A method and system for monitoring the quality of laser welding of aluminum single-panel

Abstract

This invention belongs to the field of image data processing technology, specifically relating to a method and system for monitoring the quality of laser welding of aluminum single-panel panels. The method includes: extracting the centroid displacement of the weld pool in the aluminum single-panel welding process and denoising it to obtain an oscillation velocity sequence; constructing a state vector through phase space reconstruction and calculating the trajectory deformation degree accordingly; calculating the maximum distortion threshold by combining real-time laser power, welding speed, and duration; comparing the trajectory deformation degree with the threshold, and determining instability and cutting off the laser if the threshold is exceeded. This invention utilizes the geometric characteristics of the phase space trajectory to monitor the dynamic behavior of the weld pool and introduces a dynamic compensation mechanism for accumulated enthalpy, effectively distinguishing between natural thermal fluctuations and abnormal instability, thus achieving precise monitoring and control of welding quality.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology. More specifically, this invention relates to a method and system for monitoring the quality of laser welding of aluminum single-layer panels. Background Technology

[0002] Aluminum single panels are widely used in building curtain walls and rail transit due to their lightweight, high strength and corrosion resistance. Laser welding technology (LWT) has become the main connection method due to its high energy density and deep penetration. However, the physical properties of aluminum alloys, such as high reflectivity and high thermal conductivity, result in poor keyhole stability during deep penetration welding, which easily leads to defects such as porosity and cracks.

[0003] In the early days, the industry mostly used non-destructive testing (NDT) techniques such as ultrasonic testing for offline sampling inspection. Due to the lag, it was impossible to intervene in time, resulting in a high scrap rate. Although the photodiode process monitoring introduced later had a fast response, the single-point analog signal was easily interfered with and could not characterize the spatial texture of the molten pool, and its ability to identify minute defects was insufficient.

[0004] Currently, monitoring solutions based on machine vision (MV) are gradually becoming dominant. These solutions use high-speed industrial cameras (HSIC) to capture images of the molten pool, extract Euclidean geometric features such as area or roundness, and set fixed thresholds to determine quality. This linear statistical method performs reasonably well under constant heat input scenarios, but in continuous long weld seams of aluminum panels, the flow of liquid metal exhibits strong nonlinearity and chaotic characteristics. Moreover, over time, the cumulative effect of enthalpy causes irreversible natural drift in the shape of the molten pool. Existing static threshold methods are unable to distinguish between normal process thermal fluctuations and abnormal instability precursors, which can easily lead to false alarms or missed detections in the later stages of welding. Therefore, how to accurately analyze fluid dynamics behavior under unsteady thermal conditions is a technical problem that the industry urgently needs to solve. Summary of the Invention

[0005] To address the technical problem of existing aluminum single-panel laser welding quality monitoring methods being unable to distinguish between natural thermal fluctuations and abnormal instability in the molten pool under unsteady thermal environments, resulting in low monitoring accuracy and high false alarm rate, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for monitoring the quality of laser welding of aluminum single-panel, comprising: acquiring a high-frequency molten pool image of the welding area to be tested on the aluminum single-panel and extracting the lateral displacement velocity of the centroid of the molten pool using the Otsu method and the gray-scale centroid algorithm to construct an original oscillation velocity sequence; denoising the original oscillation velocity sequence to obtain a molten pool oscillation velocity sequence; reconstructing the phase space of the molten pool oscillation velocity sequence using the optimal time delay parameter and the optimal embedding dimension to construct a phase space state vector; calculating the change in the forward direction angle and the density of neighboring reference points in the phase space trajectory based on the phase space state vector, and obtaining the trajectory deformation degree based on the change in the forward direction angle and the density of neighboring reference points; calculating the maximum distortion threshold according to the real-time laser power, the real-time welding speed, and the continuous welding time of the current welding area to be tested on the aluminum single-panel; comparing the trajectory deformation degree with the maximum distortion threshold, and determining that molten pool instability has occurred and executing a control strategy to cut off the laser energy in response to the trajectory deformation degree being greater than the maximum distortion threshold.

