Vision-guided free-teaching and free-programming laser welding system and device

Abstract

The application is suitable for the technical field of laser welding, and provides a visual guidance-based trial teaching and programming-free laser welding system and device, wherein the visual guidance-based trial teaching and programming-free laser welding system comprises a welding quality confidence assessment module under visual guidance, which is used for constructing a welding quality confidence assessment model under visual guidance according to an execution precision dynamic index, system synchronization health degree, visual perception confidence and a process stability index, and outputting welding quality confidence; and a welding speed adjustment module, which is used for comparing the welding quality confidence with a preset confidence threshold value, and controlling welding speed adjustment to a target welding speed output by a welding speed adjustment model when the welding quality confidence is less than the preset confidence threshold value. The visual guidance-based trial teaching and programming-free laser welding system provided by the application significantly improves the robustness and self-adaptive ability of the system by introducing a multi-level confidence assessment mechanism.

Description

Technical Field

[0001] This invention belongs to the field of laser welding technology, and in particular relates to a vision-guided, test-teachable, and programming-free laser welding system and device. Background Technology

[0002] Laser welding technology, with its advantages of high energy density, small heat-affected zone, and fast welding speed, is widely used in precision manufacturing fields such as electronic equipment, automotive parts, and medical device production. With increasing demands for automation, vision-guided technology has been introduced into laser welding, using industrial cameras to capture weld seam images in real time for trajectory recognition and tracking, aiming to reduce manual intervention.

[0003] Most existing systems only use vision as a simple position sensor, and the control logic is limited to driving the welding torch to follow after identifying the weld seam position, which has significant limitations.

[0004] In terms of visual perception reliability, the system relies too much on single feature indicators such as image contrast or edge sharpness for judgment. When the working conditions are complex, such as strong reflection on the metal surface, welding fumes, or oxide layer on the workpiece surface, the image quality deteriorates, the reliability of visual recognition decreases sharply, and it is very easy to cause mistracking or target loss, resulting in welding interruption.

[0005] In the process status monitoring stage, the system lacks comprehensive monitoring and quantitative evaluation of the stability of the entire welding process. Key variables such as instantaneous fluctuations in laser power, abnormal vibrations of the wire feeding mechanism, unstable protective gas flow, and optical lens contamination cannot be effectively integrated and linked to the final welding quality risks, making it difficult to identify process anomalies in a timely manner.

[0006] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0007] The purpose of this invention is to provide a vision-guided, test-teachable, and programming-free laser welding system and apparatus to solve the above-mentioned problems.

[0008] This invention is implemented as follows: a vision-guided, test-teachable, and programmable laser welding system, comprising: The visual perception confidence assessment module is used to construct a visual perception confidence assessment model based on the image contrast, image signal-to-noise ratio, feature edge sharpness, and calibrated baseline confidence of the welded area, and output the visual perception confidence. The process stability assessment module is used to construct a process stability assessment model based on the line energy fluctuation index, wire feeding stability index, protective gas stability index, and optical path health index, and output the process stability index. The vision-guided welding quality confidence assessment module is used to construct a vision-guided welding quality confidence assessment model based on the execution accuracy dynamic index, system synchronization health, visual perception confidence, and process stability index, and output the welding quality confidence. The welding speed adjustment module is used to compare the welding quality confidence with a preset confidence threshold. When the welding quality confidence is less than the preset confidence threshold, the welding speed is controlled to be adjusted to the target welding speed output by the welding speed adjustment model.

[0009] A further technical solution, in the visual perception confidence assessment model, is as follows: First, the normalized image contrast index, image signal-to-noise ratio index, and feature edge sharpness index are multiplied to obtain the first intermediate value; Then divide the first intermediate value by [1 minus the product of the preset empirical adjustment coefficient and (1 minus the first intermediate value), and perform a square root operation on the quotient result; Finally, the obtained value is multiplied by the normalized calibration baseline confidence score to obtain the visual perception confidence score. in This is an empirical adjustment coefficient. , For visual perception confidence; The normalized value of the calibration baseline confidence level is obtained by dividing the actual calibration reprojection error by the maximum allowable reprojection error value and then taking the negative exponent. The normalization method for the image contrast index is as follows: subtract the minimum allowable value of the parameter from the actual image contrast, and then divide by the difference between the maximum and minimum allowable values. The normalization method for the image signal-to-noise ratio exponent is to divide the actual signal-to-noise ratio by a reference signal-to-noise ratio value. The normalization method for the feature edge sharpness index is as follows: subtract the minimum allowed value from the actual feature edge sharpness value, and then divide by the difference between the maximum and minimum allowed values.

[0010] In a further technical solution, in the process stability assessment module, the normalization method for the line energy fluctuation index is as follows: take the coefficient of variation of the ratio of laser power to welding speed in the most recent several cycles, divide the actual coefficient of variation by the maximum allowable coefficient of variation, and then subtract the ratio from 1 to obtain the result. The normalization method for the wire feeding stability index is as follows: take the fluctuation range of the wire feeding motor current in the most recent several cycles, divide the actual current fluctuation range by the maximum allowable fluctuation range, and subtract the ratio from 1 to obtain the value. The normalization method for the protective gas stability index is as follows: divide the standard deviation of the actual flow rate of the protective gas by the maximum allowable standard deviation, and subtract the ratio from 1 to obtain the value. The normalized value of the optical path health index is obtained by dividing the absolute deviation between the actual back reflection light intensity and the theoretical value by the allowable tolerance range and then taking the negative exponent.

[0011] In a further technical solution, the process stability index in the process stability assessment model is calculated as follows: First, the geometric mean of the line energy fluctuation index, wire feeding stability index, protective gas stability index and optical path health index is used to obtain the basic stability value. Then, the minimum value is selected from the line energy fluctuation index and the optical path health index, and the product of (1 minus the minimum value) and the preset coupling weight coefficient is calculated. Finally, the process stability index is obtained by multiplying the basic stability value by (1 minus the product).

[0012] A further technical solution is that the normalization method for the dynamic index of execution accuracy is as follows: take the standard deviation of the distance deviation between the laser focus and the predetermined position on the workpiece surface measured by the laser displacement sensor within the most recent number of sampling periods, divide the standard deviation of the actual distance deviation by the calculation result of the maximum allowable standard deviation, and then take the negative exponent to obtain the result. The normalized value of the system synchronization health is obtained by dividing the cyclic communication error count and the time stamp comparison delay by the maximum allowed value, then taking the negative exponent of each, multiplying the two results, and then multiplying by the preset system synchronization health benchmark value.

[0013] A further technical solution, in the vision-guided welding quality confidence assessment model, is as follows: The visual perception confidence score, the process stability index, and a coupling term are each weighted and multiplied together. The coupling term is the larger of the product of system synchronization health and execution accuracy dynamic index and a very small positive number, where each influence coefficient is non-negative and the sum is 1.

[0014] In a further technical solution, the welding speed adjustment module constructs and outputs the target welding speed based on the welding quality confidence, the reference welding speed under the current laser power, the real-time misalignment of the welded workpiece, and the real-time gap distance of the welded workpiece. A further technical solution, the welding speed adjustment model is as follows: First, based on the preset benchmark welding speed, negative compensation is performed according to the normalized real-time misalignment index and real-time gap distance index to obtain the compensation speed, and the compensation speed is not lower than the minimum speed allowed by the system. Then, based on the compensation speed, feedback adjustment is performed according to (the difference between the preset confidence threshold and the current welding quality confidence) to finally obtain the target welding speed; The normalization method for the real-time misalignment index is: to divide the actual measured misalignment by the maximum allowable misalignment. The normalized value of the real-time gap distance index is obtained by dividing the actual measured gap distance by the maximum allowable gap distance.

