A laser cladding parameter optimization system and method based on multi-dimensional perception data

Abstract

This invention belongs to the field of laser cladding technology, specifically a laser cladding parameter optimization system and method based on multi-dimensional sensing data. It comprises five modules: multi-dimensional sensing feature extraction, quality inversion, parameter adaptive decision-making, execution feedback, and closed-loop verification. The system simultaneously acquires multi-source molten pool data (spectral, infrared, visual, and acoustic emission) and completes spatiotemporal registration, extracting four types of features to construct a cladding feature vector. It then uses a coupled analytical model of thermal, mechanical, and microstructure parameters to invert three quality indicators: grain size, residual stress, and deformation. This invention relies on multi-dimensional dynamic sensing of the molten pool, multi-physics field coupled quality inversion, and a self-correcting gradient optimization closed-loop system to dynamically optimize laser power and scanning speed. This overcomes the problems of traditional solutions, such as single sensing dimension, inaccurate quality prediction, and parameter optimization easily getting stuck in local extrema, thus stably improving the grain uniformity of the cladding layer and reducing residual stress and processing deformation.

Description

Technical Field

[0001] This invention relates to the field of laser cladding technology, specifically to a laser cladding parameter optimization system and method based on multi-dimensional sensing data. Background Technology

[0002] Laser cladding is a high-precision surface modification and component remanufacturing technology. With its small heat-affected zone, high coating bonding strength, and excellent forming accuracy, it is widely used for surface strengthening and defect repair of key components in aerospace, energy and power, precision equipment, and metallurgical machinery. The forming quality of laser cladding is highly dependent on the matching of core process parameters such as laser power and scanning speed. The dynamic adaptability of these process parameters directly determines the uniformity of the cladding layer's microstructure, the level of residual stress, and the component forming accuracy.

[0003] Currently, industrial laser cladding processes generally adopt a solidification processing mode of "preliminary parameter testing and full-process parameter fixing," relying on manual experience and a small number of orthogonal experiments to calibrate and fix process parameters, which has obvious technical shortcomings.

[0004] Firstly, the temperature, shape, and solidification state of the cladding pool fluctuate dynamically in real time during the processing. Fixed process parameters cannot adapt to the dynamic changes in the cladding pool, which easily leads to quality defects such as uneven dendrite spacing in the cladding layer, residual stress concentration, and excessive warping deformation of components. Secondly, traditional process monitoring methods are limited, relying mainly on monitoring equipment operating parameters or post-process sampling of finished products. They lack synchronous perception of multi-dimensional process information such as molten pool spectrum, temperature field, morphology, and acoustic emission, and cannot achieve full-domain, real-time, and accurate characterization of the cladding state. Third, existing process optimization algorithms mostly use fixed initial iteration values ​​to find optimal parameters. Under complex operating conditions, they are prone to getting trapped in local optima, have poor adaptive adjustment capabilities, and cannot achieve closed-loop iterative optimization of process parameters.

[0005] In summary, existing laser cladding technologies suffer from problems such as limited process perception dimensions, delayed quality prediction, fixed process parameters, easy local convergence of optimization algorithms, and lack of closed-loop adaptive correction mechanisms, making it difficult to meet the industrial-scale batch cladding processing requirements of high precision, high consistency, and high stability.

[0006] It should be noted that the above content falls within the inventor's technical knowledge and does not necessarily constitute prior art. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a laser cladding parameter optimization system and method based on multi-dimensional sensing data, which solves the current problems.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a laser cladding parameter optimization system and method based on multi-dimensional sensing data, comprising: The multidimensional perception and feature extraction module deploys a spectral sensor, an infrared thermal imager, a high-speed visual sensor, and an acoustic emission sensor to simultaneously collect molten pool spectrum, temperature field, morphology, and acoustic emission data, and completes timestamp alignment and spatial coordinate unification to obtain a spatiotemporally aligned multidimensional perception dataset. The multidimensional perception and feature extraction module has multiple feature extraction sub-modules built in, which extract the molten pool oscillation frequency, molten pool geometric parameters, temperature gradient distribution and splash features from the dataset to construct a feature vector of the cladding process. The cladding layer quality inversion module receives the feature vector of the cladding process and has a built-in thermal, mechanical, and microstructure coupled analytical model. Based on the feature vector of the cladding process, it inverts the primary dendrite spacing, residual stress distribution, and warpage deformation, and outputs the quality inversion results. The process parameter adaptive decision-making module receives the quality inversion results, internally stores a quality constraint set containing size, stress, and deformation threshold, collects laser power and scanning speed in real time, and uses the gradient descent method to combine the quality inversion results and the constraint set to complete the linkage optimization and output the optimized process parameters. The parameter execution and feedback module is connected to the process parameter adaptive decision module. It receives and sends process parameters to the corresponding controller for execution, and at the same time collects actual process parameters and new cycle sensing data to generate feedback data packets. The closed-loop adaptive verification module receives feedback data packets, calculates the quality deviation index based on the feedback data packets, and sends a reset signal to the process parameter adaptive decision module when the deviation index continues to exceed the preset tolerance, forcibly resetting the initial point of optimization.

