A laser welding penetration real-time detection method and system

Abstract

The application provides a laser welding penetration real-time detection method, which comprises the following steps: obtaining an imaging image of a target keyhole, and obtaining position information of the target keyhole according to the imaging image; introducing detection light and reference light according to the position information of the target keyhole; obtaining the detection light and the reference light reflected by the target keyhole and input to a fiber coupler, obtaining interference light signals output by the fiber coupler, transmitting the interference light signals to a data processing unit, and obtaining penetration information of the target keyhole. The method provided in the embodiment firstly obtains an imaging image of a target keyhole, positions the target keyhole according to the imaging image, accurately introduces the detection light and the reference light into the keyhole, and accurately detects the penetration information. The method is not only convenient to operate and high in detection accuracy, but also can realize real-time detection of the target keyhole and is high in detection efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical manufacturing, and particularly relates to a laser welding penetration real-time detection method and system. BACKGROUND

[0002] Welding, also known as soldering, is a manufacturing process and technology that connects metals or other thermoplastic materials such as plastics by heating, high temperature or high pressure. There are many energy sources for modern welding, including gas flame, electric arc, laser, electron beam, friction and ultrasonic wave, etc.

[0003] In the welding process of the prior art, manual visual inspection is generally used to monitor the penetration of the welding position. However, the manual visual inspection has low efficiency and the monitoring result is not accurate enough. Moreover, the manual visual inspection cannot achieve real-time control and cannot find welding defects in time to control the penetration. Therefore, the manual visual inspection of the penetration in the prior art cannot meet the requirements.

[0004] Therefore, the prior art needs to be further improved. SUMMARY

[0005] In view of the above problems in the prior art, the present application aims to provide a laser welding penetration real-time detection method and system to overcome the problem that the human eye cannot quickly and accurately monitor the welding penetration of the welding position in real time.

[0006] In a first aspect, the present application provides a laser welding penetration real-time detection method, which comprises the following steps.

[0007] An imaging image of a target keyhole is obtained, and position information of the target keyhole is obtained according to the imaging image;

[0008] A detection light and a reference light are introduced according to the position information of the target keyhole;

[0009] An interference light signal output by a fiber coupler is obtained, wherein the detection light and the reference light are reflected by the target keyhole and input to the fiber coupler;

[0010] The interference light signal is transmitted to a data processing unit to obtain penetration information of the target keyhole.

[0011] Optionally, the step of obtaining the interference light signal output by the fiber coupler, wherein the detection light and the reference light are reflected by the target keyhole and input to the fiber coupler, comprises the following steps.

[0012] Interference light signal information of a position where the target keyhole is located is collected according to a preset collection angle and a collection frequency; wherein the preset collection angle comprises a plurality of collection angles formed after superposition.

[0013] Optionally, the step of obtaining an imaging image of the target keyhole and obtaining position information of the target keyhole according to the imaging image comprises:

[0014] The imaging image of the target keyhole is obtained multiple times within a preset oscillation range, and multiple position information of the target keyhole is obtained according to each imaging image.

[0015] Optionally, before the step of transmitting the interference optical signal to a data processing unit to obtain the penetration information of the target keyhole, the method further comprises:

[0016] A preset network model generates a predicted detection welding category corresponding to a sample waveform graph in a training set according to the sample waveform graph, wherein the training set includes multiple groups of sample waveform graphs, and each group of sample waveform graphs includes a waveform graph and a detection welding category corresponding to the waveform graph; wherein the detection welding category corresponding to the waveform graph includes welding penetration data, pre-welding weld data, post-welding surface height data, and post-welding weld width data.

[0017] The preset network model corrects model parameters according to the predicted detection welding category corresponding to the sample waveform graph and the detection welding category corresponding to the sample waveform graph, and continues to perform the step of generating the predicted detection welding category corresponding to the sample waveform graph according to the waveform graph in the training set until the training condition of the preset network model meets a preset condition, to obtain the waveform graph classification model.