[0015] A vision-guided, test-less, and programmable laser welding device includes a welding table, on which a laser welding machine is mounted, and a three-axis drive device is connected between the welding table and the laser welding machine. The laser welding machine is connected to a vision monitoring system for monitoring the welding process. The device also includes: Memory, used to store executable instructions; The processor, when executing the executable instructions stored in the memory, implements the corresponding functions of the visual perception confidence assessment module, the process stability assessment module, the visually guided welding quality confidence assessment module, and the welding speed adjustment module as described in any one of claims 1-8.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: The vision-guided, no-test-teach, no-programming laser welding system provided by this invention significantly improves the system's robustness and adaptability by introducing a multi-level confidence assessment mechanism. Even if smoke or glare appears on the workpiece surface, the vision perception confidence assessment module can comprehensively evaluate multiple dimensions such as image contrast, signal-to-noise ratio, and edge sharpness, avoiding the limitations of single-index judgment and thus providing more reliable visual perception information. Simultaneously, the process stability assessment module comprehensively monitors key process parameters such as line energy, wire feeding, shielding gas, and optical path health, enabling timely detection and quantification of process anomalies, thus addressing the problem of disconnect between process status and quality assessment in traditional systems.

[0017] The present invention provides a vision-guided, no-test-teach, no-programming laser welding system. The vision-guided welding quality confidence assessment module integrates visual perception confidence, process stability index, execution accuracy dynamic index, and system synchronization health to construct a comprehensive welding quality confidence assessment model. This enables the system to conduct comprehensive risk assessment across the entire chain from "seeing" to "welding successfully," rather than relying solely on position correction.

[0018] The vision-guided, no-teaching, no-programming laser welding system provided by this invention can adaptively adjust the welding speed according to the welding speed adjustment model when the welding quality confidence level is lower than a preset threshold, in order to cope with comprehensive quality risks. This closed-loop feedback control based on global quality confidence level enables the system to proactively make decisions and optimize welding parameters when faced with dynamic changes in the workpiece, thereby ensuring the stability and consistency of welding quality without the need for manual teaching and programming. This effectively solves the core technical problem of traditional systems' inability to guarantee welding quality under dynamic working conditions. Attached Figure Description

[0019] Figure 1 A schematic diagram of a vision-guided, no-test-teach, no-programming laser welding system. Figure 2 This is a schematic diagram of a vision-guided, test-free, and programmable laser welding device.

[0020] In the attached diagram: 1. Welding table; 2. Three-axis drive device; 3. Laser welding machine. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0022] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0023] like Figure 1 As shown, a vision-guided, no-test-teach, no-programming laser welding system according to an embodiment of the present invention includes: The visual perception confidence assessment module is used to construct a visual perception confidence assessment model based on the image contrast, image signal-to-noise ratio, feature edge sharpness, and calibrated baseline confidence of the welded area, and output the visual perception confidence. The process stability assessment module is used to construct a process stability assessment model based on the line energy fluctuation index, wire feeding stability index, protective gas stability index, and optical path health index, and output the process stability index. The vision-guided welding quality confidence assessment module is used to construct a vision-guided welding quality confidence assessment model based on the execution accuracy dynamic index, system synchronization health, visual perception confidence, and process stability index, and output the welding quality confidence. The welding speed adjustment module is used to compare the welding quality confidence with a preset confidence threshold. When the welding quality confidence is less than the preset confidence threshold, the welding speed is controlled to be adjusted to the target welding speed output by the welding speed adjustment model.

[0024] In this embodiment, the vision-guided, no-teach, no-programming laser welding system refers to an automated system that can acquire welding area information through a vision sensor, autonomously evaluate the welding process status and quality based on this information, and then adaptively adjust welding parameters to achieve laser welding operations without manual teaching and programming.

[0025] The visual perception confidence assessment module receives image data from the visual monitoring system and quantifies the reliability of this image data. This module analyzes the intrinsic quality and calibration status of the images and outputs a numerical value reflecting the credibility of the visual information, namely the visual perception confidence score.

[0026] The process stability assessment module monitors key physical parameters during laser welding and quantitatively evaluates the overall stability of the welding process. This module analyzes indicators related to welding energy, material delivery, environmental control, and the optical system, outputting a numerical value reflecting process stability—the process stability index.

[0027] The vision-guided welding quality confidence assessment module is used to comprehensively evaluate the performance of the entire vision-guided welding system and the reliability of the final weld quality. This module integrates information such as execution accuracy, system synchronization, visual perception reliability, and process stability, and outputs a numerical value reflecting the reliability of the final weld quality, namely the welding quality confidence score.

[0028] The welding speed adjustment module dynamically adjusts the laser welding speed based on the welding quality confidence assessment results. This module compares the welding quality confidence score with a preset threshold and, based on the comparison result, controls the welding speed to adjust towards the target welding speed.

[0029] In a preferred embodiment of the present invention, the visual perception confidence assessment model is calculated as follows: ; in This is an empirical adjustment coefficient. , For visual perception confidence; To determine the baseline confidence level, its normalization method is as follows: the actual calibration reprojection error is divided by the maximum allowable reprojection error value and then the negative exponent is taken. The image contrast index is obtained by normalizing its value by subtracting the minimum allowable value of the parameter from the actual image contrast, and then dividing by the difference between the maximum and minimum allowable values. The image signal-to-noise ratio (SNR) index is normalized by dividing the actual SNR by a reference SNR value. The value ranges from 0 to 1; The feature edge sharpness index is obtained by normalizing the value by subtracting the minimum allowable value from the actual feature edge sharpness value and then dividing by the difference between the maximum and minimum allowable values. , , , .

[0030] In this embodiment, visual perception confidence level This value characterizes the reliability of the visual perception module's understanding and recognition of the image information of the current welding location. The higher the value, the more reliable the visual system's understanding and recognition of the image; conversely, the lower the value, the lower the reliability.

[0031] Empirical adjustment coefficient This is used to adjust the nonlinear term in the formula for calculating visual perception confidence to adapt to the different emphases of visual perception reliability assessment in various welding environments and processes. For example, this coefficient can be optimized using machine learning algorithms based on historical welding data to maximize the correlation between welding quality and visual perception confidence; or it can be manually set by experienced process engineers based on specific workpiece materials, surface treatments, and ambient lighting conditions to reflect the relative importance of various image feature indicators in a specific scenario.

[0032] Calibration of base confidence level This reflects the accuracy and stability of the vision system calibration. Besides obtaining it by dividing the actual calibration reprojection error by the maximum allowable reprojection error and taking the negative exponent, it can also be obtained through statistical analysis of the reprojection errors from multiple calibration results. For example, calculating the standard deviation of the reprojection error, comparing it with a preset threshold, and then normalizing it using the Sigmoid function; alternatively, it can be evaluated based on the stability of feature point extraction from the calibration board under different poses, reducing the base confidence level when feature point extraction is unstable.

[0033] Image contrast index The contrast ratio quantifies the brightness difference between the target area and the background area in an image, and is an important indicator for measuring image sharpness. Besides obtaining it by subtracting the minimum allowable value from the actual image contrast ratio and then dividing by the difference between the maximum and minimum allowable values, it can also be obtained by calculating the gray-level distribution range of the image histogram or by normalizing the average contrast ratio after applying a local contrast enhancement algorithm; alternatively, the root mean square contrast ratio of the image can be used as the original contrast ratio value and then normalized.

[0034] Image signal-to-noise ratio index The signal-to-noise ratio (SNR) reflects the ratio of effective signal to noise in an image and is a key parameter for measuring image quality and anti-interference capability. Besides obtaining it by dividing the actual SNR by a reference SNR value, it can also be obtained by calculating the peak SNR of the image, comparing it with a preset reference peak SNR value, and then normalizing it. Alternatively, an image denoising performance evaluation index based on wavelet transform can be used, comparing the quality of the denoised image with the original image to obtain the SNR index.

