A method and system for pressure regulation of vacuum laser welding of a weld evaporation behavior

Abstract

This invention discloses a method and system for controlling pressure in vacuum laser welding based on welding evaporation behavior, belonging to the field of laser precision control technology. The method includes establishing a multidimensional mapping function model and a model coefficient matrix; inputting process parameters into the multidimensional mapping function model to obtain a predicted local pressure evolution curve and generating a feedforward control sequence; acquiring raw spectral data; performing multi-level signal cleaning to extract actual Stark broadening; when the signal quality threshold is met, calculating the actual local pressure value, determining the pressure deviation, and generating a dynamic compensation signal based on the pressure deviation; when the signal quality threshold is not met, comparing the current actual volume with the volume threshold, and activating a graded adjustment mode based on the comparison result. This invention can, based on the initial control sequence generated from the spectral data at the moment of welding initiation, further adaptively adjust the pressure according to the signal indicators of the spectral data, thereby effectively improving the timeliness and accuracy of vacuum laser welding pressure control.

Description

Technical Field

[0001] This invention relates to the field of laser precision control technology, and more specifically to a method and system for controlling vacuum laser welding pressure based on welding evaporation behavior. Background Technology

[0002] In the fields of aerospace, new energy power batteries, and rail transit equipment manufacturing, vacuum laser welding technology for lightweight, high-strength metals such as aluminum and magnesium alloys is widely used because it can significantly reduce weld porosity and increase weld penetration. In a vacuum environment, the melting and boiling points of metals are lowered, which helps suppress porosity formation and improve weld shape. However, elements such as aluminum and magnesium have relatively low boiling points and high saturated vapor pressures. Under continuous irradiation by a high-energy laser beam, metal atoms on the molten pool surface are prone to violent evaporation and ionization, instantly forming a high-pressure, high-density metal vapor plume. This transient burst of metal vapor significantly alters the local microenvironment pressure above the molten pool keyhole, leading to high-frequency, violent fluctuations in vacuum levels.

[0003] To address the technical challenge of timely detection and proactive control of local pressure fluctuations in the welding zone during vacuum laser welding, which consequently affects welding stability and forming quality, related technologies for monitoring and control of the vacuum laser welding process have gradually become a research hotspot in recent years. However, dedicated technologies for in-situ sensing and closed-loop control of local pressure in vacuum laser welding are still relatively lacking. For example, invention patent CN119634980B proposes a vacuum laser welding penetration stabilization device, which achieves penetration stabilization by detecting the secondary current between the base material to be welded and the welding fumes and adjusting the pumping rate according to a preset threshold. However, this scheme mainly relies on the secondary current signal to indirectly characterize the welding state, and cannot directly obtain the true local pressure information of the welding zone. Furthermore, its control method is mainly based on threshold comparison, lacking the ability to finely control dynamic pressure changes under complex working conditions.

[0004] Invention patent CN119589173A discloses an online monitoring method for vacuum laser welding. This method collects secondary current signals and analyzes their correlation with weld formation and penetration depth. Although it improves the ability to identify and adjust the online welding quality status, its focus is still on weld morphology and process parameter optimization. It has not established a direct inversion and dedicated control mechanism for the local pressure changes induced by welding evaporation behavior.

[0005] Invention patent CN119772375A discloses a vacuum laser welding chamber, which mainly focuses on the design optimization of the welding chamber structure, workpiece support mechanism, laser transmission module and shielding gas delivery method. However, this solution focuses on the improvement of the welding equipment configuration and does not involve the real-time monitoring, prediction compensation and dynamic control of local pressure during the welding process.

[0006] Therefore, in view of the above-mentioned problems in the existing technology, there is an urgent need to design a vacuum laser welding pressure control method that can be actively controlled. Summary of the Invention

[0007] In view of the above problems, the present invention is proposed to provide a vacuum laser welding pressure control method and system for overcoming or at least partially solving the above problems and the welding evaporation behavior.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, embodiments of the present invention provide a method for controlling the pressure of vacuum laser welding to regulate welding evaporation behavior, comprising: Multiple historical pressure data of the vacuum chamber and multiple historical plasma spectral data of the welding area are collected, and a multidimensional mapping function model and model coefficient matrix of the welding area are established based on the historical pressure data and the historical plasma spectral data. The process parameters set at the moment of welding start are input into the multidimensional mapping function model to obtain the predicted local pressure evolution curve, and a feedforward control quantity sequence for driving the coarse adjustment of the vacuum actuator is generated. Real-time acquisition of raw spectral data of plasma in the molten pool during the welding process; The actual Stark broadening is extracted based on the original spectral data, and the signal quality index is calculated. Determine whether the signal quality index is greater than or equal to the signal quality threshold; When the signal quality threshold is greater than or equal to the actual local pressure value, the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve is determined, and a dynamic compensation signal for coordinated control is generated based on the pressure deviation. When the signal quality is less than the threshold, the actual volume of the vacuum chamber is compared with the volume threshold, and a graded adjustment mode for maintaining vacuum stability in the welding area is activated based on the comparison result.

[0010] In one embodiment, establishing a multidimensional mapping function model and model coefficient matrix of the welding area based on the pressure data and the plasma spectral data includes: The characteristic spectral lines of the plasma spectral data are determined, and the model Stark broadening of the characteristic spectral lines is extracted based on the characteristic spectral lines and the Lorentz fitting function. Determine the correspondence between the Stark broadening of the model and the pressure data; Obtain the calibration coefficients and effective interval for pressure inversion, and within the effective interval, establish the multidimensional mapping function model based on the correspondence between the model's Stark broadening and the pressure data. The multidimensional mapping function model is as follows:

[0011] in, Let be the predicted value of the local pressure in the welding area at time t. For the pressure data of the vacuum chamber, Let be the laser power at time t. Let be the welding speed at time t. For pressure conversion gain, The power influence index, This is the velocity decay index.