[0018] Optionally, the step of transmitting the interference optical signal to a data processing unit to obtain the penetration information of the target keyhole comprises:

[0019] Performing Fourier spectrum transformation on the interference optical signal to obtain a frequency spectrum waveform graph;

[0020] Inputting the frequency spectrum waveform graph into the waveform graph classification model to obtain a predicted detection welding category output by the waveform graph classification model;

[0021] Performing local optimal data weighting algorithm processing on the frequency spectrum waveform graph according to the predicted detection welding category to obtain a penetration data value corresponding to the position of the target keyhole.

[0022] Optionally, the step of performing local optimal data weighting algorithm processing on the frequency spectrum waveform graph according to the predicted detection welding category to obtain a penetration data value corresponding to the position of the target keyhole comprises:

[0023] According to a weighting value corresponding to the predicted detection welding category, weighting fitting is performed on data in the frequency spectrum waveform graph, and a least square method is used to estimate a penetration value, to obtain the penetration value of the target keyhole.

[0024] Optionally, the step of transmitting the interference optical signal to a data processing unit to obtain the penetration information of the target keyhole includes:

[0025] performing Fourier spectrum transformation on the interference optical signal to obtain a spectrum waveform diagram;

[0026] inputting the spectrum waveform diagram into the waveform diagram classification model to obtain a predicted detection welding category output by the waveform diagram classification model;

[0027] performing local optimal data weighting algorithm processing on the spectrum waveform diagram according to the predicted detection welding category to obtain a preliminary penetration detection result corresponding to the target keyhole position;

[0028] combining the preliminary penetration detection result with an imaging image of the target keyhole to obtain a penetration change curve diagram;

[0029] obtaining a penetration data value corresponding to each welding time value based on the penetration change curve diagram.

[0030] Optionally, after obtaining the penetration data value, the method further includes the steps of:

[0031] obtaining a three-dimensional point cloud image of the target keyhole based on the imaging image of the target keyhole and the penetration data value of the target keyhole, and displaying the three-dimensional point cloud image.

[0032] In a second aspect, the embodiment further provides a laser welding penetration real-time detection system, which includes:

[0033] a position acquisition module configured to acquire an imaging image of a target keyhole and obtain position information of the target keyhole according to the imaging image;

[0034] an optical signal acquisition module configured to introduce detection light and reference light according to the position information of the target keyhole, acquire an interference optical signal output by a fiber coupler, the interference optical signal being input to the fiber coupler by reflection of the detection light and the reference light on the target keyhole;

[0035] a data processing module configured to transmit the interference optical signal to a data processing unit to obtain penetration information of the target keyhole.

[0036] Beneficial effects, the present application provides a kind of laser welding penetration real-time detection method and system, by obtaining the imaging image of target spoon hole, and according to the imaging image obtains the position information of the target spoon hole;According to the position information of the target spoon hole, import detection light and reference light;The detection light and reference light are input to fiber coupler by the target spoon hole reflection, the interference light signal output by the fiber coupler, the interference light signal is transmitted to data processing unit, obtains the penetration information of the target spoon hole.The method provided in this embodiment first obtains the imaging image of target spoon hole, the position of target spoon hole is positioned according to the imaging image, to realize the accurate import spoon hole of detection light and reference light, to realize the accurate detection of penetration information, the method of this embodiment is not only easy to operate, and detection accuracy is high, and real-time detection of target spoon hole can be realized, and detection efficiency is high. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 It is the step flow chart of the laser welding penetration real-time detection method provided by the embodiment of the present application;

[0038] Figure 2 It is the device connection schematic diagram of the method in specific application of the embodiment of the present application;

[0039] Figure 3 It is the step flow chart of the specific application embodiment of the method of the embodiment of the present application;

[0040] Figure 4 It is the structure schematic diagram of the laser welding penetration real-time detection system of the embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical scheme and advantages of the present application more clear and definite, the present application is further described in detail below with reference to the drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0042] The various non-limiting embodiments of the present application will be described in detail below with reference to the drawings.