[0009] In some embodiments, the process of acquiring the multidimensional sensing dataset is as follows: The dynamic process data of laser cladding is collected in parallel and synchronously by spectral sensors, infrared thermal imagers, high-speed visual sensors and acoustic emission sensors, and the raw multi-source sensing data including molten pool spectral data, molten pool temperature field data, molten pool morphology data and acoustic emission signal data are summarized. Using the high-frequency acquisition timing of the high-speed vision sensor as a unified time reference, timestamp correction and gap filling are performed on the asynchronous sampling data of the spectral sensor, infrared thermal imager and acoustic emission sensor; Using the laser cladding equipment's processing reference coordinate system as a spatial reference, the field-of-view coordinates and molten pool spatial coordinates of all sensors are uniformly mapped and transformed to complete the spatial location matching of multi-source data, ultimately generating a spatiotemporally aligned multidimensional sensing dataset.

[0010] In some embodiments, the multi-feature extraction submodule includes a molten pool oscillation frequency extraction submodule, a molten pool geometric parameter calculation submodule, a temperature gradient analysis submodule, and a splash feature statistics submodule; The molten pool oscillation frequency extraction submodule extracts the molten pool oscillation frequency from the multidimensional sensing dataset, the molten pool geometric parameter calculation submodule extracts the molten pool width and height from the multidimensional sensing dataset, the temperature gradient analysis submodule extracts the temperature gradient distribution value from the multidimensional sensing dataset, and the spatter feature statistics submodule extracts the number and velocity distribution of spatter particles from the multidimensional sensing dataset, thus forming a feature vector of the cladding process.

[0011] In some embodiments, the cladding process feature vector construction process is as follows: The molten pool oscillation frequency extraction submodule performs time-domain and frequency-domain analysis on the acoustic emission signal data and molten pool morphology time-series data in the multidimensional sensing dataset, and extracts the inherent oscillation frequency characteristics of the molten pool by removing environmental noise interference through fast Fourier transform. The melt pool geometry parameter calculation submodule calculates the real-time melt pool width and height parameters based on the melt pool topography data, and performs image threshold segmentation and edge contour extraction on the melt pool topography data collected by the high-speed vision sensor to obtain the melt pool geometry parameter features. The temperature gradient analysis submodule decomposes the molten pool temperature field data acquired by the infrared thermal imager into a matrix grid, calculates the temperature gradient distribution values ​​from the center to the edge of the molten pool and in the front and back directions of the cladding area, and obtains the temperature gradient distribution characteristics of the molten pool. The splash feature statistics submodule performs target recognition and trajectory tracking on high-speed visual continuous frame images, counts the number of splash particles per unit time, fits the motion trajectory of splash particles to obtain the velocity distribution range, and obtains splash motion features. The extracted features of the molten pool's inherent oscillation frequency, geometric parameters, temperature gradient distribution, and spatter motion are normalized and then fused to construct a multidimensional feature vector for the cladding process. ; in, The inherent oscillation frequency characteristics of the molten pool For the geometric parameters of the molten pool, For the characteristics of the temperature gradient distribution of the molten pool, This is a characteristic of splash motion.

[0012] In some embodiments, the process of constructing the thermal, mechanical, and tissue-coupled analytical model is as follows: To match the actual working conditions of high-temperature forming and rapid solidification of laser cladding, three independent models were built: a heat conduction model of the molten pool, a stress-strain mechanics model of the cladding layer, and a grain growth kinetics model. The basic boundary parameters and material property parameters of the three types of models were calibrated, resulting in three independent single-physics field basic models. Based on the three independent single-physics field basic models obtained, a multi-physics field coupling and correlation mechanism is established. The global temperature field time series data output by the molten pool heat conduction model is used as the driving basis. The stress-strain mechanics model is associated to realize the field-coupled update of temperature gradient thermal stress and solidification shrinkage stress. At the same time, relying on the temperature field change rate and molten pool flow state, the grain structure growth kinetics model is linked to complete the dynamic evolution matching of the solidification interface. Thus, a multi-physics field coupling and correlation mechanism in which the temperature field drives the stress field and the temperature change flow field jointly regulate the grain structure evolution is constructed. Based on the established multi-physics coupling and correlation mechanism, targeted variable correlation and adaptation are performed on the feature vector of the cladding process: the molten pool temperature gradient distribution feature is used as the real-time boundary correction variable of the heat conduction model, the molten pool geometric parameter feature is used as the molten pool sintering region size constraint variable, and the molten pool inherent oscillation frequency feature is used as the molten pool flow disturbance and grain growth disturbance correction variable. The three types of standardized features are used as the real-time joint input of the model to establish the hierarchical transmission relationship of sensing feature parameters, multi-physics state and cladding quality index, and construct a nonlinear mapping relationship of multi-feature input, multi-physics linkage and multi-quality index output. By using multiple sets of process simulation calibration data and actual on-site process experimental data, the internal coupling coefficient and hysteresis compensation coefficient of the nonlinear mapping relationship were iteratively corrected, and the overall parameters of the model were solidified to obtain a thermal, mechanical, and microstructure coupled analytical model. The calculation formula for the molten pool heat conduction model is as follows: ; in, Density of the cladding material For the specific heat capacity of the material, This refers to the real-time temperature of the molten pool. For cladding time, For the thermal conductivity of the material, To provide heat flux to the laser, The heat flux lost is due to convection and radiation in the molten pool.

