A method for multi-laser scan path optimization for metal additive manufacturing

Abstract

The application provides a multi-laser scanning path optimization method for metal additive manufacturing, and relates to the technical field of additive manufacturing equipment control. The method obtains an initial scanning track and a timing beat, maps the initial scanning track and the timing beat to a logical grid to perform heat superposition and timing decay, constructs a global heat prediction atlas, and solves a thermal interference early warning index. When the index is out of limit, the forming process characteristic label of the intercepted track is extracted, adaptive reconstruction scheduling is performed to generate a control instruction, valley cold area bypass optimization is performed for a weak constraint path, and energy-speed equivalent compensation across dimensions is maintained for a strong constraint path without changing the original coordinates. The application solves the local heat field overflow problem caused by multi-laser concurrency.

Description

Technical Field

[0001] This invention relates to the field of additive manufacturing equipment control technology, specifically a multi-laser scanning path optimization method for metal additive manufacturing. Background Technology

[0002] As industrial-grade metal additive manufacturing evolves towards larger sizes and higher throughput, multi-laser high-frequency concurrent scanning technology has become a key path to overcome the physical bottlenecks in single-machine forming efficiency. Under this technological development trend, the collaborative control accuracy and thermodynamic state management of multiple execution units directly determine the final delivery quality and operational limits of high-end manufacturing equipment.

[0003] In the high-frequency concurrent processing of complex structural parts using multiple laser beams, patent document CN110918988B discloses a laser scanning path planning method and an additive manufacturing method. Although this solution reduces unfused holes at the single-layer static geometry level, it faces systemic challenges in multi-laser concurrent scenarios: existing methods fail to achieve spatiotemporal decoupling of the execution trajectory at the program control level. When multiple laser beams are concealed in space or densely converge, the residual heat generated is rapidly superimposed on the local powder bed, easily inducing excessive melting and splashing of metal powder and thermal stress warping of the formed parts; existing anti-interference strategies mostly rely on rigid geometric safety distance judgments. Once the distance is detected to be too close, the actuator is forcibly stopped and waited. This rigid control logic not only fails to quantify the risk of concealed heat accumulation, but also leads to long-term idleness of multiple wave nodes, causing a precipitous drop in the full-load concurrent production capacity of the equipment.

[0004] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-laser scanning path optimization method for metal additive manufacturing. By introducing a discrete prediction spectrum matrix with a global clock beat attenuation factor and constructing an asymmetric reconfiguration routing mechanism based on forming process constraints, this method upgrades simple geometric collision avoidance to multi-dimensional spatiotemporal and energy-coordinated scheduling, thereby addressing the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for optimizing multiple laser scanning paths in metal additive manufacturing, comprising the following steps:

[0008] Step S1: Obtain the initial scanning trajectory flow of the multi-channel laser galvanometer execution unit; synchronously obtain the hardware state control timing of the multi-channel laser galvanometer execution unit, and extract the laser scanning synchronization control cycle for controlling the synchronization of laser scanning processing based on the hardware state control timing;

[0009] Step S2: Map the initial scanning trajectory stream to the preset forming surface logic grid, perform thermal state value superposition, and combine the laser scanning synchronous control cycle to perform time-series numerical attenuation calculation to extract the global thermal prediction spectrum matrix characterizing the residual heat accumulation state of the forming surface in the powder bed region;

[0010] The initial scanning trajectory stream to be issued is subjected to discrete grid feature mapping and state indexing with the global heat prediction map matrix to extract the local path thermal interference early warning index used to characterize the degree of convergence and accumulation of the current processing heat prediction.

[0011] In response to the determination that the local path thermal interference early warning index exceeds the preset melting safety limit, the molding process feature tags associated with the intercepted trajectory are extracted, and adaptive reconstruction scheduling is performed based on the molding process feature tags and the global heat prediction map matrix to calculate the target reconstruction path data.

[0012] Step S3: Based on the target reconstruction path data, generate a scanning path reconstruction control command for the multi-channel laser galvanometer execution unit. The scanning path reconstruction control command is configured to drive the multi-channel laser galvanometer execution unit to perform the reconstructed multi-channel laser collaborative scanning processing action.

[0013] Preferably, acquiring the initial scanning trajectory flow of the multi-channel laser galvanometer execution unit includes:

[0014] Obtain the pre-parsed slice scan trajectory data stream to be executed;

[0015] Spatial coordinate extraction is performed on the slice scan trajectory data stream to be executed, and a set of two-dimensional transient coordinate time series representing the spatial transient displacement vectors of the endpoints of the multi-beam lasers is extracted;

[0016] The set of two-dimensional transient coordinate time series is established as the initial scan trajectory flow.

[0017] Preferably, the extraction of the laser scanning synchronization control cycle for controlling the synchronization of laser scanning processing includes:

[0018] The reference control cycle sequence for scanning of the multi-channel laser corresponding to the multi-channel laser galvanometer execution unit is read synchronously.

[0019] The reference control cycle sequence is converted into a globally unified digital timestamp sequence, which serves as the laser scanning synchronization control beat that triggers the periodic numerical decay of the global heat prediction spectrum matrix.

[0020] Preferably, a global heat prediction map matrix characterizing the residual heat accumulation state on the forming surface of the powder bed region is extracted; and a local path thermal interference early warning index is extracted to characterize the degree of heat convergence and accumulation in the current processing, including:

[0021] The preset heat weight is assigned and superimposed to the logical coordinate units in the preset forming surface logical grid that are hit by the initial scanning trajectory flow. In response to the periodic triggering of the laser scanning synchronization control cycle, a global synchronous subtraction dimensionality reduction process is performed on all logical coordinate units in the preset forming surface logical grid that carry the stored heat score using a preset material cooling scaling factor. The iteratively updated preset forming surface logical grid is established as the global heat prediction map matrix.

[0022] Extract the prospective scanning path segment of the initial scanning trajectory stream to be issued within the preset future processing period, and obtain the current stock evaluation score of each grid cell covered by the prospective scanning path segment in the global heat prediction map matrix.

[0023] Perform a discrete grid mean aggregation operation along the look-ahead scanning path segment on the current stock assessment score to calculate the characteristic value that reflects the risk of local thermal field energy accumulation and convergence, and configure the characteristic value as the local path thermal interference early warning index.

[0024] Preferably, the process involves extracting the forming process feature labels associated with the intercepted trajectory and calculating the target reconstruction path data used to characterize the laser avoidance direction, including:

[0025] Based on the preset slicing process feature mapping rules, the local structural attributes of the intercepted trajectory in the current three-dimensional molded part are identified, and the molding process feature label is generated. The molding process feature label includes a path weak constraint identifier for characterizing the internal solid filling and a path strong constraint identifier for characterizing the external molding contour.

[0026] From the global heat prediction map matrix, identify and extract candidate grid sets whose single grid waste heat assessment scores are lower than the preset melting safety limit, and establish them as cold zone grid sets;

[0027] In response to determining that the forming process feature label is identified as the path weak constraint identifier, the spatial distance and thermal state between the current processing endpoint of the intercepted trajectory and the cold zone grid set are evaluated, and the candidate processing landing point sequence with the shortest detour distance and in a cooling state is selected to generate the target reconstruction path data.

[0028] Preferably, based on the target reconstruction path data, a scan path reconstruction control command is generated for the multi-channel laser galvanometer execution unit, including:

[0029] Extract the path displacement deviation between the target reconstructed path data and the initial arrangement sequence contained in the initial scan trajectory stream;

[0030] Based on the path displacement deviation, trajectory rerouting information is generated to update the scanning task queue;

[0031] The trajectory rerouting information is encapsulated into the scanning path reconstruction control command to drive the multi-channel laser galvanometer execution unit to avoid the local thermal field interference region and transfer to the position corresponding to the target reconstruction path data to perform processing.

[0032] Preferably, in response to determining that the molding process feature label is identified as the path strong constraint label, the method also triggers a secondary control closed loop in parallel with spatial interleaving scheduling, specifically including:

[0033] The original spatial scanning path of the intercepted trajectory remains unchanged;

[0034] Calculate the thermal overflow deviation between the local path thermal interference early warning index and the melting safety extreme value;

[0035] Based on the thermal field overflow deviation value, a dynamic compensation instruction for process parameters is generated for the multi-channel laser galvanometer execution unit. The dynamic compensation instruction for process parameters is configured to trigger the laser emitter to dynamically reduce the laser output power or dynamically increase the scanning movement speed. The original spatial scanning path that remains unchanged is synchronously packaged with the dynamic compensation instruction for process parameters and established as the target reconstruction path data under the constraint condition for use in step S3.