[0043] In the prior art, the detection of welding penetration generally adopts manual method, but the manual naked-eye detection method is not only low in efficiency, but also cannot meet the needs of high-precision welding, and manual monitoring cannot monitor the welding penetration in real time, so it is necessary to detect the welding penetration after welding, so it is impossible to achieve timely adjustment, and therefore it is impossible to meet the needs of high efficiency and high precision.

[0044] In order to overcome the above defects, the embodiment provides a laser welding penetration real-time detection method, which is applied to real-time monitoring of welding penetration. The method can not only detect the target keyhole position after welding, but also monitor the target keyhole position during welding, so that the detection result of welding can be obtained faster, so as to adjust the welding position. The laser beam is used for welding the target keyhole position. Therefore, the method provided by the embodiment meets the requirements of high efficiency and high precision.

[0045] The method of the embodiment will be further described in detail below with reference to the drawings.

[0046] Exemplary method

[0047] The embodiment provides a laser welding penetration real-time detection method, as shown in Figure 1 The method can be applied to monitoring the penetration during or after welding. The method comprises the following steps.

[0048] In step S1, the imaging image of the target keyhole is obtained, and the position information of the target keyhole is obtained according to the imaging image.

[0049] When the target keyhole position needs to be monitored during the laser welding operation, the three-dimensional scanning data of the target keyhole position is collected by using the scanner while the laser is used to weld the target keyhole position. As shown in Figure 2 The computer 1, the system core unit 2, the optical module 3 and the 3D module 4 are combined, the driving of the motor in the 3D module 4 and the reflection of the swing mirror can guide the laser beam to the target keyhole position, so as to measure the target keyhole position. It can be considered that the method provided by the embodiment can realize the measurement of the penetration of the target keyhole position during and after welding.

[0050] In the specific implementation, the step of obtaining the imaging image of the target keyhole and obtaining the position information of the target keyhole according to the imaging image comprises the following steps.

[0051] The imaging image of the target keyhole is obtained multiple times in a preset swing range, and multiple position information of the target keyhole is obtained according to each imaging image.

[0052] Since the imaging image of the target keyhole is obtained multiple times in a certain swing range, when the welding process is in a state that the penetration is constantly changing, more detection data can be obtained, so as to more accurately position the penetration data value.

[0053] When the imaging image of the target keyhole is obtained, the position of the target keyhole is positioned according to the obtained imaging image, and the accurate position of the current target keyhole is confirmed.

[0054] Further, in order to more accurately locate the target hole, the imaging image can be a 2D image or a three-dimensional image obtained by scanning. If the imaging image is a two-dimensional image, the two-dimensional position data of the target hole is obtained. If the three-dimensional imaging of the target hole is obtained, the three-dimensional position data of the target hole is obtained. The three-dimensional data information of the hole, such as the data coordinates of the bottom of the hole and the data coordinates of the edge and the height coordinates, can be quickly located.

[0055] Specifically, in an embodiment, a scanner can be used to realize 3D scanning of the target hole position, obtain three-dimensional point cloud data of the target hole during welding, and use the three-dimensional point cloud data to reconstruct a three-dimensional model of the target hole. The three-dimensional structure of the target hole position obtained by reconstruction is displayed to enable the user to preview the welding condition of the target hole at any time.

[0056] Before scanning, the scanning galvanometer can also be parameterized. After the parameterization is completed, the scanning galvanometer scans the target hole position in all directions according to the set collection angle and collection frequency to obtain three-dimensional point cloud data of the target hole position.

[0057] Step S2, introducing detection light and reference light according to the position information of the target hole.

[0058] When the target hole position is located, the detection light and the reference light can be introduced into the target hole according to the located position information to obtain reflected interference light.

[0059] In an embodiment, the detection method of the welding depth utilizes optical coherence imaging, that is, the light emitted by a wideband light source passes through a 2*2 coupler to irradiate the sample surface and the reference mirror through the sample arm and the reference arm respectively. The reflected light in the two light paths converges in the coupler, and the optical path difference of the reflected light of the two arms interferes within the coherence length, and the interference signal is output. Fourier transform of the interference signal can obtain the depth information of the sample arm relative to the reference arm. In this step, the optical module includes the above-mentioned wideband light source, coupler and other devices. The light is irradiated to the target hole position to realize the measurement of the depth value of the target penetration position.

