Study On Defect Detection Based On Ultrasonic Synthetic Aperture Technology

Pire-inspection of the internal defects of the processed materials and the pre-inspection of the contact state of the components are the key link to ensure industrial safety production,improve product quality and improve production efficiency.By now,the problems are as follows:at first the traditional detection method is inferior to the detection of small defects in non-metallic materials,especially materials with complex internal structures and large attenuation of the energy of the detection signal,such as silicon in the photovoltaic industry.Substantial materials in the manufacture of electronic equipment,the detection means still remain in the material of the metallographic corrosion and other damage detection stages.There are several problems,such as low processing efficiency,serious cutting waste,so the detection means urgently needs update.Secondly,there is large contact structure,such as the mechanical joints.The area is large but the contact points are not evenly distributed.The contact structure between the two interfaces does not indicate the size of the defect.It is difficult to describe the coupled contact state of the joint surface by macroscopically,reflecting the change of the ultrasonic amplitude of the defect characteristics.The above two types of problems are blind areas of non-destructive testing,and there are few studies on quantitative evaluation and qualitative analysis.Therefore,updating the traditional detection methods,applying the advantages of ultrasonic non-destructive testing,researching the methods of defect classification and identification,and improving the effectiveness of ultrasonic feature extraction are the key issues for the accurate evaluation of quantitative detection and qualitative analysis.The above mentioned core issues needs to be addressed.Based on the development of laboratory defect classification and recognition imaging technology and system,this work proposes two improved detection methods and detection methods for aperture defect and contact surface.Based on the detection methods and system development requirements of different defects and contact states,the single-crystal silicon and bonding surface are taken as examples to design a synthetic aperture time domain direct imaging algorithm,and the upgrated two-dimensional and three-dimensional imaging methods are used to effectively enhance the ultrasonic image to the defect.The ability to identify,combining with the research and experiments of ultrasound image fusion technology,has effectively improved the quality of ultrasound images.The main points of this work are below:1.Detailed analysis of defect classification criteria and implementation classification includes the evaluation criteria of ultrasonic synthetic aperture technology,such as axial resolution,lateral resolution and signal-to-noidse ratio indicators.An ultrasonic synthetic aperture methods based on single-element transducer and direct time-domain imaging algorithm are proposed.A detailed algorithm implementation flow is developed to realize the reconstruction of the synthetic aperture image.The key values of the natural defect measurement and simulation are compared.The results show the quantitative analysis,specific influences of image length,depth and tilt angle on resolution and signal-to-noise ratio,which lays the theoretical foundation for the next step of ultrasonic feature extraction.2.The feature extraction and denoising methods of two types of ultrasonic echo signals are represented by single crystal silicon and bonding surface.Based on the basic theory of wavelet packet transform,a new threshold processing technique is reconstructed,as well as an optimal wavelet packet transform theory proposed to improve the existing 2D and 3D image reconstruction techniques.Compared with the traditional phased array algorithm,this work proposes an imaging algorithm based on the ultrasonic denoising signal denoising experiment,and analyzing the effect of wavelet packet decomposition layer,threshold and imaging algorithm on denoising performance and the effectiveness of reducing hardware cost.3.Establishing the theory of rough surface contact and studying the ultrasonic detection method of microscopic contact distribution features that are difficult to describe including taking the joint surface as the research object,carrying out the research work of the defect feature extraction method,combining the common types of defects in the research process and measuring the ultrasonic defect back.Wave signal is used for feature extraction.The Wiener filter algorithm is used to establish the theoretical description of the contact surface contact state.For the ultrasonic image reconstruction results,the pressure changes of different macro shape feature joint surfaces,different roughness experiments and wedge angles are discussed,and the ultrasonic testing is verified.The method is applied to the measurement accuracy of the joint surface shape error,surface roughness,pressure,surface spacing variation,and the effectiveness of the joint surface macroscopic shape error and microscopic contact distribution.4.Based on the theory of deep Boltzmann machine(DBM),the problem of intelligent identification of defects in ultrasonic testing is studied by means of image fusion technology.An ultrasonic image optimization algorithm based on neural network and adaptive learning is proposed.The experimental results show that the method does not only feature strong capture capability,but also reduced the computational complexity greatly.Compared with the existing fusion method,the superiority is more obvious,and the fusion performance and efficiency are more prioritized.5.Based on the above research,the automated ultrasonic non-destructive testing system is developed.With the ultrasonic imaging method and ultrasonic image feature extraction as the core,the ultrasonic non-destructive testing strategy for automatic defect recognition and classification detection is proposed to realize the dynamic three-dimensional imaging function during the detection process.Excellent readability and high accuracy of feature description are indicated.