[0036] Preferably, the spatial distance and thermal state between the current processing endpoint of the intercepted trajectory and the cold zone grid set are evaluated to filter out candidate processing landing point sequences with the shortest detour distance and in a cooled state, including:

[0037] For each candidate mesh node in the cold zone mesh set, extract the relative displacement distance between the node and the current processing endpoint. Based on the spatial ratio mapping relationship between the relative displacement distance and the maximum scanning stroke length of the current forming surface, extract the dimensionless normalized path offset coefficient.

[0038] Read the single-grid waste heat assessment score corresponding to each candidate grid node, and perform a division comparison between the single-grid waste heat assessment score and the preset heat dissipation benchmark reference value of the molding chamber to map out the dimensionless heat avoidance weight used to characterize the degree of local heat accumulation.

[0039] Extract the local feature fill density rate of the slice at the logical slice level of the current 3D molded part, and configure the local feature fill density rate of the slice as a dynamically optimized weight allocation.

[0040] Using the dynamic optimization allocation weight, a linear weighted summation calculation is performed on the normalized path offset coefficient and the dimensionless heat avoidance weight to calculate the candidate node comprehensive cost score, which is used to characterize the comprehensive level of physical transfer efficiency and processing safety of the current node.

[0041] In the cold zone grid set, the numerical sequence extreme value comparison is performed on the comprehensive cost score of all candidate grid nodes, and the target candidate grid nodes that meet the minimum value optimization condition are established as the candidate processing landing point sequence.

[0042] Preferably, the thermal overflow deviation value between the local path thermal interference early warning index and the melt safety extreme value is calculated; based on the thermal overflow deviation value, a dynamic compensation command for the process parameters of the multi-channel laser galvanometer execution unit is generated, including:

[0043] The thermal field overflow deviation value is linearly mapped to the preset volumetric energy density conversion coefficient to calculate the expected power reduction value used to characterize the redundant heat to be stripped from the local thermal field.

[0044] The expected power reduction value is input into the pre-configured process parameter parallel decoupling module to calculate the corresponding expected power attenuation control parameters and expected scan speed-up control parameters respectively.

[0045] Extract the preset minimum forming power threshold corresponding to the bottom layer of the multi-channel laser emitter, and extract the maximum rated scanning rate of the galvanometer at the bottom layer of the multi-channel laser galvanometer execution unit;

[0046] In response to the determination that the expected power attenuation control parameter is not lower than the preset minimum forming power threshold and the expected scanning speed-up control parameter does not exceed the maximum rated scanning rate of the galvanometer, the expected power attenuation control parameter and the expected scanning speed-up control parameter are established as safe process execution parameters, and they are combined, packaged, and output as the process parameter dynamic compensation command.

[0047] Preferably, the method further includes hardware boundary over-limit protection and parameter compensation control logic for extreme value triggering of CNC servo, specifically including:

[0048] In response to determining that the expected power attenuation control parameter is lower than the preset minimum forming power threshold, a power lower limit interception branch is triggered to limit the output laser power control parameter to the preset minimum forming power threshold.

[0049] Calculate the power cut-off difference between the theoretical down-adjustment power corresponding to the expected power reduction value and the preset minimum forming power threshold, and call the preset energy-velocity equivalent compensation conversion formula to convert the power cut-off difference into a dynamic velocity compensation increment to be superimposed on the current scanning trajectory.

[0050] Verify whether the sum of the dynamic speed compensation increment and the expected scanning speed-up control parameter is less than or equal to the maximum rated scanning rate of the galvanometer. After the verification is passed, generate the reconstructed target scanning speed. Finally, perform synchronous packaging and assembly of the limited actual output power parameter and the reconstructed target scanning speed to replace the generation of the dynamic compensation instruction for the process parameters.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] This invention, through steps S1 and S2, synchronously extracts the hardware control timing to generate a laser scanning synchronization cycle, and maps the trajectory to a discrete logic grid to perform superposition and numerical attenuation. It can proactively quantify the local path thermal interference early warning index, preventing physical overload and data flow deadlock caused by the intersection of multiple laser entities.

[0053] To address the capacity loss issue caused by the "light idling upon warning" in existing technologies, this invention extracts "weak constraint markers" from the internal solid filling and activates a multi-dimensional comprehensive cost scoring mechanism based on physical displacement, heat dissipation benchmarks, and filling density. This mechanism smoothly guides high-risk laser beams to the coldest area with the lowest overall cost for synchronous processing, eliminating the ineffective waiting time of hardware nodes. This allows the output duty cycle of multiple electromechanical execution units to approach the full-load limit, improving the efficiency of concurrent multi-laser forming.

[0054] To address the process limitation that prevents spatial avoidance of the external forming contour, a parallel secondary control closed loop is triggered using a "strong constraint identifier." While maintaining the absolute constrain of the spatial coordinate trajectory, the thermal overflow deviation is converted into a drive command for frequency reduction and amplitude limiting or speed-up compensation, and an anti-collision and meltdown check is established. This eliminates the risk of overheating and splashing without causing fracture defects, thereby improving molding yield. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the overall control architecture and spatial thermal field scheduling process of the present invention.

[0056] Figure 2 This is a time-series dynamic iteration diagram of the global heat prediction spectrum matrix of this invention.

[0057] Figure 3 This is the network diagram for weakly constrained cold zone optimization and pathfinding based on multidimensional cost scoring in this invention.

[0058] Figure 4 This is a flowchart illustrating the overall logic of the multi-laser scanning path optimization method of the present invention. Detailed Implementation

[0059] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0060] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0061] Example 1:

[0062] Please see Figures 1 to 4 The present invention provides a technical solution:

[0063] A method for optimizing multiple laser scanning paths in metal additive manufacturing, comprising the following steps:

[0064] Step S1: Acquire initial scanning trajectory data and align timing reference for multi-channel lasers. Specifically, acquire the pre-parsed scan trajectory data stream of the slice to be executed. This involves extracting the scan trajectory instruction data packet containing the expected processing path from the system interaction interface of the host computer CAM slicing software. Spatial coordinate extraction is then performed on the scan trajectory data stream of the slice to be executed. Specifically, a continuous trajectory control data stream is captured and preprocessing pipeline actions are executed to remove non-motion control logic identifiers from the data packet and extract the coordinate point matrix representing the transient spatial displacement vectors of the multi-channel laser beam endpoints on the two-dimensional plane of the physical powder bed. A single-step coordinate span tolerance threshold is set. The function of this parameter is to verify whether the parsed transient coordinate sequence is within the preset kinematic logic tolerance range, so as to filter out dirty data caused by signal fluctuations. In this embodiment, the value of this parameter is quantized and set to 0.05 mm to 0.1 mm.

[0065] The configuration includes a trajectory execution rate determination logic, and a dynamic mapping operation is performed based on the transient execution rate of the current command trajectory: specifically, the parsed scan trajectory vector is extracted in real time, and the target span distance of its single-step command within a preset control cycle is calculated to establish it as the transient rate parameter; in response to the determination that the transient rate parameter is higher than a preset jump determination threshold, in this embodiment, the jump determination threshold is quantized and set to 500 mm / s to 800 mm / s. After completing the above determination, a binary state jump action is performed, directly setting the single-step coordinate span tolerance threshold. Assign and lock the value to the preset logical tolerance boundary upper limit of 0.1 mm; in response to the determination that the transient rate parameter is lower than or equal to the jump determination threshold, indicating that the entity has entered the micro-dense filling scan condition, call the downlink smoothing operator and set the single-step coordinate span tolerance threshold. A linear downward pressure calculation is performed based on the percentage ratio of the current rate to the jump determination threshold until the preset logic tolerance boundary of 0.05 mm is reached. After the above cleaning and filtering, continuous coordinate features are output, and a two-dimensional transient coordinate time series set is generated.

