Electron beam melting multi-source collaborative scanning optimization control method and system
By constructing a multi-source heat flux coupling field model and a hierarchical collaborative optimization strategy, combined with the Kriging proxy model, the control bottleneck of the multi-electron gun melting system was solved, achieving efficient and precise melting process control and improving the compositional uniformity and production energy efficiency of titanium alloys.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BAOJI BAOTAI EQUIP TECH CO LTD
- Filing Date
- 2026-03-31
- Publication Date
- 2026-06-23
AI Technical Summary
Existing technologies in multi-electron gun electron beam melting systems face problems such as the "curse of dimensionality," the "illness" of physical field inversion, and the conflict between high-fidelity models and real-time requirements, making it difficult to achieve efficient and accurate melting process control.
A hybrid modeling architecture is adopted, combining a hierarchical collaborative optimization strategy and a Kriging surrogate model. By constructing a multi-source heat flux coupling field model, multi-objective constraint optimization is performed, and parameter adaptive rolling time-domain closed-loop control is implemented to achieve precise control of the smelting process.
It achieves nanosecond-level electron beam scanning accuracy and millisecond-level energy field optimization control, improving the compositional uniformity and production energy efficiency of titanium alloy smelting, and reducing the total energy consumption of the system.
Smart Images
Figure CN122260977A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent control technology for high-end metallurgical equipment, specifically relating to a multi-source collaborative scanning optimization control method and system for electron beam melting. Background Technology
[0002] Electron beam cold-bed melting is a key technology for preparing aerospace-grade titanium alloys and high-temperature alloys. With the development of equipment towards higher power and multi-gun designs, the control of the melting process faces significant challenges. From a control theory perspective, a multi-electron-gun melting system is a typical distributed parameter system with large time delay, strong nonlinearity, and multiple-input multiple-output (MIMO). When the high-energy electron beams emitted by each electron gun bombard the surface of the molten pool, their heat flux density distributions exhibit significant spatial overlap and coupling effects. This means that local adjustments of a single heat source can affect the global temperature field through heat conduction and convection mechanisms.
[0003] Existing control technologies face three core scientific and engineering bottlenecks when dealing with complex conditions such as multi-electron gun cooperative scanning: the "curse of dimensionality," the "ill-posedness" of physical field inversion, and the conflict between high-fidelity models and real-time requirements. Specifically, the massive parameters such as the scanning patterns, deflection angles, and power distribution of multiple electron guns constitute an extremely high-dimensional decision space, causing traditional global search algorithms to face a combinatorial optimization dilemma with exponentially increasing computational complexity, making it difficult to meet the response requirements of millisecond-level control cycles. At the same time, constrained by the thermal diffusion effect, inverting control parameters from the desired temperature field is mathematically a typical ill-conditioned inverse problem. The non-uniqueness of the solution and numerical instability make traditional control strategies prone to getting trapped in local extrema, thereby inducing energy accumulation or "cold shut-off" defects. In addition, although high-fidelity numerical simulations (FEM / FVM) based on Navier-Stokes and energy equations have high accuracy, their huge computational time and the urgent need for online real-time control present an irreconcilable spatiotemporal scale contradiction, limiting the actual performance of the control system. Summary of the Invention
[0004] To address the aforementioned problems, the present invention aims to provide a multi-source collaborative scanning optimization control method and system for electron beam melting. This method constructs a hybrid modeling architecture that integrates physical mechanisms and data-driven approaches, and utilizes a hierarchical collaborative optimization strategy to achieve order-reduction solutions for complex thermal processes. While ensuring nanosecond-level high-speed electron beam scanning accuracy, it achieves millisecond-level closed-loop optimization control of the melting energy field, thereby significantly improving the compositional uniformity and production efficiency of titanium alloy melting.