[0035] Feature edge sharpness index Characterizing the sharpness and discernibility of feature edges in an image is crucial for accurately identifying weld seam trajectories. Besides obtaining the sharpness value by subtracting the minimum allowable value from the actual feature edge sharpness value and then dividing by the difference between the maximum and minimum allowable values, it can also be obtained by calculating the average value of the image gradient magnitude or by normalizing the response intensity of edge detection algorithms such as the Laplacian operator or the Sobel operator; alternatively, it can be obtained using a Fourier transform-based spectral analysis method, evaluating edge sharpness by analyzing the energy of high-frequency components.

[0036] , , , These indices are normalized to the range of 0 to 1, ensuring their dimensional consistency and comparability when calculating visual perception confidence, avoiding the influence of different physical dimensions on the final results, and enabling each index to participate fairly in the comprehensive evaluation.

[0037] This application's solution overcomes the limitations of traditional methods that rely solely on a single image metric by constructing a comprehensive visual perception confidence assessment model. This model cleverly integrates the calibration of the base confidence level... Image contrast index Image signal-to-noise ratio index and feature edge sharpness index It is integrated into a nonlinear calculation formula. Among them, As a fundamental factor, ensuring the calibration accuracy of the vision system itself is a prerequisite for evaluating its reliability. , , The three indices quantify image quality from different dimensions, reflecting image sharpness, noise resistance, and the accuracy of feature edge recognition, respectively. These three indices appear as a product in the numerator of the formula, meaning that a decrease in any one of these indices directly leads to a reduction in visual perception reliability, reflecting the "weakest link effect." The nonlinear term in the denominator is adjusted by an empirical coefficient. The effect of this is that when the image quality index ( , , When the visual perception confidence is low, The rate of decrease is faster, resulting in a more sensitive response to visual perception risks. This nonlinear relationship allows the system to more accurately and sensitively assess the reliability of visual perception when facing complex conditions (such as strong reflections, dust interference, and poor workpiece surface conditions), avoiding mistracking or target loss caused by errors in judging a single indicator. Through this multi-dimensional, nonlinear comprehensive evaluation, the visual perception confidence assessment module can output a more realistic and comprehensive visual perception reliability index, providing more reliable input for the subsequent visually guided welding quality confidence assessment module and welding speed adjustment module, thereby significantly improving the intelligent adaptability and welding quality stability of the entire laser welding system.

[0038] In a preferred embodiment of the present invention, in the process stability assessment module, the normalization method of the line energy fluctuation index is as follows: take the coefficient of variation of the ratio of laser power to welding speed in the most recent several cycles, divide the actual coefficient of variation by the maximum allowable coefficient of variation, and then subtract the ratio from 1 to obtain the value. The normalization method for the wire feeding stability index is as follows: take the fluctuation range of the wire feeding motor current in the most recent several cycles, divide the actual current fluctuation range by the maximum allowable fluctuation range, and subtract the ratio from 1 to obtain the value. The normalization method for the protective gas stability index is as follows: divide the standard deviation of the actual flow rate of the protective gas by the maximum allowable standard deviation, and subtract the ratio from 1 to obtain the value. The normalized value of the optical path health index is obtained by dividing the absolute deviation between the actual back reflection light intensity and the theoretical value by the allowable tolerance range and then taking the negative exponent.

[0039] In this embodiment, the linear energy fluctuation index aims to quantify the stability of energy input during laser welding. Energy input is the ratio of laser power to welding speed, and its fluctuation directly affects welding quality. Normalization unifies the degree of fluctuation under different operating conditions to a comparable scale. This index can be generated by collecting laser power and welding speed data in real time from a laser power sensor and a welding speed encoder, and calculating the coefficient of variation of the ratio of these data over several recent periods. The coefficient of variation is the ratio of the standard deviation to the mean, reflecting the relative dispersion of the data. Then, the calculated actual coefficient of variation is compared with a preset maximum allowable coefficient of variation, and normalization is performed by subtracting this ratio from 1, ensuring the index value is between 0 and 1, with a larger value indicating better stability.

[0040] The wire feeding stability index is used to evaluate the stability of the wire feeding process. The fluctuation range of the wire feeding motor current directly reflects the operating status of the wire feeding mechanism, such as wire feeding resistance and the uniformity of wire feeding speed. Stable wire feeding is crucial to ensuring the quality of weld formation. The wire feeding motor current value can be read at a fixed frequency via a controller connected to the current feedback interface of the wire feeding motor driver. The maximum and minimum values ​​are recorded over several recent cycles, and the fluctuation range (maximum value minus minimum value) is calculated. This actual fluctuation range is divided by the maximum allowable fluctuation range, and then 1 is subtracted from this ratio to obtain the normalized wire feeding stability index.

[0041] The shielding gas stability index is used to monitor the stability of the shielding gas flow. In laser welding, the shielding gas plays a crucial role in isolating the weld from air, preventing oxidation, and stabilizing the arc; therefore, the stability of its flow rate is critical to weld quality. This index can be calculated by measuring the shielding gas flow rate in real time using a high-precision flow sensor and calculating its standard deviation over several recent periods. Dividing the actual flow rate standard deviation by the maximum permissible standard deviation, and then subtracting this ratio from 1, yields the normalized shielding gas stability index.

[0042] The optical path health index is used to assess the health of the laser transmission optical path, particularly the contamination or damage of optical lenses. Back-reflected light intensity is an important indicator of optical path loss and health. This index can be obtained by placing a back-reflected light sensor in the optical path to measure the back-reflected light intensity in real time. The actual measured back-reflected light intensity is compared with the theoretical value (e.g., a baseline value measured under healthy optical path conditions), and the absolute deviation is calculated. This absolute deviation is divided by the allowable tolerance range, and then the result is subjected to a negative exponentiation to obtain the optical path health index. The negative exponentiation operation allows smaller deviations to correspond to a higher health index, while larger deviations rapidly decrease the health index. Alternatively, the temperature changes of key optical components (such as focusing lenses and protective lenses) in the optical path can be monitored periodically or continuously. When lenses are contaminated, their absorbed laser energy increases, and their temperature rises. The deviation between the actual temperature and the normal operating temperature is used as an indicator of optical path health and normalized by combining it with the negative exponentiation operation.

[0043] The process stability assessment module of this application aims to comprehensively and accurately quantify the fluctuations of key process parameters during laser welding, overcoming the inaccuracies caused by single indicators or simple normalization methods in existing technologies. This module comprehensively considers four core indicators: line energy fluctuation index, wire feed stability index, shielding gas stability index, and optical path health index, employing a refined normalization method to ensure that each indicator accurately reflects the health status of the process stage it represents. The line energy fluctuation index, by calculating and normalizing the coefficient of variation of the ratio of laser power to welding speed, captures the relative change of energy input over time, rather than just the absolute value deviation, which is crucial for assessing the stability of weld penetration and width. The wire feed stability index, by monitoring and normalizing the fluctuation amplitude of the wire feed motor current, directly reflects the mechanical smoothness of the wire feed mechanism, ensuring that the welding wire is fed into the molten pool uniformly and continuously. The shielding gas stability index, by calculating and normalizing the standard deviation of the actual shielding gas flow rate, effectively monitors the uniformity of the shielding gas flow, preventing oxidation or weld defects caused by unstable gas flow. The optical path health index, normalized by performing a negative exponential operation on the absolute deviation between the actual back-reflected light intensity and the theoretical value, exhibits high sensitivity to minute contamination or loss in the optical path, ensuring efficient and stable transmission of laser energy to the workpiece. These four normalized indices together constitute a multi-dimensional and refined evaluation system for the stability of the welding process. They are not isolated but complementary, jointly depicting the overall health of the welding process. For example, when the visual perception confidence assessment module may malfunction under complex conditions, these refined indicators provided by the process stability assessment module can promptly supplement information, verifying or correcting the judgment of welding quality from a physical process perspective. In this way, the system can more comprehensively understand the current welding state, providing more reliable input for the subsequent visually guided welding quality confidence assessment module, thereby improving the accuracy and robustness of the entire system's welding quality assessment. This multi-dimensional and refined normalized evaluation method enables the system to capture potential quality risks from multiple angles, providing a solid foundation for achieving adaptive welding without trial teaching or programming.