[0066] Introduction is denoted as The set of two-dimensional transient coordinate time series represents the absolute motion displacement landing point data stream of the multi-channel laser focus on the physical powder bed plane over time. In this preferred embodiment, its spatial discrete sampling step size is set to 0.01 mm to 0.02 mm, and the set of two-dimensional transient coordinate time series is established as the initial scanning trajectory stream. The reference control cycle sequence of the multi-channel laser execution scanning corresponding to the multi-channel laser galvanometer execution unit is read synchronously. The reference clock synchronization pulse characteristics of the underlying multi-axis motion control unit are extracted through the status monitoring interface. The captured reference control cycle sequence (the timing status jump indicator fed back by the device communication interface) is used as the trigger source and input into the pre-configured system-level high-frequency phase-locked loop module. The master-slave clock synchronization topology rules are extracted and the master station card with the highest physical bus occupancy in the bus is forcibly configured as the unique global reference clock source, and the crystal oscillator interrupt signals of the underlying multi-channel heterogeneous lasers are configured as slave node signals. With the global reference clock source as the absolute reference anchor point, the following phase alignment formula with compensation factor is called to perform mapping calculation: It extracts the absolute reference digital pulse time of the current instantaneous global reference clock source through the underlying clock read protocol and establishes it as... Simultaneously, the phase drift deviation of the slave node signal relative to the master clock signal is measured at the hardware level. The deviation amount is compared with the pre-configured phase compensation scaling factor within the system. Perform logical multiplication to obtain the corrected delay compensation; add this delay compensation to the absolute reference digital pulse time to map out the calibrated and aligned absolute timestamp of the subordinate node. By iteratively performing this alignment operation on all subordinate nodes, the results are ultimately aggregated and transformed into a globally unified digital timestamp sequence. This includes the phase compensation scaling factor. In the timing alignment feature map, this refers to a smoothing adjustment parameter that characterizes the suppression of nonlinear high-frequency clock jitter; in this embodiment, the preferred range is 0.8 to 0.95, with a preferred value of 0.9. An absolute time alignment identifier is generated based on this digital timestamp sequence, thereby extracting the laser scanning synchronization control beat used to control the synchronization of laser scanning processing. In this step, a beat jitter tolerance window is set to define the hardware communication polling depth. This parameter is designed to absorb the microsecond-level physical response timing misalignment caused by bus load fluctuations when multiple underlying galvanometer devices respond to hardware interruptions. In this embodiment, the value of this parameter is set to a physical time span of 2 to 5 microseconds. Under congested conditions of large-area overlapping collaborative processing by multiple lasers, this clock jitter tolerance window is adaptively increased to prioritize ensuring the continuity of the main control heartbeat link.

[0067] Laser scanning synchronization control cycle time is recorded as Its characterization is used to lock the logic grid refresh pace and the actual light output of multiple lasers in the physical world at the same time scale as the digital heartbeat pulse. In this embodiment, its quantization period threshold is constantly set to 10 microseconds, and this is used as the benchmark for triggering the periodic numerical decay of the global heat prediction spectrum matrix.

[0068] Step S2, Execute the Discrete Mesh Feature Mapping and Thermal Interference Early Warning Assessment Step: This involves mapping the initial scan trajectory flow to a preset forming surface logical mesh and superimposing thermal state values. A two-dimensional discrete array is defined in the industrial control computer, and the preset thermal weight values ​​are directly assigned and superimposed onto the logical coordinate units physically hit by the initial scan trajectory flow in the preset forming surface logical mesh at the instantaneous coordinates. Let be an example of this process. The preset heat weight. This parameter is derived through the following logical mapping: ;in It is the rated physical output power; it characterizes the maximum safe threshold reference that the current laser emitter can output under normal hardware configuration, and in this embodiment, the preferred range is between 300 watts and 600 watts; at the same time, it is introduced as denoted as The material's preset absorption rate; it characterizes the energy attenuation coefficient of the actual thermal energy absorbed by the selected specific type of metal powder at the current working wavelength, and in this embodiment, the preferred range is 0.35 to 0.45. The rated physical output power is obtained from the device interface pool. And compare it with the material preset absorption rate in the configuration library. Perform direct hardware numerical multiplication extraction to strip away ineffective environmental energy that escapes due to surface reflection;

[0069] Simultaneously acquire the discretized reference physical volume of the preset forming surface logical mesh. and the specific heat constant of metal powder In this embodiment, the reference physical volume is discretized. This is used to characterize the absolute spatial size of a single computational unit mapped in the physical world. The acquisition logic is as follows: extract the two-dimensional side length parameter of the logical mesh (preferably 0.05 mm), and combine it with the theoretical powder-laying thickness of a single layer extracted from the external slicing process file (preferably within the range of 0.03 mm to 0.05 mm). Perform a product operation between the mesh bottom area and the powder-laying thickness to calculate the effective physical element volume of a single pulse hit, with a preferred value of approximately 7.5 × 10⁻⁶. -5 cubic millimeters; specific heat capacity constant The inherent physical quantities used to characterize the heating properties of the target metallic material are retrieved by calling a pre-set material property database. For example, for titanium alloys, the preferred range is 500 J / (kg·K) to 600 J / (kg·K), with a preferred value of 520 J / (kg·K). This embodiment uses a discretized reference physical volume. and the specific heat constant of metal powder Multiplication is performed to construct the denominator of the baseline heat dissipation capacity; finally, dimensionless calculation of the division dimension is performed to output the preset heat weight. In this preferred embodiment, the preset heat weight is 5.

[0070] Response to laser scanning synchronous control cycle The periodic triggering of the process involves using a preset material cooling scaling factor to perform synchronous subtraction dimensionality reduction on the hit logical coordinate units. Through this high-frequency alternating assignment and synchronous subtraction process, the iteratively updated preset forming surface logical grid is established as a global heat prediction map matrix characterizing the residual heat accumulation state of the forming surface in the powder bed region.

[0071] Material cooling scaling factor is denoted as It constructs an offline log replay test environment, injecting historical ambient temperature data collected from the PT100 sensor and solidification time series characteristics of real metal powder at different energy densities into the simulation verification engine as extreme test data streams. During data replay, the relative deviation between the model-predicted mesh residual score decay rate and the actual thermodynamic solidification cycle is continuously monitored. When the relative deviation is detected to have stably converged within a preset allowable error of ±2% for 10 consecutive test heartbeat cycles, a calibration termination command is triggered. The scaling value in the model memory pool at this time is extracted and solidified as the material cooling scaling factor. In this embodiment, based on the offline calibration process, the value of this parameter is a constant deduction of 0.03 to 0.08 for each clock cycle. The look-ahead scan path segment of the initial scan trajectory stream to be sent is extracted within a preset future processing period (trajectory data to be executed in the next 50 milliseconds is captured from the bus buffer); the initial scan trajectory stream to be sent is then subjected to discrete grid feature mapping and state indexing with the global heat prediction map matrix. Specifically, the following operations are performed:

[0072] For any two-dimensional logical coordinate unit The grid state is dynamically refreshed at high frequency based on the following formula, and then the current stock assessment score is output. : When the current synchronization control beat t is triggered, extract the historical score remaining from the previous heartbeat cycle for that coordinate unit. Simultaneously, read the cumulative number of overlapping hits that this unit has been physically traversed by multiple laser beams within the current clock cycle. and compare it with the aforementioned predetermined heat weight. Perform a multiplication operation to obtain transient heat injection; then, aggregate the historical score and transient heat injection, and subtract the material cooling scaling factor for a single cycle. The `max` operation of the maximum value filtering logic is invoked to truncate invalid values ​​below 0 to zero, and the final current stock evaluation score is output. The upper limit of the current stock assessment score is limited to 100, serving as an overflow protection boundary parameter. The results of the above iterations are then extracted and used as the current stock assessment scores for each grid cell to be covered by the look-ahead scan path segment.

[0073] The look-ahead scan path segment is discretized and projected onto a grid, and the feature values ​​are calculated using the following discrete aggregation derivation formula: It obtains all discrete grid sequences that are continuously hit by the look-ahead scanning path segment on the physical two-dimensional plane, and counts the total number of hit logical grids. ; Iterate through the sequence and extract the current stock evaluation score corresponding to each grid index i. Perform a pure scalar summation operation on all extracted scores; then divide the accumulated sum by the total number of logical grids. Perform mean-based calculations to output a pure numerical characteristic reflecting the risk of average thermal energy accumulation and convergence along the trajectory segment. Configure this characteristic value as a local path thermal interference early warning index. Its characterization predicts the comprehensive risk level of violent physical boiling and sputtering of the molten metal pool caused by the overlapping of multiple laser trajectories within the future processing window.

