Detection Of Hydrogen Porosity Defects In Aluminum Alloy AC GTAW Based On Arc Spectroscopy

The welding process of aluminum alloy is often accompanied by hydrogen porosity,which seriously affects the forming quality of the weld.The traditional inspection method of porosity defects has higher requirements for experimental equipment,and will cause irreversible damage to weldments.The arc spectrum signal carries a lot of element information which can reflect the dynamic process of welding,so it can establish an internal relationship with the welding hydrogen hole.At present,the key to study the spectrum of hydrogen porosity detection is to find a relatively simple method which can extract effective features from a large number of spectral information.In this paper,the composition of the arc spectrum is taken as the starting point to analyze the physical relationship between the spectral lines,so as to explore a simple and effective spectral feature extraction method,and to realize the online prediction of hydrogen pores.In this paper,based on the research background of hydrogen porosity detection in AC GTAW welding,an arc spectrum acquisition and analysis system is built on the basis of the original welding platform.In order to realize the real-time acquisition and analysis of welding arc spectrum signal,a spectrum sensing system based on Cherney Turner optical path is developed.The linear CCD signal acquisition card is used as the core of photoelectric conversion and data storage.The sensing system can realize the real-time call and analysis of the collected spectrum data.In the process of spectral data analysis,in order to better separate the feature spectral lines with lower intensity and remove the influence of continuous spectral lines on the feature spectral lines,a spectral separation algorithm based on the first derivative of spectral line intensity is used to effectively separate the continuous spectral lines and the characteristic spectral lines in linear time.In order to establish the intrinsic relationship between the feature lines and the elements existing in the welding process,and to remove the redundant information in the arc spectrum signal,the feature lines need to be extracted and classified automatically.Using the feature extraction algorithm based on Pearson correlation coefficient,the six kinds of lines extracted from the feature lines.After getting the features of the spectral samples,in order to establish the stomatal classification model,we need to get the corresponding labels of the spectral samples.Based on the X-ray image of GTAW weld,this paper establishes the Yolo v3 target detection model for the hydrogen porosity defect in the welding,realizes the end-to-end automatic marking of the porosity defect.The XGBoost classifier is established by taking the seven features obtained after two spectral separation algorithm and feature extraction algorithm as the spectral sample attributes,and the tags obtained from the target detection model as the spectral sample tags.In this paper,the algorithm flow of on-line detection of welding hydrogen porosity defect is obtained,specifically,the spectrum data collected in real time is separated into continuous spectrum and feature spectrum through two spectral separation algorithm processing;continuous spectrum gets C0 through feature conversion algorithm,feature spectrum gets C1-C6 through feature conversion algorithm;C0-C6 is sent to classifier to predict whether there will be porosity;After welding,the Xray image of the weld is sent to the target detection model to mark the porosity defects in the weld,and C0-C6 feature quantity and label are sent to the classifier for training and incremental learning.