[0044] In a preferred embodiment of the present invention, the process stability index in the process stability assessment model is calculated as follows: ; in These are the coupling weight coefficients. , It is a process stability index. The linear energy fluctuation index. The wire feeding stability index. To protect the gas stability index, For the optical path health index, , , .

[0045] In this embodiment, the method for calculating the process stability index aims to comprehensively evaluate multiple key indicators of the welding process, with particular attention to the line energy fluctuation index. He Guanglu Health Index The coupling effect between them. Among them, the geometric mean component... Used for composite linear energy fluctuation index Wire feeding stability index Protective gas stability index and optical path health index These four key process indicators are calculated using a geometric mean rather than an arithmetic mean. This better reflects the product effect of each indicator, preventing a single indicator from being too high and masking the low performance of others, thus ensuring a comprehensive and balanced evaluation. This calculation can be implemented using the processor's internal floating-point unit or software library functions, such as calling the `pow(product, 0.25)` function in C / C++. Alternatively, approximate calculations can be performed using pre-built lookup tables, especially useful in embedded systems where high computational speed is crucial; lookup tables can quickly retrieve results.

[0046] Coupling terms Aimed at capturing line energy fluctuation index Health Index of He Guanglu The coupling effect between these two indices is directly related to the stability and transmission efficiency of laser energy, playing a decisive role in welding quality. A low value in either index indicates a higher potential risk, and this coupling term dynamically reduces the overall process stability index to reflect this increased risk. The minimum function `min( , This can be implemented in software using conditional statements (such as `if-else`). The calculation of the entire coupling term is then performed using standard arithmetic operations. In some high-performance computing scenarios, parallel processing units can be used to calculate simultaneously. and Then, the minimum value is selected through hardware logic gates, and subsequent multiplication and subtraction operations are performed.

[0047] Coupling weight coefficient Used to adjust the correction strength of the coupling term to the process stability index. Its value ranges from [0, 0.2], allowing the system to flexibly adjust according to actual process requirements or empirical knowledge. and Weighting of coupling risks helps avoid excessive or insufficient penalties. It can be used as a system configuration parameter, stored in non-volatile memory, loaded into the processor at system startup, and remain unchanged during runtime. Dynamic adjustments can also be made through online learning or expert systems, for example, based on historical welding defect data and... , Relevance, real-time optimization The value is adjusted to improve the accuracy of the assessment. Linear Energy Fluctuation Index Wire feeding stability index Protective gas stability index and optical path health index These are key indicators for measuring the stability of the process. Their specific normalization methods have been detailed in the above scheme, and they are fed into the calculation model as inputs.

[0048] The scheme in this application uses the linear energy fluctuation index. Wire feeding stability index Protective gas stability index Health Index of He Guanglu Geometric averaging ensures a comprehensive and balanced consideration of all key process indicators, avoiding bias in the overall evaluation due to anomalies in a single indicator. Based on this, a coupling term is introduced, which specifically focuses on the line energy fluctuation index P1 and the optical path health index. The minimum value. Since line energy fluctuations and optical path health are the core risk factors affecting laser welding quality, when either of these indicators performs poorly, this coupling term dynamically adjusts the overall process stability index in a penalizing manner, thus more accurately reflecting the actual coupling risks. Coupling weight coefficient The introduction of this approach allows the system to flexibly adjust the strength of this coupling effect according to actual conditions, effectively capturing risks while avoiding excessive penalties, thus ensuring the balance and reliability of the evaluation results. This comprehensive calculation method enables the process stability assessment module to output a more accurate and reliable process stability index, providing high-quality input for the subsequent vision-guided welding quality confidence assessment module, thereby enhancing the adaptive control capability of the entire vision-guided, no-teach, no-programming laser welding system.

[0049] As a preferred embodiment of the present invention, the normalization method of the execution accuracy dynamic index is as follows: take the standard deviation of the distance deviation between the laser focus and the predetermined position on the workpiece surface measured by the laser displacement sensor within the most recent several sampling periods, divide the standard deviation of the actual distance deviation by the calculation result of the maximum allowable standard deviation, and then take the negative exponent to obtain the result. The normalized value of the system synchronization health is obtained by dividing the cyclic communication error count and the time stamp comparison delay by the maximum allowed value, then taking the negative exponent of each, multiplying the two results, and then multiplying by the preset system synchronization health benchmark value.

[0050] In this embodiment, the dynamic accuracy index aims to quantify the fluctuation of the distance deviation between the laser focus and a predetermined position on the workpiece surface, reflecting the dynamic positioning accuracy of the actuator during welding. The normalized value of this index, by calculating the standard deviation of the distance deviation, effectively captures the fluctuation of the position deviation, avoiding evaluation distortion caused by single measurement or instantaneous errors. Laser displacement sensors can be implemented using various technologies, such as laser sensors based on triangulation principles, laser sensors based on time-of-flight principles, or laser sensors based on confocal principles. These can measure the distance between the laser focus and the workpiece surface with high precision. The standard deviation of the distance deviation can be calculated using the sliding window method, i.e., calculating the standard deviation of data over several consecutive sampling periods, or using the exponentially weighted moving average method, assigning higher weight to recent data. The maximum allowable standard deviation can be determined based on specific welding process requirements, the accuracy level of the equipment used, or through historical data statistical analysis. Negative exponential calculation is an effective method to transform deviation values ​​into confidence-oriented indicators; for example, it can be used... In the form of, To adjust the coefficient, ensure that the exponent value is between 0 and 1, and that the larger the value, the higher the accuracy.

[0051] System synchronization health is used to comprehensively assess the health status of the entire laser welding system in terms of communication and timing, ensuring precise synchronization of data transmission and operation execution between modules. Cyclic communication error count refers to the count of error frames, timeouts, or data retransmissions occurring in industrial Ethernet or fieldbus protocols; these errors directly reflect the reliability of the communication link. Timing stamp comparison delay refers to the difference in timestamps between data acquisition or command execution between different modules within the system (e.g., the vision system and the robot controller), reflecting the accuracy of internal timing synchronization. The maximum allowable value can be set according to communication protocol specifications, overall system response time requirements, or empirical values. Similarly, negative exponential operations are used to convert errors or delays into health indicators, ensuring normalization. The preset system synchronization health baseline value can be a constant between 0 and 1, used to adjust the overall magnitude of the system synchronization health in the final multiplication operation to meet actual needs.

[0052] This application's solution provides specific normalized values ​​for the dynamic execution accuracy index and the system synchronization health index, enabling accurate and reliable quantification of these two key indicators. The dynamic execution accuracy index, calculated by negatively exponentially calculating the standard deviation of the distance deviation between the laser focus and the predetermined position on the workpiece surface, dynamically captures the positioning accuracy fluctuations of the actuator during welding, transforming it into an indicator reflecting accuracy confidence. The system synchronization health index, by integrating cyclic communication error counting and timing stamp comparison delays, multiplying them by negative exponential calculations, and then multiplying by a preset benchmark value, comprehensively assesses the reliability of internal system communication and timing. These normalized indices, serving as crucial inputs to the vision-guided welding quality confidence assessment model, ensure the accuracy and robustness of the welding quality confidence assessment. When these indices are accurately calculated and input into the welding quality confidence assessment model, they more precisely reflect the actual quality risks of the current welding process, thereby guiding the welding speed adjustment module to make more reasonable and timely adjustments to the welding speed, effectively addressing dynamic interference during the welding process and ensuring welding quality.