[0012] In one embodiment, extracting the actual Stark broadening based on the original spectral data includes: The raw spectral data is subjected to multi-stage signal cleaning. The center wavelength, maximum spectral intensity, and baseline intensity within a fitting window of the characteristic spectral lines of the cleaned spectral data are obtained. The peak amplitude is then determined based on the maximum spectral intensity and the baseline intensity. The peak amplitude is:

[0013] Where A represents the peak amplitude of the characteristic spectral line after subtracting the baseline intensity. B is the maximum spectral intensity within the fitting window, and B is the baseline intensity within the fitting window. A fitting function is constructed based on the center wavelength, the peak amplitude, and the Lorentz function. The fitting function is as follows:

[0014] in, The fitted spectral intensity at wavelength, The peak amplitude of the characteristic spectral lines after subtracting the baseline intensity. The center wavelength, b is the actual Stark broadening of the characteristic spectral line, and b is the baseline offset of the fitting function; The actual Stark broadening is obtained by solving the fitting function within the effective interval using the least squares method.

[0015] In one embodiment, the multi-stage signal cleaning of the raw spectral data includes: The original spectral data is decomposed using the Daubechies wavelet basis function to obtain wavelet coefficients; The denoising threshold is calculated based on the wavelet coefficients, and the denoising coefficient is determined based on the denoising threshold and the wavelet coefficients. The original spectral data is denoised based on the denoising coefficients to determine the denoised and reconstructed spectral intensity. The original cleaned spectral data is determined by performing a standard normal variable transformation on the denoised and reconstructed spectral intensity using a transformation function, wherein the transformation function is:

[0016] in, The spectral intensity after denoising and reconstruction. The average value of the denoised reconstructed spectral intensity within the preset standardized window. This represents the standard deviation within the baseline window.

[0017] In one embodiment, the calculation of signal quality metrics includes: Target noise bands are selected on both sides of the fitting window, and the noise standard deviation of the spectral intensity within the target noise band is determined based on the baseline intensity and the number of sampling points. The peak signal-to-noise ratio is determined based on the peak amplitude and the noise standard deviation. Obtain the fitted spectral intensity, and determine the fitting residual based on the fitted spectral intensity, the number of sampling points, and the denoised and reconstructed spectral intensity; The peak signal-to-noise ratio, maximum spectral intensity, and fitting residual are determined as the signal quality indicators.

[0018] In one embodiment, selecting target noise bands on both sides of the fitting window includes: Noise bands are set on the short-wavelength side and long-wavelength side of the fitting window according to the preset wavelength interval and noise band width; Determine the mean absolute value of the first-order difference of the spectral intensity of the noise band; Determine whether the noise band meets the noise band selection rules, wherein the noise band selection rules are that the noise band does not contain the emission peak of the characteristic spectral line and the mean absolute value of the first difference is not greater than the mean absolute value threshold. If the noise band selection rule is met, then the noise band is determined as the target noise band; If the noise band selection rules are not met, the target noise band is determined according to the preset step size.

[0019] In one embodiment, activating the graded adjustment mode for maintaining vacuum stability in the welding area based on the comparison results includes: When the comparison result indicates that the actual volume is not less than the volume threshold, a two-stage coordinated adjustment mode is activated, which uses a high-speed main vacuum pump to maintain the reference pumping speed and a small-volume auxiliary pump for fine-tuning. When the comparison result indicates that the actual volume is less than the volume threshold, a single-stage adjustment mode for adjusting pressure with a single device is activated.

[0020] In one embodiment, after determining the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve, the method further includes: Construct a cost function based on the pressure deviation; The coefficient matrix is ​​updated using the cost function and adaptive update strategy, and the update formula is as follows:

[0021] in, Let be the coefficient matrix at time t. For learning rate, The cost function is constructed from the pressure deviation. Cost function The gradient of the coefficient matrix M, The model coefficient matrix consists of pressure conversion gain, power influence index, and velocity decay index. .

[0022] In one embodiment, the raw spectral data is acquired by a spectral acquisition unit of an integrated filter assembly, the filter assembly including dual filters, wherein: the first filter is a high-density notch filter, the center stopband wavelength of the high-density notch filter is the same as the center wavelength of the welding laser source; the second filter is a bandpass filter, the band containing the characteristic emission spectral lines of the material to be welded is located within the bandpass range of the bandpass filter.

[0023] Secondly, embodiments of the present invention provide a vacuum laser welding pressure control system for welding evaporation behavior, employing the following technical solution: The multidimensional mapping model construction module is used to collect multiple historical pressure data of the vacuum chamber and multiple historical plasma spectral data of the welding area, and to establish a multidimensional mapping function model and model coefficient matrix of the welding area based on the historical pressure data and the historical plasma spectral data. The drive module is used to input the set process parameters at the moment of welding start-up into the multidimensional mapping function model to obtain the predicted local pressure evolution curve, and generate a feedforward control quantity sequence for driving the coarse adjustment of the vacuum actuator. The acquisition module is used to acquire raw spectral data of the plasma in the molten pool during the welding process in real time. The signal quality index calculation module is used to extract the actual Stark broadening based on the original spectral data and calculate the signal quality index. The judgment module is used to determine whether the signal quality index is greater than or equal to the signal quality threshold; when it is greater than or equal to the signal quality threshold, the compensation module is triggered; when it is less than the signal quality threshold, the graded adjustment module is triggered. The compensation module is used to calculate the actual local pressure value, determine the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve, and generate a dynamic compensation signal for coordinated control based on the pressure deviation. The graded adjustment module is used to compare the actual volume of the current vacuum chamber with the volume threshold, and to enable the graded adjustment mode to maintain the vacuum stability of the welding area based on the comparison result.

