A multi-phase laser welding control method and system

Abstract

The present application relates to the technical fields of laser welding, and discloses a kind of multi-phase laser welding control method and system, including collecting the molten pool radiation image data of multiple wave bands, fusion processing constructs three-dimensional temperature gradient model and extracts three-dimensional dynamic characteristic data, the three-dimensional dynamic characteristic data is compared with the preset model quantization error, generates and includes phase, power and beam deflection collaborative control instruction, the original laser is decomposed into multiple micro-beams and the phase delay of each micro-beam is adjusted, the spatial energy distribution mode of processing plane is dynamically reshaped using spatial interference superposition effect, according to instruction fine-tuning instantaneous input power and executing beam spatial deflection action.The present application eliminates temperature measurement deviation by multi-wave band fusion, uses hardware decoupling matrix to separate control instruction, cooperates with micro-beam interference to dynamically reshape energy distribution, eliminates energy compensation and time lag of beam deflection, improves energy distribution flexibility and response speed, and guarantees welding quality.

Description

Technical Field

[0001] This invention relates to the field of laser welding technology, and in particular to a multi-phase laser welding control method and system. Background Technology

[0002] Laser welding technology is widely used in the connection of metal structural components in modern industrial precision manufacturing due to its advantages such as high energy density, low heat input, and high processing precision. As the requirements for processing quality and process stability continue to increase, the requirements for the coordinated control of beam energy distribution, input power adjustment, and molten pool physical state monitoring are also becoming increasingly stringent.

[0003] Current laser welding control methods typically employ a single-band sensor to monitor the molten pool surface. This approach is susceptible to measurement errors caused by smoke and dust splashes and changes in material emissivity during welding. Furthermore, the acquired two-dimensional planar data cannot fully characterize the three-dimensional temperature gradient and hydrodynamic state within the molten pool. Regarding energy regulation, existing systems lack decoupling mechanisms for multiple physical parameters, leading to mutual interference between adjustments of various parameters. Simultaneously, traditional power regulation methods have low physical response rates, failing to effectively suppress high-frequency thermal oscillations in the molten pool. Moreover, there is a time lag between energy compensation and beam spatial deflection, making it difficult to ensure accurate correspondence between the processing position and input energy. In addition, the fixed beam spatial energy distribution pattern of traditional processing equipment cannot dynamically reshape the energy field of the processing plane at the microscopic level, limiting the spatial allocation flexibility of laser energy input and easily causing thermodynamic imbalances and welding defects.

[0004] Therefore, how to eliminate interference errors in the molten pool monitoring process to improve the comprehensiveness of physical state characterization, and to achieve flexible reshaping of spatial energy distribution patterns and precise spatiotemporal matching of composite parameters are problems that need to be solved in this field. Summary of the Invention

[0006] The purpose of this invention is to provide a multi-phase laser welding control method and system, which solves the problems of measurement deviation, mutual interference of multiple physical parameter adjustments, and lag in the spatiotemporal response of energy regulation in the prior art.

[0007] To address the above problems, the present invention provides the following technical solution:

[0008] The first aspect of this invention provides a multi-phase laser welding control method, which adopts the following technical solution:

[0009] A multi-phase laser welding control method includes the following steps:

[0010] The imaging feedback subsystem acquires molten pool radiation image data in multiple bands in the molten pool processing area, performs multi-band data fusion processing on the molten pool radiation image data to construct a three-dimensional temperature gradient model of the molten pool, and extracts three-dimensional dynamic feature data based on the three-dimensional temperature gradient model.

[0011] The three-dimensional dynamic feature data is received by the central processing subsystem. The three-dimensional dynamic feature data is compared and analyzed with the preset welding process parameter model and the current physical state error of the molten pool is quantified. Based on the physical state error of the molten pool, a collaborative control command is generated. The collaborative control command includes a phase adjustment command, a power adjustment command, and a beam deflection command.

[0012] The original single laser beam is decomposed into multiple microbeams by a laser phase array subsystem. The phase delay between each microbeam is adjusted according to the phase control command. The spatial energy distribution pattern of the processing plane is dynamically reshaped by the spatial interference superposition effect of each microbeam on the surface of the molten pool.

[0013] The composite control subsystem fine-tunes the instantaneous input power of the laser beam passing through it according to the power adjustment command, and performs beam spatial deflection according to the beam deflection command, thereby completing the closed-loop composite control of the energy input value and the beam's position.

[0014] By adopting the above technical solution, a three-dimensional temperature gradient model is constructed by multi-band data fusion processing. Combined with the spatial interference superposition effect, the spatial energy distribution pattern is dynamically reshaped. The composite execution of instantaneous input power fine-tuning and beam spatial deflection is completed in the closed loop. Therefore, the technical problem that single parameter adjustment cannot take into account the multi-dimensional dynamic fluctuations of the molten pool is solved. The spatial allocation flexibility and time response speed of laser energy input are improved, and the thermodynamic stability and processing quality of the welding process are guaranteed.

[0015] Furthermore, the step of acquiring molten pool radiation image data of multiple bands in the molten pool processing area and performing multi-band data fusion processing on the molten pool radiation image data to construct a three-dimensional temperature gradient model of the molten pool specifically includes:

[0016] A global hardware timer is used to send a global trigger pulse to synchronously collect molten pool radiation image data in the visible light band, near-infrared band, and ultraviolet band in the molten pool processing area;

[0017] The base temperature is calculated using the spectral radiant output at the first and second detection wavelengths extracted from the visible light and near-infrared bands, respectively, through a dual-band colorimetric thermometry algorithm.

[0018] The high-energy radiative exitance extracted from the ultraviolet band is used to perform synergistic correction and fusion of the base temperature to generate a two-dimensional surface temperature field distribution.

[0019] By adopting the above technical solution, radiation image data of multiple frequency bands are acquired simultaneously. The base temperature is obtained by dual-band colorimetric thermometry and coordinated correction is performed by combining it with high-energy radiation in the ultraviolet band. This eliminates the measurement deviation caused by smoke and dust interference or changes in material emissivity during single-band temperature measurement and improves the accuracy of the two-dimensional surface temperature field distribution data.

[0020] Furthermore, the step of constructing a three-dimensional temperature gradient model of the molten pool by performing multi-band data fusion processing on the molten pool radiation image data also includes:

[0021] By combining the deep-hole heat conduction model, the rate of heat attenuation and transfer from the surface to the bottom of the molten pool is calculated using the thermal diffusivity of the target metal material;

[0022] The three-dimensional temperature gradient model is constructed by mapping the two-dimensional surface temperature field distribution according to the attenuation transfer rate.

[0023] By adopting the above technical solution, the thermal conductivity physical properties of metallic materials are introduced, and the heat distribution data of the two-dimensional plane is recursively mapped along the depth direction to form a three-dimensional temperature gradient model that characterizes the thermodynamic state inside the molten pool, providing a three-dimensional temperature field data foundation for subsequent process condition evaluation.

[0024] Furthermore, the step of extracting three-dimensional dynamic feature data based on the three-dimensional temperature gradient model specifically includes:

[0025] Based on the absolute temperature gradient values ​​extracted from the three-dimensional temperature gradient model and the preset surface tension temperature coefficient, the surface tension gradient value distribution of the processing plane is calculated as surface tension fluctuation information.

[0026] Extract the grayscale abrupt edges from the molten pool radiation image data, obtain the pixel coordinates of the highlight contour of the molten pool, convert the pixel coordinate domain into the actual three-dimensional physical size coordinate domain through spatial calibration, and calculate and generate the curve of the geometric size and area of ​​the molten pool changing over time as the molten pool morphology evolution information.

[0027] The surface tension fluctuation information, the temperature field distribution information corresponding to the two-dimensional surface temperature field distribution, and the molten pool morphology evolution information are packaged and integrated into the three-dimensional dynamic feature data.