[0053] In a preferred embodiment of the present invention, the calculation method for welding quality confidence in the vision-guided welding quality confidence assessment model is as follows: ; in The visual perception influence coefficient. The process stability influence coefficient, This is the coupling effect coefficient. , , ,and ; For visual perception confidence. It is a process stability index. For the confidence level of welding quality, To implement the dynamic index of accuracy, To synchronize system health status, It is a very small positive number.

[0054] In this embodiment, the welding quality confidence assessment model aims to comprehensively evaluate multiple factors in the laser welding process to quantify the reliability of the current welding quality and provide a unified, quantifiable index for subsequent adaptive control decisions. This model can be built based on machine learning algorithms, such as support vector machines or neural networks, learning the complex nonlinear relationship between various input parameters and welding quality through training data; alternatively, it can employ expert systems or fuzzy logic systems to encode the domain expert's experiential knowledge into rules to assess the welding quality confidence. It is a normalized value between 0 and 1, used to represent the reliability or credibility of the current welding quality. The higher the value, the more reliable the welding quality. It can be used as a basis for adjusting the welding speed or other process parameters of the system.

[0055] This calculation method uses a weighted index in a product form, effectively integrating multiple independent evaluation indicators. A significant decrease in any one indicator leads to a reduction in the overall welding quality confidence level, thus achieving sensitive detection of welding risks. This calculation method can be performed in real-time by the system's built-in mathematical processing unit or a dedicated DSP chip. Visual perception influence coefficient. This is used to adjust the weight of visual perception confidence in the overall welding quality confidence assessment. For example, in welding scenarios with high visual dependence, the weight can be appropriately increased. The value can be reduced; however, in scenarios with poor visual conditions or higher requirements for process stability, the value can be lowered. This coefficient can be preset by system operators based on actual process requirements, or adaptively optimized through historical data analysis and machine learning methods. (Process stability impact coefficient) This is used to adjust the weight of the process stability index in the overall weld quality confidence assessment. For example, in precision welding where the stability of the welding process is extremely important, it can be increased... The value can be reduced; however, in scenarios where higher accuracy of visual guidance is required, the value can be reduced. This coefficient can also be set by system operators based on experience, or dynamically adjusted through online learning algorithms. Coupling Influence Coefficient Used to adjust the dynamic index of execution accuracy Synchronize health with system The weight of the product term in the overall welding quality confidence assessment reflects the degree to which the reliability of the system execution level affects the final welding quality. For example, in complex trajectory welding where strict requirements are placed on motion control accuracy and system real-time performance, it can improve... The value of this coefficient. This coefficient can be initially set based on the equipment hardware performance and control system architecture, and can be fine-tuned according to the system operating status. These weighting coefficients , , All values ​​are non-negative, and their sum is 1, which ensures the normalization of the influence of each factor, avoids imbalance in the evaluation results due to excessively large or small weights, and guarantees the rationality and comparability of the evaluation results.

[0056] Among them, visual perception confidence This reflects the reliability of the vision system in acquiring and processing image information of the welding area. Its calculation method has been detailed in the above scheme, comprehensively considering image contrast, signal-to-noise ratio, feature edge sharpness, and calibration baseline confidence. Process stability index. The stability of the laser welding process itself is reflected in the calculation method described in detail in the above scheme, which comprehensively considers the fluctuation of line energy, wire feeding stability, shielding gas stability, and optical path health. The dynamic index of execution accuracy $A$ reflects the accuracy and stability of the welding actuator during dynamic tracking. Its normalized value method has been described in detail in the above scheme, and it is quantified by the standard deviation of the distance deviation measured by the laser displacement sensor. System synchronization health. This reflects the reliability of the entire control system in terms of data communication and timing synchronization. Its normalization method has been detailed in the above scheme, and it is evaluated through cyclic communication error counting and timing stamp comparison delay. Extremely small positive numbers. Used to prevent the execution accuracy dynamic index $A$ or system synchronization health status from being affected. When the product is close to or equal to zero, the entire product term becomes too small, thus affecting the confidence level of welding quality. Abnormally low speed or unstable calculation. By taking... This ensures that the term always has a minimum effective contribution, enhancing the robustness of the model. The value can be set according to the system's minimum execution reliability requirements, for example, it can be set to 0.001 or 0.00001.

[0057] This application's solution addresses the shortcomings of traditional assessment methods, such as being one-sided and easily overlooking key factors, by constructing a comprehensive welding quality confidence assessment model. The core of this model lies in using a weighted exponential product to calculate the welding quality confidence level. Specifically, it will assess the confidence level of visual perception. Process stability index and execution accuracy dynamic index Synchronize health with system The product of these three dimensions serves as the three main evaluation dimensions. The advantage of this product form is that a significant decrease in any one dimension will affect the final weld quality confidence level. This significantly reduces the risk, allowing for the sensitive detection of potential welding quality risks. These three dimensions are not simply added together, but rather weighted by their respective coefficients. , , Weighting is applied, and the factors are multiplied exponentially. This exponential weighting method makes the influence of each factor on the final confidence level non-linear, better simulating the coupling effect of various risk factors in actual welding processes. Simultaneously, the sum of all weight coefficients is 1, ensuring the normalization and reasonableness of the evaluation results. Furthermore, to enhance the robustness of the model, dynamic exponents of execution accuracy are processed. Synchronize health with system When multiplying, a very small positive number is introduced. By taking Even in Even when the value is very close to zero, it ensures that this item makes a minimal but effective contribution to the overall confidence score, avoiding the problem of the entire confidence score calculation failing due to a single extreme value. Through the above mechanism, the solution in this application can effectively integrate multi-source heterogeneous information from visual perception, process flow, and system execution to form a comprehensive, objective, and robust quantitative indicator of welding quality risk. This comprehensive assessment capability enables the system to overcome the limitations of a single indicator, more accurately determine the reliability of the current welding state, and provide a solid data foundation for subsequent adaptive welding speed adjustments.

[0058] In a preferred embodiment of the present invention, the welding speed adjustment module constructs and outputs a target welding speed based on the welding quality confidence, the reference welding speed under the current laser power, the real-time misalignment of the welded workpiece, and the real-time gap distance of the welded workpiece. In this embodiment, the welding speed adjustment module is a key functional unit in the system. Its core function is to dynamically adjust the laser welding speed based on a comprehensive evaluation of the welding process status. This can be achieved through a software algorithm module that receives input from the evaluation unit and outputs control commands to the welding actuator. The welding speed adjustment model is the core algorithm or logical structure used to calculate and determine the target welding speed. It does not simply perform linear adjustments but comprehensively considers multiple input parameters and outputs a suggested speed that optimizes welding quality through preset mathematical relationships, lookup tables, expert system rules, or machine learning algorithms. For example, the model can be a multivariate regression model, fitting each input parameter as an independent variable and the target welding speed as a dependent variable; or it can be a fuzzy logic-based decision system that infers based on a fuzzy set of input parameters and preset rules. The welding quality confidence score is a quantitative assessment of the system's reliability of the current welding quality. Its value is typically between 0 and 1, with higher values ​​indicating more reliable welding quality. It comprehensively reflects information from multiple dimensions, including visual perception, process stability, and system synchronization health.

[0059] In the welding speed adjustment model, welding quality confidence is an important input used to indicate the overall health of the current welding process, thereby guiding the model to make more conservative or more aggressive speed adjustments when the quality risk is high.

[0060] The reference welding speed under the current laser power refers to the initial or reference welding speed that achieves good welding results under the current laser power conditions, determined through pre-calibration or experience. It provides a basic, process-matched starting point for welding speed adjustment models. This reference speed can be obtained through offline experiments, process manual lookups, or calculations based on empirical formulas for material type and thickness.