[0024] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following: A multidimensional mapping function model is established by collecting multiple historical data. Then, based on the process parameters at the moment of welding start-up and the multidimensional mapping function model, a pressure evolution curve and a feedforward control sequence for initial coarse adjustment are generated to enable timely control of vacuum pressure at the moment of welding start-up. Next, the raw spectral data during the welding process is acquired in real time and processed to obtain interference-free spectral data. This interference-free spectral data is used to determine signal quality indicators for more accurate judgment results. Based on the accurate judgment results, targeted adaptive closed-loop adjustment and compensation are performed according to actual pressure monitoring data and the volume of the vacuum chamber, thereby effectively achieving local adaptive and precise control of the welding area. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0026] Figure 1 This invention provides a method for controlling the pressure of vacuum laser welding to observe the evaporation behavior of the weld. Figure 2 This is a comparison diagram of the spectral signal before and after signal cleaning provided in this embodiment of the invention. Figure 3 The image shows the Stark broadening variation of characteristic spectral lines under different vacuum pressures and the Lorentz fitting diagram provided in this embodiment of the invention. Figure 4 This is a schematic diagram of the welding pressure system provided in an embodiment of the present invention; wherein, 1-laser emitting head; 2-laser beam; 3-molten pool; 4-workpiece to be welded; 5-metal vapor plume; 6-optical acquisition probe; 7-transmission optical fiber; 8-spectrometer; 9-control host; 10-vacuum pump group; 11-vacuum regulating valve; 12-vacuum gauge; 13-vacuum connection pipeline; 14-notch filter; 15-bandpass filter; 16-vacuum chamber; Figure 5 This is a schematic diagram of the welding pressure control logic provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the vacuum laser welding pressure control system for welding evaporation behavior provided in an embodiment of the present invention. Detailed Implementation

[0027] 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 embodiments of the present invention, and not all embodiments. 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.

[0028] This invention discloses a method for controlling the pressure in vacuum laser welding to regulate welding evaporation behavior, such as... Figure 1 The diagram shown is a flowchart illustrating a vacuum laser welding pressure control method for welding evaporation behavior according to an embodiment of the present invention. The method includes steps S101, S102, S103, S104, S105, S106, and S107, wherein: Step S101: Collect multiple historical pressure data of the vacuum chamber and multiple historical plasma spectral data of the welding area, and establish a multidimensional mapping function model and model coefficient matrix of the welding area based on the historical pressure data and historical plasma spectral data.

[0029] Specifically, a vacuum gauge is installed on the wall of the vacuum chamber to collect pressure data, while the plasma spectral data is collected by the spectral acquisition unit.

[0030] The process of establishing the multidimensional mapping function model includes: collecting pressure data at the vacuum chamber wall and plasma spectral data at the center of the welding area during the pre-set exploratory experimental phase; determining characteristic spectral lines and extracting the model Stark broadening using the Lorentz function; determining the correspondence between the extracted model Stark broadening and the pressure data; determining the pressure interval corresponding to when the fitting residual does not exceed the residual threshold as the inversion coefficients and effective working interval for pressure inversion; establishing a multidimensional mapping function model between the pressure data of the welding area and the laser power and welding speed within the effective working interval; initializing the model coefficient matrix of the multidimensional mapping function model; and storing the multidimensional mapping function model in the experience base for subsequent updates. The expression of the multidimensional mapping function model is as follows:

[0031] in, Let t be the predicted pressure value of the welding area. For stress data, Let be the laser power at time t. Let be the welding speed at time t. For pressure conversion gain, The power influence index, This is the velocity decay index.

[0032] Understandably, by introducing an online-evolvable nonlinear function model and using real-time monitored residuals to correct the model coefficients, it can automatically adapt to welding conditions under different workpiece materials, different laser aging levels, and different environmental backgrounds, thus possessing self-evolution capabilities.

[0033] Step S102: Input the set process parameters at the moment of welding start-up into the multidimensional mapping function model to obtain the predicted local pressure evolution curve, and generate a feedforward control quantity sequence for driving the coarse adjustment of the vacuum actuator.

[0034] Specifically, the welding process parameters are the laser power W and the welding speed v during the welding process. These parameters are input into a multidimensional mapping function model. The model calculates the local pressure evolution curve corresponding to the welding process parameters and generates a feedforward control sequence based on this curve. This feedforward control sequence drives the variable frequency speed of the vacuum pump unit and the opening of the electronically controlled throttle valve, eliminating the calculation delay during system cold start and achieving initial adjustment of the vacuum pressure, ensuring that the local pressure change directly approximates the target trajectory. In this embodiment, the laser power can be set to 8000 W, and the welding speed can be set to v = 1.5 m / min.

[0035] Step S103: Real-time acquisition of raw spectral data of plasma in the molten pool during the welding process.

[0036] Specifically, the raw spectral data of the metal vapor plasma above the molten pool during the welding stage can be acquired in real time using a pre-set dual-filter optical interference filter assembly. The first layer of the optical interference filter assembly is a trap filter with an optical density of OD6, matched to the center wavelength of the welding laser, used to physically block the welding laser from entering the spectrometer. The second layer is a bandpass filter covering the wavelength band containing the characteristic emission lines of the target element. The bandpass range covers the wavelength band containing the characteristic emission lines of the target element in the material to be welded, used to suppress thermal radiation and stray light interference. In this embodiment, the passband range is 200nm to 500nm, and the spectrometer sampling frequency is 1kHz.