[0007] This invention utilizes phase space reconstruction technology to transform the oscillation velocity sequence of the molten pool in the welding area of ​​the aluminum single-panel into a phase space state vector reflecting the dynamic behavior. It also combines the change in the angle of the forward direction with the density of neighboring reference points to obtain the trajectory deformation degree. By comparing the trajectory deformation degree with the dynamic maximum distortion threshold that includes the thermal sensitivity coefficient and welding time, abnormal fluctuations in the fluid dynamics behavior inside the molten pool can be identified. This allows for timely interruption of laser energy when molten pool instability is detected during the laser welding process of the aluminum single-panel, thereby reducing the production of defective products.

[0008] Preferably, the step of extracting the lateral displacement velocity of the molten pool centroid using the Otsu method and the gray-scale centroid algorithm to construct the original oscillation velocity sequence, and then denoising the original oscillation velocity sequence to obtain the molten pool oscillation velocity sequence, includes: extracting the binarized mask of the bright molten pool region in the high-frequency molten pool image using the Otsu method; calculating the sub-pixel coordinates of the molten pool centroid using the gray-scale centroid algorithm; calculating the lateral displacement velocity of the molten pool centroid in the direction perpendicular to the welding path to obtain the original oscillation velocity sequence; and denoising the original oscillation velocity sequence using wavelet denoising to obtain the molten pool oscillation velocity sequence.

[0009] This invention utilizes the Otsu method to extract a binarized mask of the high-brightness molten pool region and combines it with a gray-scale centroid algorithm to calculate the sub-pixel coordinates of the molten pool centroid. Subsequently, wavelet denoising is used to process the original oscillation velocity sequence. This multi-processing method can filter out electromagnetic interference and mechanical vibration noise in the industrial field, extracting a molten pool oscillation velocity sequence that only contains fluid dynamics characteristics, thereby improving the signal-to-noise ratio and accuracy of aluminum single-panel laser welding monitoring data.

[0010] Preferably, the phase space state vector satisfies the expression: In the formula, express The phase space state vector at time t; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; Indicates the optimal time delay parameter; Indicates the optimal embedding dimension; This represents the transpose of a vector.

[0011] Preferably, the optimal time delay parameter is calculated using the mutual information method; the optimal embedding dimension is calculated using the spurious neighbor method.

[0012] Preferably, the trajectory deformation degree satisfies the expression: In the formula, express The degree of trajectory deformation at any given moment; express The change in the angle of the direction of travel at any given moment; Indicates the number of neighboring reference points; express The phase space state vector at time t; Indicates distance in historical trajectory The most recent A neighborhood phase space state point; Indicates the Gaussian kernel width; Represents the Euclidean distance operation; This represents an exponential function with the natural constant e as its base. This represents the cosine function.

[0013] This invention characterizes the abrupt change in the direction of the molten pool motion by calculating the change in the angle of the forward direction in the phase space trajectory, and uses the density of neighborhood reference points to reflect the distribution of the trajectory in the phase space. By combining these two dimensions, the trajectory deformation degree is obtained, enabling the system to sensitively capture the weak precursor signals when the molten pool changes from stable flow to chaotic instability during the laser welding of aluminum single plates, thereby improving the sensitivity of judging welding quality anomalies.

[0014] Preferably, the Gaussian kernel width is equal to the average Euclidean distance of the stable data within the previous 0.5-second time window.

[0015] Preferably, the maximum distortion threshold satisfies the expression: In the formula, express The maximum distortion threshold at any given time; Indicates the basic statistical threshold; Indicates the thermal sensitivity coefficient; Indicates real-time laser power; Indicates the real-time welding speed; This indicates the duration of welding in the area of ​​the aluminum panel to be tested. This represents the thermal saturation time constant of the aluminum panel. Represents the natural logarithm operation.