[0028] By adopting the above technical solution, the surface tension fluctuation state is derived from the temperature gradient, and the objective physical geometric size change law is obtained by combining image contour coordinate transformation. The thermodynamic characteristics, fluid dynamic characteristics and macroscopic morphological characteristics are integrated into the data, which improves the comprehensiveness of the physical state characterization of the molten pool.

[0029] Furthermore, the step of comparing and analyzing the three-dimensional dynamic feature data with a preset welding process parameter model and quantifying the current physical state error of the molten pool, and generating collaborative control commands based on the physical state error of the molten pool specifically includes:

[0030] The three-dimensional dynamic feature data is converted into the actual three-dimensional dynamic feature vector at the current moment. The actual three-dimensional dynamic feature vector at the current moment is then subjected to matrix difference operation with the ideal state reference vector at the current moment output by the preset welding process parameter model to calculate the state error vector.

[0031] Combining the proportional-integral-differential (PI-DI) operation mechanism, based on the instantaneous value, historical cumulative amount, and trend of the state error vector, the overall control input vector at the current moment required to offset the deviation is calculated and output.

[0032] The overall control input vector is separated by a linear transformation using a hardware decoupling mapping matrix, and then decomposed and mapped into the phase adjustment command, the power adjustment command, and the beam deflection command.

[0033] By adopting the above technical solution, multidimensional features are converted into vectors for matrix difference operations. The transient error and steady-state deviation are handled by combining proportional-integral-differential operation mechanism. Variable separation is achieved through hardware decoupling mapping matrix, eliminating strong coupling interference between multiple physical parameters in the control process and ensuring the independence and accuracy of various control command outputs.

[0034] Furthermore, the step of using a hardware decoupling mapping matrix to perform a linear transformation and separate the overall control input vector, decomposing and mapping it into the phase adjustment command, the power adjustment command, and the beam deflection command specifically includes:

[0035] The low-frequency demands for minor local morphological mutations and spatial reshaping are allocated to the phase modulation commands;

[0036] The high-frequency surface tension oscillations and the high-frequency energy transient demand for heat loss or overload in the overall temperature field are allocated to the power regulation command.

[0037] The macroscopic positional offset of the molten pool and the asymmetric requirements of the heat-affected zone are assigned to the beam deflection command.

[0038] By adopting the above technical solution, based on the frequency domain characteristics of the molten pool fluctuation state and the macroscopic physical requirements, the control task is split and assigned to control branches with different hardware response characteristics. This fully utilizes the physical advantages of each actuator and avoids process loss of control caused by insufficient response bandwidth of a single actuator.

[0039] Furthermore, the step of decomposing the original single laser beam into multiple microbeams through the laser phase array subsystem, adjusting the phase delay between each microbeam according to the phase modulation command, and dynamically reshaping the spatial energy distribution pattern of the processing plane using the spatial interference superposition effect of each microbeam on the molten pool surface specifically includes:

[0040] The original single-beam laser is physically divided into a two-dimensional microbeam array with uniform energy to form each microbeam, and the initial electric field complex amplitude distribution equation of each microbeam in the spatial physical coordinate system is established.

[0041] The phase modulation command is parsed using the drive control circuit inside the spatial light modulator and converted into a corresponding pixel-level drive voltage matrix.

[0042] The pixel-level driving voltage matrix is ​​used to control the deflection of liquid crystal molecules, thereby changing the optical refractive index of the region where the liquid crystal molecules are located, and injecting an independent dynamic phase delay into each of the micro-beams.

[0043] Each of the microbeams carrying different dynamic phase delay values ​​is projected onto the processing plane through the subsequent optical path, and forms an alternating high and low energy interference pattern in space based on the interference superposition effect of the modulated complex amplitude.

[0044] By adopting the above technical solution, the refractive index is changed by the pixel-level driving voltage output by the spatial light modulator, and the dynamic phase injection of the independent micro beam is completed. The energy distribution is reshaped by the spatial interference effect after the micro beam is projected, and the non-mechanical high-speed optical distribution reconstruction is completed, which meets the requirement of flexible distribution of surface energy field in the processing area.

[0045] Furthermore, the step of fine-tuning the instantaneous input power of the laser beam passing through the composite control subsystem according to the power adjustment command specifically includes:

[0046] The acousto-optic modulator in the composite control subsystem parses the power adjustment command and converts the power adjustment command into a radio frequency ultrasonic signal acting on the acousto-optic crystal;

[0047] The radio frequency ultrasonic signal is used to generate a dynamic refractive index grating inside the acousto-optic crystal. Based on the physical diffraction efficiency of the acousto-optic crystal, the instantaneous output power of the laser beam passing through the composite control subsystem and outputting to the first-order light direction is finely adjusted at high frequency through the acousto-optic diffraction effect.

[0048] By adopting the above technical solution, a refractive index grating is constructed inside the acousto-optic crystal using radio frequency ultrasonic signals. The energy transmission ratio of the beam is directly controlled by the diffraction efficiency of the crystal. Compared with the method of adjusting the pump source current of the laser, the physical response rate of the power adjustment action is significantly improved, and the high-frequency thermal oscillation on the surface of the molten pool is effectively suppressed.

[0049] Furthermore, the step of executing the spatial deflection of the beam according to the beam deflection command to complete the closed-loop composite control of the energy input value and the beam's position specifically includes:

[0050] The high-speed galvanometer assembly in the composite control subsystem drives the dual-axis servo motor to perform the spatial deflection of the beam according to the beam deflection command.

[0051] The instantaneous power output status parameters of the acousto-optic modulator are collected in real time by the power detection unit, and the real-time deflection angle parameters of the high-speed galvanometer assembly are continuously collected by the deflection angle detection unit and fed back to the central processing subsystem.

[0052] By relying on a hardware timer and a high-speed field-programmable gate array for clock synchronization, while the high-speed galvanometer assembly guides the laser beam to the target offset region, the acousto-optic modulator synchronously performs energy compensation for heat loss or overload, thus completing the closed-loop composite control of the energy input value and the beam's position.

[0053] By adopting the above technical solution, a synchronous monitoring feedback loop for beam position and output power is established. Under the same clock reference, the position deflection action of the servo motor and the energy adjustment action of the acousto-optic crystal are synchronously triggered, eliminating the time lag between spatial displacement and energy injection, and ensuring the accurate correspondence between processing position and compensation energy.

[0054] A second aspect of the present invention also provides a multi-phase laser welding control system, which adopts the following technical solution:

[0055] A multi-phase laser welding control system, comprising:

[0056] The basic electromechanical auxiliary subsystem is used to provide physical support for the multi-phase laser welding process and monitor the operating environment status.

[0057] An imaging feedback subsystem is used to acquire molten pool radiation image data of multiple bands in the molten pool processing area, perform multi-band data fusion processing on the molten pool radiation image data to construct a three-dimensional temperature gradient model of the molten pool, and extract three-dimensional dynamic feature data based on the three-dimensional temperature gradient model.

[0058] The central processing subsystem is used to receive the three-dimensional dynamic feature data, compare and analyze the three-dimensional dynamic feature data with the preset welding process parameter model and quantify the current physical state error of the molten pool, and generate collaborative control instructions based on the physical state error of the molten pool. The collaborative control instructions include phase adjustment instructions, power adjustment instructions and beam deflection instructions.

[0059] A laser phase array subsystem is used to decompose the original single laser beam into multiple microbeams, adjust the phase delay between each microbeam according to the phase modulation command, and dynamically reshape the spatial energy distribution pattern of the processing plane by utilizing the spatial interference superposition effect of each microbeam on the surface of the molten pool.

[0060] The composite control subsystem is used to fine-tune the instantaneous input power of the laser beam passing through the composite control subsystem according to the power adjustment command, and to perform beam spatial deflection action according to the beam deflection command, thereby completing the closed-loop composite control of energy input value and beam position.