[0061] Real-time misalignment of a welded workpiece refers to the lateral deviation between the laser focus and the predetermined weld centerline during the welding process. It reflects the real-time error in workpiece assembly accuracy or fixture positioning. Real-time misalignment can be accurately measured by analyzing weld images in real time using a vision sensor (such as a CCD camera) and combining this with image processing algorithms (such as edge detection and feature matching).

[0062] The real-time gap distance refers to the actual size of the gap between the workpieces being welded during the welding process. Fluctuations in the gap distance can significantly affect weld formation and penetration depth. The real-time gap distance can be measured using images acquired by vision sensors, such as by analyzing the distance between the two edges of the weld, or by using a laser displacement sensor for non-contact measurement.

[0063] The target welding speed is the final welding speed calculated by the welding speed adjustment model based on all input parameters, used to guide the actual operation of the laser welding equipment. It is a dynamically adjusted value designed to ensure that welding quality remains at an optimal level under various operating conditions. This target welding speed is sent to the motion control system of the laser welding machine via a control interface to adjust the movement speed of the welding head in real time.

[0064] The operational logic of this scheme is as follows: First, the system continuously acquires and evaluates the welding quality confidence level. This confidence level comprehensively reflects information from multiple dimensions, including visual perception, process stability, and system synchronization health, providing a comprehensive quantitative indicator for welding quality. Simultaneously, the system acquires a baseline welding speed at the current laser power as an initial reference for speed adjustment. Furthermore, through a visual monitoring system or dedicated sensors, the system acquires in real-time the misalignment and gap distance of the welded workpiece; these geometric parameters directly affect the weld formation and quality.

[0065] The welding speed adjustment model takes the aforementioned welding quality confidence level, reference welding speed, real-time misalignment, and real-time gap distance as inputs, and performs comprehensive calculations and judgments through its internally preset algorithm logic. This model does not simply adjust based on a single factor, but rather couples and analyzes these interrelated factors. For example, when the welding quality confidence level is low, the model may tend to reduce the welding speed to ensure penetration and forming; when the misalignment or gap distance is large, the model may also adjust the speed accordingly to compensate for the impact of these geometric defects. Ultimately, the model outputs an optimized target welding speed that better adapts to the current complex welding conditions. This speed is then sent to the motion control system of the laser welding machine via the welding speed adjustment module, thereby achieving refined and adaptive control of the welding process.

[0066] In this way, the solution proposed in this application, based on the basic welding quality confidence assessment, further incorporates consideration of real-time geometric parameters and baseline process conditions, making the adjustment of welding speed more precise and intelligent. This effectively addresses the problem that traditional solutions, which rely solely on a single confidence index, cannot adequately adapt to real-time geometric changes in the workpiece.

[0067] As a preferred embodiment of the present invention, the welding speed adjustment model is as follows: ; ; in This is the offset compensation coefficient. This is the gap distance compensation coefficient. , , The minimum allowable welding speed of the system is set according to the welding process requirements. To preset the baseline welding speed, To compensate for welding speed, For the confidence level of welding quality, To preset the reliability threshold, For feedback gain coefficient, , Target welding speed; The real-time misalignment index is normalized by dividing the actual measured misalignment by the maximum permissible misalignment. ; The real-time gap distance index is obtained by normalizing the actual measured gap distance by dividing it by the maximum allowable gap distance. .

[0068] In this embodiment, the welding speed adjustment model is a mathematical or algorithmic framework used to dynamically calculate and output an optimized target welding speed based on real-time monitored welding status parameters. Its core lies in establishing a quantitative relationship between input parameters (such as welding quality confidence, misalignment, gap distance, etc.) and output parameters (target welding speed). This model can be implemented using pre-trained machine learning algorithms (such as neural networks or support vector machines), learning the complex nonlinear relationship between input and output through a large amount of welding data; alternatively, the model can also be implemented using an analytical model built based on physical principles and empirical formulas, directly calculating the target welding speed through a series of mathematical expressions.

[0069] formula Used to calculate feedforward compensated welding speed It achieves this by adjusting the preset reference welding speed. Corrections are made to proactively address the impact of workpiece geometric deviations (misalignment and gap distance) on weld quality. The `max` function ensures that the calculated speed does not fall below the system's minimum allowable welding speed. This ensures the stability and safety of the welding process. In the controller, this formula can be directly calculated in real time via the floating-point unit, processing the data collected by the sensors. and Substitute the value and combine it with the preset value. , , , Parameters, quickly obtained Alternatively, the calculation logic of the formula can be embedded in the firmware of a programmable logic controller (PLC) or digital signal processor (DSP) to achieve real-time feedforward adjustment of the welding speed through periodic sampling and calculation.

[0070] formula Used to calculate the final target welding speed It compensates for welding speed via feedforward. Based on this, a confidence level based on welding quality was introduced. The feedback adjustment mechanism. When the confidence level of actual welding quality... Below the preset confidence threshold At that time, the formula will be based on the feedback gain coefficient. Reducing the welding speed allows more time for energy input or defect repair, thereby improving weld quality; conversely, if... Higher than If necessary, the speed can be appropriately increased to improve efficiency. This formula can be implemented in the main controller or motion controller by receiving feedforward compensation welding speed. and welding quality confidence and combined with preset and The value is calculated in real time to obtain the final value. Alternatively, the feedback adjustment logic can be implemented as an independent functional module at the software level, running in parallel with the feedforward control module, and its output adjustment factor can be applied to the feedforward speed to form the final target speed command.

[0071] The (misalignment compensation coefficient) is a dimensionless empirical parameter used to quantify the misalignment index. The degree of influence on welding speed. Its value ranges from [0, 0.3], representing the percentage reduction in welding speed that should occur as misalignment increases. This coefficient can be calibrated through offline experiments and expert experience. Welding tests are conducted under different misalignment conditions to observe changes in welding quality, and a suitable coefficient is determined based on experience. The value; or, the coefficient can also be optimized using an online adaptive algorithm, for example, dynamically adjusted based on feedback from the welding quality confidence level during the welding process. The value is adjusted to better adapt to the current working conditions.

[0072] The gap compensation coefficient is a dimensionless empirical parameter used to quantify the influence of the gap exponent $g$ on welding speed. Its value ranges from [0, 0.3], representing the proportion by which the welding speed should decrease as the gap distance increases. similar, Welding experiments under different gap distances can be conducted, and the process engineer can empirically set the parameters based on the welding quality assessment results; or, It can also be dynamically adjusted through adaptive control strategies based on fuzzy logic or reinforcement learning to cope with the differences in gap sensitivity of workpieces of different materials and thicknesses.

[0073] (The minimum allowed welding speed) is a lower limit set according to specific welding process requirements and equipment capabilities. It ensures that even under the most unfavorable conditions, the welding speed will not fall below this value to avoid problems such as excessive heat input, burn-through, or weld pool instability caused by excessively low speed. This value is typically determined by welding process experts through experimental verification and process specifications based on factors such as material type, plate thickness, and laser power, and is fixed as a system parameter in the controller; or, in some advanced systems, It can dynamically retrieve information from a preset process database based on the material, thickness, and other information of the current welding task.

[0074] (Preset reference welding speed) is an initial welding speed preset according to process requirements under ideal or standard welding conditions. It is the starting point for feedforward compensation and feedback adjustment, representing the expected welding speed in the absence of external interference or quality risks. Typically, process engineers set the parameters offline or based on experience, considering the workpiece material, thickness, joint type, and desired welding quality and efficiency; or... It can also be optimized through simulation software, which simulates the welding process in a virtual environment to find the benchmark speed that can achieve the best welding effect under specific conditions.