[0016] This invention takes into account the thermo-rheological properties of aluminum and calculates the maximum distortion threshold based on real-time laser power, real-time welding speed, and the duration of continuous welding. By introducing a thermal sensitivity coefficient and the thermal saturation time constant of the aluminum single plate, the judgment criteria are dynamically compensated. This mechanism can adapt to the physical phenomenon that the natural fluctuation amplitude of the molten pool increases due to the accumulation of enthalpy in the later stage of welding, and reduces false alarms caused by normal thermal fluctuations of the molten pool during the quality monitoring of laser welding of aluminum single plates.

[0017] Preferably, the basic statistical threshold is obtained by adding three times the standard deviation to the mean of the stable data in the initial stage of statistical welding.

[0018] Preferably, the control strategy of determining that molten pool instability has occurred and executing the laser energy cutoff includes: in response to the trajectory deformation degree being greater than the maximum distortion threshold and the duration exceeding a preset time, determining that molten pool instability has occurred at the current position of the welding area to be tested on the aluminum single plate; sending a cutoff signal to the laser main control card to cut off the laser energy output; recording and marking the coordinate position of the defect point fed back by the current encoder; and triggering an alarm.

[0019] Upon determining that molten pool instability has occurred, this invention immediately sends a cutoff signal to the laser main control card to cut off the laser energy output, records the coordinates of the defect point fed back by the encoder, and triggers an alarm. This closed-loop control strategy not only enables immediate intervention in the early stages of defect formation to prevent damage from spreading, but also provides accurate location information for subsequent manual re-inspection and problem tracing, ensuring the safe and efficient operation of the aluminum single-panel laser welding production line.

[0020] Secondly, the present invention provides an aluminum single-panel laser welding quality monitoring system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned aluminum single-panel laser welding quality monitoring method is implemented.

[0021] By adopting the above technical solution, a computer program for monitoring the quality of laser welding of aluminum single-panel is generated and stored in a memory so that it can be loaded and executed by a processor. A terminal device can then be made based on the memory and the processor for convenient use.

[0022] The beneficial effects of this invention are as follows:

[0023] This invention applies phase space reconstruction theory to restore a one-dimensional time series to a geometric trajectory in a high-dimensional space, transforming the invisible dynamic instability of the laser welding pool of aluminum single-panel into a visible geometric trajectory distortion for evaluation. This solves the problem that traditional time-domain or frequency-domain analysis methods cannot distinguish between random fluctuations and chaotic anomalies, providing a nonlinear dynamic monitoring perspective for the welding quality of aluminum single-panel.

[0024] This invention establishes an adaptive threshold adjustment mechanism based on energy input and time accumulation. By calculating the maximum distortion threshold including the thermal sensitivity coefficient in real time, it follows the changes in the thermal field during the welding process. This enables the monitoring system to distinguish between the increase in natural fluctuations caused by the decrease in the viscosity of the molten aluminum and the true instability of the molten pool, thereby improving the robustness and adaptability of the system under different welding process parameters and continuous welding duration.

[0025] This invention combines the Otsu method, gray-scale centroid algorithm, and wavelet denoising to perform deep feature extraction and denoising processing on high-frequency molten pool images. It can accurately capture micron-level lateral displacement information of the molten pool centroid from complex industrial environments, ensuring the purity and accuracy of the basic data on which subsequent dynamic analysis depends, thus laying a data foundation for high-precision judgment of the quality of laser welding of aluminum single plates. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a laser welding quality monitoring method for aluminum single-panel according to the present invention;

[0027] Figure 2 This is a schematic diagram illustrating the geometric characteristics of the phase space trajectory;

[0028] Figure 3 This is a schematic diagram illustrating the trajectory deformation.

[0029] Figure 4 This is a schematic diagram illustrating the determination of the maximum distortion threshold. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] This invention discloses a method for monitoring the quality of laser welding of aluminum single-layer panels, referring to... Figure 1This includes steps S1-S5:

[0033] S1. Obtain a high-frequency molten pool image of the welding area to be tested on the aluminum single plate and use the Otsu method and gray-scale centroid algorithm to extract the lateral displacement velocity of the molten pool centroid to construct the original oscillation velocity sequence; denoise the original oscillation velocity sequence to obtain the molten pool oscillation velocity sequence.