[0061] By adopting the above technical solutions, the system integrates equipment environment monitoring, multi-band image processing, centralized computing control, and dual optical execution hardware, constructing a hardware architecture from status acquisition and multi-dimensional decoupled calculation to optical field distribution reshaping and power position composite fine-tuning. With the cooperation of data communication and clock synchronization mechanisms, each hardware subsystem completes the joint control of optical energy in the temporal and spatial domains, improving its adaptability to complex working planes.

[0062] In summary, the present invention has at least one of the following beneficial technical effects:

[0063] 1. This invention utilizes a hardware decoupling mapping matrix to linearly transform and separate the overall control input vector. In conjunction with the laser phase array subsystem, it dynamically reshapes the spatial energy distribution pattern by utilizing the spatial interference superposition effect of each microbeam on the processing plane. This eliminates the mutual interference between multiple physical parameters, improves the spatial allocation flexibility and time response speed of laser energy input, and ensures the thermodynamic stability and processing quality of the welding process.

[0064] 2. This invention utilizes an acousto-optic modulator to construct a refractive index grating using radio frequency ultrasonic signals to perform high-frequency fine-tuning of instantaneous input power. Furthermore, it relies on a hardware timer and a high-speed field-programmable gate array to synchronize the energy adjustment action with the spatial deflection action of the high-speed galvanometer assembly. This improves the physical response rate of the power adjustment action, smooths out high-frequency thermal oscillations, and eliminates the time lag between spatial displacement and energy injection to ensure the correspondence between the processing position and the compensation energy.

[0065] 3. This invention synchronously acquires molten pool radiation image data in multiple bands through an imaging feedback subsystem, obtains the base temperature using a dual-band colorimetric thermometry algorithm, and performs collaborative correction by combining it with high-energy radiation in the ultraviolet band. It also constructs a three-dimensional temperature gradient model by mapping down the deep molten hole heat conduction model, thereby eliminating measurement deviations caused by smoke interference or changes in material emissivity. It integrates thermodynamic characteristics with macroscopic morphological characteristics to improve the comprehensiveness of the physical state characterization of the molten pool. Attached Figure Description

[0066] Figure 1 This is a structural block diagram of a multi-phase laser welding control system according to an embodiment of the present invention;

[0067] Figure 2 This is a flowchart of a multi-phase laser welding control method according to an embodiment of the present invention;

[0068] Figure 3 This is a comparison curve of molten pool temperature changes according to an embodiment of the present invention;

[0069] Figure 4 This is a graph showing the instantaneous output power response of an embodiment of the present invention;

[0070] Figure 5 This is a bar chart comparing weld flatness errors according to an embodiment of the present invention. Detailed Implementation

[0071] The technical solutions in 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.

[0072] See attached document Figure 1 The present invention provides a multi-phase laser welding control system, which may include: a basic electromechanical auxiliary subsystem, a laser phase array subsystem, a composite control subsystem, an imaging feedback subsystem, and a central processing subsystem.

[0073] The basic electromechanical auxiliary subsystem includes the machine tool bed assembly, water-cooled unit, gas supply circuit, and basic optical lens assembly. The machine tool bed assembly provides the physical support surface and controlled movement mechanism. The water-cooled unit is connected to the working equipment for circulating heat dissipation and is equipped with water pressure and temperature sensors to collect real-time data on operating water pressure and temperature. The gas supply circuit includes compressed air and laser gas pipelines, with pressure sensors collecting gas pressure values ​​for the corresponding pipelines. The basic optical lens assembly is equipped with contamination monitoring sensors to detect surface contamination status data on the reflecting and focusing mirrors. The basic electromechanical auxiliary subsystem transmits the collected operating water pressure, operating water temperature, gas pressure, and surface contamination status data to the central processing subsystem for status monitoring.

[0074] The laser phase array subsystem, arranged sequentially along the optical path, includes a laser emitter, a microbeam splitter assembly, a spatial light modulator, and a phase detection and adjustment unit. The laser emitter outputs a raw single-beam laser. This raw single-beam laser is incident on the microbeam splitter assembly for splitting and then projected onto the working surface of the spatial light modulator.

[0075] The spatial light modulator receives phase modulation commands from the central processing subsystem and performs pixel-level phase delay modulation on the projected original single-beam laser, decomposing the original single-beam laser into multiple independently controllable micro-beams. The phase detection and adjustment unit is connected to the spatial light modulator. The phase detection and adjustment unit collects the real-time phase values ​​of each micro-beam and feeds the real-time phase values ​​back to the spatial light modulator to adjust the phase difference between the micro-beams.

[0076] The composite control subsystem includes an acousto-optic modulator, a high-speed galvanometer assembly, a power detection unit, and a deflection angle detection unit. The acousto-optic modulator is located in the main optical path between the laser emitter and the micro-beam splitter assembly. It receives power adjustment commands from the central processing subsystem and adjusts the input power of the original single-beam laser passing through it via the acousto-optic effect. The high-speed galvanometer assembly is located in the subsequent optical path of the spatial light modulator. It receives beam deflection commands from the central processing subsystem and controls the beam's deflection angle and trajectory in space.

[0077] The power detection unit acquires the instantaneous power output status parameters of the acousto-optic modulator. The deflection angle detection unit acquires the real-time deflection angle parameters of the high-speed galvanometer assembly. The composite control subsystem continuously feeds back the instantaneous power output status parameters and the real-time deflection angle parameters to the central processing subsystem.

[0078] The imaging feedback subsystem includes a multispectral imaging camera, an optical filtering component, a 3D data reconstruction unit, and a temperature gradient analysis unit. The multispectral imaging camera is equipped with an optical filtering component at its front end and is arranged coaxially or off-axis with the processing optical path. The multispectral imaging camera simultaneously acquires molten pool radiation image data in the visible, near-infrared, and ultraviolet bands of the working area and transmits the acquired molten pool radiation image data to the 3D data reconstruction unit and the temperature gradient analysis unit.

[0079] The 3D data reconstruction unit and the temperature gradient analysis unit perform multi-band data fusion on the molten pool radiation image data to construct a 3D temperature gradient model of the molten pool. Based on the 3D temperature gradient model, the 3D data reconstruction unit and the temperature gradient analysis unit extract surface tension fluctuation information, temperature field distribution information, and molten pool morphology evolution information of the molten pool. The imaging feedback subsystem transmits the extracted dynamic feature data to the central processing subsystem.

[0080] The central processing subsystem is equipped with a communication interface and a bus architecture. Through this bus architecture, the central processing subsystem connects to the basic electromechanical auxiliary subsystem, the laser phase array subsystem, the composite control subsystem, and the imaging feedback subsystem. The central processing subsystem receives dynamic feature data transmitted from the imaging feedback subsystem and physical state feedback data transmitted from the basic electromechanical auxiliary subsystem.

[0081] The central processing subsystem analyzes and processes the received data according to a preset welding process algorithm, generating joint collaborative control commands. These commands include phase modulation commands, power adjustment commands, and beam deflection commands. The central processing subsystem sends the phase modulation command to the laser phase array subsystem, and simultaneously sends the power adjustment and beam deflection commands to the composite control subsystem, driving each subsystem to execute closed-loop control steps.

[0082] See attached document Figure 2 This invention provides a multi-phase laser welding control method, comprising the following steps:

[0083] S100 and the basic electromechanical auxiliary subsystem monitor the operating water temperature and pressure of the water-cooled unit and the gas pressure of the gas supply circuit in real time throughout the entire welding process. They also acquire surface contamination status data of the basic optical lens group. When the underlying operating environment data deviates from the preset safety threshold, the underlying safety interlock protection mechanism is triggered.

[0084] The S200 imaging feedback subsystem uses a multispectral imaging camera to simultaneously acquire molten pool radiation image data in the visible light, near-infrared, and ultraviolet bands of the molten pool processing area. Through the three-dimensional data reconstruction unit and the temperature gradient analysis unit, the molten pool radiation image data is processed by multi-band data fusion to eliminate dynamic emissivity interference, construct a three-dimensional temperature gradient model of the molten pool, and extract the surface tension fluctuation information, temperature field distribution information, and molten pool morphology evolution information of the molten pool as three-dimensional dynamic feature data based on the three-dimensional temperature gradient model.