[0037] Step S104: Extract the actual Stark broadening based on the original spectral data and calculate the signal quality index.

[0038] Specifically, the original spectral data is subjected to multi-stage signal cleaning, and the actual Stark broadening of the cleaned spectral data is extracted. The signal quality index of the cleaned spectral data is then calculated.

[0039] In this embodiment of the invention, signal quality indicators include: noise standard deviation, peak signal-to-noise ratio (PSNR), and fitting residual. It is understood that less underlying interference and a smaller noise standard deviation indicate a cleaner background; a higher PSNR indicates a more prominent effective signal; and a smaller fitting residual indicates a smaller error in Stark broadening. Therefore, judging signal quality based on noise standard deviation, PSNR, and fitting residual allows for a comprehensive assessment across three different dimensions, resulting in more reliable and accurate signal quality judgments. Furthermore, enabling different control modes based on accurate signal quality judgments allows for more precise adjustment of vacuum pressure.

[0040] The process of multi-level signal cleaning of raw spectral data includes: selecting Daubechies wavelet basis functions to perform multi-scale decomposition of the spectral signal to obtain wavelet coefficients W, and then calculating the denoising threshold based on the wavelet coefficients. The calculation formula is as follows:

[0041] in, The noise reduction threshold, These are wavelet coefficients.

[0042] Then, the denoising coefficient is obtained by shrinking according to the preset soft threshold rule. In this embodiment of the invention, when < hour, ,when > hour, And obtain the denoised and reconstructed spectral intensity. .

[0043] Then, perform a standard normal transformation on the spectral intensity. The transformation formula is as follows:

[0044] in, The spectral intensity after denoising and reconstruction. The average value of the denoised reconstructed spectral intensity within the preset standardized window. This represents the standard deviation within the window.

[0045] like Figure 2 The image shown is a comparison of the spectral signal before and after signal cleaning provided in this embodiment of the invention. Signal cleaning can remove noise and invalid data from the original spectral data, resulting in cleaner and more effective feature data.

[0046] The specific process for extracting actual Stark broadening includes: obtaining the center wavelength, maximum spectral intensity, and baseline intensity within the fitting window of the characteristic spectral line; and determining the peak amplitude based on the maximum spectral intensity and baseline intensity. The peak amplitude is:

[0047] Where A represents the peak amplitude of the characteristic spectral line after subtracting the baseline intensity. B is the maximum spectral intensity within the fitting window, and B is the baseline intensity within the fitting window. Then, a fitting function is constructed based on the center wavelength, peak amplitude, and Lorentz function. The fitting function is as follows:

[0048] in, Let be the fitted spectral intensity at wavelength , and A be the peak amplitude of the characteristic spectral line after subtracting the baseline offset b. The center wavelength, b is the actual Stark broadening of the characteristic spectral lines, and b is the baseline offset of the fitted function; Then, within the effective inversion interval, the actual Stark broadening is obtained by solving the fitting function using the inversion formula and the least squares method. It can be understood that by solving the actual Stark broadening in this embodiment of the invention, the problems of slow response and large spatial position deviation of traditional vacuum gauges can be solved, and millisecond-level in-situ sensing of the microenvironment pressure above the molten pool can be achieved.

[0049] The specific process for calculating the signal quality index of the cleaned spectral data includes: at the center wavelength of the spectral data... Selecting a fitting window near the All spectral intensities within the fitting window are arranged in ascending order, and the spectral intensity in the middle or the last position is determined as the baseline intensity of the fitting window. When there are two spectral intensity values ​​in the middle order, the average spectral intensity is taken as the baseline intensity.

[0050] Obtain the number of sampling points, and determine the noise standard deviation based on the number of sampling points, baseline strength, and standard deviation calculation formula.

[0051] The peak signal-to-noise ratio (PSNR) is determined based on the peak amplitude and the standard deviation of the noise level. The calculation formula is as follows:

[0052] Where A is the peak amplitude. This represents the standard deviation of noise.

[0053] The fitting residual is determined based on the fitted spectral intensity, sampling points, and the denoised reconstructed spectral intensity. The calculation formula is as follows:

[0054] in, The residuals are the fitting values, where N is the number of sampling points and A is the peak amplitude. The spectral intensity after denoising and reconstruction. To fit the spectral intensity.

[0055] Furthermore, the specific process of selecting the noise band includes: within the fitting window The first noise band is set on both the shortwave and longwave sides. Second noise band Furthermore, the wavelength interval between the noise band and the fitting window is not less than the wavelength interval threshold. Set the noise band width to Furthermore, the noise band does not contain characteristic spectral emission peaks, and the mean absolute value of the first-order difference of light intensity within the noise band is determined to be no higher than the mean absolute value threshold. This is to eliminate the influence of adjacent spectral lines and strong background fluctuations on noise estimation. Furthermore, when the noise band... and If the noise band contains characteristic spectral emission peaks or its absolute mean value is higher than the absolute mean threshold, then the noise band is shifted away from the fitting window by a preset step size until the condition is met. This embodiment of the invention is not applicable. , , Limit the preset step size.

[0056] Step S105: Determine whether the signal quality index is greater than or equal to the signal quality threshold.