[0005] To achieve the above objectives, the present invention provides a multi-source cooperative scanning optimization control method for electron beam melting, comprising the following steps: Step S1: Based on the electron beam scanning trajectory and power parameters, construct a multi-source heat flux coupled field model to determine the spatiotemporal distribution of energy on the melting surface; Step S2: Construct a multi-objective constrained optimization problem (MOP): With the global objectives of minimizing the total energy consumption of the system and optimizing the statistical uniformity of the energy field, define a set of feasible domain constraints that include energy uniformity, melting quality window, and interface gradient smoothness. Step S3: For the multi-objective constrained optimization problem MOP in step S2, perform a two-layer decoupled collaborative optimization based on the Kriging surrogate model to solve it, and obtain the optimal control parameters through hierarchical iteration; Step S4: Implement parameter-adaptive rolling time-domain closed-loop control. Acquire real-time energy distribution data of the melting surface using an infrared thermal imager. Employ a rolling time-domain estimation algorithm to identify and update the model's thermal efficiency coefficient online based on the prediction deviation within the sliding window. Use the updated model parameters for optimization at the next moment and map the solution results into scanning control commands to the actuator for precise control of the electron beam melting process.
[0006] In step S1, a multi-source heat flux coupled field model is constructed. The specific process is as follows: S11: Regarding the strong nonlinear coupling effect caused by the overlap of multiple electron gun scanning regions, assuming the first... i Electron gun at all times t If the heat flux distribution satisfies the asymmetric bielliptic Gaussian heat source function, then the mathematical expression characterizing the energy tailing property of the electron beam during high-speed scanning is: In the formula: For the first i The input power of the electron gun; The electron beam thermal conversion efficiency coefficient; This is the effective semi-axis length of the light spot in the vertical scanning direction; , These represent the effective semi-axis lengths of the light spot in front of and behind the scanning trajectory, respectively. For the first i deflection angle of the electron gun Determined instantaneous scan center coordinates; S12: Discretize the melting surface into The spatial grid, based on the principle of linear superposition, distributes the total energy of the system. Characterized as about the power vector Linear transformation: In the formula: Configure matrices for the deflection angles of all electron guns; The coupling coefficient matrix is configured to change dynamically with angle, and its elements represent the unit power thermal gain of each electron gun to each spatial grid cell. The background dissipation field vector includes ambient radiation and substrate thermal conductivity.
[0007] In step S2, a multi-objective constrained optimization problem (MOP) is constructed: with the global objectives of minimizing the total energy consumption of the system and optimizing the statistical uniformity of the energy field, a functional extremum problem is constructed. In the formula, These are the weighting coefficients for each sub-objective; The L1 norm of the power vector is used to characterize the total energy consumption of the system and induce the sparsity of the solution; The preset process target energy field; The square of the Euclidean norm of the tracking error; Z This refers to the number of process zones; For the first k The coefficient of variation of energy distribution within a process zone is defined as the ratio of the standard deviation of energy in that zone to the mean. At the same time, define the set of feasible region constraints, including: Input saturation constraints: ,in This represents the physical power limit of the electron gun; Overlap rate constraint: the degree of overlap between adjacent electron gun scanning regions ,in This is the preset minimum overlap threshold; Zoned energy constraints: Energy density threshold ranges set separately for the roughing zone and the refining zone.
[0008] In step S3, a two-layer decoupling collaborative optimization based on the Kriging surrogate model is performed to solve the problem. The specific process is as follows: S31: Top-level topology decision: for non-convex scan angle variables An improved genetic algorithm is used to perform a global search to determine the optimal spatial cover topology; S32: Bottom-level flux distribution: Topology at a defined angle Next, the global objective function in step S2 is transformed into a standard-form convex quadratic programming subproblem, and the optimal power allocation vector is solved using the interior-point method. : In the formula, is the Hessian matrix; f is the linear coefficient vector; C,d are the linear constraint parameter matrix and vector composed of saturation constraints and partitioned energy constraints; st indicates that it is constrained by the constraint conditions; To find the optimal power allocation vector solution obtained by solving the above minimization problem; S33: Accelerating the surrogate model: Constructing a Kriging Gaussian process regression model It replaces time-consuming full-physics calculations and uses the expectation-improvement criterion (EI) to guide sampling, prioritizing exploration of regions with large prediction variances.