[0085] The S300 and central processing subsystem receive three-dimensional dynamic feature data and low-level operating environment data fed back by the basic electromechanical auxiliary subsystem in real time. They compare and analyze the three-dimensional dynamic feature data with the preset welding process parameter model, call the adaptive multivariable control algorithm to quantify the current physical state error of the molten pool, and generate collaborative control instructions based on the physical state error. The collaborative control instructions include phase adjustment instructions, power adjustment instructions, and beam deflection instructions for different hardware decoupling.

[0086] The S400 laser phase array subsystem receives phase control commands from the central processing subsystem and adjusts the phase delay between the decomposed microbeams in real time through the spatial light modulator. It uses the spatial interference superposition effect of the microbeams on the surface of the molten pool to dynamically reshape the spatial energy distribution pattern of the processing plane, so as to counteract the surface tension fluctuations of the molten pool and maintain the stability of the molten pool shape.

[0087] The S500 and composite control subsystem synchronously receive power adjustment and beam deflection commands from the central processing subsystem, drive the acousto-optic modulator to dynamically fine-tune the instantaneous input power of the passing laser beam in a high-frequency response state, and coordinate with the high-speed galvanometer assembly to synchronously perform high-precision beam spatial deflection, thus completing the closed-loop composite control of energy input value and beam position.

[0088] The following section elaborates on the specific technical implementation details of each step, using a concrete physical model and control mechanism.

[0089] To ensure the stable operation of subsequent high-frequency optical path modulation and multi-dimensional data closed-loop processing, the multi-phase laser welding control system, before performing complex optical modulation and parameter decisions, uses a basic electromechanical auxiliary subsystem to perform real-time monitoring and logical judgment of the underlying physical carrier state throughout the entire processing cycle. The specific implementation process of step S100 is divided into the following sub-steps:

[0090] S101, the basic electromechanical auxiliary subsystem performs status polling on the water-cooled unit and gas supply circuit at a preset sampling frequency to obtain real-time operating parameters of the fluid medium. The central processing subsystem has a built-in state threshold matrix, and the real-time operating parameters include operating water temperature, operating water pressure, and gas pressure values.

[0091] The central processing subsystem continuously determines the numerical relationship between the acquired real-time operating parameters and the state threshold matrix. Specifically, the central processing subsystem determines whether the operating water pressure is lower than the preset safety lower limit of 8.5 bar, whether the operating water temperature is within the preset operating temperature range of 16±1℃, and whether the gas pressure is lower than the preset safety lower limit of 4.5 bar.

[0092] S102, The pollution monitoring sensor configured in the basic optical lens group collects the surface pollution status data of the reflecting mirror and the focusing mirror in real time, and transmits the acquired surface pollution status data to the central processing subsystem to calculate the current scattering loss rate.

[0093] In principle, pollution monitoring sensors utilize the principle of surface light scattering to assess the degree of dust or oil contamination by collecting the intensity of diffusely reflected light overflowing from the surface of a mirror. In actual processing environments, when high-energy lasers pass through or are reflected, dust particles or oil stains adhering to the surface of the optical mirror disrupt the microscopic optical flatness of the mirror, causing photons that should be directionally transmitted to undergo spatial deflection, forming diffusely reflected light. The more severe the contamination of the optical lens, the greater the intensity of the deflected diffusely reflected light, thus directly reflecting the light transmission health of the lens.

[0094] After receiving the surface contamination status data, the central processing subsystem calculates the scattering loss rate, expressed by the following formula:

[0095] ;

[0096] In the formula, This refers to the scattering loss rate; The real-time intensity of scattered light detected by pollution monitoring sensors; To calibrate the reference light intensity under pollution-free conditions. The reference light intensity value is fixed and calibrated based on substrate test data of the optical lens group in a pollution-free standard environment.

[0097] After the calculation is completed, the central processing subsystem logically compares the calculated scattering loss rate with the preset scattering loss threshold. The preset scattering loss threshold is set to a range of 3% to 5%, and its specific value is determined based on a comprehensive safety calibration of the maximum power limit of the processing laser and the thermal damage threshold of the basic optical lens group.

[0098] S103. When the real-time operating parameters deviate from the set threshold range, or the scattering loss rate is greater than the preset scattering loss threshold, the central processing subsystem determines that the underlying physical operating environment is abnormal and immediately triggers the underlying safety interlock protection action.

[0099] Once the triggering conditions are met, the central processing subsystem sends a high-level interrupt signal through the hardware bus to directly cut off the power supply circuit of the laser emitter. At the same time, it stops the servo drive source of the controlled moving mechanism in the machine tool bed assembly and simultaneously sends the equipment shutdown status code to the peripheral port of the multi-phase laser welding control system.

[0100] The underlying safety interlock protection action achieves physical isolation through a hardware relay group. Upon receiving a high-level interrupt signal, the internal electromagnetic coil of the hardware relay group instantly flips its state, causing the mechanical main contacts to physically open, thereby blocking energy output within a preset control period of 1 to 5 milliseconds. This prevents damage to the processing equipment due to basic heat dissipation failure or severe contamination of the optical lenses. The specific hardware connection structure and photoelectric conversion circuit implementation for extracting scattered light from the optical mirror in step S100 are well-known technologies in the field and will not be described further here.

[0101] Under the premise of maintaining a stable basic operating environment, the multi-phase laser welding control system overcomes the errors caused by the dynamic changes in the emissivity of the molten metal in traditional single-wavelength temperature measurement. It uses an imaging feedback subsystem to accurately extract and reconstruct the physical characteristics of the molten pool in three dimensions. The specific implementation process of step S200 is divided into the following sub-steps:

[0102] S201. The imaging feedback subsystem controls the multispectral imaging camera to simultaneously acquire molten pool radiation image data in the visible, near-infrared, and ultraviolet bands of the working area. To ensure absolute consistency of data across multiple bands in the temporal dimension, the imaging feedback subsystem uses an internal global hardware timer to send global trigger pulses to each independent sensor array within the multispectral imaging camera, enabling the image sensors of the three bands to perform exposure operations of a uniform duration at the same time. In real-world physical scenarios, a high-speed moving laser beam can cause significant changes in the morphology and temperature distribution of the molten pool surface. If the exposures of different bands of the multispectral camera are not synchronized, the molten state at the same physical moment will result in positional misalignment and morphological deviations in images from different bands, compromising the accuracy of subsequent data fusion.

[0103] In terms of physical principles, the energy distribution of electromagnetic waves radiated outward from a molten metal pool in a high-temperature molten state follows Planck's law of radiation. The three-dimensional data reconstruction unit extracts the grayscale value of each pixel in the molten pool radiation image data and converts it into spectral radiant exitance at a specific detection wavelength. The specific physical equation is expressed as follows:

[0104] ;

[0105] In the formula, Spectral radiative exitance at a specific detection wavelength; The dynamic spectral emissivity of the metallic material at this specific detection wavelength; For a specific detection wavelength; The absolute temperature of the molten pool; It is the first radiation constant; is the second radiation constant. Wherein, and These are all standard physical constants, and those skilled in the art can directly take their values ​​based on Planck's radiation law.

[0106] S202, the three-dimensional data reconstruction unit and the temperature gradient analysis unit receive multi-band molten pool radiation image data and perform multi-band data fusion processing on the molten pool radiation image data to eliminate the temperature measurement error caused by dynamic spectral emissivity changes, thereby constructing a three-dimensional temperature gradient model of the molten pool.