[0013] In some embodiments, the process of solving the mechanical and tissue indices using the coupled thermal, mechanical, and tissue analytical model includes: The real-time global temperature field data output by the coupled thermal, mechanical, and structural analytical model is called, and thermal stress and solidification shrinkage stress are superimposed to obtain the global basic stress data of the cladding layer. Combined with the actual tooling boundary constraints of the component, the global basic stress data of the cladding layer is solved by global integration to obtain the global residual stress distribution data of the cladding layer. Matching the overall stiffness parameters of the component, the warping deformation of the component is calculated synchronously based on the obtained global residual stress distribution data of the cladding layer. By correlating the temperature gradient distribution characteristics of the molten pool, the inherent oscillation frequency characteristics of the molten pool, and the real-time solidification rate of the molten pool, the coupling influence of the three factors is quantified, and the multi-parameter coupling correlation coefficient of the molten pool is obtained. Based on the obtained multi-parameter coupling correlation coefficient of the molten pool, the disturbance amplitude of the molten pool flow and temperature fluctuation on the grain growth interface is quantified, and the quantitative relationship of grain growth disturbance is obtained. Combining the inherent laws of metal solidification kinetics, a quantitative calculation model of parameters and primary dendrite spacing is constructed based on the obtained quantitative relationship of grain growth disturbance, and the primary dendrite spacing of the cladding layer is obtained by solving.

[0014] In some embodiments, the process parameter adaptive decision module completes the process parameter optimization process by including: Combining the process specifications and finished product quality access standards of the laser cladding industry, the upper limit thresholds for primary dendrite spacing, residual stress, and warpage deformation are preset, and all hard control indicators are summarized to obtain the cladding process quality constraint set. With the minimum primary dendrite spacing, minimum residual stress, and minimum warpage deformation as multiple optimization objectives, a multi-objective loss function is constructed based on the obtained cladding process quality constraint set, resulting in the multi-objective loss function for the cladding process. Real-time acquisition of two core operating parameters of the equipment: current laser power and scanning speed, to obtain the equipment's real-time raw process parameters; Substitute the quality inversion results and the obtained real-time original process parameters of the equipment into the multi-objective loss function of the cladding process, solve for the gradient change value of the current process parameters, iterate and update the process parameters along the optimal direction based on the gradient descent method until the loss function converges and all indicators meet the requirements of the cladding process quality constraint set, and output the optimized process parameters. The expression for the multi-objective loss function is: ; in, The total loss value, To invert the primary dendrite spacing in real time, This represents the upper limit of the primary dendrite spacing. For real-time residual stress, The residual stress threshold, This represents the real-time warping deformation. The threshold for warping deformation. and These are the weighting coefficients corresponding to each quality indicator.

[0015] In some embodiments, the process by which the parameter execution and feedback module generates a feedback data packet includes: The system receives optimized process parameters from the adaptive decision-making module and sends them to the laser controller and motion controller to execute the cladding operation, generating a process parameter execution record. The generated process parameter execution records are collected, and the actual laser power and scanning speed of the equipment during operation are collected in real time as actual process parameters. The generated process parameter execution record and the actual process parameters are retrieved, and the complete set of multi-dimensional sensing data corresponding to the next complete cladding processing cycle is collected simultaneously. The three types of data are bound and encapsulated to generate a feedback data packet.

[0016] In some embodiments, the deviation calculation and adaptive reset process of the closed-loop adaptive verification module includes: The feedback data packet is analyzed to extract key data such as the actual geometric dimensions of the cladding product and the actual cooling rate of the molten pool. The system's preset standard process quality parameters are retrieved, and the difference between the two types of parameters is normalized to calculate the real-time quality deviation index. A preset quality deviation tolerance threshold is set, and the real-time quality deviation index changes over multiple cladding cycles are continuously monitored. If the real-time quality deviation index continues to exceed the preset tolerance, it is determined that the gradient descent optimization has fallen into a local optimum. For the locally optimal solution obtained by the determination, an optimization reset signal is generated and pushed to the process parameter adaptive decision module to forcibly reset the gradient descent optimization initial point, restart the parameter iteration optimization, and realize the closed-loop adaptive optimization of the cladding process.

[0017] Another technical problem to be solved by this invention is to propose a laser cladding parameter optimization method based on multi-dimensional sensing data, comprising the following steps: Step 1: Simultaneously collect four types of sensor data—spectral, infrared, visual, and acoustic emission—through the multi-dimensional perception and feature extraction module, complete the spatiotemporal registration process of timestamp alignment and spatial coordinate unification, extract the molten pool oscillation frequency, geometric parameters, temperature gradient, and spatter characteristics, and fuse them to construct a standardized cladding process feature vector. Step 2: Using the built-in thermal, mechanical, and microstructure coupled analytical model of the cladding layer quality inversion module, input the cladding process feature vector to perform multi-physics field coupling deduction, and invert to obtain the cladding layer quality indicators such as primary dendrite spacing, residual stress distribution, and warping deformation. Step 3: The preset quality constraint set is retrieved through the process parameter adaptive decision module, and the gradient descent method is used to perform multi-objective linkage optimization of laser power and scanning speed, and output the optimal process parameters that are adapted to the real-time cladding conditions. Step 4: The optimized process parameters are issued through the parameter execution and feedback module to complete the cladding operation, and the actual process parameters of the equipment and the multi-dimensional sensing data of the new cycle are collected simultaneously and packaged to generate a feedback data package; Step 5: The feedback data packet is parsed and the quality deviation index is calculated through the closed-loop adaptive verification module. When the deviation index continues to exceed the preset tolerance, the initial point of optimization is reset and the process is repeated in a loop to achieve closed-loop adaptive optimization of laser cladding process parameters.

