Research On Ultrasonic Inspection System Design And Defect Recognition Of Stainless Steel Sheet Defects

Stainless steel sheet is widely used in defense,construction,water conservancy and mechanical secondary processing and manufacturing.The quality of the original sheet is directly related to the quality of the project and product performance.Nowadays,the inspection of thin plates in practical engineering applications not only requires the presence or absence of defect detection,but also requires the acquisition of more valuable defect information for the test pieces,which serves as a guideline for the project quality,engineering service life,etc.This paper designs a set of ultrasound hardware detection system based on FPGA.At the same time,it proposes a feature extraction model based on EMD and principal component analysis to analyze the feature of flaw echo signals,realizing the classification of defect types.For the research object of this paper,analysis and comparison of several commonly used thin sheet defect detection technology methods,determine the use of ultrasonic Lamb wave detection technology to complete the defect detection.From the Rayleigh-Lamb wave equation to derive the velocity,phase velocity and frequency thickness product of the thin plate group for the relationship between them,drawing dispersion curve and sound field diagram to determine the type and specific parameters of Lamb wave excitation probe in the system.Ultrasonic excitation circuit,ultrasonic receiver circuit,signal conditioning circuit,signal acquisition circuit and data transmission circuit are designed in sequence from the overall hardware detection system.In this system,FPGA completes the trigger control and data acquisition control of the front-end circuit signals,and at the same time builds the system kernel based on Nios Ⅱ,and uses the detection algorithm to realize the initial detection and determination of the echo signals.For the defect echo signal,non-stationary and non-linear,introduced a feature extraction model based on EMD and principal component analysis.The EMD decomposition can adaptively decompose echo signals to obtain multi-order IMF components,which is determined by the eigenvalue contribution rate in principal component analysis.Signal components of the original defect signal can be characterized,reducing the analysis process data complexity and data redundancy,feature extraction of selected feature quantities.Experimental results show that the ultrasound hardware detection system based on FPGA can determine the presence or absence of defects through the system’s external defect devices,the three types of typical flaw echo signals obtained by the acquisition are based on EMD and principal component analysis.After the feature extraction model,six types of typical signal feature information are selected to represent the original defect echo signal.The BP neural network model is used as a classifier to encode the output defect type code value.By comparing the actual output code value with the ideal output code value,this method achieves good reliability and stability while identifying and classifying defects.