Research On Phased Array Ultrasonic Testing And Automated Evaluation Methods For Internal Defects In Metal Laminates

Metal laminates are multilayer structures formed by the bonding of two or more different metals.They possess advantages such as high strength-to-weight ratio,lightweight properties,and resistance to fatigue,making them widely utilized in aerospace,automotive manufacturing,construction,energy,and medical device industries.However,factors such as production processes,environmental conditions like humidity and temperature,and fatigue loading during usage can result in defects within the laminate layers and at the interfaces,such as voids,cracks,inclusions,and delamination,thereby affecting their reliability and safety.Therefore,it is imperative to employ effective non-destructive testing methods for comprehensive and accurate inspection and assessment of metal laminates during manufacturing and in-service processes.Ultrasonic phased array defect imaging technology is a widely used,highly efficient,and accurate non-destructive testing method.However,in the detection of multi-layer heterogeneous structures,phenomena such as reflection,refraction,and attenuation occur as sound waves traverse interfaces between different media,affecting the resolution of imaging.Existing ultrasonic phased array detection methods still cannot meet the demand for highresolution real-time detection and automated assessment of multi-layer structures.To address these issues,this paper aims to achieve efficient,high-resolution imaging and automated assessment of metal laminates using ultrasonic phased array technology.Firstly,it summarizes the failure modes and non-destructive testing methods of metal laminates,as well as the research status of ultrasonic phased array detection and automated assessment methods both domestically and abroad.Then,it clarifies the sound wave propagation mechanism of ultrasonic phased array in the detection of multi-layer structures.Finally,it focuses on the key technologies for efficient imaging of flat interface and complex interface multi-layer structures using ultrasonic phased array,and integrates machine learning methods to achieve automated qualitative,quantitative,and positional assessment of defects in ultrasonic images.The main research contents of this article are as follows:(1)The study investigated the principles of ultrasonic phased array flaw detection and beam control,clarifying the propagation mechanism of sound waves in multi-layer structures with both flat and complex interfaces.It analyzed the acoustic field characteristics of ultrasonic phased array in bilayer structures,as well as the influence of various phased array probe parameters on the radiated acoustic field.This provided a solid theoretical foundation for subsequent research endeavors.(2)In response to the issue of poor real-time performance of ultrasonic phased array in detecting multi-layer structures with flat interfaces,a sparse full-focusing imaging method based on an adaptive differential evolution algorithm is proposed.This method constructs a fitness function using the main lobe width and side lobe peak of the phased array transducer as constraints.It employs the adaptive differential evolution algorithm to obtain the distribution of sparse array element positions.Subsequently,the optimal sparse array data and root mean square velocity principle are utilized for sparse full-focusing imaging of multi-layer media.Simulation and experimental results demonstrate that this method significantly reduces the amount of imaging data,improves computational efficiency,and achieves measurement errors within 0.2 mm for detecting circular void defects with a diameter of 1 mm.(3)A multi-layer structure full-focusing imaging method based on arrival time difference interface estimation is proposed,efficiently and with high resolution achieving ultrasonic phased array detection of complex interface multi-layer structures.This method first estimates unknown interfaces based on the arrival time difference principle.Then,it calculates the delay time of each array element in the multi-layer structure full-focusing imaging based on interface contour information and the root mean square velocity principle,and performs imaging through delay superposition.Results indicate that the proposed method offers faster computation time and superior contour extraction accuracy compared to ray-based full matrix full-focusing methods.It reduces array performance metrics,enhances computational efficiency,and achieves measurement errors within 0.5 mm for detecting circular void defects with a diameter of 1 mm,as opposed to ray-based methods.(4)A deep learning-based automated classification and localization method for defects in ultrasonic phased array images is proposed.This method systematically employs state-of-theart deep learning models to benchmark the dataset,and subsequently improves upon the bestperforming YOLOv5 s model through lightweight modifications.Additionally,data cleansing techniques are utilized to enhance the quality of data labels.Experimental results demonstrate that the proposed method achieves a precision rate of 97.8%.Compared to the YOLOv5 s model,it reduces parameter count by 43.2%.(5)A machine learning-based automated quantitative method for defects in ultrasonic phased array images is proposed.Firstly,a signal processing method based on windowed averaging weighting is employed to eliminate the entrance wave and bottom echo.Then,improvements are made to the random forest algorithm.An adaptive window adjustment strategy based on mean absolute deviation is implemented at multi-scale scans to enhance feature extraction flexibility and reduce the computational time complexity of random forest.Within the cascade forest,a weighted method optimized by genetic algorithm is utilized to mitigate the impact of irrelevant attributes in the root node’s random feature selection on algorithm performance,thus enhancing classification accuracy and generalization capability.Experimental results show that the proposed method achieves an accuracy of 97.5%,precision of 97.26%,recall of 96.63%,and F1 score of 96.92,demonstrating higher detection accuracy compared to other models while reducing time costs.The research presented in this paper has achieved non-destructive testing during the manufacturing,processing,and service stages of metal laminates,along with automated assessment of defects.It has enhanced the accuracy and efficiency of ultrasonic phased array detection and evaluation,thereby providing significant practical value for ensuring the safety of critical equipment.