[0060] Step S3, obtaining the interference light signal output by the optical fiber coupler when the detection light and the reference light are reflected by the target hole and input into the optical fiber coupler.

[0061] In this step, the interference light signal output by the optical fiber coupler is received, Fourier transform of the interference light signal is performed to obtain a frequency spectrum waveform corresponding to the interference light signal, and analysis is performed based on the frequency spectrum waveform to obtain the penetration data of the target hole.

[0062] Step S4, transmitting the interference optical signal to a data processing unit to obtain the target keyhole's penetration information.

[0063] The interference optical signal obtained in the above step S3 is analyzed, the interference optical signal is subjected to Fourier transform to obtain a frequency spectrum waveform corresponding to the interference optical signal, and the target keyhole's penetration data is obtained based on the frequency spectrum waveform.

[0064] Specifically, the step of transmitting the interference optical signal to the data processing unit to obtain the target keyhole's penetration information includes:

[0065] performing Fourier spectrum transform on the interference optical signal to obtain a frequency spectrum waveform;

[0066] inputting the frequency spectrum waveform into the waveform graph classification model to obtain a predicted detection welding category output by the waveform graph classification model;

[0067] performing local optimal data weighting algorithm processing on the frequency spectrum waveform according to the predicted detection welding category to obtain a penetration data value corresponding to the target keyhole position.

[0068] In addition, in order to obtain more accurate penetration data and improve data processing, the step of transmitting the interference optical signal to the data processing unit to obtain the penetration information includes:

[0069] performing Fourier spectrum transform on the interference optical signal to obtain a frequency spectrum waveform;

[0070] inputting the frequency spectrum waveform into the waveform graph classification model to obtain a predicted detection welding category output by the waveform graph classification model;

[0071] obtaining the penetration data value according to the imaging image information of the target keyhole position and the real-time temperature value of the target keyhole position.

[0072] Further, the step of obtaining the penetration data value according to the imaging image information of the target keyhole position and the real-time temperature value of the target keyhole position includes:

[0073] performing Fourier spectrum transform on the interference optical signal to obtain a frequency spectrum waveform;

[0074] inputting the frequency spectrum waveform into the waveform graph classification model to obtain a predicted detection welding category output by the waveform graph classification model;

[0075] performing local optimal data weighting algorithm processing on the frequency spectrum waveform according to the predicted detection welding category to obtain a preliminary detection result of the penetration corresponding to the target keyhole position;

[0076] combining the preliminary detection result of the penetration depth with the imaging image of the target keyhole, to obtain a penetration depth change curve;

[0077] obtaining a penetration depth data value corresponding to each welding time value based on the penetration depth change curve.

[0078] In an embodiment, after the penetration depth data value is obtained, the method further comprises the steps of:

[0079] obtaining a three-dimensional point cloud image of the target keyhole based on the image information of the target keyhole and the penetration depth data value of the target keyhole, and displaying the three-dimensional point cloud image.

[0080] When the penetration depth data value of the current welding is calculated, the three-dimensional point cloud image of the current keyhole can be obtained by combining the two-dimensional or three-dimensional imaging data monitored by the target keyhole, and the three-dimensional solid image of the target keyhole can be constructed according to the three-dimensional point cloud image, so that the user can watch the welding state of the target keyhole in real time.

[0081] The method provided in the embodiment uses online coherent imaging (IC I) to provide a higher level of detail and accuracy for laser weld monitoring. ICI uses a low-power infrared laser beam to measure distance. The measurement beam is emitted through the same optics as the welding laser to the bottom of the keyhole and records its depth in real time. A direct geometric measurement of the weld penetration is obtained during the welding process. This measurement method does not damage the parts, and the results can be obtained immediately.

