CN-115880265-B - Welding defect detection method and system integrating magneto-optical imaging and infrared thermal imaging

CN115880265BCN 115880265 BCN115880265 BCN 115880265BCN-115880265-B

Abstract

The invention discloses a welding defect detection method and a welding defect detection system integrating magneto-optical imaging and infrared thermal imaging, which relate to the technical field of welding defect detection and comprise the steps of acquiring magneto-optical images and infrared images corresponding to welding defect types, respectively training a built magneto-optical image sub-classifier and an infrared image sub-classifier after pretreatment, acquiring magneto-optical images and infrared images of all position points at a welding seam of a to-be-detected welding piece, correspondingly inputting the trained magneto-optical image sub-classifier and the trained infrared image sub-classifier to acquire a magneto-optical classification decision vector and an infrared classification decision vector, finally integrating the magneto-optical classification decision vector and the infrared classification decision vector, calculating an integrated decision vector, and determining the final welding defect type corresponding to the position point. The invention combines the advantages of magneto-optical imaging and infrared imaging, combines the unique characteristic information of the magneto-optical imaging and the infrared imaging, increases the anti-interference capability and has high detection precision.

Inventors

GAO XIANGDONG
XIE YUEXUAN
GAO PENGYU

Assignees

广东工业大学

Dates

Publication Date: 20260508
Application Date: 20221227

Claims (7)