[0009] In step S4, adaptive rolling time-domain closed-loop control is implemented based on measured energy distribution data fed back from the high dynamic thermal imager. The model parameters are corrected online using rolling time-domain estimation. Based on the sliding window data of the most recent N sampling times, the thermal efficiency coefficient is identified and updated. : In the formula, This is the estimated value of the thermal efficiency coefficient at the current moment; To correct the gain; This is the predicted energy field calculated based on the current model; This represents the vector inner product. The updated parameters are used as the initial state to enter the rolling optimization window at the next time step, thereby achieving adaptive and robust control against fluctuations in raw material properties and environmental drift.
[0010] An electron beam melting multi-source collaborative scanning optimization control system, used to implement the above control method, includes: Actuator module: includes multiple electron guns and their matching high-frequency deflection coils. The electron guns are used to generate high-energy electron beams to bombard the melting surface, and the high-frequency deflection coils are used to control the deflection angle and scanning trajectory of the electron beams according to the received scanning control commands. The sensing feedback module includes a high dynamic range infrared thermal imager, which is installed on the top of the melting furnace. The field of view covers the entire melting surface, the sampling frequency is ≥50Hz, the spatial resolution is not less than M×N grid density, the temperature measurement range covers 800℃~2000℃, and the measurement accuracy is ≤±5℃. It is used to collect the two-dimensional temperature field distribution T(x,y,t) of the melting surface in real time and convert it into energy distribution data through a radiation calibration model. The computing platform module is equipped with a Kriging Gaussian process regression surrogate model and an improved genetic algorithm optimization engine. The solution results are mapped into scanning control commands and sent to the actuators for precise control of the electron beam melting process.
[0011] The beneficial effects of this invention are as follows: 1. By constructing a two-layer decoupling architecture of "macro-topology decision-micro-flux allocation" and a Kriging surrogate model based on the expectation enhancement criterion, this invention effectively overcomes the bottlenecks of "curse of dimensionality" and "illness of conformity" in multi-source collaborative control, avoids the massive computational time consumption of high-fidelity finite element simulation, and realizes a time scale leap from offline calculation to millisecond-level online response; 2. This invention establishes a physical-data dual-driven adaptive robust control system using rolling time-domain estimation. By identifying and correcting key parameters such as thermal efficiency online to offset raw material property fluctuations, and combining L1 sparse regularization and coefficient of variation constraints, it effectively reduces the total energy consumption of the system while improving the energy statistical uniformity of the smelting area, achieving a dual improvement in the process quality stability and production energy efficiency of high-end titanium alloy smelting. Attached Figure Description
[0012] Figure 1 This is a flowchart of a multi-source collaborative scanning optimization control method for electron beam melting according to the present invention.
[0013] Figure 2 This is a diagram of the two-layer decoupled collaborative optimization architecture based on the Kriging proxy model of this invention.
[0014] Figure 3 This is a structural block diagram of an electron beam melting multi-source collaborative scanning optimization control system according to the present invention.
[0015] Figure 4 This is a schematic diagram illustrating the partitioning and spatial distribution constraints of energy density in the electron beam cold bed melting process of the present invention. Detailed Implementation
[0016] To make the content, technical solution, and advantages of the present invention clearer, the present invention will be further described in detail below according to specific embodiments, wherein: Example 1 like Figures 1-2 As shown, a multi-source collaborative scanning optimization control method for electron beam melting includes the following steps: Step S1: Based on the electron beam scanning trajectory and power parameters, construct a multi-source heat flux coupled field model to determine the spatiotemporal distribution of energy on the melting surface; Step S2: Construct a multi-objective constrained optimization problem (MOP): With the global objectives of minimizing the total energy consumption of the system and optimizing the statistical uniformity of the energy field, define a set of feasible domain constraints that include energy uniformity, melting quality window, and interface gradient smoothness. Step S3: For the multi-objective constrained optimization problem MOP in step S2, perform a two-layer decoupled collaborative optimization based on the Kriging surrogate model to solve it, and obtain the optimal control parameters through hierarchical iteration; Step S4: Implement parameter-adaptive rolling time-domain closed-loop control. Acquire real-time energy distribution data of the melting surface using an infrared thermal imager. Employ a rolling time-domain estimation algorithm to identify and update the model's thermal efficiency coefficient online based on the prediction deviation within the sliding window. Use the updated model parameters for optimization at the next moment and map the solution results into scanning control commands to the actuator for precise control of the electron beam melting process.