[0034] It should be noted that during the laser welding process of aluminum single-panel aluminum, the transient rupture of the alumina film typically lasts only a few milliseconds and manifests as extremely weak morphological fluctuations rather than significant brightness changes. To capture this microscopic oscillation information, this invention must utilize high-sampling-rate imaging equipment and effectively filter out strong electromagnetic interference in the industrial environment, thereby extracting high signal-to-noise ratio data containing only hydrodynamic characteristics.

[0035] Specifically, this invention installs a high-speed industrial camera with a sampling frequency of 2000 Hz at a 45-degree angle to the side of the laser head to continuously capture images of the molten pool in the welding area of ​​the aluminum single plate to be tested. The system uses the Otsu method to extract the binarized mask of the bright molten pool area in each frame of the image and applies the gray-scale centroid algorithm to calculate the sub-pixel coordinates of the molten pool centroid. The system calculates the lateral displacement velocity of the molten pool centroid in the direction perpendicular to the welding path to form the original oscillation velocity sequence. This invention uses wavelet denoising to refine the original oscillation velocity sequence. The system selects the db4 wavelet basis to decompose the signal to the third layer, accurately identifies and forces the node coefficients representing high-frequency electromagnetic noise to zero, and then reconstructs the signal to output a pure molten pool oscillation velocity sequence.

[0036] It should be further noted that the camera sampling frequency is set to 2000 Hz; if the camera sampling frequency is too low, the Nyquist sampling theorem cannot be satisfied, and the high-frequency transient characteristics when the oxide film breaks will be lost; at the same time, the extraction range of the lateral displacement velocity in this invention is usually within ±0.5 mm; if the extraction range of the lateral displacement velocity is too large, it will introduce the shaking noise of the robotic arm itself; if the extraction range of the lateral displacement velocity is too small, it cannot cover the physical oscillation of the edge of the molten pool; in this embodiment, it is set to ±0.2 mm, which can accurately capture the surface tension wave of liquid metal.

[0037] S2. The phase space of the molten pool oscillation velocity sequence is reconstructed using the optimal time delay parameter and the optimal embedding dimension to construct the phase space state vector.

[0038] It should be noted that one-dimensional velocity waveform data is difficult to intuitively reflect the complex fluid dynamics within the molten pool, and it is also difficult to distinguish between normal random fluctuations and abnormal chaotic instability. This invention utilizes phase space reconstruction technology to restore a one-dimensional time series to a geometric trajectory in a high-dimensional space, thereby transforming the instability of the molten pool into geometric shape distortion, and converting the difficult-to-observe temporal logic into an easily observable geometric shape.

[0039] Specifically, this invention reads the oscillation velocity sequence of the molten pool, uses the sliding window technique to extract historical data, and constructs a phase space state vector containing system dynamics information.

[0040] The phase space state vector satisfies the expression:

[0041]

[0042] In the formula, express The phase space state vector at time t; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; Indicates the optimal time delay parameter; Indicates the optimal embedding dimension; This represents the transpose of a vector.

[0043] In the formula, the phase space state vector is introduced by the optimal time delay parameter. The dynamic information of historical moments is folded into the current phase space state vector; when the aluminum plate is in a stable welding state, the molten aluminum flows regularly, and the phase space state vector... The connection in space will form a smooth, closed circular structure; when a defect occurs, the flow of molten aluminum loses its constraint, and the phase space geometric trajectory will be knotted, twisted or broken.

[0044] It is necessary to further explain the optimal time delay parameters. and optimal embedding dimension Manual arbitrary setting is strictly prohibited; the optimal time delay parameter is calculated using the mutual information method, with an empirical value of 4, representing the length of the system's memory; the optimal embedding dimension is calculated using the spurious neighbor method, with an empirical value of 5, representing the degrees of freedom of the system's motion; optimal embedding dimension The choice of embedding dimension is crucial; if the optimal embedding dimension is less than 3, complex melt pool trajectories will overlap in low-dimensional space, leading to misjudgment of the state; if the optimal embedding dimension is greater than 8, the computational load increases exponentially, failing to meet real-time control requirements. This embodiment selects 5 dimensions, achieving fast computation while ensuring the integrity of the topology.