[0074] To further explain step S2, the asymmetric adaptive reconfiguration scheduling decoupling step is performed: Let be an example. The melting safety extreme value is defined as the maximum residual heat accumulation value that the current metal powder can withstand at the physical level without liquid metal splashing and keyhole collapse. In this embodiment, its quantization threshold is preferably set to 85. This embodiment constructs an offline data playback environment for the logic layer, injecting an extreme test trajectory stream containing multiple laser densely interspersed into the prediction spectrum simulation module; continuously increasing the virtual grid heat accumulation value in the background and simultaneously aligning and comparing the virtual predicted value with the real metal droplet splashing moment captured by a high-speed industrial camera (visual monitoring node with a frame rate exceeding 10000fps) on the physical machine tool; when the visual judgment shows keyhole boiling characteristics that damage the metal layer formation on the surface of the physical molten pool, the transient accumulation score of the corresponding grid in the prediction spectrum at this time is immediately locked. By extracting such critical scores from at least hundreds of destructive test samples and performing Gaussian distribution expectation value calculation, it is finally solidified as a benchmark. This benchmark is further introduced as the melting safety extreme value, denoted as . It represents the maximum residual assessment limit that the current metal powder can accumulate without liquid metal splashing. In this embodiment, its preferred threshold is set to a value in the range of 80 to 85.

[0075] Response to the determination of local path thermal interference early warning index Exceeding the preset melt safety limit Immediately block the transmission of control data streams targeting the current high-risk trajectory and extract the forming process feature tags associated with the intercepted trajectory. Extract the original process identifier bits attached to the instruction data packets corresponding to the intercepted trajectory using the slice semantic parsing component. Use this as a preset slice process feature mapping rule to perform feature state mapping comparison. Based on the comparison results, attach the forming process feature tag to the high-risk intercepted trajectory. The forming process feature tag is used as a binary enumerated state identifier parameter, and its initial default state is uniformly set to the safe release state. Once extreme value interception is triggered, specific constraint states will be extracted from the data dictionary and injected: when attribute identifiers such as "Hatch" or "Infill" (internal solid mesh filling) are extracted, the tag mapping transitions to a weak constraint identifier representing a path with high physical fault tolerance, allowing for break-print and jump optimization; when attribute identifiers such as "Contour" or "Perimeter" (external precision forming contour) are extracted, the tag transitions to a strong constraint identifier representing a path that absolutely prohibits trajectory break-chain jumps. When the aforementioned weak constraint flag is determined to be in effect, the subsequent optimization and avoidance link is triggered and allowed; when the strong constraint flag is determined to be in effect, the spatial physical location change action is blocked, and the underlying energy compensation and regulation link is activated.

[0076] A global scan is performed on the global heat prediction map matrix to identify and extract candidate grid sets whose single-grid residual heat assessment scores are lower than the preset melting safety limit, and these sets are established as cold zone grid sets. Adaptive reconstruction scheduling is then performed based on molding process feature labels and the global heat prediction map matrix. Specifically, the adaptive reconstruction scheduling is executed through the following adaptive decoupling routes:

[0077] Weakly constrained routing branches trigger multidimensional heterogeneous optimization dimensional self-consistent closed loop: In response to the determination that the forming process feature label is identified as a path weak constraint label, the spatial distance and thermal state between the current processing endpoint of the intercepted trajectory and the cold zone grid set are evaluated, and the candidate processing landing point sequence with the shortest detour distance and in a cooling state is selected.

[0078] For each candidate mesh node in the cold zone mesh set, extract its relative displacement distance with the current processing endpoint. Specifically, use the following linear mapping formula to generate the normalized path offset coefficient. : It extracts the two-dimensional Euclidean distance between the current processing endpoint and the target candidate mesh node from the optimization matrix, and uses it as the relative displacement distance between the nodes. Synchronously read the pre-configured physical limit span of the machine tool guideway and use it as the maximum scanning stroke length. Dividing the former by the latter directly strips away physical units of length such as "millimeters," mapping it to a relative proportional value. This normalized path offset coefficient... The output is limited to the closed interval [0,1].

[0079] Normalized path offset coefficient After removing the absolute length constraint, the relative mechanical wear and time loss costs required for the laser focus to physically jump and avoid obstacles are characterized; further, a relative safety assessment based on the dynamic environmental noise floor is performed. Specifically, the dimensionless heat avoidance weight is calculated using the following comparison formula. : It reads the current single-grid waste heat assessment score of the candidate grid node after the aforementioned logical steps have been updated. The specific value of the single-grid waste heat assessment score is directly mapped and represented by the transient calculation result of the iterative formula of the global heat prediction map matrix extracted in the previous steps at the physical coordinates of the current candidate grid node; the environmental reference temperature data of the powder bed boundary area collected in real time by the environmental temperature sensing node is obtained, and the environmental reference temperature data is input into the pre-configured heat dissipation conversion operator to perform mathematical comparison, specifically calling the following linear scaling formula:

[0080]

[0081] The real-time environmental Celsius variable extracted from the environmental reference temperature data is denoted as... It represents the measured state value inside the current physical chamber; the preset upper limit benchmark for the safe temperature of the molding chamber is denoted as... In this embodiment, 40 degrees Celsius is preferred; the standard room temperature reference value is denoted as... In this embodiment, the preferred temperature is 20 degrees Celsius; the environmental heat exchange conversion coefficient is denoted as... In this embodiment, a dimensionless scaling constant of 1 is preferred. The difference between the upper limit of the safe temperature and the real-time environmental reference temperature data is calculated, divided by the fixed difference between the upper limit of the safe temperature and the standard ambient temperature to achieve dimensional normalization, and then multiplied by the environmental heat exchange conversion coefficient. This is then mapped to a purely numerical parameter characterizing the passive heat dissipation capacity of the processing area, and established as the heat dissipation reference value for the molding chamber. As a general engineering implementation example, the ambient temperature sensing node uses conventional temperature acquisition components. An analog-to-digital converter transforms the analog temperature into ambient reference temperature data, which is then input into the system's heat dissipation calculation logic. A division comparison of the two values ​​generates a pure numerical value characterizing the relative saturation level of local heat accumulation. In this embodiment, this dimensionless heat avoidance weight... It is confined to the convergence interval [0,1].

[0082] For the currently executed logical slice level, the two-dimensional geometric center coordinates of the current processing endpoint are obtained, and the reference laser spot diameter parameter, which is fixed at the bottom layer of the general machine tool, is extracted. This spot diameter parameter is then multiplied by a pre-configured sampling scale amplification factor to obtain a definite macroscopic physical side length. This precise physical side length is used to define the bounding rectangle of the local feature sampling centered on the current processing endpoint. The sampling scale amplification factor is used as a reference. To characterize the macroscopic spatial metric parameter of the density measurement field of view, its preferred quantization value in this embodiment is calibrated to be 5 to 10 times. All logical coordinate cells within the circumscribed rectangle are traversed, and the total number of cells marked as entity filling attributes is counted. This number is then divided by the total capacity of the logical coordinate cells within the circumscribed rectangle to obtain the slice local feature filling density rate ranging from [0, 100%]. The percentage value of this slice local feature filling density rate is directly extracted as the input source for calculation, and a pre-configured linear interpolation operator is called to perform cross-interval proportional mapping calculation. Specifically, this utilizes... ;in The numerical value of continuous feature density in the range [0,1] used to characterize the preceding statistical output; The physical lower limit benchmark for the preset optimization weight is used to characterize the weight, and in this embodiment, the fixed value is 0.2. The physical upper limit benchmark for the preset optimization weight is fixed at 0.8 in this embodiment. By calculating the product of the density rate value and the difference between the two extremes, and then superimposing the basic lower limit value, the final value that smoothly transitions within the value range is calculated, and this value is configured as the dynamic optimization allocation weight. In this embodiment, the control boundary of the dynamic optimization allocation weight is limited to [0.2, 0.8]. If the weight parameter exceeds the preset upper limit of 1 due to data contamination, it will lead to errors in the core state machine mapping and decision offset. At this time, when evaluating the cost, it will tend to focus on spatial distance and ignore the actual residual heat state of the grid. This will cause the laser focus to directly plunge into the nearby boiling high-risk zone in order to pursue the shortest jump time, triggering physical interference collision. In the actual processing game, when facing a large area of ​​dense filling (high density rate) working condition, it will adaptively move towards the upper limit of 0.8, prioritizing the shortening of the avoidance jump distance to ensure production efficiency; when facing a sparse lattice skeleton (low density rate) working condition, it will smoothly transition to the lower limit of 0.2, giving the heat avoidance weight absolute dominance, and achieving a dynamic balance between quality and efficiency under different constraints.