[0057] Specifically, when the peak signal-to-noise ratio of the spectral data of the continuous dissociation frames is not lower than the peak signal-to-noise ratio threshold, the fitting residual is not higher than the residual threshold, and the saturation criterion is not triggered, the signal quality index is greater than or equal to the signal quality index threshold, and step S106 is executed; otherwise, it is determined to be less than the signal quality index threshold, at which point closed-loop compensation is exited and the feedforward control quantity is maintained, and closed-loop compensation is resumed when the signal quality index meets the signal quality index threshold.

[0058] The saturation criterion is: the maximum spectral intensity is greater than the full scale of the spectrometer. Alternatively, whether the proportion of saturated points within the fitting window is greater than the saturated point proportion threshold; where a saturated point is a sampling point whose spectral intensity is greater than the maximum spectral intensity; the peak signal-to-noise ratio threshold, residual threshold, and saturated point proportion threshold are preset and are not limited in this embodiment of the invention.

[0059] Step S106: When the signal quality threshold is met, calculate the actual local pressure value, determine the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve, and generate a dynamic compensation signal based on the pressure deviation.

[0060] Specifically, the actual pressure value is then solved using the inversion relation and the inversion effective interval, and it is determined whether the actual pressure value is within the inversion effective interval. If it is within the inversion effective interval, the actual pressure value is determined to be valid.

[0061] The inversion relation is:

[0062] Where C is the inversion coefficient.

[0063] In this embodiment of the invention, C is determined during the process of establishing the multidimensional mapping function model. For example... Figure 3 The figure shows the Stark broadening variation of characteristic spectral lines under different vacuum pressures according to an embodiment of the present invention, along with a Lorentz fitting diagram. The horizontal axis represents wavelength, and the vertical axis represents normalized spectral intensity. Different curves correspond to different vacuum pressure conditions. As the vacuum pressure changes, the full width at half maximum (FWHM) of the characteristic spectral lines changes. The Lorentz fitting curve is used to extract the actual Stark broadening and determine the calibration coefficients and effective inversion interval for pressure retrieval. The pressure deviation is calculated based on the actual pressure value and the predicted pressure value. The pressure deviation is:

[0064] in, This is the actual pressure value. To predict pressure values.

[0065] Furthermore, a dynamic compensation signal is generated based on the pressure deviation value, and combined with the feedforward control sequence to coordinate the control of the vacuum pump unit speed and the throttle valve opening.

[0066] Furthermore, after determining the pressure deviation, the coefficient matrix of the multidimensional mapping model needs to be updated again using an adaptive scheduling strategy. The update process includes: constructing a cost function based on the pressure deviation value, where the cost function is... Construct a coefficient matrix based on the model's empirical coefficients. The coefficient matrix is ​​as follows: Then update according to the update formula, which is:

[0067] in, Let be the coefficient matrix at time t. For learning rate, The cost function is constructed from the pressure deviation. Cost function The gradient of the coefficient matrix M, The model coefficient matrix consists of pressure conversion gain, power influence index, and velocity decay index. It is understandable that by modifying the multidimensional mapping model, the model can be made to adaptively update along with the welding process, thus more closely approximating the current real physical state.

[0068] Step S107: When the signal quality threshold is not met, compare the actual volume of the current vacuum chamber with the volume threshold, and enable the graded adjustment mode to maintain the vacuum stability of the welding area based on the comparison result.

[0069] Specifically, the volume threshold is determined based on the correspondence between pressure response time and the equivalent volume of the vacuum chamber. Specifically, it includes: the time required for the vacuum chamber pressure to reach 63.2% of the steady-state change after applying a step adjustment is defined as the pressure response time, and the volume of the vacuum chamber corresponding to the pressure response time exceeding the pressure response time threshold is determined as the volume threshold. In this embodiment of the invention, the preferred volume threshold is 0.8. The vacuum chamber volume is 1.2. .

[0070] The graded adjustment modes include single-stage adjustment mode and two-stage adjustment mode; the volume of the vacuum chamber can be detected by the corresponding sensors.

[0071] When the signal quality does not meet the signal quality threshold, the closed-loop compensation based on the inversion pressure is stopped, and the feedforward control is maintained. The vacuum chamber volume and the volume threshold are compared, and a single-stage adjustment mode or a two-stage adjustment mode is activated based on the comparison result to drive the vacuum system to maintain the stability of the vacuum level in the welding area.

[0072] When the actual volume of the vacuum chamber is less than the volume threshold, it indicates that the current system is in a low-inertia operating condition. At this time, a single-stage adjustment command is generated to implement single-stage adjustment mode, controlling the frequency conversion of a single main vacuum pump or the opening of a single-stage valve. This avoids adjustment oscillations caused by the coupling of multiple actuators in a low-volume system and improves system stability. When the vacuum chamber volume is not less than the volume threshold, it indicates that the current system is in a high-inertia operating condition, and a two-stage coordinated control command is generated. This involves controlling the high-speed main vacuum pump to maintain the reference negative pressure, while simultaneously using a small-volume, high-frequency-response auxiliary pump or piezoelectric regulating valve for dynamic fine-tuning compensation. This compensates for the pressure response hysteresis of the large-volume vacuum system and improves the dynamic response speed and control accuracy of local pressure regulation. The piezoelectric regulating valve can perform dynamic fine-tuning compensation according to the dynamic compensation signal.

[0073] Understandably, adopting single-stage and two-stage hierarchical control strategies to address the differences in fluid inertia among vacuum chambers of different volumes can suppress oscillations caused by multi-actuator coupling in small chambers and improve response hysteresis through two-stage synergy in large chambers, thereby enhancing engineering applicability.