1. A welding defect detection method integrating magneto-optical imaging and infrared thermal imaging is characterized by comprising the following steps: S1, acquiring a magneto-optical image and an infrared image corresponding to welding defect types, wherein the welding defect types comprise defects, cracks, pits, air holes and unfused; S2, preprocessing a magneto-optical image and an infrared image corresponding to the welding defect type to obtain a preprocessed magneto-optical image and a preprocessed infrared image; S3, training the constructed magneto-optical image sub-classifier by using the preprocessed magneto-optical image, training the constructed infrared image sub-classifier by using the preprocessed infrared image until the correspondingly set magneto-optical image classification loss function and the infrared image classification loss function are converged, and obtaining a trained magneto-optical image sub-classifier and a trained infrared image sub-classifier; s4, acquiring magneto-optical images and infrared images of all position points of a weld joint of a weldment to be detected, wherein the specific method comprises the following steps: sequentially numbering the position points at the weld joint of the weldment to be detected The No. 0 position point is the initial position of the welding line of the weldment to be detected, The number position point is the tail position of the welding line of the weldment to be detected, and the distance between the adjacent position points is d; Magneto-optical image for obtaining position point number 0 at moment , Magneto-optical image for acquiring position point No. 1 at moment And infrared image of position point No. 0 , Time of day acquisition Magneto-optical image of number location points And Infrared image of number location point Sequentially acquiring all position points at the weld joint of the to-be-detected weldment until Time of day acquisition Infrared image of number location point In the formula (I), in the formula (II), ; S5, inputting the magneto-optical images of all the position points at the weld joint of the weldment to be detected into a trained magneto-optical image sub-classifier, and outputting magneto-optical classification decision vectors of the position points; The magneto-optical classification decision vector and the infrared classification decision vector are vectors comprising five dimensions and dimension scores, wherein the same dimensions of the magneto-optical classification decision vector and the infrared classification decision vector correspond to the same welding defect type, and each dimension score represents magneto-optical probability scores and infrared probability scores of the corresponding welding defect types; s6, fusing magneto-optical classification decision vectors and infrared classification decision vectors corresponding to all position points at the welding seam of the to-be-detected welding piece to obtain fused decision vectors of all position points at the welding seam of the to-be-detected welding piece; And S7, determining the final welding defect type corresponding to each position point according to the fusion decision vector of each position point at the welding line of the weldment to be detected.
2. The welding defect detection method of fused magneto-optical imaging and infrared thermal imaging of claim 1, wherein the preprocessing operation comprises a graying operation, a cutting operation, a rotating operation, and a flipping operation.
3. The welding defect detection method for fusing magneto-optical imaging and infrared thermal imaging as claimed in claim 1, wherein network structures of the magneto-optical image sub-classifier and the infrared image sub-classifier are the same, and are constructed based on the existing residual neural network.
4. The welding defect detection method for fusing magneto-optical imaging and infrared thermal imaging according to claim 1, wherein the specific method for obtaining the fusion decision vector of each position point at the weld joint of the weldment to be detected is as follows: in the formula, Representing the first weld joint of the weldment to be measured The fusion decision vector of the number position points, Representing the first weld joint of the weldment to be measured Magneto-optical classification decision vector of number position points, Representing the first weld joint of the weldment to be measured The infrared classification decision vector of the number location point, Representing the weighting parameters; The fusion decision vector is a vector comprising five dimensions and dimension scores, each dimension of the fusion decision vector is the same as the welding defect type corresponding to the dimension of the corresponding magneto-optical classification decision vector or infrared classification decision vector, and each dimension score represents the fusion probability score of the welding defect type.
5. The welding defect detection method of fusion magneto-optical imaging and infrared thermal imaging according to claim 1, wherein the specific method for determining the final welding defect type corresponding to each position point of the weld joint of the weldment to be detected according to the fusion decision vector of the position point is as follows: and comparing the magnitude of each dimension fraction of the fusion decision vector for the fusion decision vector of any position point at the weld joint of the weldment to be detected, wherein the welding defect type corresponding to the dimension fraction with the largest numerical value is the final welding defect type of the position point.
6. A welding defect detection system for fusing magneto-optical imaging and infrared thermal imaging, for realizing the welding defect detection method for fusing magneto-optical imaging and infrared thermal imaging according to any one of claims 1 to 5, comprising: The training image acquisition module is used for acquiring magneto-optical images and infrared images corresponding to the welding defect types; The training image preprocessing module is used for preprocessing the magneto-optical image and the infrared image corresponding to the welding defect type to obtain a preprocessed magneto-optical image and an preprocessed infrared image; The sub-classifier training module trains the constructed magneto-optical image sub-classifier by utilizing the preprocessed magneto-optical image, trains the constructed infrared image sub-classifier by utilizing the preprocessed infrared image until the correspondingly arranged magneto-optical image classifying loss function and the infrared image classifying loss function are converged, and acquires a trained magneto-optical image sub-classifier and a trained infrared image sub-classifier; The magneto-optical infrared detection equipment is used for acquiring magneto-optical images and infrared images of all position points at the weld joint of the weldment to be detected; The sub-classification decision module is used for inputting the magneto-optical image of each position point at the weld joint of the weldment to be detected into the trained magneto-optical image sub-classifier and outputting the magneto-optical classification decision vector of the position point; the fusion decision module is used for fusing the magneto-optical classification decision vector and the infrared classification decision vector corresponding to each position point at the welding seam of the to-be-detected welding piece to obtain a fusion decision vector of each position point at the welding seam of the to-be-detected welding piece; And the final defect identification module is used for determining the final welding defect type corresponding to each position point of the weld joint of the weldment to be detected according to the fusion decision vector of the position point.
7. The welding defect detection system integrating magneto-optical imaging and infrared thermal imaging according to claim 6, wherein the magneto-optical infrared detection device comprises a box body (1), a controller (2), a magneto-optical film (3), a polarized light source (4), an electromagnet (5), a magneto-optical probe (6), an infrared sensor (7), an induction coil (8) and a processor (9); the controller (2) is arranged on the side surface inside the box body (1) and is electrically connected with the polarized light source (4), the electromagnet (5) and the induction coil (8) respectively; the magneto-optical film (3) is arranged on the bottom surface in the box body (1), the electromagnet (5) is arranged above the magneto-optical film (3), the polarized light source (4) and the magneto-optical probe (6) are respectively arranged on two sides of the electromagnet (5), the polarized light source (4) emits polarized light to the magneto-optical film (3), the polarized light enters the magneto-optical probe (6) after being refracted by the magneto-optical film (3), and the magneto-optical probe (6) generates a magneto-optical image to be transmitted to the processor (9); The infrared sensor (7) is correspondingly arranged on the top surface in the box body (1) and corresponds to the induction coil (8), and the infrared sensor (7) collects infrared images and transmits the infrared images to the processor (9); the processor (9) is provided with a trained magneto-optical image sub-classifier and a trained infrared image sub-classifier.