[0082] Further, before the step of obtaining the three-dimensional point cloud data of the target keyhole position, the method further comprises the steps of:

[0083] determining whether the laser focal length of the laser is abnormal, and if abnormal, adjusting the laser focal length to make the maximum position of the laser power correspond to the target keyhole position.

[0084] In order to obtain more accurate three-dimensional scanning data, before scanning the target keyhole position, it is necessary to determine whether the laser focal length of the laser is abnormal, and if abnormal, the laser focal length needs to be adjusted so that the maximum value of the laser power of the adjusted laser corresponds to the target keyhole position. In a specific embodiment, the laser focal length can be manually adjusted according to the measurement information feedback of the current laser focal length, so that the laser focal length is normal, and the data information is recorded.

[0085] Further, in order to obtain better detection results, before the step of transmitting the interference light signal to the data processing unit to obtain the penetration depth information of the target keyhole, the method further comprises the steps of:

[0086] The preset network model generates a predicted detection welding category corresponding to the sample waveform graph according to the sample waveform graph, wherein the training set includes multiple groups of sample waveform graphs, and each group of sample waveform graphs includes a waveform graph and a detection welding category corresponding to the waveform graph; and the category of the waveform graph data includes welding penetration data, pre-welding weld data, post-welding surface height data, and post-welding weld width data.

[0087] The preset network model corrects model parameters according to the predicted detection welding category corresponding to the sample waveform graph and the detection welding category corresponding to the sample waveform graph, and continues to perform the step of generating the predicted detection welding category corresponding to the sample waveform graph according to the waveform graph in the training set until the training condition of the preset network model meets a preset condition, so as to obtain the waveform graph classification model.

[0088] When the obtained frequency spectrum waveform graph is analyzed, the detection type thereof is also classified, and whether the detection type meets a preset welding penetration requirement is compared according to the analyzed detection type, so that the welding adjustment effect is more accurate after adjustment.

[0089] In an embodiment, a waveform graph classification model is trained by using a deep learning algorithm. The waveform graph classification model is trained based on different waveform graph sample data, and quickly classifies an input waveform graph. Specifically, welding frequency spectrum waveform graphs corresponding to different detection types are collected. Since the welding categories are different, the differences between the data values are different, and the corresponding waveform graphs are different. Therefore, identifying the welding category to which the current detection belongs can improve the accuracy of data processing.

[0090] In order to obtain accurate detection results, the step of obtaining the three-dimensional point cloud data of the target keyhole position also includes:

[0091] The near-infrared camera is started to obtain image information of the target keyhole position.

[0092] The image information of the target keyhole position is obtained by using the near-infrared camera, and the two-dimensional image and the three-dimensional point cloud data are combined to monitor the welding penetration of the target keyhole, which can improve the detection efficiency. In an embodiment, the welding penetration of the target keyhole position is preliminarily judged according to the obtained image information. If the welding penetration is obviously not up to the requirements, the welding penetration can be directly adjusted. When the welding penetration needs to be further and more accurately judged, the scanned three-dimensional point cloud data is analyzed, thereby reducing the data processing amount and improving the analysis efficiency.

[0093] In an embodiment, the collected image of the target keyhole position can also be analyzed by using a deep learning algorithm, the image is trained, and the final analysis result is obtained.

[0094] Optionally, the step of transmitting the interference optical signal to a data processing unit to obtain the penetration information of the target keyhole includes:

[0095] performing Fourier spectrum transformation on the interference optical signal to obtain a spectrum waveform diagram;

[0096] inputting the spectrum waveform diagram into the waveform diagram classification model to obtain a predicted detection welding category output by the waveform diagram classification model;

[0097] performing local optimal data weighting algorithm processing on the spectrum waveform diagram according to the predicted detection welding category to obtain the penetration data value of the target keyhole position.

[0098] Specifically, the step of performing local optimal data weighting algorithm processing on the spectrum waveform diagram according to the predicted detection welding category to obtain the penetration data value of the target keyhole position includes:

[0099] performing weighted fitting on data in the spectrum waveform diagram according to the weighting value corresponding to the predicted detection welding category, and estimating the penetration value by using the least square method to obtain the penetration value of the target keyhole.