[0075] (Welding quality confidence) is a comprehensive indicator that reflects the overall health of the current welding process and the reliability of the final weld quality. It is usually derived from a comprehensive evaluation of multiple sub-indicators, including visual perception confidence, process stability index, execution accuracy dynamic index, and system synchronization health. The results are calculated and output in real time by a dedicated evaluation module (such as the vision-guided welding quality confidence evaluation module in the system described above), which integrates multiple sensor data and evaluation models; or, The calculation can be based on multi-sensor fusion technology, which uses algorithms such as Kalman filtering or particle filtering to fuse information from different sensors in order to obtain a more accurate and robust confidence assessment.

[0076] (Preset confidence threshold) is a pre-defined critical value used to determine the confidence level of the current welding quality. Is it within an acceptable range? When Below At that time, the system considers there to be a risk to the welding quality and requires intervention and adjustment. The settings are determined by process experts based on tolerance for welding quality, product standards, and production efficiency requirements; or, Alternatively, it can be determined through statistical analysis of historical welding data, for example, by identifying factors that will lead to defective products. Use the value as a reference and set a threshold that is slightly higher than that value.

[0077] The feedback gain coefficient is a positive dimensionless parameter used to adjust the confidence level of weld quality. Compared with the preset confidence threshold The difference between them affects the target welding speed The intensity of feedback adjustment. Larger. A higher value means that the system responds more quickly and adjusts more drastically to changes in quality confidence. The proportional gain parameter of the PID controller (proportional-integral-derivative controller) can be set, and a suitable value that achieves both fast response and avoids overshoot can be found through experimental debugging; or, Dynamic adjustments can also be made through adaptive control algorithms; for example, using a smaller [control] parameter at system startup. To ensure stability, the value should be appropriately increased when a large quality fluctuation is detected. Values ​​are used to expedite the response.

[0078] The target welding speed is the final output of the welding speed adjustment model, used to guide the actual operation of the laser welding machine. It is the optimal speed after feedforward compensation and feedback adjustment, designed to ensure stable welding quality under dynamic operating conditions. As a command signal, it is sent to the motion controller via industrial Ethernet or CAN bus, and the motion controller drives the laser welding head to move at that speed; or... It can also directly control the motion axis of the welding equipment through analog output or pulse signal.

[0079] The (Real-time Misalignment Index) is a normalized representation of the real-time misalignment, with a value range of [0,1]. It is obtained by dividing the actual measured misalignment by the maximum allowable misalignment, allowing misalignment data of different dimensions to participate uniformly in the model calculation. The actual misalignment can be measured in real time by a vision sensor (such as an industrial camera with image processing algorithms), and then the controller performs normalization calculations; alternatively, the misalignment can also be measured by a laser displacement sensor or a structured light sensor to measure the workpiece edge position, and then the actual misalignment is calculated and normalized.

[0080] The (real-time gap distance index) is a normalized representation of the real-time gap distance, with a value range of [0,1]. It is obtained by dividing the actual measured gap distance by the maximum allowable gap distance, allowing gap distance data of different dimensions to participate uniformly in the model calculation. The actual gap distance can be measured in real time by a vision sensor (such as an industrial camera with image processing algorithms), for example, through edge detection and distance calculation; alternatively, the gap distance can also be measured by a laser triangulation sensor or a confocal displacement sensor to measure the distance between the two sides of the weld, and then the actual gap distance can be calculated and normalized.

[0081] The welding speed adjustment model in this application achieves adaptive and precise control of welding speed by combining feedforward compensation and feedback adjustment mechanisms, effectively addressing the challenges posed by workpiece misalignment, gap fluctuations, and dynamic changes in welding quality. The model first uses a feedforward compensation stage to adjust the welding speed based on a preset benchmark welding speed. Real-time misalignment index and real-time gap distance index To calculate feedforward compensated welding speed Specifically, when the vision system detects misalignment or gaps in a workpiece, the model will use a misalignment compensation coefficient. and gap distance compensation coefficient For reference speed A deceleration correction is implemented. This correction is a proactive, predictive adjustment designed to increase laser energy input by reducing speed before welding defects occur, effectively compensating for the risk of insufficient energy due to geometric deviations. Simultaneously, a feedforward compensation welding speed is ensured through the `max` function. Always maintain a welding speed no lower than the minimum allowed by the system. This provides basic safety assurance for the welding process, preventing new process problems such as molten pool instability or burn-through caused by excessive deceleration. Building on this, the model further incorporates a feedback adjustment mechanism, utilizing feedforward compensation to adjust the welding speed. and real-time welding quality confidence To calculate the final target welding speed Welding quality confidence level It is a comprehensive indicator that integrates information from multiple aspects, including visual perception confidence level, process stability index, execution accuracy dynamic index, and system synchronization health. When the system assesses the welding quality confidence level... Below the preset confidence threshold When this occurs, it indicates a potential risk to the current welding quality, and the model will adjust the feedback gain coefficient accordingly. This further reduces welding speed. This feedback mechanism enables the system to respond and correct complex quality problems that may occur during the actual welding process but cannot be fully predicted or compensated for by the feedforward mechanism in real time. For example, when visual perception is reduced due to smoke interference, or when abnormalities occur in the process (such as wire feeding, shielding gas)... When the speed is reduced, the system can decelerate in a timely manner, providing a longer effective time for the laser energy to act, thereby improving welding quality. Through this combined feedforward and feedback strategy, the welding speed adjustment model of this application can not only anticipate known geometric deviations but also respond in real time to unknown or dynamically changing quality risks. The feedforward compensation mechanism provides a fast and direct response, while the feedback adjustment mechanism provides global, closed-loop control based on comprehensive quality assessment. This combination of mechanisms enables the system to highly adaptively adjust the welding speed without manual teaching or programming, thus significantly improving the stability and reliability of welding quality under dynamically changing conditions.

[0082] like Figure 2As shown, a vision-guided, test-less, and programmable laser welding device includes a welding table 1, on which a laser welding machine 3 is mounted. A three-axis drive device 2 is connected to the welding table 1 to move the laser welding machine 3 in the xyz directions. The laser welding machine is connected to a vision monitoring system for monitoring the welding process. The device also includes: Memory, used to store executable instructions; The processor, when executing the executable instructions stored in the memory, implements the corresponding functions of the visual perception confidence assessment module, the process stability assessment module, the visually guided welding quality confidence assessment module, and the welding speed adjustment module as described in any one of claims 1-8.

[0083] In this embodiment, the welding station provides stable physical support, ensuring that the laser welding machine and the vision monitoring system operate in a consistent environment, avoiding external interference that could affect welding accuracy. The laser welding machine performs the welding task, and its connected vision monitoring system continuously captures image data of the welding area, providing raw input for subsequent evaluation. The executable instructions stored in the memory contain the system algorithm logic, enabling the device to be programmable and adaptive. When the processor executes these instructions, it first analyzes the received image data through the visual perception confidence assessment module, combining image contrast, image signal-to-noise ratio, feature edge sharpness, and calibration baseline confidence to output a visual perception confidence score. Specifically, image contrast is obtained by calculating the variance of the image grayscale histogram, image signal-to-noise ratio is estimated by comparing signal strength and noise intensity, feature edge sharpness is determined by detecting edge gradient strength using a gradient operator, and the calibration baseline confidence score is set based on historical calibration data. These parameters are weighted and averaged to form the visual perception confidence score, effectively overcoming the limitations of judging by a single indicator.