[0045] For example, Figure 2This is a schematic diagram of the geometric characteristics of the phase space trajectory, showing different geometric forms of the molten pool oscillation signal in phase space during the welding of aluminum single plates. Among them, the trajectory with a regular and closed loop structure represents the stable welding stage, indicating that the molten pool flows in an orderly manner; while the trajectory with twisted, divergent and escape characteristics represents the unstable stage of the molten pool, reflecting the chaotic anomalies in the internal dynamic behavior of the molten pool.

[0046] S3. Calculate the change in the forward direction angle and the density of neighboring reference points in the phase space trajectory based on the phase space state vector, and obtain the trajectory deformation degree based on the change in the forward direction angle and the density of neighboring reference points.

[0047] It should be noted that this invention requires a scale to measure the degree of distortion of the phase space geometric trajectory in order to capture the signs of instability in a short period of time with extreme sensitivity; at the moment the oxide film breaks, the direction of the molten pool movement will change drastically, which is geometrically manifested as a sharp turn and escape of the trajectory.

[0048] Specifically, this invention calculates the degree of geometric distortion of the current phase space state vector to obtain the trajectory deformation degree. The trajectory deformation degree satisfies the expression:

[0049]

[0050] In the formula, express The degree of trajectory deformation at any given moment; express The change in the angle of the direction of travel at any given moment; Indicates the number of neighboring reference points; express The phase space state vector at time t; Indicates distance in historical trajectory The most recent A neighborhood phase space state point; Indicates the Gaussian kernel width; Represents the Euclidean distance operation; This represents an exponential function with the natural constant e as its base. This represents the cosine function.

[0051] In the formula, the numerator of the trajectory deformation degree represents a sudden change in direction, and the denominator represents the density of state points in the neighborhood phase space; when the aluminum single plate shows signs of instability in the welding area to be tested, the pressure imbalance inside the molten pool causes the trajectory to suddenly reverse, and the included angle The value increases, causing the molecular term to increase; simultaneously, the trajectory escapes into a sparse region of phase space, and the distance from the reference point... An increase in the numerator leads to a decrease in the exponential decay term, which in turn decreases the denominator; an increase in the numerator and a decrease in the denominator result in a decrease in the trajectory deformation. It increases exponentially, thus becoming a sensitive indicator of anomalies.

[0052] It is necessary to further add the number of neighborhood reference points. Set to 20; Gaussian kernel width It is dynamically set based on the average Euclidean distance of the stable data 0.5 seconds before welding, ensuring the adaptability of the trajectory deformation to different welding process parameters.

[0053] For example, Figure 3 This is a schematic diagram of the trajectory deformation, showing the evolution of the trajectory deformation calculated based on phase space characteristics over welding time. In the stable stages of the initial and middle stages of welding, the trajectory deformation value remains at a low level with slight fluctuations; when entering the unstable stage, the value rises sharply and instantaneously, exhibiting a significant pulse-like abrupt change, which intuitively characterizes the deterioration of the molten pool state.

[0054] S4. Calculate the maximum distortion threshold based on the real-time laser power, real-time welding speed, and the continuous welding time of the current aluminum panel to be tested welding area.

[0055] It should be noted that aluminum has significant thermo-rheological properties. As the welding process continues, the enthalpy of the aluminum plate accumulates, causing the temperature of the molten pool to rise and the viscosity of the molten aluminum to decrease. This means that in the later stages of welding, the natural fluctuation range of the molten pool will physically increase. If a fixed threshold is used, it will lead to a large number of false alarms. Therefore, this invention needs to eliminate the interference of physical heat accumulation on the monitoring accuracy.

[0056] Specifically, this invention calculates the maximum distortion threshold in real time based on the energy input. The maximum distortion threshold satisfies the following expression:

[0057]

[0058] In the formula, express The maximum distortion threshold at any given time; Indicates the basic statistical threshold; This represents the thermal sensitivity coefficient, with units of meters per joule. Indicates real-time laser power; Indicates the real-time welding speed; This indicates the duration of welding in the area of ​​the aluminum panel to be tested. This represents the thermal saturation time constant of the aluminum panel. Represents the natural logarithm operation.