[0100] The local optimal data weighting algorithm is to perform polynomial weighted fitting on a to-be-fitted point based on local observation data, and estimate by using the least square method. In this embodiment, different weighting weights are set according to different detection welding categories, so the data in the spectrum waveform diagram is weighted fitted according to the predicted welding category and the preset corresponding weighting weight, the penetration value is estimated by using the least square method, and the penetration value of the welding is obtained.

[0101] In order to improve the analysis efficiency, the step of obtaining the three-dimensional point cloud data of the monitoring target keyhole position further includes:

[0102] starting a temperature control component to obtain a real-time temperature value of the target keyhole position.

[0103] The temperature value of the target keyhole position is obtained by using the temperature control component, and it is judged whether the welding of the target keyhole position meets the conditions based on the temperature value. If the temperature value exceeds a certain range, the welding of the welding position may not meet the requirements, so the welding penetration can be judged according to the welding temperature value of the target keyhole position in the specific implementation process.

[0104] In another embodiment, the step of obtaining the penetration data value according to the imaging image information of the target keyhole position and the real-time temperature value of the target keyhole position includes:

[0105] performing Fourier spectrum transformation on the interference optical signal to obtain a spectrum waveform diagram;

[0106] inputting the spectrum waveform diagram into the waveform diagram classification model to obtain a predicted detection welding category output by the waveform diagram classification model;

[0107] performing local optimal data weighting algorithm processing on the spectrum waveform diagram according to the predicted detection welding category to obtain a preliminary detection result of the target keyhole position corresponding to the target keyhole position;

[0108] combining the preliminary detection result of the target keyhole position with the imaging image of the target keyhole to obtain a change curve diagram of the target keyhole position;

[0109] obtaining a melting depth data value corresponding to each welding time value based on the change curve diagram of the target keyhole position.

[0110] In this step, the three-dimensional point cloud image obtained from the imaging image of the target keyhole and the melting depth value obtained after algorithm processing are analyzed to obtain a change curve diagram of the melting depth, so as to provide display information of the dynamic change of the melting depth to the user, thereby more conveniently providing the user with the convenience of adjusting the keyhole according to the current melting depth.

[0111] Optionally, after obtaining the melting depth data value based on the imaging image information of the target keyhole position and the real-time temperature value of the target keyhole position, the method further includes the steps of:

[0112] obtaining a three-dimensional point cloud image of the target keyhole based on the image information of the target keyhole and the melting depth data value of the target keyhole, and displaying the three-dimensional point cloud image.

[0113] In combination with Figure 3 The method given in the embodiment is further explained more accurately.

[0114] First, start the system, judge whether the laser focal length is normal, if abnormal, measure the distance manually through laser feedback, confirm the best position of the laser power, and record, if normal, execute the next step.

[0115] Perform function selection, judge whether to start near-infrared camera, 3D line laser or temperature control equipment, according to the need of detection accuracy, the above three can be started at the same time, or one of the three functions can be started.

[0116] Obtain visual data by using the near-infrared camera, and perform 2D image analysis on the visual data, and perform 2D image training, and output the analysis result of the 2D image training.

[0117] The 3D point cloud data of the target keyhole position is collected by using a laser scanning galvanometer, the 3D point cloud data is converted into three-dimensional point cloud matrix data, the three-dimensional point cloud matrix data is subjected to Fourier frequency domain transformation to obtain a frequency spectrum waveform diagram, and the waveform diagram is trained to obtain a training result of the waveform diagram. Specifically, in this step, the frequency spectrum waveform diagram is input into a waveform diagram classification model to obtain a predicted detection welding category output by the waveform diagram classification model.

[0118] The waveform diagram training result is combined with the temperature waveform diagram, and a local optimal data weighting is used to obtain an analysis result of the waveform diagram, and the analysis result of the waveform diagram is combined with the analysis result of the 2D diagram training to obtain a final detection result of the penetration data.

