The Simulation And The Orientation Of Ultrasonic Detection Flaw Of The Aluminum Plate Based On Array Probe

In the process of manufacture, forming and service, the internal defects of Aluminumplate will inevitably produce all kinds of damage, causative its mechanical performancedegradation, even result in serious catastrophic accidents. At present, as the Aluminum iswidely used in the field of the mechanical parts manufacturing, aerospace and shipbuildingindustries, Aluminum as the backbone of raw materials for the development of moderneconomy and high technology, its importance quality is outstanding. Therefore, it is necessaryto carry out the defect of Aluminum parts nondestructive testing. In recent years, developednondestructive testing technology is one of an important part of in product quality control.The research based on ultrasonic testing technology, single-launch-multiple-receiveultrasonic array probe is designed, combined with principal component analysis andprobabilistic neural network, realized the localization of the Aluminum plate in different depthof21kinds of defect. Study first by finite element simulation software for every defect inmodeling and simulation experiment that was carried out respectively, then the six receivingsensor array unit receives all defects of time domain waveform information analysis andprocessing, to get the initial time domain information, to get the initial amplitudecharacteristic information for spectrum analysis; Again, using the principal componentanalysis to carry out the initial amplitude characteristic information dimension reduction toextract the defect area feature vector. Finally, by the receiving probe array unit to extractfeature vector for the unit, based on probabilistic neural network is used to identify the defectposition. Meanwhile, based on the same feature vector, using support vector machine (SVM)to locate the defect area analysis. The number of units and analyzes the receiving sensor arrayand receives the same probe array unit and different unit distribution, which affect thepositioning accuracy of the defect area.The results showed that, using the probabilistic neural network to the excitation sourcedistance is120mm,160mm and200mm respectively the three major flaws area contains thelocation identification, the average rate of correct recognition of each defect area were100%,100%,82.14%. Using support vector machine (SVM) of the recognition accuracy of thedefect area respective were92.89%,89.29%,75%. The results proved to each feature vector extracted by the receiving probe array unit to unit. Combining with the defect of principalcomponent analysis and probabilistic neural network contains the defect area of differentdistance of position recognition effect, and analyzes the number receiving probe array unitand a receiving probe array unit the same number in different distribution, effect of each subregion contained defects correctly identify the defect areas of different depth ratio.