[0107] During high-temperature laser welding, the metal transitions from a solid to a liquid and then to a gaseous state, resulting in significant fluctuations in its dynamic spectral emissivity. The three-dimensional data reconstruction unit selects a first detection wavelength corresponding to the visible light band, with a preset range of 400nm to 700nm (e.g., using a 532nm characteristic spectral line). The second detection wavelength corresponding to the near-infrared band is preset to a range of 800nm ​​to 1100nm (e.g., using a 1064nm characteristic spectral line). Using a dual-band colorimetric thermometry algorithm, assuming that the dynamic spectral emissivity of the material tends to be equal within similar wavelength ranges, the fundamental temperature calculation equation, independent of the unknown emissivity, is derived by solving for the ratio of spectral radiative exitance at the two wavelengths.

[0108] ;

[0109] In the formula, To calculate the absolute temperature of the molten pool; The first detection wavelength; This is the second detection wavelength; The spectral radiative exitance at the first detection wavelength; The spectral radiative exitance at the second detection wavelength; is the second radiation constant.

[0110] Based on this, the 3D data reconstruction unit utilizes the high-energy radiative exitance extracted from the third detection wavelength corresponding to the ultraviolet band (preset range between 200nm and 400nm, for example, using the 355nm characteristic spectral line) to perform high-frequency temperature abrupt change collaborative correction and fusion on the aforementioned base temperature. The 3D data reconstruction unit performs temperature calculation and correction on each effective pixel within the image area, generating a high-precision two-dimensional surface temperature field distribution. Subsequently, the 3D data reconstruction unit combines the deep-pore heat conduction model of the metallic material to map the two-dimensional surface temperature field distribution along the depth direction. During this physical mapping process, the 3D data reconstruction unit uses the known thermal diffusivity of the target metallic material and calculates the rate of heat attenuation and transfer from the surface to the bottom of the molten pool according to Fourier's law of thermal conduction, thereby expanding the two-dimensional surface features into three-dimensional spatial internal data and constructing a complete three-dimensional temperature gradient model of the molten pool.

[0111] S203, the temperature gradient analysis unit extracts the surface tension fluctuation information, temperature field distribution information and molten pool morphology evolution information of the molten pool based on the three-dimensional temperature gradient model, and packages these three parts of information as three-dimensional dynamic feature data.

[0112] Based on the Marangoni effect, the surface tension distribution of a fluid is strongly correlated with its surface temperature; temperature differences directly drive convective oscillations in the liquid metal within the molten pool. The temperature gradient analysis unit calculates the surface tension gradient at any spatial coordinate point using a three-dimensional temperature gradient model. The specific physical equation is as follows:

[0113] ;

[0114] In the formula, This represents the surface tension gradient value; is the surface tension temperature coefficient, which is selected by looking up a table or experiment based on the basic thermophysical parameters of the metal material being processed, and represents the change in surface tension caused by a unit temperature change. This represents the absolute temperature gradient value.

[0115] The temperature gradient analysis unit calculates the surface tension gradient distribution across the entire processing plane, serving as information on surface tension fluctuations. Simultaneously, the image processing unit within the imaging feedback subsystem extracts the highlight contour pixel coordinates of the molten pool based on the grayscale abrupt changes at the edges of each frame. After spatial calibration, using the camera's preset intrinsic and extrinsic parameter matrices, it converts the two-dimensional pixel coordinate domain into the actual three-dimensional physical dimension coordinate domain, generating curves showing the molten pool's geometric dimensions and area changing over time, serving as information on the molten pool's morphological evolution.

[0116] Finally, the imaging feedback subsystem transmits the reconstructed 3D dynamic feature data to the central processing subsystem. The calculation process for the deep fused hole heat conduction model and the specific algorithm code for image edge contour extraction are well-known technologies in this field and will not be elaborated upon here.

[0117] After acquiring and reconstructing the high-precision multi-dimensional physical state of the molten pool, the multi-phase laser welding control system enters the core data analysis and decision-making stage. The central processing subsystem quantifies the errors in the dynamic characteristic data and decomposes the complex correction requirements into physical action instructions suitable for execution by different underlying hardware. The specific implementation process of step S300 is divided into the following sub-steps:

[0118] S301, the central processing subsystem integrates the received surface tension fluctuation information, temperature field distribution information, and molten pool morphology evolution information into a real three-dimensional dynamic feature vector for the current moment. The central processing subsystem internally stores a preset welding process parameter model, which outputs an ideal process benchmark corresponding to the spatial location and time node, serving as the ideal state benchmark vector for the current moment. In practical applications, the preset welding process parameter model is constructed based on historical benchmark datasets recorded under standard defect-free welding experiments for different specific metal materials. Its output ideal process benchmark specifically includes the standard ideal molten pool width, ideal weld depth, and ideal surface temperature field scalar for that processing stage.

[0119] The central processing subsystem performs matrix difference operations to calculate the state error vector. The specific calculation equation is as follows:

[0120] ;

[0121] In the formula, This is the state error vector at the current moment; The ideal state reference vector at the current moment; This represents the actual three-dimensional dynamic feature vector at the current moment. This refers to the current moment.

[0122] S302. After obtaining the current state error vector, the central processing subsystem inputs it into the built-in adaptive multivariable control algorithm. The adaptive multivariable control algorithm, combining proportional-integral-derivative (PID) arithmetic, calculates the overall control input vector required to offset the deviation at the current moment based on the instantaneous value, historical cumulative amount, and trend of the current state error vector. In the physical principle of molten pool control, the proportional term directly dampens any instantaneous morphological distortion or temperature change; the integral term eliminates the slow drift of the background temperature caused by long-term heat accumulation; and the derivative term captures the high-frequency oscillation trend of the molten pool surface tension, injecting reverse suppression energy in advance to prevent liquid metal splashing.

[0123] Its specific physical equations are expressed as follows:

[0124] ;

[0125] In the formula, This is the overall control input vector at the current moment; This is the proportional gain matrix; This is the integral gain matrix; It is the differential gain matrix; Let the integral variable be at time; Let be the state error vector at time step of integration. The parameter values ​​of the proportional gain matrix, integral gain matrix, and differential gain matrix are empirically calibrated and fixed in the central processing subsystem by performing a step energy excitation test on a static molten pool and recording its thermodynamic response curve, using conventional empirical tuning rules.

[0126] Through these calculations, the central processing subsystem quantifies the total amount of external intervention energy and spatial distribution adjustment required to maintain the stability of the molten pool.

[0127] S303. The central processing subsystem executes a multi-hardware collaborative control instruction decoupling mechanism, decomposing and mapping the overall control input vector into specific instruction sequences recognizable by different execution subsystems. Based on the physical response bandwidth and operating mechanism of different hardware components, the central processing subsystem uses a hardware decoupling mapping matrix for linear transformation separation. The specific transformation equation is expressed as:

[0128] ;

[0129] In the formula, This is the phase control command vector at the current moment; This is the power adjustment command vector at the current moment; This is the beam deflection command vector at the current moment; For hardware decoupling mapping matrix; This is the matrix transpose notation. The hardware decoupling mapping matrix is ​​a pre-constructed frequency and spatial weighting matrix. The matrix elements are determined based on the physical limits of each hardware component: that is, high-frequency weights are assigned to high-frequency, inertial-free hardware; low-frequency spatial weights are assigned to spatial array hardware; and low-frequency and macroscopic spatial weights are assigned to mechanical motion hardware.

[0130] In terms of specific decoupling logic, the central processing subsystem allocates the low-frequency demands of minor local morphological mutations and spatial distribution reshaping to the phase modulation command vector at the current moment to drive the spatial light modulator; it allocates the high-frequency surface tension oscillations and the high-frequency energy transient demands of overall temperature field heat loss or overload to the power adjustment command vector at the current moment to drive the high-frequency inertia-free acousto-optic modulator to change the energy input; and it allocates the macroscopic positional offset of the molten pool and the asymmetric demands of the heat-affected zone to the beam deflection command vector at the current moment to drive the high-speed galvanometer assembly to adjust the beam landing point.

[0131] Through this instruction decoupling and distribution, the central processing subsystem synchronously sends joint and coordinated control instructions to the laser phase array subsystem and the composite control subsystem. The specific parameter tuning process of the gain matrix in the adaptive multivariable control algorithm and the compilation of the underlying code for matrix operations are well-known technologies in this field and will not be elaborated upon here.