[0119] The method disclosed in the embodiment can quickly obtain the detection result of the penetration by using a computer to analyze the collected interference optical signal, the laser scanner galvanometer can acquire the three-dimensional point cloud data of the target keyhole position in real time, so that the user can obtain the 3D imaging of the target keyhole in real time, preview the target welding position, and the method disclosed in the embodiment is easy to implement and operate, has high monitoring efficiency, and has accurate detection result, so that it can be applied to quantitative analysis of welding penetration data, screening of defective products according to the penetration size, and welding depth quality evaluation or welding seam width detection, and can be used to conveniently, quickly and intuitively calibrate and correct problems in welding, thereby providing technical support for improving the qualified rate of welding products.

[0120] Exemplary apparatus

[0121] The embodiment also provides a laser welding penetration real-time detection system, as shown in Figure 4 The system comprises:

[0122] The position acquisition module 100 is configured to acquire an imaging image of a target keyhole and obtain position information of the target keyhole according to the imaging image; and the function thereof is as described in step S1.

[0123] The optical signal acquisition module 200 is configured to introduce detection light and reference light according to the position information of the target keyhole, acquire interference optical signals output by a fiber coupler from the detection light and the reference light reflected by the target keyhole and input to the fiber coupler; and the function thereof is as described in steps S1 and S2.

[0124] The data processing module 300 is configured to transmit the interference optical signals to a data processing unit to obtain penetration information of the target keyhole; and the function thereof is as described in step S3.

[0125] The application provides a laser welding penetration real-time detection method and system, the method comprising the following steps: obtaining an imaging image of a target keyhole, and obtaining position information of the target keyhole according to the imaging image; introducing detection light and reference light according to the position information of the target keyhole; obtaining the detection light and the reference light reflected by the target keyhole and input to a fiber coupler, obtaining interference light signals output by the fiber coupler, transmitting the interference light signals to a data processing unit, and obtaining penetration information of the target keyhole. The method provided in the embodiment firstly obtains an imaging image of a target keyhole, and positions the target keyhole according to the imaging image, so as to accurately introduce the detection light and the reference light into the keyhole, and accurately detect the penetration information. The method is not only convenient to operate and high in detection accuracy, but also can realize real-time detection of the target keyhole and is high in detection efficiency.

[0126] It can be understood that, for those skilled in the art, equivalent replacements or changes can be made according to the technical solutions and the inventive concept of the application, and all the changes or replacements shall belong to the protection scope of the claims appended to the application.

Claims

1. A method for real-time detection of laser welding penetration depth, characterized in that, include: Acquire an image of the target keyhole, and obtain the position information of the target keyhole based on the image; The location information is the three-dimensional location data of the target keyhole; a scanner is used to perform a 3D scan of the target keyhole location to obtain the three-dimensional point cloud data of the target keyhole during welding, and the three-dimensional point cloud data is used to reconstruct the three-dimensional model of the target keyhole, and the three-dimensional structure of the reconstructed target keyhole location is displayed so that the user can preview the welding status of the target keyhole. Based on the location information of the target keyhole, probe light and reference light are introduced; The probe light and reference light are reflected by the target keyhole and input to the fiber coupler, and the interference light signal is output by the fiber coupler; The interference optical signal is transmitted to the data processing unit to obtain the melting depth information of the target keyhole; The step of transmitting the interference optical signal to the data processing unit to obtain the melting depth information of the target keyhole includes: Perform a Fourier transform on the interference optical signal to obtain the spectral waveform. The spectral waveform is input into the waveform classification model to obtain the predicted detection welding category output by the waveform classification model; Based on the weighted value corresponding to the predicted welding category, the data in the spectrum waveform is weighted and fitted, and the penetration depth is estimated using the least squares method to obtain the penetration depth of the target keyhole. Different weighting weights are set for different welding categories. The data in the spectrum waveform is weighted and fitted according to the weighting weights corresponding to the predicted welding categories. The penetration depth is estimated by using the least squares method to obtain the penetration depth of the welding.

