Research On Ultrasonic Quantitative Evaluation Technology Based On BP Neural Network

With the development of science and technology,the safety requirements of materials and structures in all walks of life are increasing.Due to its material defects and the long-term impact of various loads,materials and structures will produce various types of defects during production and service,such as cracks,holes,and inclusions,leading to significant safety accidents,causing casualties and deaths.Therefore,it is essential to develop effective non-destructive testing technology.Ultrasonic detection technology is widely used because of its high sensitivity,strong penetrating ability,accurate defect location,simple operation,and harmlessness to the human body.However,it is still difficult to locate and quantify the defects whose shapes are complex based on ultrasonic testing.The amount of data required for ultrasonic imaging is relatively large and the efficiency is relatively low.Based on the ultrasonic A-scan technology,this paper carries out numerical simulation and experimental testing for defects of different shapes,obtaining the change characteristics of the ultrasonic signal,and the sensitive damage characteristic parameters are extracted using signal processing technology.Finally,this paper researches on the inversion imaging of single defect and multiple defects by using neural networks,image processing technology,and data fusion technology.The specific research content and innovations of this article are as follows:(1)The finite element model of ultrasonic propagation is established by COMSOL software,and the interaction relationship between ultrasonic propagation and triangular and circular defects in 2024 aluminum is analyzed.Then an ultrasonic inspection system is set up to detect aluminum block samples containing through-hole defects of different shapes.Both simulation results and experimental results show differences in ultrasonic signals between various defects,such as signal amplitude,width,and symmetry.The differences lay the foundation for the further extraction of defect features;simultaneously,the consistency between the simulation results and the experimental results indicates the correctness of the finite element simulation.(2)Nineteen features are extracted from the three aspects of time domain,frequency domain and morphology by using signal processing technology,and eleven features that are sensitive to defects are obtained by analyzing the law between features and defects,including four time-domain characteristics: peak value,amplitude value Reduced time,0.1Im linear duration,0.5Im linear duration;three frequency domain characteristics: spectrum peaks,low-frequency components,-1d B bandwidth;four morphological characteristics: shape coefficient,standard deviation,normalized energy,amplitude Mean.It lays the foundation for defect inversion.(3)Using BP neural network and DS evidence theory fusion technology,the inversion of a triangle,circular and multiple defects are realized,which provides a new method for the inversion of complex shape defects;At the same time,the inversion results of the signals collected at different positions of the same defect are fused.Compared with the inversion results of the signal at a single position,It makes full use of the signals reflected from different directions of defects,thereby improving the accuracy and reliability of defect inversion.