Research On Super Resolution Imaging Method Based On Multiple Signal Classification

With the development of sound wave theory and the advancement of radar technology,ultrasound has also been widely used in the field of non-destructive testing.The use of ultrasonic Lamb waves to detect damage to the structure can systematically know whether there is damage inside the structure,so as to maintain the structure,make the operation of the structure safer,and reduce maintenance costs.Therefore,in the field of damage detection in aerospace,shipbuilding,infrastructure construction and other fields,ultrasonic Lamb wave technology is widely used.Ultrasonic Lamb wave is required to image damage to achieve accurate judgment of damage.There are many ways to use ultrasound Lamb waves for damage imaging.Among them,ultrasonic phased array imaging has attracted wide attention in the field of ultrasonic imaging due to its advantages of good flexibility,high detection efficiency and high imaging accuracy.Among the existing ultrasonic phased array imaging methods,a more widely used method is an algorithm for delay superimposing time-domain signals with fixed weight values.Although this method has high imaging efficiency,it can not effectively suppress the noise of non-scanning azimuth,so side lobe artifacts will occur and the imaging resolution is not high.There are other damage detection methods such as time reversal and ellipse positioning,but these methods cannot achieve super-resolution imaging of damage because they cannot break through the Rayleigh limit.Therefore,based on the ultrasound Lamb wave phased array imaging,this paper developed a multiple signal classification algorithm.In this method,damage is detected by multiple phased arrays,and the received signal is processed by a specific algorithm to obtain super-resolution imaging.The signal is excited by M excitation elements and then received by N receiving elements.The total signal can form an M×N matrix.After singular value decomposition of the collected signal matrix,T significant eigenvalues can be obtained,and the number of significant eigenvalues represents the number of damages.The singular vector after singular value decomposition contains the information of the noise space.Using the randomness of the noise subspace,the imaging index is formulated to restore the spatial frequency greater than 2K to achieve super-resolution imaging.In this method,as long as the minimum value of M and N is greater than the number of damages,each damage canbe successfully distinguished at super-resolution.Therefore,an appropriate number of matrix elements should be selected for different damage situations.In order to verify whether the method can achieve super-resolution in imaging effect,this paper has conducted numerical simulation and experiment for the method.First,a basic model of Lamb wave propagation in isotropic material plates is established on COMSOL.After the excitation signal is sent by the excitation element,the echo signal is received by using a uniformly arranged linear array.The received echo signal is excluding the direct wave and the boundary reflection wave to form a covariance matrix,and then the matrix is processed with a multiple signal classification algorithm to obtain an imaging result.Then,the obtained imaging results are compared with the traditional imaging results,and it is verified that the multiple signal classification algorithm can achieve super-resolution.Then,a mixed detection experimental platform of PZT-SLDV was built to perform damage detection on the boards with single damage and double damage preset respectively.For the response signal collected in the experiment,the direct wave and the boundary reflected wave are eliminated to form a covariance matrix.The multiple signal classification algorithm is also used to resolve the two damages less than half the wavelength on the final damage imaging map.Super resolution is successfully achieved Damage imaging.The above research results show that the multiple signal classification algorithm overcomes the limitation of imaging resolution brought by Rayleigh limit by using the randomness of noise subspace,and provides a new method for achieving super-resolution damage imaging.