[0018] Compared with existing technologies, this invention provides a laser cladding parameter optimization system and method based on multi-dimensional sensing data, which has the following beneficial effects: 1. This laser cladding parameter optimization system and method based on multi-dimensional sensing data synchronously collects multi-source process data of molten pool spectrum, temperature field, morphology and acoustic emission through multiple sensors, and combines spatiotemporal registration and normalized feature fusion technology to construct a standardized feature vector that comprehensively characterizes the dynamic evolution state of the molten pool. This overcomes the shortcomings of traditional laser cladding monitoring, such as single dimension, one-sided representation of working conditions and misalignment of data time sequence and space. Through refined, multi-dimensional, and highly synchronous process perception capabilities, it provides an accurate, reliable, and comprehensive data foundation for subsequent quality inversion and process parameter optimization, effectively improving the completeness and real-time performance of cladding condition identification; 2. This laser cladding parameter optimization system and method based on multi-dimensional sensing data replaces the traditional single-physics empirical estimation and post-inspection mode by building a coupled analytical model of thermal, mechanical, and microstructure multi-physics fields. It relies on the feature vector of the cladding process to realize the synchronous real-time inversion of multiple core quality indicators such as dendrite spacing, residual stress distribution, and component warping deformation. It effectively solves the problems of lagging quality feedback and inability to intervene in defects in advance in traditional processes, and greatly improves the controllability and stability of laser cladding forming quality. 3. This laser cladding parameter optimization system and method based on multi-dimensional sensing data constructs a multi-objective loss function by combining a cladding quality constraint set. It employs a gradient descent algorithm to adaptively optimize laser power and scanning speed in conjunction with the cladding quality status, dynamically correcting process parameters based on real-time cladding quality conditions. This allows it to adapt to dynamic fluctuations in the molten pool and changes in operating conditions, overcoming the problems of poor adaptability, uneven microstructure, stress concentration, and out-of-tolerance forming in traditional fixed-parameter processes. Under the premise of meeting rigid process constraints, it can achieve multi-objective collaborative optimization of microstructure, residual stress, and forming accuracy, significantly improving the consistency of finished products. 4. This laser cladding parameter optimization system and method based on multi-dimensional sensing data adds a multi-cycle quality deviation monitoring and adaptive reset mechanism for the initial optimization point, effectively solving the technical drawbacks of traditional gradient optimization algorithms that are prone to getting trapped in local optima, drifting in long-term processing conditions, and decay of optimization accuracy, and improving the global search capability and iterative stability of process optimization. This invention forms a closed-loop adaptive intelligent optimization system for the entire process, realizing unmanned, dynamic, and intelligent control of laser cladding process, significantly reducing the defect rate of mass production, and adapting to complex and variable working condition precision cladding processing scenarios. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments and 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.

[0021] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0022] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0024] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0025] Please see Figure 1 In this implementation scheme: a laser cladding parameter optimization system based on multidimensional sensing data includes a multidimensional sensing and feature extraction module, a cladding layer quality inversion module, a process parameter adaptive decision-making module, a parameter execution and feedback module, and a closed-loop adaptive verification module.

[0026] The multi-dimensional sensing and feature extraction module is responsible for the synchronous acquisition, spatiotemporal registration, and feature extraction of multi-source sensor data. The specific execution process is as follows: In the laser cladding processing area, spectral sensors, infrared thermal imagers, high-speed vision sensors, and acoustic emission sensors are deployed simultaneously. These four types of sensors work in parallel to collect real-time spectral data, temperature field data, morphology data, and acoustic emission signal data of the molten pool during the dynamic laser cladding process. All the original collected data are then summarized to obtain the original multi-source sensing data.

[0027] To address the data timing misalignment issue caused by asynchronous acquisition from multiple sensors, this embodiment uses the high-speed vision sensor with the highest sampling frequency as a unified time reference. Timestamp correction is applied to the asynchronous sampling data from the other three types of sensors. Linear interpolation is used to fill in data gaps caused by sampling intervals, ensuring complete timing alignment of all data. Simultaneously, the processing reference coordinate system calibrated at the factory of the laser cladding equipment is used as a spatial reference. A unified mapping and transformation is performed on the field-of-view coordinates of each sensor and the real-time spatial coordinates of the molten pool, eliminating field-of-view biases between multiple devices and achieving precise spatial matching of multi-source data. This ultimately generates a multi-dimensional sensing dataset with temporal and spatial alignment.