2. The method according to claim 1, characterized in that, The step of acquiring the interference light signal output by the fiber optic coupler after the detection light and reference light are reflected by the target keyhole includes: According to the preset acquisition angle and acquisition frequency, the interference light signal information at the location of the target keyhole is acquired; wherein, the preset acquisition angle includes multiple acquisition angles that are superimposed to form an all-round acquisition angle.

3. The method according to claim 2, characterized in that, The step of acquiring an image of the target keyhole and obtaining the location information of the target keyhole based on the image includes: Multiple imaging images of the target keyhole are acquired within a preset swing range, and multiple positional information of the target keyhole is obtained based on each imaging image.

4. The method according to claim 1, characterized in that, Before the step of transmitting the interference optical signal to the data processing unit to obtain the melting depth information of the target keyhole, the method further includes: A preset network model generates a predicted detection welding category corresponding to the sample waveforms based on the sample waveforms in the training set. The training set includes multiple sets of sample waveforms, each set of sample waveforms including: the waveform and the corresponding detection welding category. The detection welding category corresponding to the waveform includes: weld penetration data, pre-weld weld data, post-weld surface height data, and post-weld weld width data. The preset network model corrects the model parameters based on the predicted detection welding category corresponding to the sample waveform and the detected welding category corresponding to the sample waveform, and continues to execute the step of generating the predicted detection welding category corresponding to the sample waveform based on the waveforms in the training set, until the training status of the preset network model meets the preset conditions, so as to obtain the waveform classification model.

5. The method according to claim 4, characterized in that, The step of transmitting the interference optical signal to the data processing unit to obtain the melting depth information of the target keyhole includes: Perform a Fourier transform on the interference optical signal to obtain the spectral waveform. The spectral waveform is input into the waveform classification model to obtain the predicted detection welding category output by the waveform classification model; Based on the predicted welding category, the spectrum waveform is processed by a local optimal data weighting algorithm to obtain a preliminary detection result of the weld depth corresponding to the target keyhole position; The preliminary detection results of the melt depth are combined with the imaging image of the target keyhole to obtain a melt depth variation curve. Based on the aforementioned melt depth variation curve, the melt depth data values ​​corresponding to each welding time value are obtained.

6. The method according to claim 1, characterized in that, After obtaining the melt depth data value, the process further includes the following steps: Based on the imaging image of the target keyhole and the melting depth data of the target keyhole, a three-dimensional point cloud image of the target keyhole is obtained, and the three-dimensional point cloud image is displayed.

7. A real-time laser welding penetration detection system, characterized in that, include: The location acquisition module is used to acquire an imaging image of the target keyhole and obtain the location information of the target keyhole based on the imaging image; The location information is the three-dimensional location data of the target keyhole; a scanner is used to perform a 3D scan of the target keyhole location to obtain the three-dimensional point cloud data of the target keyhole during welding, and the three-dimensional point cloud data is used to reconstruct the three-dimensional model of the target keyhole, and the three-dimensional structure of the reconstructed target keyhole location is displayed so that the user can preview the welding status of the target keyhole. The optical signal acquisition module is used to import probe light and reference light according to the position information of the target keyhole, and acquire the interference light signal output by the optical fiber coupler after the probe light and reference light are reflected by the target keyhole. The data processing module is used to transmit the interference optical signal to the data processing unit to obtain the melting depth information of the target keyhole; The data processing module is used to perform Fourier spectrum transformation on the interference light signal to obtain a spectrum waveform; input the spectrum waveform to a waveform classification model to obtain the predicted detection welding category output by the waveform classification model; perform weighted fitting on the data in the spectrum waveform according to the weighted value corresponding to the predicted detection welding category, and estimate the penetration depth value using the least squares method to obtain the penetration depth value of the target keyhole. Different weighting weights are set for different welding categories. The data in the spectrum waveform is weighted and fitted according to the weighting weights corresponding to the predicted welding categories. The penetration depth is estimated by using the least squares method to obtain the penetration depth of the welding.

Images