[0059] In the formula, the term within parentheses of the maximum distortion threshold simulates the heat accumulation effect; as the welding time of the current aluminum panel under test welding area continues... As it increases, the logarithmic value increases; simultaneously, the linear energy density... The increased size of the pool enhances its liquidity; both of these factors combined lead to a larger compensation coefficient, driving the maximum distortion threshold. Increased; this means that the system automatically relaxed the judgment criteria in the later stages of welding, offsetting the increase in natural fluctuations caused by the thinning of the molten aluminum.

[0060] It is necessary to further supplement the basic statistical threshold. The thermal saturation time constant of the aluminum single panel was obtained by adding three times the standard deviation to the mean of 1000 frames of stable data during the initial welding stage. The thermal sensitivity coefficient depends on the thickness of the aluminum panel. For a 3mm thick aluminum panel, the empirically set thermal saturation time constant is 5 seconds. If the thermal saturation time constant is set too small, the compensation will be too fast, leading to missed detections. If the thermal saturation time constant is set too large, the compensation will be insufficient, and false alarms cannot be eliminated. The thermal sensitivity coefficient is obtained as follows: During the equipment calibration stage, three different ratios of laser power to welding speed are selected for defect-free welding tests. The average drift of the trajectory deformation after welding relative to the initial welding stage is recorded. The average drift is linearly fitted to the linear energy density using the least squares method, and the slope of the fitted line is the thermal sensitivity coefficient. In this embodiment, the thermal sensitivity coefficient was measured to be 0.05 m / joule after calibration.

[0061] S5. Compare the trajectory deformation degree with the maximum distortion threshold. If the trajectory deformation degree is greater than the maximum distortion threshold, determine that the molten pool is unstable and execute the control strategy of cutting off the laser energy.

[0062] It should be noted that the ultimate goal of monitoring is to intervene in the early stages of defect formation, transforming algorithm results into actual industrial control signals to prevent the generation of waste.

[0063] Specifically, the present invention will calculate the trajectory deformation in real time. With the maximum distortion threshold Compare; if the trajectory deformation degree Greater than the maximum distortion threshold If the duration exceeds 1.5 milliseconds, it is determined that the molten pool has become unstable at the current position of the welding area to be tested on the aluminum panel; the system then executes the following control strategy: sends a high-level cutoff signal to the laser main control card to cut off the laser energy output; records the coordinate position of the defect point fed back by the encoder and marks the area corresponding to the coordinate position of the defect point as a red alarm zone on the upper computer monitoring interface; at the same time, triggers the audible and visual alarm to prompt the operator to perform manual re-inspection.

[0064] For example, Figure 4This is a schematic diagram of the maximum distortion threshold determination, showing the real-time comparison between the trajectory deformation degree and the maximum distortion threshold. The maximum distortion threshold curve shows a slow upward trend as welding time progresses to compensate for the increase in natural fluctuations caused by the accumulation of enthalpy; when the trajectory deformation degree curve exceeds the maximum distortion threshold curve, their intersection point is the instability determination point, thus achieving accurate monitoring in unsteady thermal environments.

[0065] This invention also discloses an aluminum single-panel laser welding quality monitoring system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an aluminum single-panel laser welding quality monitoring method according to the present invention is implemented.