[0083] Dynamically optimize weight allocation This characterizes the adaptive adjustment leverage effect of the current physical scan segment's packing density on the overall optimization direction. Further, the following linearly weighted objective function is called to calculate the comprehensive cost score of the candidate nodes. : It extracts the normalized path offset coefficients obtained from the preceding calculations. Combine it with dynamic optimization to allocate weights Perform multiplication calculations to obtain the spatial distance cost penalty term; simultaneously extract the dimensionless heat avoidance weight. Combine it with the weight complement. Multiplication is performed to obtain a heat hazard penalty term. The two penalty terms are then summed and aggregated to output a comprehensive cost score for the candidate node, reflecting its overall disadvantage. After traversing and scoring the entire cold zone grid set, all scores are input into a minimum value retrieval engine. Through numerical comparison and sorting, the target candidate grid node with the smallest absolute score (minimum combined distance cost and heat cost) is established as the optimal landing point, thus generating a candidate processing landing point sequence. This embodiment sets a timeout threshold for a single optimization calculation (absolute physical calculation time must not exceed 20 microseconds). If the time taken for a single traversal exceeds this preset logical tolerance boundary due to an excessively large candidate pool, the optimization process will be forcibly interrupted, and the node with the smallest score in the currently traversed sequence will be directly taken as the suboptimal solution and forcibly distributed to complete the calculation of the target reconstruction path data.

[0084] Setting up a strongly constrained routing branch: When faced with an external contour segment, the forming process feature label is identified as a strongly constrained path, triggering a secondary control loop in parallel with spatial interleaving scheduling. Under the control of this secondary control loop, the original spatial scanning path (XY spatial coordinate group) of the intercepted trajectory is forcibly maintained to ensure metallurgical continuity, and avoidance optimization actions that change the spatial physical coordinates are no longer performed. Specifically, the algebraic difference between the local path thermal interference early warning index and the molten safety extreme value is extracted and established as the thermal field overflow deviation value. ; Calculate the expected power reduction value : It obtains the pre-configured volumetric energy density conversion coefficient. The thermal field overflow deviation value A direct scalar multiplication is performed with this volumetric energy density conversion factor to output the dimensionless danger index as the equivalent physical energy amplitude (watts) cut off from the laser's output power. Volumetric energy density conversion factor The specific value depends on the absorbance of the base processing material and the thickness of the powder bed. Specifically, this coefficient is determined during the pre-configuration stage by calling an offline calibration program. This involves obtaining the pre-configured theoretical melting enthalpy of the metal powder and performing multiple single-track scanning tests on a test substrate; extracting the optimal actual output power and actual scanning speed that do not produce spatter and can form a continuous and stable melt path through feature analysis; dividing the optimal actual output power by the product of the actual scanning speed and the preset spot area to calculate the actual volumetric energy density consumed by a single track; comparing this actual consumed volumetric energy density with the theoretical melting enthalpy to map out a dimensionless proportionality coefficient used to characterize the degree of energy transfer loss, and establishing this as the volumetric energy density conversion coefficient. In this embodiment, its value is preferably set to the range of 1.5 to 2.

[0085] Expected power reduction Characterized as the equivalent physical watt (W) energy amplitude theoretically removed from the laser end to mitigate impending droplet splash defects. The expected power reduction value is extracted. It then calls the energy decoupling allocation ratio extracted from the general system configuration library to perform algebraic allocation deduction between the power control domain and the speed control domain. The specific formula is as follows: ; Expected power reduction value Energy decoupling distribution ratio Perform scalar multiplication to directly calculate the expected power attenuation control parameters for the first stage. ;

[0086] Extract remaining energy and reduce weight Compared with the expected power reduction value The mathematical product of these values ​​generates the power margin that needs to be compensated for in the second stage. This product margin is then used as the numerator input of a pre-defined energy-velocity equivalent compensation conversion formula, specifically calling the following compensation equation:

[0087]

[0088] This allows for the reverse mapping of the expected scan speedup control parameters for the second stage. Among them, the expected scan speed control parameters This means that to dissipate the aforementioned power margin, the underlying servo motor must add a physical speed increase increment on top of the original scan rate reference, the value of which is preferably converged and limited within the safe operating range of 0 mm / s to 1500 mm / s; target line energy density reference. With laser spot diameter parameter The product of is used to characterize the base heat quota that can be passed through per unit distance of movement.

[0089] Among them, the energy decoupling allocation ratio This parameter represents the trade-off between prioritizing power reduction and speed increase when the system faces interference. In this embodiment, its preferred value is constantly calibrated to 0.6 to 0.8, with the laser power being directly reduced by 60%-80% first, and the remaining dangerous heat being absorbed by the galvanometer speed-up.

[0090] Extract the preset minimum forming power threshold corresponding to the bottom layer of the multi-channel laser emitter, and extract the maximum rated scanning rate of the galvanometer at the bottom layer of the multi-channel laser galvanometer execution unit.

[0091] Furthermore, two physical boundary parameters are injected into the upper-layer logic via the device configuration bus:

[0092] One is the preset minimum forming power threshold. This represents the minimum energy supply required to maintain the complete liquefaction phase transition of metal particles, which is quantified as 150 watts in this embodiment. Regarding this control parameter, if the command issued by the control system exceeds this lower limit of 150W, it will cause the energy flow at the physical metallurgical level to be interrupted, resulting in a physical discontinuity between the metal layers that should have been continuously fused.

[0093] Secondly, the maximum rated scanning rate of the galvanometer. This represents the maximum servo response limit of a conventional drive motor without command slippage, which is quantized to 3000 mm / s in this embodiment. If the calculated speed exceeds this tolerance limit, it will cause the internal interpolation queue of the underlying servo motor to become disordered and lose steps, resulting in the subsequent laser scanning landing point deviating from the design coordinates, thereby contaminating the three-dimensional structural features of the entire formed part.

[0094] In response to the determination that the expected power attenuation control parameter is not lower than the preset minimum forming power threshold and the expected scanning speed-up control parameter does not exceed the maximum rated scanning rate of the galvanometer, the expected power attenuation control parameter and the expected scanning speed-up control parameter are established as safe process execution parameters, and the two are combined and packaged for output.

[0095] In response to the hot deposition condition, if the expected power attenuation control parameter is determined to be lower than the preset minimum forming power threshold, the power lower limit interception branch is triggered, and the output laser power control parameter is limited to the preset minimum forming power threshold.

[0096] This embodiment further extracts the expected power reduction value, subtracts it from the preset minimum forming power threshold, and identifies the power cut-off difference that cannot be absorbed by power reduction. The dynamic velocity compensation increment is calculated by applying the following energy-velocity equivalent compensation conversion formula. : It extracts the target linear energy density benchmark from the pre-cured material. and the macroscopic diameter parameters of the laser spot The product of the two is used as the compensatory denominator. The target line energy density benchmark is... The index used to characterize the effective thermal energy injected per unit physical length along the scanning trajectory is preferably configured in the range of 0.2 J / mm to 0.5 J / mm in this embodiment, with a preferred value of 0.3 J / mm; the macroscopic diameter parameter of the laser spot The preferred range for characterizing the physical interaction scale of the laser beam on the powder bed surface is 0.05 mm to 0.15 mm, with a preferred value of 0.08 mm. Simultaneously, the aforementioned intercepted power truncation difference is considered. As a molecule, a division mapping calculation is performed to convert the uneliminable watt-level energy surplus into millimeter-level scan speed increments that are additionally superimposed on the servo motor.

[0097] Dynamic speed compensation increment This characterizes the nonlinear speed-up compensation command value used to force the dissipation of retained heat by increasing the deflection frequency of the galvanometer motor across physical dimensions under the engineering dead end where the output power can be reduced to the lowest possible level.