[0028] The module incorporates four feature extraction sub-modules, each performing targeted feature extraction: the molten pool oscillation frequency extraction sub-module performs time-domain and frequency-domain analysis on acoustic emission signals and molten pool morphology time-series data, filters out environmental noise and equipment vibration interference noise through fast Fourier transform, and accurately extracts the inherent oscillation frequency of the molten pool; the molten pool geometric parameter calculation sub-module performs threshold segmentation and edge contour extraction on molten pool morphology images acquired by high-speed vision, and accurately calculates the real-time molten pool width and height geometric parameters; the temperature gradient analysis sub-module decomposes the infrared temperature field data into a grid, calculates the temperature gradient values ​​from the center of the molten pool to the edge and before and after the cladding advance direction grid by grid, and obtains the global temperature gradient distribution characteristics; the spatter feature statistics sub-module performs target recognition and trajectory tracking on continuous frame visual images, counts the number of spatter particles per unit time, fits the particle motion trajectory to obtain the velocity distribution range, and obtains the spatter motion characteristics.

[0029] The extracted features of the molten pool's inherent oscillation frequency, geometric parameters, temperature gradient distribution, and spatter motion are normalized and then fused to construct a multidimensional feature vector for the cladding process. ; in, The inherent oscillation frequency characteristics of the molten pool For the geometric parameters of the molten pool, For the characteristics of the temperature gradient distribution of the molten pool, This is a characteristic of splash motion.

[0030] The cladding layer quality inversion module completes the quality index inversion based on a coupled thermal, mechanical, and microstructure analytical model. The specific process is as follows: Three types of single-physics field basic models were pre-built, and parameter calibration was completed by combining the physical properties of the cladding material and the boundary conditions of equipment processing. These models include the molten pool heat conduction model, the stress-strain mechanics model, and the grain growth kinetics model. Driven by the molten pool heat conduction model, the boundary conditions of the heat conduction model were corrected in real time using the aforementioned molten pool temperature gradient characteristics. Based on the global dynamic temperature field output by the heat conduction model, the stress-strain mechanics model was linked to update the temperature gradient thermal stress and solidification shrinkage stress in real time. Combined with the component tooling constraints, the global residual stress distribution and component warping deformation were solved. Simultaneously, based on the temperature field change rate and molten pool oscillation disturbance characteristics, the disturbance effect of molten pool flow on the grain solidification interface was quantified. Combined with the laws of metal solidification kinetics, a primary dendrite spacing quantification calculation model was constructed to accurately solve the grain structure indicators of the cladding layer.

[0031] The heat conduction model of the molten pool uses the following formula: ; in, Density of the cladding material For the specific heat capacity of the material, This refers to the real-time temperature of the molten pool. For cladding time, For the thermal conductivity of the material, To provide heat flux to the laser, The heat flux lost is due to convection and radiation in the molten pool.

[0032] It accurately characterizes the dynamic balance between heat input, conduction, and heat dissipation loss in the molten pool. By iteratively correcting the model coupling coefficient through multiple sets of simulation data and field experimental data, it ensures that the residual stress, warpage deformation, and primary dendrite spacing output by the model are highly matched with the actual processing quality.

[0033] The adaptive decision-making module for process parameters enables intelligent optimization of core process parameters. The specific process is as follows: Based on the quality standards and processing requirements of the laser cladding industry, upper limit thresholds for primary dendrite spacing, residual stress, and warpage deformation are preset to construct a set of process quality constraints as hard boundaries for parameter optimization. A multi-objective loss function is constructed with the minimization of these three quality indicators as the optimization objective. ; in, The total loss value, To invert the primary dendrite spacing in real time, This represents the upper limit of the primary dendrite spacing. For real-time residual stress, The residual stress threshold, This represents the real-time warping deformation. The threshold for warping deformation. and These are the weighting coefficients corresponding to each quality indicator.

[0034] The laser power and scanning speed parameters of the equipment are collected in real time. The real-time process parameters and quality inversion results are substituted into the loss function to solve for the parameter gradient change value. The gradient descent method is used to iteratively update the process parameters along the direction of minimum loss value. The iteration continues until the loss function converges and all quality indicators meet the constraint threshold requirements. The optimal laser power and scanning speed parameters under the current operating conditions are output.

[0035] The parameter execution and feedback module realizes the parameter execution and closed-loop feedback of operating data. The specific process is as follows: The system receives optimized process parameters from the decision-making module and distributes them to the laser controller and equipment motion controller to drive the equipment to complete the cladding process. Simultaneously, it records parameter execution logs to form a process parameter execution record. Real-time acquisition of the equipment's actual laser power and scanning speed yields accurate process parameter data. After the current cladding cycle ends, a complete set of multi-dimensional sensing data for the next full processing cycle is collected. The process execution record, actual process parameters, and new cycle sensing data are bound and encapsulated to generate a standardized feedback data package, providing data support for subsequent deviation verification.