[0066] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method of monitoring the quality of laser welding of an aluminum single plate, characterized by, include: High-frequency molten pool images of the welding area to be tested on the aluminum single plate were acquired, and the lateral displacement velocity of the centroid of the molten pool was extracted using the Otsu method and the gray-scale centroid algorithm to construct the original oscillation velocity sequence. The original oscillation velocity sequence is denoised to obtain the molten pool oscillation velocity sequence; The phase space of the molten pool oscillation velocity sequence is reconstructed using the optimal time delay parameter and the optimal embedding dimension to construct a phase space state vector. Based on the phase space state vector, the change in the forward direction angle and the density of neighboring reference points in the phase space trajectory are calculated. The trajectory deformation is then obtained based on the change in the forward direction angle and the density of neighboring reference points. The trajectory deformation satisfies the following: ； represents the trajectory deformation degree at the time point; represents the advancing direction angle change amount at the time point; represents the number of neighborhood reference points; represents the phase space state vector at the time point; represents the distance closest to the neighborhood phase space state point in the history trajectory; represents the Gaussian kernel width; represents the Euclidean distance operation; represents the exponential function with the natural constant e as the base; represents the cosine function; The maximum distortion threshold is calculated based on the real-time laser power, real-time welding speed, and the continuous welding time of the current aluminum panel welding area to be tested. The trajectory deformation degree is compared with the maximum distortion threshold. In response to the trajectory deformation degree being greater than the maximum distortion threshold, it is determined that the molten pool is unstable and a control strategy to cut off the laser energy is executed.

2. The method for monitoring the quality of laser welding of aluminum single-layer panels according to claim 1, characterized in that, The lateral displacement velocity of the molten pool centroid is extracted using the Otsu method and the gray-scale centroid algorithm to construct the original oscillation velocity sequence; The original oscillation velocity sequence is denoised to obtain the molten pool oscillation velocity sequence, including: The Otsu method was used to extract the binarized mask of the bright molten pool region in the high-frequency molten pool image; the gray-scale centroid algorithm was used to calculate the sub-pixel coordinates of the molten pool centroid; the lateral displacement velocity of the molten pool centroid in the direction perpendicular to the welding path was calculated to obtain the original oscillation velocity sequence; wavelet denoising was used to denoise the original oscillation velocity sequence to obtain the molten pool oscillation velocity sequence.

3. The method for monitoring the quality of laser welding of aluminum single-layer panels according to claim 1, characterized in that, The phase space state vector satisfies the expression: ； In the formula, express The phase space state vector at time t; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; express The amplitude of the oscillation velocity at any given moment; Indicates the optimal time delay parameter; Indicates the optimal embedding dimension; This represents the transpose of a vector.

4. The method for monitoring the quality of laser welding of aluminum single-layer panels according to claim 3, characterized in that, The optimal time delay parameter is calculated using the mutual information method; the optimal embedding dimension is calculated using the spurious neighbor method.

5. The method for monitoring the quality of laser welding of aluminum single-layer panels according to claim 1, characterized in that, The Gaussian kernel width is equal to the average Euclidean distance of the stable data within the previous 0.5-second time window.

6. The method for monitoring the quality of laser welding of aluminum single-layer panels according to claim 1, characterized in that, The maximum distortion threshold satisfies the expression: ； In the formula, express The maximum distortion threshold at any given time; Indicates the basic statistical threshold; Indicates the thermal sensitivity coefficient; Indicates real-time laser power; Indicates the real-time welding speed; This indicates the duration of welding in the area of ​​the aluminum panel to be tested. This represents the thermal saturation time constant of the aluminum panel. Represents the natural logarithm operation.

7. The method for monitoring the quality of laser welding of aluminum single-layer panels according to claim 6, characterized in that, The basic statistical threshold is obtained by adding three times the standard deviation to the mean of the stable data in the initial stage of welding.

8. The method for monitoring the quality of laser welding of aluminum single-layer panels according to claim 1, characterized in that, The control strategy for determining when molten pool instability occurs and executing the laser energy cutoff includes: In response to the trajectory deformation degree being greater than the maximum distortion threshold and the duration exceeding the preset time, it is determined that the current position of the weld pool instability has occurred in the aluminum single panel welding area to be tested; a cutoff signal is sent to the laser main control card to cut off the laser energy output; the coordinate position of the defect point fed back by the encoder is recorded and marked; and the alarm is triggered to sound an alarm.

9. A laser welding quality monitoring system for aluminum single-panel panels, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a method for monitoring the quality of laser welding of aluminum single-panel according to any one of claims 1-8.

Images