[0017] In step S1, a multi-source heat flux coupled field model is constructed. The specific process is as follows: S11: Regarding the strong nonlinear coupling effect caused by the overlap of multiple electron gun scanning regions, assuming the first... i Electron gun at all times t If the heat flux distribution satisfies the asymmetric bielliptic Gaussian heat source function, then the mathematical expression characterizing the energy tailing property of the electron beam during high-speed scanning is: In the formula: For the first i The input power of the electron gun; The electron beam thermal conversion efficiency coefficient; This is the effective semi-axis length of the light spot in the vertical scanning direction; , These represent the effective semi-axis lengths of the light spot in front of and behind the scanning trajectory, respectively. For the first i deflection angle of the electron gun Determined instantaneous scan center coordinates; S12: Discretize the melting surface into The spatial grid, based on the principle of linear superposition, distributes the total energy of the system. Characterized as about the power vector Linear transformation: In the formula: Configure matrices for the deflection angles of all electron guns; The coupling coefficient matrix is configured to change dynamically with angle, and its elements represent the unit power thermal gain of each electron gun to each spatial grid cell. The background dissipation field vector includes ambient radiation and substrate thermal conductivity.
[0018] In step S2, a multi-objective constrained optimization problem (MOP) is constructed: with the global objectives of minimizing the total energy consumption of the system and optimizing the statistical uniformity of the energy field, a functional extremum problem is constructed. In the formula, These are the weighting coefficients for each sub-objective; The L1 norm of the power vector is used to characterize the total energy consumption of the system and induce the sparsity of the solution; The preset process target energy field; The square of the Euclidean norm of the tracking error; Z This refers to the number of process zones; For the first k The coefficient of variation of energy distribution within a process zone is defined as the ratio of the standard deviation of energy in that zone to the mean. At the same time, define the set of feasible region constraints, including: Input saturation constraints: ,in This represents the physical power limit of the electron gun; Overlap rate constraint: the degree of overlap between adjacent electron gun scanning regions ,in This is the preset minimum overlap threshold; Zoned energy constraints: Energy density threshold ranges set separately for the roughing zone and the refining zone.
[0019] In step S3, a two-layer decoupling collaborative optimization based on the Kriging surrogate model is performed to solve the problem. The specific process is as follows: S31: Top-level topology decision: for non-convex scan angle variables An improved genetic algorithm is used to perform a global search to determine the optimal spatial cover topology; S32: Bottom-level flux distribution: Topology at a defined angle Next, the global objective function in step S2 is transformed into a standard-form convex quadratic programming subproblem, and the optimal power allocation vector is solved using the interior-point method. : In the formula, is the Hessian matrix; f is the linear coefficient vector; C,d are the linear constraint parameter matrix and vector composed of saturation constraints and partitioned energy constraints; st indicates that it is constrained by the constraint conditions; To find the optimal power allocation vector solution obtained by solving the above minimization problem; S33: Accelerating the surrogate model: Constructing a Kriging Gaussian process regression model It replaces time-consuming full-physics calculations and uses the expectation-improvement criterion (EI) to guide sampling, prioritizing exploration of regions with large prediction variances.
[0020] In step S4, adaptive rolling time-domain closed-loop control is implemented based on measured energy distribution data fed back from the high dynamic thermal imager. The model parameters are corrected online using rolling time-domain estimation. Based on the sliding window data of the most recent N sampling times, the thermal efficiency coefficient is identified and updated. : In the formula, This is the estimated value of the thermal efficiency coefficient at the current moment; To correct the gain; This is the predicted energy field calculated based on the current model; This represents the vector inner product. The updated parameters are used as the initial state to enter the rolling optimization window at the next time step, thereby achieving adaptive and robust control against fluctuations in raw material properties and environmental drift.