[0132] Upon receiving the coordinated control command generated by the decoupling of the central processing subsystem, the laser phase array subsystem dynamically reshapes the static energy distribution of the processing plane based on the physical laws of optical phase modulation and multi-beam interference, thereby achieving high-frequency optimization of the spatial energy distribution pattern. The specific implementation process of step S400 is divided into the following sub-steps:

[0133] S401. The laser emitter outputs a raw single laser beam. After passing through the acousto-optic modulator, the raw single laser beam is incident on the microbeam splitter and is physically divided into multiple microbeams. In terms of specific hardware implementation, the microbeam splitter uses diffractive optical elements or microlens arrays. It utilizes a micron-scale periodic relief structure etched on the surface to spatially divide the original raw single laser beam, which originally had a single Gaussian distribution, into a two-dimensional array of microbeams with uniform energy.

[0134] In the spatial physical coordinate system, the laser phase array subsystem establishes the initial electric field complex amplitude distribution equations for each microbeam.

[0135] Assuming that the original single laser beam is physically divided into a specific number of microbeams, the specific number is preset to 16 beams (4x4 array) or 64 beams (8x8 array). The specific number is matched and calibrated based on the maximum output power of the laser emitter and the spot resolution required for processing.

[0136] No. The initial complex amplitude distribution equation of the electric field of a microbeam in the spatial physical coordinate system is specifically expressed as follows:

[0137] ;

[0138] In the formula, For the first The initial complex amplitude of the electric field of a microbeam; For the first The initial amplitude of the microbeam; The imaginary unit; For the first The initial phase of each microbeam; The horizontal spatial coordinates of the machining plane; The longitudinal spatial coordinates of the machining plane; This is the sequence number of the microbeam.

[0139] S402. The physically segmented microbeams are projected onto the working surface of the spatial light modulator. The spatial light modulator receives the phase modulation command vector for the current moment from the central processing subsystem. The built-in drive control circuit of the spatial light modulator analyzes the phase modulation command vector for the current moment and converts it into the corresponding pixel-level drive voltage matrix.

[0140] In a spatial light modulator, liquid crystal molecules on the working surface are deflected under the influence of a pixel-level driving voltage matrix, altering the local optical refractive index of the working surface and thus injecting an independent dynamic phase delay into each projected microbeam. In general physics, the long and short axes of nematic liquid crystal molecules have different refractive indices. When different gradient driving voltages are applied, the liquid crystal molecules tilt at different angles, causing light waves passing through the pixel region to experience different optical paths, thereby achieving pure phase modulation without changing the amplitude of the light wave.

[0141] After injecting dynamic phase delay, the first The equation for the complex amplitude distribution of the modulated microbeam is expressed as:

[0142] ;

[0143] In the formula, For the first The modulated complex amplitude of a microbeam at the current moment; For the spatial light modulator at the current moment, the first The dynamic phase delay of each microbeam injection is directly mapped from the phase modulation command vector at the current moment. This refers to the current moment.

[0144] During the phase injection process, the phase detection and adjustment unit continuously collects the real-time phase values ​​of each microbeam for closed-loop feedback to compensate for hardware response errors and ensure the accurate execution of dynamic phase delay.

[0145] S403. Each microbeam carrying a different dynamic phase delay is projected onto the working plane of the molten pool through the subsequent optical path, resulting in spatial interference and superposition. According to the principle of coherent superposition of optical fields, the total optical field intensity distribution on the working plane of the molten pool is determined by the superposition of the modulated complex amplitudes of each microbeam. The specific formula for the total optical field intensity distribution is as follows:

[0146] ;

[0147] In the formula, This represents the total light field intensity distribution on the working plane of the molten pool at the current moment. The total number of microbeams; For the first The modulated complex amplitude of a microbeam at the current moment; The horizontal spatial coordinates of the machining plane; The longitudinal spatial coordinates of the machining plane; The current moment; This is the sequence number of the microbeam.

[0148] By updating the dynamic phase delay in real time at high frequency through a spatial light modulator, the laser phase array subsystem dynamically reshapes the total optical field intensity distribution on the working plane of the molten pool. This total optical field intensity distribution forms a specific high- and low-energy alternating interference pattern in space. The local thermal expansion driving force generated in the high-energy region couples with the cooling contraction force in the low-energy region, generating a specific spatial surface tension gradient on the molten pool surface. This spatial surface tension gradient precisely cancels out the surface tension fluctuations extracted by the imaging feedback subsystem, thereby suppressing the convective oscillations of the liquid metal inside the molten pool and maintaining the stability of the molten pool morphology.

[0149] The specific structural design of the microbeam splitter and the microscopic physical process of liquid crystal molecule deflection drive inside the spatial light modulator are well-known technologies in this field and will not be elaborated here.

[0150] While the laser phase array subsystem completes the microscopic optimization of spatial energy distribution, the composite control subsystem simultaneously performs decoupling control of the total macroscopic energy input and the spatial position of the beam, as well as microsecond-level spatiotemporal precise matching. The specific implementation process of step S500 is divided into the following sub-steps:

[0151] In the S501 composite control subsystem, the acousto-optic modulator is located in the main optical path and receives the current power adjustment command vector issued by the central processing subsystem. The built-in radio frequency drive unit of the acousto-optic modulator parses the current power adjustment command vector and converts it into a radio frequency ultrasonic signal acting on the acousto-optic crystal.

[0152] Radio frequency (RF) ultrasonic signals generate a dynamic refractive index grating within an acousto-optic crystal, enabling high-frequency fine-tuning of the instantaneous input power of a laser beam passing through an acousto-optic modulator through acousto-optic diffraction. In general physics, when RF ultrasonic waves propagate within a crystal, they cause periodic elastic deformation and density variations in the crystal medium, equivalent to the formation of a moving phase grating. When an incident laser beam enters the crystal at a specific Bragg angle, most of the light energy is diffracted to the first-order direction. The higher the amplitude of the RF ultrasonic wave, the more drastic the refractive index change, and the greater the proportion of power diffracted to the first-order direction, thus achieving high-speed, inertia-free control of the output optical power.

[0153] The specific physical equation for the instantaneous output power of the acousto-optic modulator at the current moment is expressed as:

[0154] ;

[0155] In the formula, The instantaneous output power at the current moment; The constant input reference power before the laser beam enters the acousto-optic modulator; This represents the maximum physical diffraction efficiency of an acousto-optic crystal. The acousto-optic drive conversion constant; This is the power adjustment command vector at the current moment; The driving voltage amplitude calculated from the power regulation command vector at the current moment; This refers to the current moment. Through this mechanism, the acousto-optic modulator can dynamically change the beam modulation depth, achieving significant suppression of energy fluctuations.

[0156] S502, the high-speed galvanometer assembly in the composite control subsystem is located in the post-optical path of the acousto-optic modulator, synchronously receiving the beam deflection command vector at the current moment from the central processing subsystem. The dual-axis servo motor inside the high-speed galvanometer assembly performs spatial deflection actions according to the beam deflection command vector at the current moment.

[0157] After deflection, the light beam is projected onto the machining plane. The spatial geometric mapping equation between its dynamic deflection angle and the target position in the two-dimensional physical coordinate system of the machining plane is specifically expressed as follows:

[0158] ;

[0159] ;

[0160] In the formula, The coordinates of the horizontal target on the machining plane at the current moment; The coordinates of the longitudinal target on the machining plane at the current moment; The equivalent focal length of the optical focusing field lens configured for the high-speed galvanometer assembly; This refers to the lateral scanning deflection angle; This refers to the longitudinal scanning deflection angle; This refers to the current moment.