[0098] The verification process checks whether the sum of the dynamic speed compensation increment and the expected scan speed-up control parameters is less than or equal to the galvanometer's maximum rated scan rate. If the verification passes, a target reconstruction scan rate is generated. Conversely, in response to extreme heat buildup causing the verification to fail (the sum exceeding the galvanometer's maximum rated scan rate), this embodiment clamps and locks the target reconstruction scan rate to the galvanometer's maximum rated scan rate. Simultaneously, for heat that cannot be properly compensated and absorbed due to physical limitations, a forced shutdown signal (Laser-Off) is directly thrown to the control interface of the multi-channel laser galvanometer execution unit. In this embodiment, the control system calls the standard digital input / output (I / O) interface, switches the pre-configured enable state bit from the enabled state to the disabled state, and outputs the corresponding digital level signal to the enable circuit of the laser emitter, thereby blocking the drive circuit of the laser pump source and creating a cooling window. After triggering the above-mentioned forced cooling window, a closed-loop re-entry and dynamic wake-up flow logic for the shutdown state is built in. At the moment the forced shutdown signal (Laser-Off) is thrown, the control system synchronously activates the pre-configured system recovery timer and stores the current spatial coordinate point that caused the overload into the breakpoint location cache sequence. Within each microsecond-level clock tick of the system recovery timer, the aforementioned global heat prediction spectrum matrix is ​​continuously called to perform a high-frequency thermal field secondary polling operation on the target grid coordinates in the breakpoint location cache sequence to extract the material cooling scaling factor over time. The transient inventory score after continuous decay. This is determined in response to the transient inventory score falling below the melt safety threshold. When the preset safety margin lower bound is reached, the system immediately intercepts and resets the system recovery timer, and simultaneously sends a laser reset start command (Laser-On) containing the breakpoint location cache sequence coordinates to the multi-channel laser galvanometer execution unit. This command is configured to wake up the physical execution unit and drive it to restart oscillation processing at the original breakpoint coordinates with the restored valid process parameters. The preset safety margin lower bound is defined as a hysteresis comparison threshold to prevent high-frequency oscillation start-stop deadlock caused by grid heat at extreme value edges. In this embodiment, its preferred quantized value is defined as the melt safety extreme value. The actual output power parameter, after being limited, is synchronously packaged and assembled with the target scanning speed for reconstruction, replacing the generation of dynamic compensation instructions for process parameters. These dynamic compensation instructions are configured to trigger the laser emitter to dynamically reduce the laser output power or dynamically increase the scanning speed. The original spatial scanning path, which remains unchanged, is synchronously packaged with the generated dynamic compensation instructions for process parameters, establishing the target reconstruction path data under this constraint for use in step S3.

[0099] Step S3: Obtain the aforementioned calculated target reconstruction path data (cold zone space avoidance coordinates produced by the weak constraint branch, or synchronous packaged data containing the original space scan path and process parameter dynamic compensation instructions produced by the strong constraint branch).

[0100] Extract the physical coordinate vector of the safe target pointed to by the target reconstruction path data generated by the preceding solution. Simultaneously extract the coordinate vectors of high-risk original landing points that are about to undergo thermal field interference in the initial scan trajectory flow. Perform a direct vector subtraction operation on the two within the two-dimensional feature domain. By separating the distance and direction difference between the two on the physical axis, the path displacement deviation can be extracted. Path displacement deviation The two-dimensional spatial geometric span characterizing the forced physical jump of multiple laser beams from their original high-risk execution point to a safe landing point to avoid physical impact or local heat overflow is, in this preferred embodiment, limited to a closed interval with an absolute value not exceeding 15 mm. This prevents excessively long physical spans during single correction from causing the galvanometer motor to consume too much time during the polarization addressing process, which could lead to metallurgical defects such as interlayer cold shuts in the pre-scanned but not fully solidified molten metal pool due to excessive cooling time. The instruction cache queue of the underlying motion control component is manipulated based on the path displacement deviation to generate trajectory rerouting information for updating the scanning task queue. The newly calculated deviation vector and safety compensation parameters directly overwrite and replace the original high-risk processing instruction sequence nodes.

[0101] Trajectory rerouting information is recorded as The reconstructed update message data packet, which has been cleaned and corrected and conforms to the direct-read format of the underlying machine tool driver board, is quantized in this embodiment to have a fixed 64-byte control frame structure containing a priority mandatory flag. This physical size is set to match the single-pass microsecond-level full-load data throughput bandwidth of mainstream industrial field high-speed communication buses, preventing millisecond-level communication blockage and crashes caused by the forced segmentation of excessively long custom messages by the underlying communication interface, thereby ensuring the real-time performance of the avoidance action. The previously acquired trajectory rerouting information is imported into the network packet encapsulation component. The standard message header features required by the general physical bus protocol (including the target device addressing identifier and message length marker) are extracted and sequentially concatenated with the trajectory rerouting information. The standard cyclic redundancy check (CRC) algorithm is executed on the concatenated data block to calculate the check bit, which is then appended to the end of the data frame for final encapsulation and output as a scan path reconstruction control command. .

[0102] In a preferred embodiment, a microsecond-level hardware state maintenance mechanism for CRC check failures is constructed at the receiving end (within the galvanometer control card of the multi-channel laser galvanometer execution unit). Specifically, when the communication parsing module of the galvanometer control card determines that the CRC check result of the current frame data does not match, it directly discards the damaged segment and prevents it from entering the kinematic interpolation buffer. Simultaneously, it instructs the galvanometer servo driver to forcibly maintain the effective deflection voltage of the previous microsecond cycle (maintaining the original transient displacement vector inertial motion) and triggers the counter of the communication fault state machine to increment. This embodiment introduces a communication frame loss tolerance counting threshold; its meaning is to characterize the maximum number of consecutive failed messages that the hardware can continuously withstand before the servo is disabled; it is limited to a discrete set of positive integers, preferably 3 to 5 times in this embodiment.

[0103] If the consecutive frame loss count exceeds the communication frame loss tolerance threshold, the galvanometer control card will directly cut off the power level of the galvanometer motor, bypassing the host computer. If the total length of the message generated in the final encapsulation stage exceeds the single-frame capacity limit of the standard communication bus, the excessively long signaling will be forcibly segmented and truncated by the bus transmission node during transmission, causing severe millisecond-level communication congestion. This will result in a large number of timing-aligned execution instructions being misaligned and delayed, destroying the timeliness and safety of multi-laser electromechanical collaborative processing.

[0104] Scan path reconstruction control instructions The representation ultimately refers to the electromechanical drive command sequence that can penetrate the upper-level software abstraction architecture and be directly output to physical devices through the lower-level control interface. In this preferred embodiment, the command entity explicitly includes a digital power control signal for directly controlling the output wattage of the laser pump source power supply, and a deflection control reference waveform for directly controlling the rotational torque of the coil inside the galvanometer deflection motor. The scanning path reconstruction control command is sent to the peripheral communication port via the industrial field high-speed data bus. The multi-channel laser galvanometer execution unit generates a physical action closed loop within a microsecond-level clock cycle after receiving the command. Specifically, when facing a weakly constrained solid filling area, it physically avoids the local residual thermal field interference area and instantly moves to the trough cold area position corresponding to the target reconstruction path data to perform interleaving processing; when facing a strongly constrained edge contour area, it maintains the original spatial coordinate path without deviation and directly performs energy smoothing processing actions of frequency reduction, amplitude limiting, and speed-up compensation in situ.

[0105] Example 2:

[0106] When the local path thermal interference early warning index When the value approaches the theoretical minimum of 0, it indicates that the residual heat from the previous stage within the predicted look-ahead grid region has been completely dissipated by the environment, and the current multiple laser trajectories do not spatially surround or densely intersect. At this point, the actuator achieves physical full-load high-speed movement, representing the system in an absolutely safe state of maximum processing efficiency.

[0107] When the local path thermal interference early warning index Approaching and exceeding the melt safety limit At this time, the input energy density in the local physical region is rapidly increasing, about to break the surface tension limit of the liquid metal, causing boiling at the keyhole and splashing of solid metal droplets. This triggers the hardware defense logic, converting the purely numerical output parameters into a specific sequence of restrictive electromechanical control commands.

[0108] Limited by the maximum rated scanning rate of the galvanometer When dynamic speed compensation increment A larger value indicates a greater thermal overflow deviation. The larger the value, the greater the theoretically expected power reduction required, which exceeds the minimum forming power threshold needed to maintain the physical solid shape. The more;

[0109] Volumetric energy density conversion factor Compared with the expected power reduction value There is a positive correlation. In physical processing, there are physical differences in the absorbance and enthalpy of melting of different grades of metal powders. (Volume energy density conversion factor) The dimensionless local path thermal interference warning index is mapped to physical output energy in the watt (W) range. Its positive correlation design ensures that as the warning index increases, the attenuation command fed back to the laser can proportionally and realistically strip away the corresponding physical residual heat energy.