[0021] Example 2 like Figure 3 As shown, an electron beam melting multi-source collaborative scanning optimization control system, used to implement the above-mentioned control method, includes: Actuator module: includes multiple electron guns and their matching high-frequency deflection coils. The electron guns are used to generate high-energy electron beams to bombard the melting surface, and the high-frequency deflection coils are used to control the deflection angle and scanning trajectory of the electron beams according to the received scanning control commands. The sensing feedback module includes a high dynamic range infrared thermal imager, which is installed on the top of the melting furnace. The field of view covers the entire melting surface, the sampling frequency is ≥50Hz, the spatial resolution is not less than M×N grid density, the temperature measurement range covers 800℃~2000℃, and the measurement accuracy is ≤±5℃. It is used to collect the two-dimensional temperature field distribution T(x,y,t) of the melting surface in real time and convert it into energy distribution data through a radiation calibration model. The computing platform module is equipped with a Kriging Gaussian process regression surrogate model and an improved genetic algorithm optimization engine. The solution results are mapped into scanning control commands and sent to the actuators for precise control of the electron beam melting process.
[0022] Example 3 This embodiment employs a multi-source collaborative scanning optimization control method for electron beam melting as described in Embodiment 1 to optimize the control of a 3000kW electron beam cold hearth furnace (EBCHR) equipped with seven high-power electron guns. Figure 4 As shown, the specific structure of the smelting furnace is as follows: (1) Actuator: Each electron gun has a rated power of 450kW, is equipped with a high-frequency deflection coil, supports bidirectional scanning of X / Y axes, and has a maximum scanning frequency of 10kHz; (2) Sensing feedback: A high dynamic range infrared thermal imager with a sampling frequency of 50Hz is installed on the furnace top to collect the two-dimensional temperature field distribution on the surface of the molten pool in real time. ; (3) Computing platform: The control system adopts a two-layer architecture of "industrial control computer (IPC) + embedded controller". The IPC is responsible for running the proxy model and global optimization algorithm described in this invention (upper layer slow dynamics), and the embedded controller is responsible for executing millisecond-level power allocation and spot trajectory generation (lower layer fast dynamics). The optimization control process is as follows: Step 1: Construct a multi-source heat flux coupling field mechanism model, focusing on the surface region of the cold hearth melting process. (size is) A heat flux density distribution model is established, taking into account the integral effect and tailing phenomenon of high-speed electron beam scanning, and the first... i Electron gun at all times t The heat flux is equivalent to an asymmetric double-elliptic Gaussian heat source distribution: in, For power, Due to the deflection angle Determined instantaneous scan center coordinates; This is the effective semi-axis length of the light spot in the vertical scanning direction; , These represent the effective semi-axis lengths of the light spot in front of and behind the scanning trajectory, respectively, and the total energy distribution of the system. In this embodiment, the superposition and convolution of heat dissipation from each heat source and the environment is achieved through discretization of the mesh ( This physical process is abstracted as a matrix mapping: in It is a vector configured according to angle. The changing coupling coefficient matrix reflects the "strong coupling" characteristic; To warm up the background; Step 2: Define a multi-objective optimization problem and set the optimization objective function. J This includes a weighted sum of energy consumption and quality indicators: in This is the L1 regularization term (characterizing total energy consumption). Let be the coefficient of variation of energy distribution in the roughing / refining zone, with the following constraints: (1) Energy density target for the refining zone: ; (2) Energy density target for refining zone: ; (3) Single-gun power constraint: ; (4) Scanning angle overlap rate constraint: overlap of adjacent gun scanning areas (Ensure there are no cold spots at the junction); Step 3: Execute the two-layer proxy collaborative optimization strategy. For the above non-convex and nonlinear problems, the system runs the following closed-loop process: 3.1 Offline training phase: 500 sets of "angle-power" samples were generated using Latin hypercube sampling (LHS), and the corresponding energy field distribution was calculated through finite element simulation to construct a Kriging Gaussian process regression model; 3.2 Online hierarchical optimization stage: Upper layer (topology search): An improved genetic algorithm (GA) is used to search for the optimal angle configuration. The population size is set to 50, and the number of generations is 100. Lower layer (flux distribution): For each generation of individuals produced by GA (i.e., a fixed angle) The system solves a convex quadratic programming (QP) subproblem: The interior-point method can be used to quickly obtain the globally optimal power allocation at this angle. ; Active learning and updating: By introducing the expected improvement criterion EI, the system will automatically identify the "most uncertain" region of