Description

Welding defect detection method and system integrating magneto-optical imaging and infrared thermal imaging Technical Field The invention relates to the technical field of welding defect detection, in particular to a welding defect detection method and system integrating magneto-optical imaging and infrared thermal imaging. Background Welding technology is one of the most important material forming and processing technologies in modern manufacturing industry, and is widely applied to the industrial manufacturing fields of ship transportation, automobile manufacturing, petroleum industry, aerospace and the like. During the welding process, various defects such as cracks, pits, air holes, unfused parts and the like can appear on the welding parts due to uncertain factors such as improper welding parameter adjustment, complex environment and the like, and the welding quality is seriously affected. The quality of the weld is an important factor in the proper functioning of the associated equipment. The nondestructive testing methods commonly used for welding defect detection include a magnetic powder testing method, an ultrasonic method, a ray testing method and the like. The magnetic powder detection method is to uniformly spread magnetic powder on the surface of a weldment, apply strong current on two sides of the weldment to be detected, and change the distribution of the magnetic powder due to a leakage magnetic field at the defect position so as to present the defect. The ultrasonic method utilizes the energy changes such as reflection, refraction, attenuation and the like which occur in the process of ultrasonic wave propagation in a medium, and the difference in acoustic physical properties exists between a defect area and a plate, so that the internal defect of a weldment is detected. The ray method uses the attenuation characteristics of rays (such as X-rays, gamma rays and the like) on different structures to image the difference in penetration amount, so that the defect is imaged and detected. As shown in figure 1, the magneto-optical detection method is realized based on Faraday magneto-optical effect and magnetic leakage principle, and the sample to be tested is magnetized by using external excitation, and because the magnetic permeability of the sample is different from that of air, a defective area in the sample can generate a magnetic leakage field and diffuse in the air. Light emitted by a light source in the magneto-optical sensor passes through the polarizer to form polarized light, the polarized light deflects in a leakage magnetic field of a sample according to Faraday magneto-optical effect, and the deflected polarized light shows the intensity difference of the light after passing through the polarization analyzer, so that the polarized light is collected by a CMOS camera in the magneto-optical sensor to form a magneto-optical image containing defect information of the sample. As shown in figure 2, the eddy current thermal imaging method is an active infrared thermal imaging detection method, and according to Faraday electromagnetic induction law, pulse current is introduced into an electrified coil to enable an adjacent metal conductor to generate eddy current induction, so that induction heating is realized on a sample to be detected. When defects exist in a sample to be detected, the defects can influence the distribution of eddy current and the conduction of induction heat, so that the heat distribution on the surface of the sample is uneven, and infrared rays with uneven intensity are radiated outwards. Infrared images containing sample defect information can be obtained by utilizing the infrared sensor to collect infrared radiation sequences or infrared radiation of specific frames in a period of time in the heat transmission process. The decision fusion method is a multi-source information fusion technology, comprehensively judges detection tasks by integrating data from different data sources, and has the advantages of strong adaptability, strong anti-interference capability and single sensor error elimination. The prior art discloses a workpiece welding defect detection device, a method and a computer readable storage medium. The workpiece welding defect detection device comprises a communicator, a processor, a logic processing assembly and a display unit, wherein the communicator is used for receiving a first image from an image acquisition module, the processor is coupled with the communicator and used for extracting welding information according to the first image, and the processor is used for conveying the welding information to the logic processing assembly to form a welding defect and form display information of the welding defect so as to display an image or a characteristic value of the welding defect according to the display information. The method only uses a single image, has error defects and poor anti-interference capability, obtains the whole image of the welded wo