properly resolved with the hybrid approach used in the present investigation. The calculations showed that the spectra for the body and the propeller were separately resulted from feeding into the FWH formulations. Moreover, contributions of the rudder, the cap, the hub, the blade tip on the noise were examined using power spectral density. On the other hand, vibroacoustic is very dominant for the vehicle-propagated sound. Therefore, some synthetic excitations of displacement are produced representing different types of failure on bearings and misalignment on shaft. Structural response of those generated excitations are determined directly coupled with acoustical solutions. Without considering the mode shapes, acoustical velocity on the excited surfaces are used as a boundary condition in the formulation of sound propagation mechanism. Moreover, the vibroacoustical characteristics are illustrated in the octave band spectrum which shows that there are high amplitudes in the most of the frequency range. Therefore, it might be concluded that synthetic displacements are so high that they can not be damped in structural modes. On the other hand, damping is not modeled in the simulation process, which might be the reason of high amplitude peaks appeared in high frequency ranges. In terms of digital sonar design, CFD and acoustical calculations are both used to simulate the acoustic source which has a specified orientation. The main aim of the design is to build algorithm based on conventional beamforming method while considering all the required digital signal processes. At a starting point, two types of simple sound wave from different sources is sent to planer sonar array. With the aim of orientation detection of the underwater vehicle, beamforming algorithm is built and tested with those simple sound waves. The high accuracy on the target orientation has been observed in the sonar simulations. The design is, however, limited to the consideration of the physical aspects and the beamforming algorithm coupled with the numerical hydro/vibroacoustical simulations without considering electronical subparts. The filtering, signal conversions and wave vectors are processed in the range of 3 kHz due to the limited capacity of the hydrodynamical calculations. However, the algorithm would be more efficient at the higher frequency ranges by decreasing the sonar size and the damping the sublobes in the beamforming. By this way, the sound source might be replaced by a real sonar data or with the CFD pressure signal in the higher frequency range obtained with a much lower time-step size.