the surrogate model (i.e. the point with the largest EI value), trigger a real physical verification or high-precision calculation, and incorporate new data into the training set to correct the Kriging model parameters online; Step 4: Execute the closed-loop issuance and performance verification of the optimal control command. After the online iterative algorithm converges, the IPC issues the optimal command to the PLC to drive the 7 electron guns to work together and dynamically adjust the scanning waveform. After optimization, the total system input power of the electron gun array is significantly reduced, the statistical uniformity of energy density distribution on the melting surface (especially the refining zone) is greatly improved, and the energy gradient at the interface is smooth and without abrupt changes. Step 5: Implement adaptive robust control under dynamic disturbances. For abnormal conditions during the smelting process where fluctuations in the moisture content and particle size of the raw titanium sponge lead to an increase in latent heat of fusion, resulting in a significant drop in the average temperature of the roughing zone detected by the infrared thermal imager (warning of the risk of the cooling bed "freezing"), the system executes the following adaptive correction procedure: (1) Anomaly monitoring: The monitoring module detects that the temperature field deviation exceeds the threshold. ; (2) Model adaptive correction: The system immediately starts rolling time-domain estimation (RHE); based on the sliding window bias data of the most recent 10 sampling times (N=10), the thermal efficiency coefficient in the mechanism model is updated. : (3) Fast replanning: Based on the corrected model, local hot start optimization is triggered. Since the optimal solution at the previous moment is used as the initial value, the algorithm converges within 5 iterations. (4) Execution result: The system automatically increases the power of electron guns No. 1, 2 and 3 (corresponding to the roughing zone) and finely adjusts the angle of gun No. 4 to deflect towards the roughing zone to provide thermal compensation, and finally pulls the molten pool temperature back to the process window without disrupting the stability of the refining zone.
[0023] Obviously, the specific embodiments described above are only some of the embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the protection scope of the present invention.
Claims
1. A method for optimizing control of multi-source collaborative scanning in electron beam melting, characterized in that: Includes the following steps: Step S1: Based on the electron beam scanning trajectory and power parameters, construct a multi-source heat flux coupled field model to determine the spatiotemporal distribution of energy on the melting surface; Step S2: Construct a multi-objective constrained optimization problem (MOP): With the global objectives of minimizing the total energy consumption of the system and optimizing the statistical uniformity of the energy field, define a set of feasible domain constraints that include energy uniformity, melting quality window, and interface gradient smoothness. Step S3: For the multi-objective constrained optimization problem MOP in step S2, perform a two-layer decoupled collaborative optimization based on the Kriging surrogate model to solve it, and obtain the optimal control parameters through hierarchical iteration; Step S4: Implement parameter-adaptive rolling time-domain closed-loop control. Acquire real-time energy distribution data of the melting surface using an infrared thermal imager. Employ a rolling time-domain estimation algorithm to identify and update the model's thermal efficiency coefficient online based on the prediction deviation within the sliding window. Use the updated model parameters for optimization at the next moment and map the solution results into scanning control commands to the actuator for precise control of the electron beam melting process.
2. The method for multi-source collaborative scanning optimization control of electron beam melting according to claim 1, characterized in that: In step S1, a multi-source heat flux coupled field model is constructed. The specific process is as follows: S11: Regarding the strong nonlinear coupling effect caused by the overlap of multiple electron gun scanning regions, assuming the first... i Electron gun at all times t If the heat flux distribution satisfies the asymmetric bielliptic Gaussian heat source function, then the mathematical expression characterizing the energy tailing property of the electron beam during high-speed scanning is: In the formula: For the first i The input power of the electron gun; The electron beam thermal conversion efficiency coefficient; This is the effective semi-axis length of the light spot in the vertical scanning direction; , These are the effective semi-axis lengths of the light spot in front of and behind the scanning trajectory, respectively; For the first i deflection angle of the electron gun Determined instantaneous scan center coordinates; S12: Discretize the melting surface into The spatial grid, based on the principle of linear superposition, distributes the total energy of the system. Characterized as about the power vector Linear transformation: In the formula: Configure matrices for the deflection angles of all electron guns; The coupling coefficient matrix is configured to change dynamically with angle, and its elements represent the unit power thermal gain of each electron gun to each spatial grid cell. The background dissipation field vector includes ambient radiation and substrate thermal conductivity.
