Research On Approaches For Super Resolution Imaging Using Ultrasonic Phased Array

The development of areas such as aerospace, spaceflight, railway, bridge, nuclear power plant, new material, has led to the progress of Nondestructive Testing(NDT) techniques. On the other hand, it also is a huge challenge for NDT. Imaging based on ultrasonic phased array is one of the most prominent NDT techniques, which efficiently combines the ultrasonic phased array and ultrasonic imaging techniques together. It can scan a large sample quickly, and the result could be displayed in the form of image, which is intuitionistic and easy to be quantitied. However, the imaging resolution is diffraction limited. In order to achieve high resolution, short wavelength need to be employed. However, the shorter the wavelength is, the higher the attenuation of the ultrasonic wave becomes, and the shallower the volume that can be imaged. The thesis is focused on the time-reversal-based super resolution imaging for enhanced resolved capacity, and then achieving a higher level image, when the frequency of ultrasonic wave holds. The main conclusions are summarized:Firstly, the sound field of ultrasonic phased array in solid medium is explored, and the Point Spread Function(PSF) of ultrasonic imaging system is constructed, as the indicators to consider the array parameters. The influence of array parameters on imaging resolution is studied quantitatively. The sound field of array is analyzed based on the Huygen’s principle and the far field directivity function of a single array element. It is shown that when different delay laws are applied to the array, the synthetical beam can be focused and steered, achieving imaging by scanning. The PSF can be obtained by combining the Full Matrix Capture(FMC) and Total Focusing Method(TFM), in which FMC is used to capture the ultrasonic array data, TFM is used to image the ideal point scatterer. According to the PSF, the influence of array parameters on the imaging resolution is followed. When the element pitch is equal to half wavelength, the grating lobes could be suppressed. The duration of ultrasonic pulse affects the axial resolution. The shorter the pulse is, the higher the axial resolution. The array element affects the lateral resolution. The more the array element number is, the higher the lateral resolution.Secondly, the time reversal with multiple signal classification, termed TR-MUSIC, is introduced, which is based on the theory of ultrasonic time reversal. The influence of noise on the imaging resolution is considered. Under the weakly noisy cases, TR-MUSIC can break the diffraction limit and achieve super resolution imaging, resolving the problem between the imaging resolution and detected depth. The model including the multiple scattering between different targets is constructed, and its eigenvalue system as well as TR-MUSIC imaging function is discussed. Additionally the influence of noise on the imaging resolution is considered. It is shown that TR-MUSIC can break diffraction limit and achieve super resolution imaging under weakly noisy cases. In the experiment, a window function is applied to the time domain signals for extracting the scattered signals reflected from defect, and a threshold criterion is provided to distinguish the subspaces under various noisy conditions. The experiment implemented in the steel sample shows that TR-MUSIC can distinguish the two side-drilled holes(SDHs) in the sample, with the central distance d smaller than d_R, in which d_R is the smallest resolved distance of the ultrasonic imaging system, and it can be calculated according to Rayleigh criterion.Thirdly, the multi-frequency time reversal with multiple signal classification, termed MF-TR-MUSIC, is presented, and it can solve the problem that CF-TR-MUSIC is sensitive to noise. In strongly noisy cases, CF-TR-MUSIC fails. Therefore, MF-TR-MUSIC is presented, and its parameters are chosen as: increased step Δω =100k Hz, as well as frequency bandwidth Ω =3MHz ～7MHz, equally to the-6d B bandwidth of ultrasonic pulse. MF-TR-MUSIC is tested with simulated and experimental array data, each containing different noise levels. It is shown that MF-TR-MUSIC can distinguish the two SDHs with central distance d smaller than d_R, achieving super resolution imaging. On the other hand, compared with CF-TR-MUSIC, MF-TR-MUSIC is robust to noise. It can suppress the spurious targets shown in the CF-TR-MUSIC image induced by noise of copper, leading to the stable imaging result.Lastly, the coherent point spread function(CPSF) of linear array is studied, and the compensated phase-coherent time reversal with multiple signal classification, termed PC-MUSIC, is introduced, which is used to solve the problem that the axial resolution of TR-MUSIC is limited by the axial elongation. The CPSF of 1-D linear array is defined to explain the axial elongation exhibiting in the TR-MUSIC images. The singular value decomposition of array response matrix is studied and shown to be arbitrary. The imaging function of PC-MUSIC has advantage of cancelling the arbitrariness and keeping phase information. PC-MUSIC is utilized to distinguish the two targets distributed along the axial direction. It is shown that PC-MUSIC over a given frequency bandwidth can diminish the axial elongation presenting in the TR-MUSIC images, and has the potential of improving the axial resolution. And the ability of PC-MUSIC to locate the targets is enhanced by compensating the phase.