[0161] S503. During physical execution, the composite control subsystem implements dual decoupled composite closed-loop control of energy input values ​​and beam position. The power detection unit collects the instantaneous power output status parameters of the acousto-optic modulator in real time, while the deflection angle detection unit continuously collects the real-time deflection angle parameters of the high-speed galvanometer assembly using a high-resolution photoelectric encoder. The instantaneous power output status parameters and real-time deflection angle parameters are continuously fed back to the central processing subsystem for closed-loop timing verification.

[0162] The composite control subsystem utilizes microsecond-level timing coordination logic, specifically relying on hardware timers and high-speed field-programmable gate arrays within the central processing subsystem to achieve nanosecond-level clock synchronization. Simultaneously, while the high-speed galvanometer assembly guides the beam to the macroscopic position offset region or the asymmetric region of the heat-affected zone, the acousto-optic modulator synchronously performs precise energy compensation for heat loss or overload. This spatiotemporal matching mechanism effectively suppresses welding defects such as keyhole collapse, spatter, and uneven weld formation caused by local energy accumulation or sudden reduction. The hardware design of the acousto-optic modulator's RF drive circuit and the specific underlying circuit implementation for signal extraction from the photoelectric encoder of the high-speed galvanometer assembly are well-known technologies in this field and will not be elaborated upon here.

[0163] Specific application examples:

[0164] To understand the technical solution of this invention, the invention will be described in detail below with reference to specific application scenarios and accompanying drawings. This specific application embodiment is based on a butt welding scenario of aluminum alloy plates with a thickness of 5mm. The multi-phase laser welding control system is responsible for collecting the thermodynamic state and spatial energy distribution of the molten pool, and performing multi-phase joint control in conjunction with the optical physical model calculation module.

[0165] During the baseline temperature acquisition phase, the imaging feedback subsystem needs to perform a dual-band colorimetric thermometry calculation step to determine the absolute temperature of the molten pool at that moment. The parameters are set as follows: the rated output power of the laser emitter configured in the processing equipment is 3000W, and the first detection wavelength is extracted. The spectral radiative exitance at the first detection wavelength is 0.532 μm. The second detection wavelength was extracted at a value of 1500 W / m²·μm. The spectral radiative exitance at the second detection wavelength is 1.064 μm. The second radiation constant is extracted from 8000 W / m²·μm. The absolute temperature of the molten pool is 14388 μm·K. The imaging feedback subsystem is used to calculate this absolute temperature using the formula. :

[0166] ;

[0167] The numerical calculation process is as follows:

[0168] ;

[0169] The calculation results show the absolute temperature of the molten pool at that moment. The temperature was 2631 K. The calculation process eliminated the interference caused by the dynamic fluctuation of the emissivity of the aluminum alloy in the liquid state, and obtained the temperature value.

[0170] During the thermodynamic state assessment phase, the central processing subsystem needs to perform a surface tension gradient calculation step to determine the surface tension gradient value. The parameters are set as follows: A three-dimensional temperature gradient model is obtained, and the absolute temperature gradient values ​​in a certain measurement area at the edge of the molten pool are extracted. The surface tension temperature coefficient is 50 K / mm, and is extracted based on the fixed selection of aluminum alloy physical properties. The value is -0.43 mN / m·K. The central processing subsystem substitutes the values ​​into the formula to calculate the surface tension gradient. :

[0171] ;

[0172] The numerical calculation process is as follows:

[0173] ;

[0174] The calculation results show the surface tension gradient value of the region. The value is -21.5 mN / m·mm, which represents the convection tendency that the system determines to spread outward from the inside of the molten pool based on the surface tension gradient distribution.

[0175] During the power dynamic compensation phase, the composite control subsystem needs to execute the calculation steps of the specific physical equations for the instantaneous output power at the current moment in order to determine the instantaneous output power at the current moment. The parameters are set as follows: Due to the detection of localized thermal overload, the central processing subsystem issues a power adjustment command. The constant input reference power before the laser beam enters the acousto-optic modulator is obtained. At 3000W, the maximum physical diffraction efficiency of the acousto-optic crystal was extracted. The conversion constant of the acousto-optic drive is 85%. The value is 0.5 rad / V, and the driving voltage amplitude is calculated from the power regulation command vector at the current moment. The voltage is 1.5V. The composite control subsystem is substituted into the formula to calculate the instantaneous output power at the current moment. :

[0176] ;

[0177] The numerical calculation process is as follows:

[0178] ;

[0179] The calculation results show the instantaneous output power at the current moment under the fine-tuning action. The value is 1184.7W, which represents the result of the composite control subsystem using fine-tuning to reduce high-frequency energy injection and suppress convective oscillations caused by heat accumulation.

[0180] To evaluate the effectiveness of the multi-phase laser welding control system, experimental verification and performance comparison are provided. The experiment compares the fixed-power single-beam laser welding method with the multi-phase joint control method of this invention. Both sets of experiments used the same substrate size and reference moving speed.

[0181] See attached document Figure 3 , Figure 3 The horizontal axis represents time, with the direction of the horizontal axis indicating the increase of time, and the vertical axis represents temperature. The solid line in the figure represents the single-beam control method, which, due to the lack of closed-loop power feedback, experiences surface temperature fluctuations with heat accumulation, dropping to a minimum of 2220K and rising to a maximum of 2930K. The dotted line in the figure represents the control method of this invention, where the temperature remains within a range fluctuating around the set baseline of 2600K. The results verify that this invention can maintain the stability of the molten pool's thermodynamic state and avoid porosity and spatter defects caused by fluctuations.

[0182] See attached document Figure 4 , Figure 4 The horizontal axis represents time, with the direction of the horizontal axis indicating time increment, and the vertical axis represents output power. The dotted line in the figure represents the target reference trajectory, which exhibits a step decrease when heat exceeding the limit is detected at 10 μs. The solid line with square markings represents the actual output power, which coincides with the target reference trajectory and adjusts from 2800 W to converge around 2200 W within the following 10 μs. The results verify that the composite execution hardware of this invention completed the energy compensation action without overshoot caused by mechanical inertia.

[0183] See attached document Figure 5 , Figure 5The horizontal axis represents the experimental group, with the direction of the horizontal axis indicating an increase in the number of experimental groups. The vertical axis represents the flatness error. The dark-colored bars in the figure represent the single-beam control method, which cannot dynamically reconstruct the spatial energy distribution. The flatness error of each group ranges from 0.31 mm to 0.48 mm, with the corresponding dashed mean line at the 0.4 mm level. The light-colored bars in the figure represent the spatial interference reshaping introduced by the method of this invention using multiple micro-beams. The flatness error is controlled within the range of 0.06 mm to 0.12 mm, with the corresponding dashed mean line at the level below 0.1 mm. The numerical markings above the bars in the figure represent specific flatness error values ​​and are located above the bars. The results verify that the spatiotemporal decoupling control of this invention improves the macroscopic forming quality of the weld, balances the spatial stress distribution, and avoids undercut and height fluctuations.

Claims

1. A multi-phase laser welding control method, comprising the following steps: The imaging feedback subsystem acquires molten pool radiation image data in multiple bands in the molten pool processing area, performs multi-band data fusion processing on the molten pool radiation image data to construct a three-dimensional temperature gradient model of the molten pool, and extracts three-dimensional dynamic feature data based on the three-dimensional temperature gradient model. The three-dimensional dynamic feature data is received by the central processing subsystem. The three-dimensional dynamic feature data is compared and analyzed with the preset welding process parameter model and the current physical state error of the molten pool is quantified. Based on the physical state error of the molten pool, a collaborative control command is generated. The collaborative control command includes a phase adjustment command, a power adjustment command, and a beam deflection command. The original single laser beam is decomposed into multiple microbeams by a laser phase array subsystem. The phase delay between each microbeam is adjusted according to the phase control command. The spatial energy distribution pattern of the processing plane is dynamically reshaped by the spatial interference superposition effect of each microbeam on the surface of the molten pool. The composite control subsystem fine-tunes the instantaneous input power of the laser beam passing through it according to the power adjustment command, and performs beam spatial deflection according to the beam deflection command, thereby completing the closed-loop composite control of the energy input value and the beam's position.