[0074] In one possible implementation of the present invention, taking the butt welding of 5A06 aluminum alloy plates as an example, the local pressure based on the welding evaporation behavior during the vacuum laser welding process is inverted and controlled online, and the model parameters are updated online and the experience base is solidified on the same device; such as Figure 4 The diagram shown is a schematic of the welding pressure system provided in an embodiment of the present invention, including a laser emitter 1, a laser beam 2, a molten pool 3, a workpiece to be welded 4, a metal vapor plume 5, an optical acquisition probe 6, a transmission optical fiber 7, a spectrometer 8, a control host 9, a vacuum pump group 10, a vacuum regulating valve 11, a vacuum gauge 12, a vacuum connection pipeline 13, a notch filter 14, a bandpass filter 15, and a vacuum chamber 16.

[0075] During the welding process, the pressure inside the vacuum chamber 16 is stably controlled at 0.1 Pa. A 10kW fiber laser welding head 1 is installed at the top of the chamber, and the acquisition probe 6 is fixedly installed on the side of the welding head 1. The laser emitting head 1 emits a laser beam 2 into the molten pool 3 of the workpiece 4 to be welded, and generates a metal vapor plume 5. The optical acquisition probe 6 collects data through a notch filter 14 and a bandpass filter 15, and transmits the data to a spectrometer 8 through a transmission fiber 7. The spectrometer 8 analyzes the data and sends the analysis results to the control host 9. At the same time, the vacuum machine 12 collects vacuum data. The vacuum regulating valve 11 and the vacuum pump group adjust the vacuum according to the control signal of the control host 9.

[0076] Furthermore, such as Figure 5 The diagram shown is a schematic of the welding pressure control logic provided in an embodiment of the present invention. A bulkhead vacuum gauge is fixedly installed on the side wall of the bulkhead to collect background pressure data. Evacuate and stabilize the background pressure inside the chamber to =1000Pa. Subsequently, exploratory experiments were conducted, scanning the laser power W and welding speed v according to a predetermined combination: W was successively set to 5000W, 7000W, and 9000W; v was successively set to 1.0m / min, 1.5m / min, and 2.0m / min. During the exploratory phase, multiple sets of background pressure reference conditions were superimposed, and the background pressure of the bulkhead vacuum gauge was simultaneously collected under each set of conditions. Spectral data of plasma emission from the welding area were obtained. The 383.8 nm magnesium atom spectral line was used as a characteristic spectral line, and the full width at half maximum (FWHM) was extracted using Lorentz fitting to obtain the Stark broadening. .right and Least square calibration was performed to obtain the inversion coefficients C, and the pressure range where the fitting residuals did not exceed a threshold was taken as the effective inversion range. Within the effective inversion range, the inverted local pressure was used as the calibration value to establish a mapping function model between the local pressure and the process parameters. After the exploratory experiment was completed, the solution was obtained. , and The system is initialized, and the initialized coefficients are written into the experience base in the form of a coefficient matrix, which serves as the benchmark model for subsequent welding start-up feedforward control.

[0077] Upon entering the welding start-up phase, the process parameters for this phase are input into a pre-set mapping model to obtain the vacuum pressure corresponding to the process parameters. A feedforward control sequence is then generated based on the vacuum pressure to achieve initial adjustment. Next, spectral data during the welding process is acquired, and the data undergoes signal cleaning and physical feature extraction. The signal quality of the spectral data is then calculated, and it is determined whether the signal quality meets the signal quality threshold.

[0078] If the signal quality threshold is met, the inverted pressure is calculated, and the pressure deviation between the predicted pressure value and the actual pressure value of the mapping model is calculated. A dynamic compensation signal is generated based on the pressure deviation value, and then coordinated control is performed based on the feedforward control sequence and the dynamic compensation signal. At the same time, the coefficient matrix of the mapping model is updated in real time based on the pressure deviation value.

[0079] If the signal quality threshold is not met, the current feedforward control is maintained, and the vacuum chamber volume is compared with the volume threshold. When the vacuum chamber volume is less than the volume threshold, a single-stage adjustment command is generated to control the frequency conversion of a single main pump or the opening of a single-stage valve. When the vacuum chamber volume is greater than the volume threshold, a two-stage adjustment command is generated to control the high-speed main pump to maintain the reference frequency conversion and to control the fine-tuning of the small-volume auxiliary pump. Then, a control command is output to control the vacuum system. During the welding process, the full width at half maximum (FWHM) of the 383.8 nm magnesium atom spectral line was monitored. The pressure spiked from 0.2 nm to 0.5 nm, and the inversion analysis showed that the local pressure in the welding area increased to over 0.25 Pa; at this time, the vacuum gauge reading on the bulkhead had not yet shown a synchronous abrupt change. The process control host immediately output a compensation command, increasing the frequency of the exhaust pump by 20 Hz and simultaneously adjusting the opening of the electronically controlled throttle valve. Approximately 100 ms later, the pressure disturbance was transmitted to the bulkhead location, and the pressure fluctuation within the chamber was limited to within ±0.02 Pa, maintaining a stable welding process.

[0080] Furthermore, after welding, the updated model coefficient matrix is ​​updated to the experience base for subsequent feedforward control initialization under the same material and working conditions, thereby achieving continuous evolution and stable reuse of the pressure control model. By employing a combined feedforward and feedback drive strategy, both the response during the startup phase and the control accuracy during the steady-state phase are considered. The system sets a spectral signal quality criterion; when the signal quality falls below a threshold, closed-loop compensation is exited while feedforward control is maintained, reducing the risk of runaway due to sensor failure.

[0081] All hardware is connected to the external process control host via the EtherCAT bus. The host performs feedforward prediction, spectral processing, pressure inversion, closed-loop compensation, and online model updates.