[0110] Target Line Energy Density Benchmark With dynamic speed compensation increment They show a negative correlation. This represents the fixed power truncation difference returned by the hardware deadlock. The target linear energy density benchmark that the processed material can withstand. The larger the value, the smaller the additional physical displacement velocity required by the galvanometer deflection motor. By setting this benchmark parameter as the denominator of the cross-dimensional compensation mapping and generating negative correlation constraint results, it is ensured that the compensation action is controlled within the heat absorption limit of the metal material, thus preventing secondary runaway and system crashes caused by the control center calculating excessive speed.

[0111] This embodiment relies on a dual-beam selective laser melting solid processing monitoring platform. This platform connects to an RTC series multi-axis galvanometer control card and a full-frame molten pool feature physical camera assembly via an EtherCAT industrial bus. The platform is configured to perform the forming task of multi-laser concurrent large-area dense solids intersecting with thin-walled edge contours. This condition exhibits typical characteristics of strong constraints and high-frequency thermal field interference. The physical boundary reference parameters of the solid testing platform are set as follows:

[0112] Molten Safety Extreme =85; Current basic physical output power of the laser emitter =200W; Minimum forming power threshold =150W; Volumetric energy density conversion factor =1.8; Energy decoupling allocation ratio =0.7; Target linear energy density benchmark =0.3J / mm; Macroscopic diameter parameter of light spot =0.08mm; Current base scanning rate of the galvanometer =1000mm / s; Maximum rated scanning rate of the galvanometer =3000mm / s; In this embodiment, the monitoring node continuously injects a constantly increasing local path thermal interference early warning index into the central control unit. This induces the control system to cross a slight interference boundary from normal stable operation, until it breaks through the lower power limit deadlock line of 150W, and finally touches the maximum physical speed limit of 3000mm / s for the galvanometer motor. By capturing the linkage evolution values ​​of various physical parameters within the bus, the execution capability of this scheme in cross-dimensional interference resolution under electromechanical constraints is demonstrated, as shown in the table below:

[0113] Table 1: Energy decoupling and limit compensation calculation data under a multi-beam concurrent strongly constrained solid contour scanning platform

[0114]

[0115] Define the dynamic velocity compensation increment actually executed by the galvanometer control card after entering the extreme deadlock state. This is then compared sequentially with the target line energy density reference and the macroscopic diameter parameter of the light spot. Perform continuous scalar multiplication to calculate the absolute energy dissipated by the physical platform through acceleration; then calculate the difference between this absolute energy value and the power cutoff value intercepted at the light source. A division correlation operation is performed to map out a percentage value representing the absorption efficiency. The minimum value filtering logic is invoked to compare this percentage calculation result with a 100% upper limit constraint benchmark and extract the minimum value as the final spillover rate. This spillover rate is forcibly reset to zero when the target velocity exceeds the galvanometer limit and triggers hardware-level interception. This parameter represents the percentage of successful recovery and dissipation of the destructive high-temperature load from the physical powder bed by squeezing the remaining mechanical potential of the servo motor across dimensions when encountering a physical electromechanical impasse where the laser output power can be reduced to its lowest possible level.

[0116] In scenarios 2 and 3, the control platform is based on the energy decoupling allocation ratio. The set leverage constant of 0.7 distributes the physical thermal field reduction index of 9W or 45W, which would normally be sufficient to cause molding fracture when applied unidirectionally to a single piece of hardware, across the light source and servo ends; at this point, the power cutoff difference... Strictly maintain 0W, and the processing command flow will be smooth.

[0117] This solution has been definitively validated in scenarios 4 and 5: Local path thermal interference early warning index When the voltage climbed to 125 and continued to jump to 135, the theoretically expected power attenuation parameters predicted by the system's internal logic reached 50.4W and 63W respectively. Since the maximum safe reduction allowed by the current physical hardware is only 50W, this solution triggers an over-limit compensation branch at this node, firmly locking the actual attenuation command at 50W, and directly extracting the unresolved 0.4W and 13W truncation difference into the cross-dimensional compensation link; under the strict premise of maintaining the original coordinate scanning path in scenario 5, the servo motor is driven to add a compensation increment of 541.7mm / s, successfully pushing the overall operating composite speed to 2666.7mm / s.

[0118] Figure 1 This diagram presents the mapping mechanism and step architecture of multi-laser collaborative optimization in the additive manufacturing control system. The physical powder bed on the left, the multi-channel laser galvanometer execution unit above, and the central digital control motherboard on the side correspond to the hardware execution environment in step S1 for acquiring the initial scanning trajectory flow and extracting the laser scanning synchronization control cycle. The three-dimensional logic grid superimposed on the upper layer of the powder bed and its internal staggered columnar totems map the trajectory to the forming surface grid in step S2, performing numerical attenuation and extracting the global heat prediction spectrum matrix state of the "local path thermal interference early warning index". The technical roadmap on the right, containing four rectangles, summarizes the core elements of the invention and ultimately leads to step S3, demonstrating the complete closed loop of encapsulating logical reconstruction data into control commands to drive the safe collaborative processing of the physical laser and servo motor.

[0119] Figure 2 This represents the generation mechanism of the global heat prediction map matrix. Specifically, through the flow between panels A and B, it clarifies how, within the two-dimensional discrete array defined within the additive manufacturing control system, the preset heat weights are directly applied to the logical coordinate units physically hit by the initial scan trajectory flow in the instantaneous coordinates. The numerical assignment and superposition operation. Panel C mapping responds to the laser scanning synchronization control cycle. Periodic triggering, utilizing a preset material cooling scaling factor. A global synchronous subtraction dimensionality reduction process is performed on all logical coordinate units carrying the existing heat value, thereby establishing the iteratively updated logical grid output in the final diagram as the global heat prediction map matrix.

[0120] Figure 3 This represents the adaptive reconfiguration scheduling logic for weakly constrained routes. Specifically, the interception breakpoints on the left side of the screen clearly define the response index to the local path thermal interference warning. Exceeding the preset melt safety limit The determination state is then mapped. Based on this, it maps how the control system identifies single-grid waste heat assessment scores below a certain threshold from the global heat prediction map matrix when the molding process feature label is identified as a weak constraint identifier representing the internal solid filling path. The cold zone mesh set. The algorithm derivation area on the right shows the specific calculation process: extracting the relative displacement distance of nodes and performing spatial scaling with the maximum scan stroke length to extract the normalized path offset coefficient. Reading the waste heat assessment score and mapping it to the environmental noise floor reveals a dimensionless heat avoidance weight. And it utilizes dynamic optimization to allocate weights based on the fill density rate configuration of local slice features. The comprehensive cost score of the candidate node is calculated by performing a linear weighted summation on the above coefficients. Finally, by comparing the extreme values ​​of the numerical sequence, the candidate processing landing point sequence of the target is established and the target reconstruction path data is generated.

[0121] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, min-max-normalization and Z-score standardization.

[0122] The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random-access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of this invention.

[0123] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for optimizing multiple laser scanning paths in metal additive manufacturing, characterized in that, The specific steps include: Step S1: Obtain the initial scanning trajectory flow of the multi-channel laser galvanometer execution unit; synchronously obtain the hardware state control timing of the multi-channel laser galvanometer execution unit, and extract the laser scanning synchronization control cycle for controlling the synchronization of laser scanning processing based on the hardware state control timing; Step S2: Map the initial scanning trajectory stream to the preset forming surface logic grid, perform thermal state value superposition, and combine the laser scanning synchronous control cycle to perform time-series numerical attenuation calculation to extract the global thermal prediction spectrum matrix characterizing the residual heat accumulation state of the forming surface in the powder bed region; The initial scanning trajectory stream to be issued is subjected to discrete grid feature mapping and state indexing with the global heat prediction map matrix to extract the local path thermal interference early warning index used to characterize the degree of convergence and accumulation of the current processing heat prediction. In response to the determination that the local path thermal interference early warning index exceeds the preset melting safety limit, the molding process feature tags associated with the intercepted trajectory are extracted, and adaptive reconstruction scheduling is performed based on the molding process feature tags and the global heat prediction map matrix to calculate the target reconstruction path data. Step S3: Based on the target reconstruction path data, generate a scanning path reconstruction control command for the multi-channel laser galvanometer execution unit. The scanning path reconstruction control command is configured to drive the multi-channel laser galvanometer execution unit to perform the reconstructed multi-channel laser collaborative scanning processing action.

2. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 1, characterized in that: The acquisition of the initial scanning trajectory stream of the multi-channel laser galvanometer execution unit includes: Obtain the pre-parsed slice scan trajectory data stream to be executed; Spatial coordinate extraction is performed on the slice scan trajectory data stream to be executed, and a set of two-dimensional transient coordinate time series representing the spatial transient displacement vectors of the endpoints of the multi-beam lasers is extracted; The set of two-dimensional transient coordinate time series is established as the initial scan trajectory flow.

3. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 2, characterized in that: The extraction of the laser scanning synchronization control cycle for controlling the synchronization of laser scanning processing includes: The reference control cycle sequence for scanning of the multi-channel laser corresponding to the multi-channel laser galvanometer execution unit is read synchronously. The reference control cycle sequence is converted into a globally unified digital timestamp sequence, which serves as the laser scanning synchronization control beat that triggers the periodic numerical decay of the global heat prediction spectrum matrix.

4. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 3, characterized in that: A global heat prediction map matrix characterizing the residual heat accumulation state on the forming surface of the powder bed region is extracted; and a local path thermal interference early warning index characterizing the degree of heat convergence and accumulation in the current processing is extracted, including: The preset heat weight is assigned and superimposed to the logical coordinate units in the preset forming surface logical grid that are hit by the initial scanning trajectory flow. In response to the periodic triggering of the laser scanning synchronization control cycle, a global synchronous subtraction dimensionality reduction process is performed on all logical coordinate units in the preset forming surface logical grid that carry the stored heat score using a preset material cooling scaling factor. The iteratively updated preset forming surface logical grid is established as the global heat prediction map matrix. Extract the prospective scanning path segment of the initial scanning trajectory stream to be issued within the preset future processing period, and obtain the current stock evaluation score of each grid cell covered by the prospective scanning path segment in the global heat prediction map matrix. Perform a discrete grid mean aggregation operation along the look-ahead scanning path segment on the current stock assessment score to calculate the characteristic value that reflects the risk of local thermal field energy accumulation and convergence, and configure the characteristic value as the local path thermal interference early warning index.

5. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 4, characterized in that: Extracting the forming process feature labels associated with the intercepted trajectory, and calculating the target reconstruction path data used to characterize the laser avoidance direction, including: Based on the preset slicing process feature mapping rules, the local structural attributes of the intercepted trajectory in the current three-dimensional molded part are identified, and the molding process feature label is generated. The molding process feature label includes a path weak constraint identifier for characterizing the internal solid filling and a path strong constraint identifier for characterizing the external molding contour. From the global heat prediction map matrix, identify and extract candidate grid sets whose single grid waste heat assessment scores are lower than the preset melting safety limit, and establish them as cold zone grid sets; In response to determining that the forming process feature label is identified as the path weak constraint identifier, the spatial distance and thermal state between the current processing endpoint of the intercepted trajectory and the cold zone grid set are evaluated, and the candidate processing landing point sequence with the shortest detour distance and in a cooling state is selected to generate the target reconstruction path data.

6. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 5, characterized in that: Based on the target reconstruction path data, a scan path reconstruction control command is generated for the multi-channel laser galvanometer execution unit, including: Extract the path displacement deviation between the target reconstructed path data and the initial arrangement sequence contained in the initial scan trajectory stream; Based on the path displacement deviation, trajectory rerouting information is generated to update the scanning task queue; The trajectory rerouting information is encapsulated into the scanning path reconstruction control command to drive the multi-channel laser galvanometer execution unit to avoid the local thermal field interference region and transfer to the position corresponding to the target reconstruction path data to perform processing.

7. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 6, characterized in that: In response to the determination that the forming process feature label is identified as the path strong constraint label, the method also triggers a secondary control closed loop in parallel with spatial interleaving scheduling, specifically including: The original spatial scanning path of the intercepted trajectory remains unchanged; Calculate the thermal overflow deviation between the local path thermal interference early warning index and the melting safety extreme value; Based on the thermal field overflow deviation value, a dynamic compensation instruction for process parameters is generated for the multi-channel laser galvanometer execution unit. The dynamic compensation instruction for process parameters is configured to trigger the laser emitter to dynamically reduce the laser output power or dynamically increase the scanning movement speed. The original spatial scanning path that remains unchanged is synchronously packaged with the dynamic compensation instruction for process parameters and established as the target reconstruction path data under the constraint condition for use in step S3.

8. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 7, characterized in that: Assess the spatial distance and thermal state between the current processing endpoint of the intercepted trajectory and the cold zone grid set, and filter out the candidate processing landing point sequence with the shortest detour distance and in a cooled state, including: For each candidate mesh node in the cold zone mesh set, extract the relative displacement distance between the node and the current processing endpoint. Based on the spatial ratio mapping relationship between the relative displacement distance and the maximum scanning stroke length of the current forming surface, extract the dimensionless normalized path offset coefficient. Read the single-grid waste heat assessment score corresponding to each candidate grid node, and perform a division comparison between the single-grid waste heat assessment score and the preset heat dissipation benchmark reference value of the molding chamber to map out the dimensionless heat avoidance weight used to characterize the degree of local heat accumulation. Extract the local feature fill density rate of the slice at the logical slice level of the current 3D molded part, and configure the local feature fill density rate of the slice as a dynamically optimized weight allocation. Using the dynamic optimization allocation weight, a linear weighted summation calculation is performed on the normalized path offset coefficient and the dimensionless heat avoidance weight to calculate the candidate node comprehensive cost score, which is used to characterize the comprehensive level of physical transfer efficiency and processing safety of the current node. In the cold zone grid set, the numerical sequence extreme value comparison is performed on the comprehensive cost score of all candidate grid nodes, and the target candidate grid nodes that meet the minimum value optimization condition are established as the candidate processing landing point sequence.

9. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 8, characterized in that: Calculate the thermal overflow deviation between the local path thermal interference early warning index and the melting safety extreme value; Based on the thermal overflow deviation value, dynamic compensation instructions for process parameters of the multi-channel laser galvanometer execution unit are generated, including: The thermal field overflow deviation value is linearly mapped to the preset volumetric energy density conversion coefficient to calculate the expected power reduction value used to characterize the redundant heat to be stripped from the local thermal field. The expected power reduction value is input into the pre-configured process parameter parallel decoupling module to calculate the corresponding expected power attenuation control parameters and expected scan speed-up control parameters respectively. Extract the preset minimum forming power threshold corresponding to the bottom layer of the multi-channel laser emitter, and extract the maximum rated scanning rate of the galvanometer at the bottom layer of the multi-channel laser galvanometer execution unit; In response to the determination that the expected power attenuation control parameter is not lower than the preset minimum forming power threshold and the expected scanning speed-up control parameter does not exceed the maximum rated scanning rate of the galvanometer, the expected power attenuation control parameter and the expected scanning speed-up control parameter are established as safe process execution parameters, and they are combined, packaged, and output as the process parameter dynamic compensation command.

10. The method for optimizing multiple laser scanning paths in metal additive manufacturing according to claim 9, characterized in that: The method also includes hardware boundary over-limit protection and parameter compensation control logic for extreme value triggering of CNC servos, specifically including: In response to determining that the expected power attenuation control parameter is lower than the preset minimum forming power threshold, a power lower limit interception branch is triggered to limit the output laser power control parameter to the preset minimum forming power threshold. Calculate the power cut-off difference between the theoretical down-adjustment power corresponding to the expected power reduction value and the preset minimum forming power threshold, and call the preset energy-velocity equivalent compensation conversion formula to convert the power cut-off difference into a dynamic velocity compensation increment to be superimposed on the current scanning trajectory. Verify whether the sum of the dynamic speed compensation increment and the expected scanning speed-up control parameter is less than or equal to the maximum rated scanning rate of the galvanometer. After the verification is passed, generate the reconstructed target scanning speed. Finally, perform synchronous packaging and assembly of the limited actual output power parameter and the reconstructed target scanning speed to replace the generation of the dynamic compensation instruction for the process parameters.

Images