[0036] The closed-loop adaptive verification module completes the monitoring and optimization reset correction of operating condition deviations. The specific process is as follows: The system analyzes feedback data packets to extract key actual operating condition data such as the geometric dimensions of the cladding product and the cooling rate of the molten pool. This data is then compared with the pre-stored standard process quality parameters using normalized difference calculations to obtain the real-time quality deviation index. The system presets a fixed deviation tolerance threshold and continuously monitors the deviation index changes over multiple cladding cycles. If the deviation index exceeds the tolerance range for several consecutive cycles, the gradient descent optimization algorithm is deemed to have entered a local optimum and cannot continue iterative optimization. At this point, an optimization reset signal is automatically generated and pushed to the process parameter adaptive decision module, forcibly resetting the initial point of the gradient descent iteration, restarting the parameter optimization process, breaking the local optimum limitation, and achieving continuous adaptive optimization of process parameters.

[0037] Correspondingly, the optimization method of this invention, through the coordinated operation of the above five modules, completes the fully automated iterative operation of the entire process of "multi-dimensional perception - feature extraction - quality inversion - parameter optimization - execution feedback - closed-loop verification", realizing dynamic adaptive optimization of laser cladding process parameters, effectively improving the cladding processing accuracy and finished product quality stability. The specific steps are as follows: Step 1: Simultaneously collect four types of sensor data—spectral, infrared, visual, and acoustic emission—through the multi-dimensional perception and feature extraction module, complete the spatiotemporal registration process of timestamp alignment and spatial coordinate unification, extract the molten pool oscillation frequency, geometric parameters, temperature gradient, and spatter characteristics, and fuse them to construct a standardized cladding process feature vector. Step 2: Using the built-in thermal, mechanical, and microstructure coupled analytical model of the cladding layer quality inversion module, input the cladding process feature vector to perform multi-physics field coupling deduction, and invert to obtain the cladding layer quality indicators such as primary dendrite spacing, residual stress distribution, and warping deformation. Step 3: The preset quality constraint set is retrieved through the process parameter adaptive decision module, and the gradient descent method is used to perform multi-objective linkage optimization of laser power and scanning speed, and output the optimal process parameters that are adapted to the real-time cladding conditions. Step 4: The optimized process parameters are issued through the parameter execution and feedback module to complete the cladding operation, and the actual process parameters of the equipment and the multi-dimensional sensing data of the new cycle are collected simultaneously and packaged to generate a feedback data package; Step 5: The feedback data packet is parsed and the quality deviation index is calculated through the closed-loop adaptive verification module. When the deviation index continues to exceed the preset tolerance, the initial point of optimization is reset and the process is repeated in a loop to achieve closed-loop adaptive optimization of laser cladding process parameters.

[0038] In the application, several formulas are calculated by removing dimensions and taking their numerical values. The formulas are established by collecting a large amount of data and simulating the most recent real situation. Some coefficients or weights in the formulas are set by those skilled in the art according to the actual situation, so they will not be elaborated here.

[0039] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0040] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A laser cladding parameter optimization system based on multi-dimensional sensing data, characterized in that, include: The multidimensional perception and feature extraction module deploys a spectral sensor, an infrared thermal imager, a high-speed visual sensor, and an acoustic emission sensor to simultaneously collect molten pool spectrum, temperature field, morphology, and acoustic emission data, and completes timestamp alignment and spatial coordinate unification to obtain a spatiotemporally aligned multidimensional perception dataset. The multidimensional perception and feature extraction module has multiple feature extraction sub-modules built in, which extract the molten pool oscillation frequency, molten pool geometric parameters, temperature gradient distribution and splash features from the dataset to construct a feature vector of the cladding process. The cladding layer quality inversion module receives the feature vector of the cladding process and has a built-in thermal, mechanical, and microstructure coupled analytical model. Based on the feature vector of the cladding process, it inverts the primary dendrite spacing, residual stress distribution, and warpage deformation, and outputs the quality inversion results. The process parameter adaptive decision-making module receives the quality inversion results, internally stores a quality constraint set containing size, stress, and deformation threshold, collects laser power and scanning speed in real time, and uses the gradient descent method to combine the quality inversion results and the constraint set to complete the linkage optimization and output the optimized process parameters. The parameter execution and feedback module is connected to the process parameter adaptive decision module. It receives and sends process parameters to the corresponding controller for execution, and at the same time collects actual process parameters and new cycle sensing data to generate feedback data packets. The closed-loop adaptive verification module receives feedback data packets, calculates the quality deviation index based on the feedback data packets, and sends a reset signal to the process parameter adaptive decision module when the deviation index continues to exceed the preset tolerance, forcibly resetting the initial point of optimization.

2. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 1, characterized in that, The process of acquiring the multidimensional sensing dataset is as follows: The dynamic process data of laser cladding is collected in parallel and synchronously by spectral sensors, infrared thermal imagers, high-speed visual sensors and acoustic emission sensors, and the raw multi-source sensing data including molten pool spectral data, molten pool temperature field data, molten pool morphology data and acoustic emission signal data are summarized. Using the high-frequency acquisition timing of the high-speed vision sensor as a unified time reference, timestamp correction and gap filling are performed on the asynchronous sampling data of the spectral sensor, infrared thermal imager and acoustic emission sensor; Using the laser cladding equipment's processing reference coordinate system as a spatial reference, the field-of-view coordinates and molten pool spatial coordinates of all sensors are uniformly mapped and transformed to complete the spatial location matching of multi-source data, ultimately generating a spatiotemporally aligned multidimensional sensing dataset.

3. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 2, characterized in that, The multi-feature extraction submodule includes a molten pool oscillation frequency extraction submodule, a molten pool geometric parameter calculation submodule, a temperature gradient analysis submodule, and a spatter feature statistics submodule; The molten pool oscillation frequency extraction submodule extracts the molten pool oscillation frequency from the multidimensional sensing dataset, the molten pool geometric parameter calculation submodule extracts the molten pool width and height from the multidimensional sensing dataset, the temperature gradient analysis submodule extracts the temperature gradient distribution value from the multidimensional sensing dataset, and the spatter feature statistics submodule extracts the number and velocity distribution of spatter particles from the multidimensional sensing dataset, thus forming a feature vector of the cladding process.

4. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 3, characterized in that, The process of constructing the feature vector for the cladding process is as follows: The molten pool oscillation frequency extraction submodule performs time-domain and frequency-domain analysis on the acoustic emission signal data and molten pool morphology time-series data in the multidimensional sensing dataset, and extracts the inherent oscillation frequency characteristics of the molten pool by removing environmental noise interference through fast Fourier transform. The melt pool geometry parameter calculation submodule calculates the real-time melt pool width and height parameters based on the melt pool topography data, and performs image threshold segmentation and edge contour extraction on the melt pool topography data collected by the high-speed vision sensor to obtain the melt pool geometry parameter features. The temperature gradient analysis submodule decomposes the molten pool temperature field data acquired by the infrared thermal imager into a matrix grid, calculates the temperature gradient distribution values ​​from the center to the edge of the molten pool and in the front and back directions of the cladding area, and obtains the temperature gradient distribution characteristics of the molten pool. The splash feature statistics submodule performs target recognition and trajectory tracking on high-speed visual continuous frame images, counts the number of splash particles per unit time, fits the motion trajectory of splash particles to obtain the velocity distribution range, and obtains splash motion features. The extracted features of the molten pool's inherent oscillation frequency, geometric parameters, temperature gradient distribution, and spatter motion are normalized and then fused to construct a multidimensional feature vector for the cladding process. ; in, The inherent oscillation frequency characteristics of the molten pool For the geometric parameters of the molten pool, For the characteristics of the temperature gradient distribution of the molten pool, This is a characteristic of splash motion.

5. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 1, characterized in that, The process of constructing the thermal, mechanical, and tissue-coupled analytical model is as follows: To match the actual working conditions of high-temperature forming and rapid solidification of laser cladding, three independent models were built: a heat conduction model of the molten pool, a stress-strain mechanics model of the cladding layer, and a grain growth kinetics model. The basic boundary parameters and material property parameters of the three types of models were calibrated, resulting in three independent single-physics field basic models. Based on the three independent single-physics field basic models obtained, a multi-physics field coupling and correlation mechanism is established. The global temperature field time series data output by the molten pool heat conduction model is used as the driving basis. The stress-strain mechanics model is associated to realize the field-coupled update of temperature gradient thermal stress and solidification shrinkage stress. At the same time, relying on the temperature field change rate and molten pool flow state, the grain structure growth kinetics model is linked to complete the dynamic evolution matching of the solidification interface. Thus, a multi-physics field coupling and correlation mechanism in which the temperature field drives the stress field and the temperature change flow field jointly regulate the grain structure evolution is constructed. Based on the established multi-physics coupling and correlation mechanism, targeted variable correlation and adaptation are performed on the feature vector of the cladding process: the molten pool temperature gradient distribution feature is used as the real-time boundary correction variable of the heat conduction model, the molten pool geometric parameter feature is used as the molten pool sintering region size constraint variable, and the molten pool inherent oscillation frequency feature is used as the molten pool flow disturbance and grain growth disturbance correction variable. The three types of standardized features are used as the real-time joint input of the model to establish the hierarchical transmission relationship of sensing feature parameters, multi-physics state and cladding quality index, and construct a nonlinear mapping relationship of multi-feature input, multi-physics linkage and multi-quality index output. By using multiple sets of process simulation calibration data and actual on-site process experimental data, the internal coupling coefficient and hysteresis compensation coefficient of the nonlinear mapping relationship were iteratively corrected, and the overall parameters of the model were solidified to obtain a thermal, mechanical, and microstructure coupled analytical model. The calculation formula for the molten pool heat conduction model is as follows: ； in, Density of the cladding material For the specific heat capacity of the material, This refers to the real-time temperature of the molten pool. For cladding time, For the thermal conductivity of the material, To provide heat flux to the laser, The heat flux lost is due to convection and radiation in the molten pool.

6. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 5, characterized in that, The process of solving the mechanical and tissue indices using the coupled thermal, mechanical, and tissue analytical model includes: The real-time global temperature field data output by the coupled thermal, mechanical, and structural analytical model is called, and thermal stress and solidification shrinkage stress are superimposed to obtain the global basic stress data of the cladding layer. Combined with the actual tooling boundary constraints of the component, the global basic stress data of the cladding layer is solved by global integration to obtain the global residual stress distribution data of the cladding layer. Matching the overall stiffness parameters of the component, the warping deformation of the component is calculated synchronously based on the obtained global residual stress distribution data of the cladding layer. By correlating the temperature gradient distribution characteristics of the molten pool, the inherent oscillation frequency characteristics of the molten pool, and the real-time solidification rate of the molten pool, the coupling influence of the three factors is quantified, and the multi-parameter coupling correlation coefficient of the molten pool is obtained. Based on the obtained multi-parameter coupling correlation coefficient of the molten pool, the disturbance amplitude of the molten pool flow and temperature fluctuation on the grain growth interface is quantified, and the quantitative relationship of grain growth disturbance is obtained. Combining the inherent laws of metal solidification kinetics, a quantitative calculation model of parameters and primary dendrite spacing is constructed based on the obtained quantitative relationship of grain growth disturbance, and the primary dendrite spacing of the cladding layer is obtained by solving.

7. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 1, characterized in that, The process parameter adaptive decision-making module completes the process parameter optimization process by including: Combining the process specifications and finished product quality access standards of the laser cladding industry, the upper limit thresholds for primary dendrite spacing, residual stress, and warpage deformation are preset, and all hard control indicators are summarized to obtain the cladding process quality constraint set. With the minimum primary dendrite spacing, minimum residual stress, and minimum warpage deformation as multiple optimization objectives, a multi-objective loss function is constructed based on the obtained cladding process quality constraint set, resulting in the multi-objective loss function for the cladding process. Real-time acquisition of two core operating parameters of the equipment: current laser power and scanning speed, to obtain the equipment's real-time raw process parameters; Substitute the quality inversion results and the obtained real-time original process parameters of the equipment into the multi-objective loss function of the cladding process, solve for the gradient change value of the current process parameters, iterate and update the process parameters along the optimal direction based on the gradient descent method until the loss function converges and all indicators meet the requirements of the cladding process quality constraint set, and output the optimized process parameters. The expression for the multi-objective loss function is: ； in, The total loss value, To invert the primary dendrite spacing in real time, This represents the upper limit of the primary dendrite spacing. For real-time residual stress, The residual stress threshold, This represents the real-time warping deformation. The threshold for warping deformation. and These are the weighting coefficients corresponding to each quality indicator.

8. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 1, characterized in that, The process by which the parameter execution and feedback module generates feedback data packets includes: The system receives optimized process parameters from the adaptive decision-making module and sends them to the laser controller and motion controller to execute the cladding operation, generating a process parameter execution record. The generated process parameter execution records are collected, and the actual laser power and scanning speed of the equipment during operation are collected in real time as actual process parameters. The generated process parameter execution record and the actual process parameters are retrieved, and the complete set of multi-dimensional sensing data corresponding to the next complete cladding processing cycle is collected simultaneously. The three types of data are bound and encapsulated to generate a feedback data packet.

9. The laser cladding parameter optimization system based on multi-dimensional sensing data according to claim 1, characterized in that, The deviation calculation and adaptive reset process of the closed-loop adaptive verification module includes: The feedback data packet is analyzed to extract key data such as the actual geometric dimensions of the cladding product and the actual cooling rate of the molten pool. The system's preset standard process quality parameters are retrieved, and the difference between the two types of parameters is normalized to calculate the real-time quality deviation index. A preset quality deviation tolerance threshold is set, and the real-time quality deviation index changes over multiple cladding cycles are continuously monitored. If the real-time quality deviation index continues to exceed the preset tolerance, it is determined that the gradient descent optimization has fallen into a local optimum. For the locally optimal solution obtained by the determination, an optimization reset signal is generated and pushed to the process parameter adaptive decision module to forcibly reset the gradient descent optimization initial point, restart the parameter iteration optimization, and realize the closed-loop adaptive optimization of the cladding process.

10. A method for optimizing laser cladding parameters based on multidimensional sensing data, characterized in that, The system applied to claim 1 includes the following steps: Step 1: Simultaneously collect four types of sensor data—spectral, infrared, visual, and acoustic emission—through the multi-dimensional perception and feature extraction module, complete the spatiotemporal registration process of timestamp alignment and spatial coordinate unification, extract the molten pool oscillation frequency, geometric parameters, temperature gradient, and spatter characteristics, and fuse them to construct a standardized cladding process feature vector. Step 2: Using the built-in thermal, mechanical, and microstructure coupled analytical model of the cladding layer quality inversion module, input the cladding process feature vector to perform multi-physics field coupling deduction, and invert to obtain the cladding layer quality indicators such as primary dendrite spacing, residual stress distribution, and warping deformation. Step 3: The preset quality constraint set is retrieved through the process parameter adaptive decision module, and the gradient descent method is used to perform multi-objective linkage optimization of laser power and scanning speed, and output the optimal process parameters that are adapted to the real-time cladding conditions. Step 4: The optimized process parameters are issued through the parameter execution and feedback module to complete the cladding operation, and the actual process parameters of the equipment and the multi-dimensional sensing data of the new cycle are collected simultaneously and packaged to generate a feedback data package; Step 5: The feedback data packet is parsed and the quality deviation index is calculated through the closed-loop adaptive verification module. When the deviation index continues to exceed the preset tolerance, the initial point of optimization is reset and the process is repeated in a loop to achieve closed-loop adaptive optimization of laser cladding process parameters.

Images