3. The method for multi-source collaborative scanning optimization control of electron beam melting according to claim 2, characterized in that: In step S2, a multi-objective constrained optimization problem (MOP) is constructed: with the global objectives of minimizing the total energy consumption of the system and optimizing the statistical uniformity of the energy field, a functional extremum problem is constructed. In the formula, These are the weighting coefficients for each sub-objective; The L1 norm of the power vector is used to characterize the total energy consumption of the system and induce the sparsity of the solution; The preset process target energy field; The square of the Euclidean norm of the tracking error; Z This refers to the number of process zones; For the first k The coefficient of variation of energy distribution within a process zone is defined as the ratio of the standard deviation of energy in that zone to the mean. At the same time, define the set of feasible region constraints, including: Input saturation constraints: ,in This represents the physical power limit of the electron gun; Overlap rate constraint: the degree of overlap between adjacent electron gun scanning regions ,in This is the preset minimum overlap threshold; Zoned energy constraints: Energy density threshold ranges set separately for the roughing zone and the refining zone.
4. The method for multi-source collaborative scanning optimization control of electron beam melting according to claim 3, characterized in that: In step S3, a two-layer decoupling collaborative optimization based on the Kriging surrogate model is performed to solve the problem. The specific process is as follows: S31: Top-level topology decision: for non-convex scan angle variables An improved genetic algorithm is used for global search to determine the optimal spatial cover topology; S32: Bottom-level flux distribution: Topology at a defined angle Next, the global objective function in step S2 is transformed into a standard-form convex quadratic programming subproblem, and the optimal power allocation vector is solved using the interior-point method. : In the formula, is the Hessian matrix; f is the linear coefficient vector; C,d are the linear constraint parameter matrix and vector composed of saturation constraints and partitioned energy constraints; st indicates that it is constrained by the constraint conditions; To find the optimal power allocation vector solution obtained by solving the above minimization problem; S33: Accelerating the surrogate model: Constructing a Kriging Gaussian process regression model It replaces time-consuming full-physics calculations and uses the expectation-improvement criterion (EI) to guide sampling, prioritizing exploration of regions with large prediction variances.
5. The electron beam melting multi-source collaborative scanning optimization control method according to claim 4, characterized in that: In step S4, adaptive rolling time-domain closed-loop control is implemented based on measured energy distribution data fed back from the high dynamic thermal imager. The model parameters are corrected online using rolling time-domain estimation. Based on the sliding window data of the most recent N sampling times, the thermal efficiency coefficient is identified and updated. : In the formula, This is the estimated value of the thermal efficiency coefficient at the current moment; To correct the gain; This is the predicted energy field calculated based on the current model; This represents the vector inner product. The updated parameters are used as the initial state to enter the rolling optimization window at the next time step, thereby achieving adaptive and robust control against fluctuations in raw material properties and environmental drift.
6. A multi-source collaborative scanning optimization control system for electron beam melting, used to implement the method described in claim 5, characterized in that, include: Actuator module: includes multiple electron guns and their matching high-frequency deflection coils. The electron guns are used to generate high-energy electron beams to bombard the melting surface, and the high-frequency deflection coils are used to control the deflection angle and scanning trajectory of the electron beams according to the received scanning control commands. The sensing feedback module includes a high dynamic range infrared thermal imager, which is installed on the top of the melting furnace. The field of view covers the entire melting surface, the sampling frequency is ≥50Hz, the spatial resolution is not less than M×N grid density, the temperature measurement range covers 800℃~2000℃, and the measurement accuracy is ≤±5℃. It is used to collect the two-dimensional temperature field distribution T(x,y,t) of the melting surface in real time and convert it into energy distribution data through a radiation calibration model. The computing platform module is equipped with a Kriging Gaussian process regression surrogate model and an improved genetic algorithm optimization engine. The solution results are mapped into scanning control commands and sent to the actuators for precise control of the electron beam melting process.