2. The multi-phase laser welding control method according to claim 1, characterized in that, The steps of acquiring molten pool radiation image data in multiple bands within the molten pool processing area and performing multi-band data fusion processing on the molten pool radiation image data to construct a three-dimensional temperature gradient model of the molten pool specifically include: A global hardware timer is used to send a global trigger pulse to synchronously collect molten pool radiation image data in the visible light band, near-infrared band, and ultraviolet band in the molten pool processing area; The base temperature is calculated using the spectral radiant output at the first and second detection wavelengths extracted from the visible light and near-infrared bands, respectively, through a dual-band colorimetric thermometry algorithm. The high-energy radiative exitance extracted from the ultraviolet band is used to perform synergistic correction and fusion of the base temperature to generate a two-dimensional surface temperature field distribution.

3. The multi-phase laser welding control method according to claim 2, characterized in that, The step of constructing a three-dimensional temperature gradient model of the molten pool by performing multi-band data fusion processing on the molten pool radiation image data further includes: By combining the deep-hole heat conduction model, the rate of heat attenuation and transfer from the surface to the bottom of the molten pool is calculated using the thermal diffusivity of the target metal material; The three-dimensional temperature gradient model is constructed by mapping the two-dimensional surface temperature field distribution according to the attenuation transfer rate.

4. The multi-phase laser welding control method according to claim 3, characterized in that, The steps for extracting three-dimensional dynamic feature data based on the three-dimensional temperature gradient model specifically include: Based on the absolute temperature gradient values ​​extracted from the three-dimensional temperature gradient model and the preset surface tension temperature coefficient, the surface tension gradient value distribution of the processing plane is calculated as surface tension fluctuation information. Extract the grayscale abrupt edges from the molten pool radiation image data, obtain the pixel coordinates of the highlight contour of the molten pool, convert the pixel coordinate domain into the actual three-dimensional physical size coordinate domain through spatial calibration, and calculate and generate the curve of the geometric size and area of ​​the molten pool changing over time as the molten pool morphology evolution information. The surface tension fluctuation information, the temperature field distribution information corresponding to the two-dimensional surface temperature field distribution, and the molten pool morphology evolution information are packaged and integrated into the three-dimensional dynamic feature data.

5. The multi-phase laser welding control method according to claim 1, characterized in that, The step of comparing and analyzing the three-dimensional dynamic feature data with the preset welding process parameter model and quantifying the current physical state error of the molten pool, and generating collaborative control commands based on the physical state error of the molten pool specifically includes: The three-dimensional dynamic feature data is converted into the actual three-dimensional dynamic feature vector at the current moment. The actual three-dimensional dynamic feature vector at the current moment is then subjected to matrix difference operation with the ideal state reference vector at the current moment output by the preset welding process parameter model to calculate the state error vector. Combining the proportional-integral-differential (PI-DI) operation mechanism, based on the instantaneous value, historical cumulative amount, and trend of the state error vector, the overall control input vector at the current moment required to offset the deviation is calculated and output. The overall control input vector is separated by a linear transformation using a hardware decoupling mapping matrix, and then decomposed and mapped into the phase adjustment command, the power adjustment command, and the beam deflection command.

6. The multi-phase laser welding control method according to claim 5, characterized in that, The step of using a hardware decoupling mapping matrix to perform a linear transformation and separation of the overall control input vector, and decomposing and mapping it into the phase adjustment command, the power adjustment command, and the beam deflection command specifically includes: The low-frequency demands for minor local morphological mutations and spatial reshaping are allocated to the phase modulation commands; The high-frequency surface tension oscillations and the high-frequency energy transient demand for heat loss or overload in the overall temperature field are allocated to the power regulation command. The macroscopic positional offset of the molten pool and the asymmetric requirements of the heat-affected zone are assigned to the beam deflection command.

7. The multi-phase laser welding control method according to claim 1, characterized in that, The steps of decomposing the original single laser beam into multiple microbeams through a laser phase array subsystem, adjusting the phase delay between each microbeam according to the phase modulation command, and dynamically reshaping the spatial energy distribution pattern of the processing plane by utilizing the spatial interference superposition effect of each microbeam on the molten pool surface specifically include: The original single-beam laser is physically divided into a two-dimensional microbeam array with uniform energy to form each microbeam, and the initial electric field complex amplitude distribution equation of each microbeam in the spatial physical coordinate system is established. The phase modulation command is parsed using the drive control circuit inside the spatial light modulator and converted into a corresponding pixel-level drive voltage matrix. The pixel-level driving voltage matrix is ​​used to control the deflection of liquid crystal molecules, thereby changing the optical refractive index of the region where the liquid crystal molecules are located, and injecting an independent dynamic phase delay into each of the micro-beams. Each of the microbeams carrying different dynamic phase delay values ​​is projected onto the processing plane through the subsequent optical path, and forms an alternating high and low energy interference pattern in space based on the interference superposition effect of the modulated complex amplitude.

8. The multi-phase laser welding control method according to claim 1, characterized in that, The step of fine-tuning the instantaneous input power of the laser beam passing through the composite control subsystem according to the power adjustment command specifically includes: The acousto-optic modulator in the composite control subsystem parses the power adjustment command and converts the power adjustment command into a radio frequency ultrasonic signal acting on the acousto-optic crystal; The radio frequency ultrasonic signal is used to generate a dynamic refractive index grating inside the acousto-optic crystal. Based on the physical diffraction efficiency of the acousto-optic crystal, the instantaneous output power of the laser beam passing through the composite control subsystem and outputting to the first-order light direction is finely adjusted at high frequency through the acousto-optic diffraction effect.

9. The multi-phase laser welding control method according to claim 8, characterized in that, The step of executing the spatial deflection of the beam according to the beam deflection command to complete the closed-loop composite control of the energy input value and the beam's position specifically includes: The high-speed galvanometer assembly in the composite control subsystem drives the dual-axis servo motor to perform the spatial deflection of the beam according to the beam deflection command. The instantaneous power output status parameters of the acousto-optic modulator are collected in real time by the power detection unit, and the real-time deflection angle parameters of the high-speed galvanometer assembly are continuously collected by the deflection angle detection unit and fed back to the central processing subsystem. By relying on a hardware timer and a high-speed field-programmable gate array for clock synchronization, while the high-speed galvanometer assembly guides the laser beam to the target offset region, the acousto-optic modulator synchronously performs energy compensation for heat loss or overload, thus completing the closed-loop composite control of the energy input value and the beam's position.

10. A multi-phase laser welding control system, characterized in that, For performing the multi-phase laser welding control method according to any one of claims 1-9, comprising: The basic electromechanical auxiliary subsystem is used to provide physical support for the multi-phase laser welding process and monitor the operating environment status. An imaging feedback subsystem is used to acquire molten pool radiation image data of multiple bands in the molten pool processing area, perform multi-band data fusion processing on the molten pool radiation image data to construct a three-dimensional temperature gradient model of the molten pool, and extract three-dimensional dynamic feature data based on the three-dimensional temperature gradient model. The central processing subsystem is used to receive the three-dimensional dynamic feature data, compare and analyze the three-dimensional dynamic feature data with the preset welding process parameter model and quantify the current physical state error of the molten pool, and generate collaborative control instructions based on the physical state error of the molten pool. The collaborative control instructions include phase adjustment instructions, power adjustment instructions and beam deflection instructions. A laser phase array subsystem is used to decompose the original single laser beam into multiple microbeams, adjust the phase delay between each microbeam according to the phase modulation command, and dynamically reshape the spatial energy distribution pattern of the processing plane by utilizing the spatial interference superposition effect of each microbeam on the surface of the molten pool. The composite control subsystem is used to fine-tune the instantaneous input power of the laser beam passing through the composite control subsystem according to the power adjustment command, and to perform beam spatial deflection action according to the beam deflection command, thereby completing the closed-loop composite control of energy input value and beam position.

Images