[0084] Furthermore, the process stability assessment module monitors key parameters such as laser power, wire feed motor current, shielding gas flow rate, and back-reflected light intensity, calculating the line energy fluctuation index, wire feed stability index, shielding gas stability index, and optical path health index, and outputting the process stability index. For example, the line energy fluctuation index is obtained by real-time monitoring of the fluctuation in the ratio of laser power to welding speed; the wire feed stability index is assessed by feedback from the wire feed motor encoder to evaluate wire feed uniformity; the shielding gas stability index is determined by the fluctuation output from the flow sensor; and the optical path health index is indirectly assessed by the back-reflected light intensity. The vision-guided welding quality confidence assessment module integrates visual perception confidence, process stability index, execution accuracy dynamic index, and system synchronization health to generate a welding quality confidence score. The execution accuracy dynamic index is obtained by calculating trajectory deviation using a laser displacement sensor, and the system synchronization health is determined by monitoring communication delays between modules. These indices are linearly combined to form a comprehensive quality assessment.

[0085] When the welding quality confidence level falls below a preset confidence threshold, the welding speed adjustment module invokes the welding speed adjustment model. Based on the welding quality confidence level, the baseline welding speed at the current laser power, the real-time workpiece misalignment, and the real-time gap distance, it calculates the target welding speed and controls the welding speed adjustment. For example, if visual perception confidence decreases due to fumes during welding, or if process stability index decreases due to wire feeding jamming, the system will automatically reduce the welding speed to increase the laser energy application time and compensate for the impact of dynamic interference factors. Through the above technical solution, this application achieves autonomous confidence assessment and proactive parameter optimization for the entire chain from "seeing" to "welding well." In complex working conditions such as workpiece surface reflection, fume interference, or fluctuations in shielding gas flow, the system can comprehensively quantify quality risks using multi-dimensional indicators and adaptively adjust the welding speed. This ensures welding quality stability under dynamically changing working conditions without the need for manual trial teaching and programming, effectively solving the technical limitations of traditional systems.

[0086] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A vision-guided, test-teachable, and programmable laser welding system, characterized in that, include: The visual perception confidence assessment module is used to construct a visual perception confidence assessment model based on the image contrast, image signal-to-noise ratio, feature edge sharpness, and calibrated baseline confidence of the welded area, and output the visual perception confidence. The process stability assessment module is used to construct a process stability assessment model based on the line energy fluctuation index, wire feeding stability index, protective gas stability index, and optical path health index, and output the process stability index. The vision-guided welding quality confidence assessment module is used to construct a vision-guided welding quality confidence assessment model based on the execution accuracy dynamic index, system synchronization health, visual perception confidence, and process stability index, and output the welding quality confidence. The welding speed adjustment module is used to compare the welding quality confidence with a preset confidence threshold. When the welding quality confidence is less than the preset confidence threshold, the welding speed is controlled to be adjusted to the target welding speed output by the welding speed adjustment model.

2. The vision-guided, test-less, and programming-free laser welding system according to claim 1, characterized in that, In the visual perception confidence assessment model, the visual perception confidence is calculated as follows: First, the normalized image contrast index, image signal-to-noise ratio index, and feature edge sharpness index are multiplied to obtain the first intermediate value; Then divide the first intermediate value by [1 minus the product of the preset empirical adjustment coefficient and (1 minus the first intermediate value), and perform a square root operation on the quotient result; Finally, the obtained value is multiplied by the normalized calibration baseline confidence score to obtain the visual perception confidence score. in This is an empirical adjustment coefficient. For visual perception confidence; The normalized value of the calibration baseline confidence level is obtained by dividing the actual calibration reprojection error by the maximum allowable reprojection error value and then taking the negative exponent. The normalization method for the image contrast index is as follows: subtract the minimum allowable value of the parameter from the actual image contrast, and then divide by the difference between the maximum and minimum allowable values. The normalization method for the image signal-to-noise ratio exponent is to divide the actual signal-to-noise ratio by a reference signal-to-noise ratio value. The normalization method for the feature edge sharpness index is as follows: subtract the minimum allowed value from the actual feature edge sharpness value, and then divide by the difference between the maximum and minimum allowed values.

3. The vision-guided, test-teachable, and programming-free laser welding system according to claim 1, characterized in that, In the process stability assessment module, the normalization method for the line energy fluctuation index is as follows: take the coefficient of variation of the ratio of laser power to welding speed in the most recent several cycles, divide the actual coefficient of variation by the maximum allowable coefficient of variation, and then subtract the ratio from 1 to obtain the value. The normalization method for the wire feeding stability index is as follows: take the fluctuation range of the wire feeding motor current in the most recent several cycles, divide the actual current fluctuation range by the maximum allowable fluctuation range, and subtract the ratio from 1 to obtain the value. The normalization method for the protective gas stability index is as follows: divide the standard deviation of the actual flow rate of the protective gas by the maximum allowable standard deviation, and subtract the ratio from 1 to obtain the value. The normalized value of the optical path health index is obtained by dividing the absolute deviation between the actual back reflection light intensity and the theoretical value by the allowable tolerance range and then taking the negative exponent.

4. The vision-guided, no-test-teach, no-programming laser welding system according to claim 3, characterized in that, In the process stability assessment model, the process stability index is calculated as follows: First, the geometric mean of the line energy fluctuation index, wire feeding stability index, protective gas stability index and optical path health index is used to obtain the basic stability value. Then, the minimum value is selected from the line energy fluctuation index and the optical path health index, and the product of (1 minus the minimum value) and the preset coupling weight coefficient is calculated. Finally, the process stability index is obtained by multiplying the basic stability value by (1 minus the product).

5. The vision-guided, test-less, and programming-free laser welding system according to claim 1, characterized in that, The normalization method for the dynamic index of execution accuracy is as follows: take the standard deviation of the distance deviation between the laser focus and the predetermined position on the workpiece surface measured by the laser displacement sensor within the most recent sampling period, divide the standard deviation of the actual distance deviation by the calculated result of the maximum allowable standard deviation, and then take the negative exponent to obtain the result. The normalized value of the system synchronization health is obtained by dividing the cyclic communication error count and the time stamp comparison delay by the maximum allowed value, then taking the negative exponent of each, multiplying the two results, and then multiplying by the preset system synchronization health benchmark value.

6. The vision-guided, no-test-teach, no-programming laser welding system according to claim 5, characterized in that, In the visually guided welding quality confidence assessment model, the welding quality confidence is calculated as follows: The visual perception confidence score, the process stability index, and a coupling term are each weighted and multiplied together. The coupling term is the larger of the product of system synchronization health and execution accuracy dynamic index and a very small positive number, where each influence coefficient is non-negative and the sum is 1.

7. The vision-guided, no-test-teach, no-programming laser welding system according to claim 1, characterized in that, In the welding speed adjustment module, the welding speed adjustment model is constructed and outputs the target welding speed based on the welding quality confidence, the reference welding speed under the current laser power, the real-time misalignment of the welded workpiece, and the real-time gap distance of the welded workpiece.

8. The vision-guided, no-test-teach, no-programming laser welding system according to claim 7, characterized in that, The welding speed adjustment model is as follows: First, based on the preset benchmark welding speed, negative compensation is performed according to the normalized real-time misalignment index and real-time gap distance index to obtain the compensation speed, and the compensation speed is not lower than the minimum speed allowed by the system. Then, based on the compensation speed, feedback adjustment is performed according to (the difference between the preset confidence threshold and the current welding quality confidence) to finally obtain the target welding speed; The normalization method for the real-time misalignment index is: to divide the actual measured misalignment by the maximum allowable misalignment. The normalized value of the real-time gap distance index is obtained by dividing the actual measured gap distance by the maximum allowable gap distance.

9. A vision-guided, test-less, and programmable laser welding device, comprising a welding table, a laser welding machine mounted on the welding table, a three-axis drive device connected between the laser welding machine and the welding table, and a vision monitoring system for monitoring the welding process connected to the laser welding machine, characterized in that... Also includes: Memory, used to store executable instructions; The processor, when executing the executable instructions stored in the memory, implements the corresponding functions of the visual perception confidence assessment module, the process stability assessment module, the visually guided welding quality confidence assessment module, and the welding speed adjustment module as described in any one of claims 1-8.

Images