[0082] Based on the same inventive concept, this invention also provides a vacuum laser welding pressure control system for welding evaporation behavior. Since the principle of solving the problem by these systems and clients is similar to that of the aforementioned vacuum laser welding pressure control method for welding evaporation behavior, the implementation of this system and client can refer to the implementation of the aforementioned method, and the repeated parts will not be described again.

[0083] This invention provides a vacuum laser welding pressure control system for welding evaporation behavior, the system as follows: Figure 6 As shown, it can specifically include: The multidimensional mapping model construction module 201 is used to collect multiple historical pressure data of the vacuum chamber and multiple historical plasma spectral data of the welding area, and to establish a multidimensional mapping function model and model coefficient matrix of the welding area based on the historical pressure data and historical plasma spectral data. The drive module 202 is used to input the set process parameters at the moment of welding start-up into the multidimensional mapping function model to obtain the predicted local pressure evolution curve, and generate a feedforward control quantity sequence for driving the vacuum actuator for coarse adjustment. The acquisition module 203 is used to acquire raw spectral data of plasma in the molten pool during the welding process in real time. The signal quality index calculation module 204 is used to extract the actual Stark broadening based on the original spectral data and calculate the signal quality index. The judgment module 205 is used to determine whether the signal quality index is greater than or equal to the signal quality threshold; when it is greater than or equal to the signal quality threshold, the compensation module 206 is triggered; when it is less than the signal quality threshold, the graded adjustment module 207 is triggered. The compensation module 206 is used to calculate the actual local pressure value, determine the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve, and generate a dynamic compensation signal for coordinated control based on the pressure deviation. The graded adjustment module 207 is used to compare the actual volume of the current vacuum chamber with the volume threshold, and to enable the graded adjustment mode to maintain the vacuum stability of the welding area based on the comparison result.

[0084] Specifically, the multidimensional mapping model construction module 201 collects multiple pressure data from the vacuum chamber and plasma spectral data obtained during welding during the experimental exploration phase, and constructs a multidimensional mapping function model. The drive module 202 collects and sets the process parameters at the moment of welding start-up, and generates a pressure evolution curve and a feedforward control quantity sequence for driving the vacuum actuator adjustment based on the multidimensional mapping function model constructed by the multidimensional mapping model construction module 201.

[0085] Simultaneously, the acquisition module 203 acquires the raw spectral data of the plasma during the welding process; the signal quality index calculation module 204 performs multi-level signal cleaning, extracts Stark broadening, and calculates multiple signal quality indices on the raw spectral data; the judgment module 205 judges whether each signal quality index meets the threshold of each signal quality index; when the signal quality index threshold is met, the compensation module 206 is triggered; otherwise, the graded adjustment module 207 is triggered.

[0086] The compensation module 206 calculates the actual local pressure value in real time and generates a dynamic compensation signal based on the pressure deviation between the actual local pressure value and the predicted local pressure to achieve fine-grained adjustment.

[0087] The graded adjustment module 207 compares the actual volume of the vacuum chamber with the volume threshold. When the actual volume of the vacuum chamber is less than the volume threshold, a single-stage adjustment command is generated, which controls the frequency conversion of a single main vacuum pump or the opening of a single-stage valve. When the actual volume of the vacuum chamber is not less than the volume threshold, a two-stage coordinated control command is generated, which controls the high-speed main vacuum pump to maintain the reference negative pressure, while using a small-volume, high-frequency response auxiliary pump or piezoelectric valve for dynamic fine-tuning compensation.

[0088] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0089] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for controlling the pressure in vacuum laser welding based on welding evaporation behavior, characterized in that, include: Multiple historical pressure data of the vacuum chamber and multiple historical plasma spectral data of the welding area are collected, and a multidimensional mapping function model and model coefficient matrix of the welding area are established based on the historical pressure data and the historical plasma spectral data. The process parameters set at the moment of welding start are input into the multidimensional mapping function model to obtain the predicted local pressure evolution curve, and a feedforward control quantity sequence for driving the coarse adjustment of the vacuum actuator is generated. Real-time acquisition of raw spectral data of plasma in the molten pool during the welding process; The actual Stark broadening is extracted based on the original spectral data, and the signal quality index is calculated. Determine whether the signal quality index is greater than or equal to the signal quality threshold; When the signal quality threshold is greater than or equal to the actual local pressure value, the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve is determined, and a dynamic compensation signal for coordinated control is generated based on the pressure deviation. When the signal quality is less than the threshold, the actual volume of the vacuum chamber is compared with the volume threshold, and a graded adjustment mode for maintaining vacuum stability in the welding area is activated based on the comparison result.

2. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 1, characterized in that, The step of establishing a multidimensional mapping function model and model coefficient matrix for the welding area based on the pressure data and the plasma spectral data includes: The characteristic spectral lines of the plasma spectral data are determined, and the model Stark broadening of the characteristic spectral lines is extracted based on the characteristic spectral lines and the Lorentz fitting function. Determine the correspondence between the Stark broadening of the model and the pressure data; Obtain the calibration coefficients and effective interval for pressure inversion, and within the effective interval, establish the multidimensional mapping function model based on the correspondence between the model's Stark broadening and the pressure data. The multidimensional mapping function model is as follows: in, Let be the predicted value of the local pressure in the welding area at time t. For the pressure data of the vacuum chamber, Let be the laser power at time t. Let be the welding speed at time t. For pressure conversion gain, The power influence index, This is the velocity decay index.

3. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 2, characterized in that, The step of extracting the actual Stark broadening based on the original spectral data includes: The raw spectral data is subjected to multi-stage signal cleaning. The center wavelength, maximum spectral intensity, and baseline intensity within a fitting window of the characteristic spectral lines of the cleaned spectral data are obtained. The peak amplitude is then determined based on the maximum spectral intensity and the baseline intensity. The peak amplitude is: Where A represents the peak amplitude of the characteristic spectral line after subtracting the baseline intensity. B is the maximum spectral intensity within the fitting window, and B is the baseline intensity within the fitting window. A fitting function is constructed based on the center wavelength, the peak amplitude, and the Lorentz function. The fitting function is as follows: in, The fitted spectral intensity at wavelength, The peak amplitude of the characteristic spectral lines after subtracting the baseline intensity. The center wavelength, The actual Stark broadening of the characteristic spectral line is given by b, where b is the baseline offset of the fitting function. The actual Stark broadening is obtained by solving the fitting function within the effective interval using the least squares method.

4. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 3, characterized in that, The multi-level signal cleaning of the raw spectral data includes: The original spectral data is decomposed using the Daubechies wavelet basis function to obtain wavelet coefficients; The denoising threshold is calculated based on the wavelet coefficients, and the denoising coefficient is determined based on the denoising threshold and the wavelet coefficients. The original spectral data is denoised based on the denoising coefficients to determine the denoised and reconstructed spectral intensity. The original cleaned spectral data is determined by performing a standard normal variable transformation on the denoised and reconstructed spectral intensity using a transformation function, wherein the transformation function is: in, The spectral intensity after denoising and reconstruction. The average value of the denoised reconstructed spectral intensity within the preset standardized window. This represents the standard deviation within the baseline window.

5. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 3, characterized in that, The calculated signal quality metrics include: Target noise bands are selected on both sides of the fitting window, and the noise standard deviation of the spectral intensity within the target noise band is determined based on the baseline intensity and the number of sampling points. The peak signal-to-noise ratio is determined based on the peak amplitude and the noise standard deviation. Obtain the fitted spectral intensity, and determine the fitting residual based on the fitted spectral intensity, the number of sampling points, and the denoised and reconstructed spectral intensity; The peak signal-to-noise ratio, maximum spectral intensity, and fitting residual are determined as the signal quality indicators.

6. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 5, characterized in that, Selecting target noise bands on both sides of the fitting window includes: Noise bands are set on the short-wavelength side and long-wavelength side of the fitting window according to the preset wavelength interval and noise band width; Determine the mean absolute value of the first-order difference of the spectral intensity of the noise band; Determine whether the noise band meets the noise band selection rules, wherein the noise band selection rules are that the noise band does not contain the emission peak of the characteristic spectral line and the mean absolute value of the first difference is not greater than the mean absolute value threshold. If the noise band selection rule is met, then the noise band is determined as the target noise band; If the noise band selection rules are not met, the target noise band is determined according to the preset step size.

7. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 1, characterized in that, After determining the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve, the method further includes: Construct a cost function based on the pressure deviation; The coefficient matrix is ​​updated using the cost function and adaptive update strategy, and the update formula is as follows: in, Let be the coefficient matrix at time t. For learning rate, The cost function is constructed from the pressure deviation. Cost function The gradient of the coefficient matrix M, The model coefficient matrix consists of pressure conversion gain, power influence index, and velocity decay index. .

8. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 1, characterized in that, The step of activating a graded adjustment mode to maintain vacuum stability in the welding area based on the comparison results includes: When the comparison result indicates that the actual volume is not less than the volume threshold, a two-stage coordinated adjustment mode is activated, which uses a high-speed main vacuum pump to maintain the reference pumping speed and a small-volume auxiliary pump for fine-tuning. When the comparison result indicates that the actual volume is less than the volume threshold, a single-stage adjustment mode for adjusting pressure with a single device is activated.

9. The vacuum laser welding pressure control method for welding evaporation behavior as described in claim 1, characterized in that, The raw spectral data is acquired by the spectral acquisition unit of the integrated filter assembly, which includes dual filters: the first filter is a high-density notch filter, the center stopband wavelength of which is the same as the center wavelength of the welding laser source; the second filter is a bandpass filter, the band containing the characteristic emission spectral lines of the material to be welded is located within the bandpass range of the bandpass filter.

10. A vacuum laser welding pressure control system for welding evaporation behavior, characterized in that, include: The multidimensional mapping model construction module is used to collect multiple historical pressure data of the vacuum chamber and multiple historical plasma spectral data of the welding area, and to establish a multidimensional mapping function model and model coefficient matrix of the welding area based on the historical pressure data and the historical plasma spectral data. The drive module is used to input the set process parameters at the moment of welding start-up into the multidimensional mapping function model to obtain the predicted local pressure evolution curve, and generate a feedforward control quantity sequence for driving the coarse adjustment of the vacuum actuator. The acquisition module is used to acquire raw spectral data of the plasma in the molten pool during the welding process in real time. The signal quality index calculation module is used to extract the actual Stark broadening based on the original spectral data and calculate the signal quality index. The judgment module is used to determine whether the signal quality index is greater than or equal to the signal quality threshold; when it is greater than or equal to the signal quality threshold, the compensation module is triggered; when it is less than the signal quality threshold, the graded adjustment module is triggered. The compensation module is used to calculate the actual local pressure value, determine the pressure deviation between the actual local pressure value and the predicted local pressure in the predicted local pressure evolution curve, and generate a dynamic compensation signal for coordinated control based on the pressure deviation. The graded adjustment module is used to compare the actual volume of the current vacuum chamber with the volume threshold, and to enable the graded adjustment mode to maintain the vacuum stability of the welding area based on the comparison result.

Images