A multi-laser SLM layer vector data multi-source collaborative dynamic allocation method

Abstract

The present application relates to the field of manufacturing and production process management, in particular to a multi-laser SLM layer vector data multi-source collaborative dynamic allocation method. The method comprises: according to a production order containing business constraints, processing a three-dimensional model to generate a hierarchical process topology set carrying process rules; fusing the static attributes of the topology set and the real-time process state to generate a dynamic collaborative matching matrix, and dynamically scheduling the multi-laser production resources according to the matrix; by comparing the real-time process state with a preset space-time process potential field diagram, calling a correction strategy, and realizing closed-loop management of the manufacturing task. The present application solves the defect that the static allocation strategy in the background technology cannot cope with dynamic working conditions by constructing a dynamic decision model, realizes intelligent optimization configuration of production resources under the condition of meeting the business constraints of the order, and improves the quality, efficiency and cost-effectiveness of complex component manufacturing.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of manufacturing and production process management, in particular to a multi-laser SLM layer vector data multi-source collaborative dynamic allocation method. BACKGROUND

[0002] Under the background of digital transformation of high-end manufacturing, multi-laser selective laser melting (SLM) equipment as a key flexible production resource, its efficient operation management and scheduling is the core to realize the maximization of investment return. However, the current mainstream manufacturing execution system or enterprise resource planning system, when planning production tasks and scheduling resources for such additive manufacturing equipment, still generally uses the traditional linear production model based on static work order.

[0003] The core defect of such traditional scheduling system lies in the limitation of its information processing architecture. The decision logic of the system is only based on the static geometric data in the production order and the preset fixed working hours, which is essentially an open-loop management mode with poor information. It lacks the ability to collect, process and fuse the massive and heterogeneous data streams generated by multi-laser equipment during task execution, including real-time sensor data reflecting the physical state of the equipment, dynamic cost data representing resource consumption, and scheduling conflict data affecting delivery cycle.

[0004] Due to the inability to integrate these multi-source dynamic data into the decision-making cycle, a series of management and business problems have been caused. First, the production cost becomes uncontrollable, the system cannot perform dynamic cost accounting according to real-time energy consumption and potential waste risk, making it difficult to guarantee project profit margins. Second, the production cycle becomes unreliable, unable to dynamically rearrange schedules in response to changes in device state or production line bottlenecks, resulting in frequent violations of delivery cycle commitments in service level agreements. Finally, this low operational management efficiency and business risk prediction ability makes it impossible for enterprises to provide accurate dynamic pricing and reliable delivery guarantees for complex, high-value manufacturing services, thus losing their core advantage in fierce market competition.

[0005] Therefore, a multi-laser SLM layer vector data multi-source collaborative dynamic allocation method is proposed. SUMMARY

[0006] The purpose of the present application is to provide a multi-laser SLM layer vector data multi-source collaborative dynamic allocation method, aiming to realize dynamic allocation and closed-loop risk control of laser resources in the manufacturing process.

[0007] To achieve the above purpose, the present application provides the following technical solutions:

[0008] A multi-laser SLM layer vector data multi-source collaborative dynamic allocation method, comprising:

[0009] According to a manufacturing task comprising a three-dimensional model and associated production order information, the three-dimensional model is sliced to generate initial layer vector data, multi-dimensional feature processing is performed by a vector feature engine to generate a hierarchical process topology set carrying independent process rules of vector objects;

[0010] Static attributes of each vector object in the hierarchical process topology set are fused with real-time process state data, a matching degree between each object to be processed and idle production resources is quantitatively calculated to generate a dynamic collaborative matching matrix, and multi-laser production resources are dynamically scheduled according to an optimal matching value in the dynamic collaborative matching matrix;

[0011] A space-time process potential field map is established through offline simulation and historical process data analysis, and the space-time process potential field map is used to define an ideal process state of each spatial position of a part at a processing time and associate a process strategy for correcting state deviation;

[0012] A processing task assigned is executed, real-time process states are compared with ideal states defined by the space-time process potential field map, a correction strategy is called, and closed-loop management and control of the manufacturing task are realized on the premise of meeting cost budget and delivery cycle.

[0013] Preferably, the step of multi-dimensional feature processing by the vector feature engine includes: geometric and topological analysis of the initial layer vector data, identification and marking of vector clusters representing key shapes of a part; according to the results of the geometric and topological analysis, and in combination with quality levels defined in the production order information, the vector clusters are classified and constructed into vector objects of different levels. The vector objects of different levels include: core entity objects for efficient filling, fine contour objects for ensuring dimensional accuracy and surface quality, and multi-laser overlapping objects for ensuring metallurgical bonding strength, thereby constituting the hierarchical process topology set.

[0014] Preferably, the independent process rules carried by the vector objects cover multiple dimensions, including: geometric constraints for ensuring forming accuracy, the geometric constraints including limits of wall thickness and hole diameter; and process parameter constraints for guiding energy input, the process parameter constraints including preset ranges of scanning speed and laser power.

[0015] Preferably, the real-time process state data is derived from a multi-modal sensing system, and the sensing system includes: a visual sensing unit for identifying molten pool morphology and powder bed geometric abnormalities, and a thermal flow sensing unit for monitoring molten pool temperature and thermal field distribution.

[0016] Preferably, the method for constructing the dynamic collaborative matching matrix is: for the pairing between the vector object to be processed and the idle production resource, a quantitative matching degree is calculated according to a multi-objective function; the multi-objective function weightedly combines the following indexes: a risk index for evaluating the quantitative value of the risk of local heat accumulation, thermal stress concentration and splashing interference process caused after the production resource is assigned to the vector object; a benefit index for evaluating the contribution to improving the metallurgical quality of the lap joint area and ensuring the key quality target of fine profile forming accuracy after the index is executed; and a cost index for evaluating the time cost and energy consumption cost required for the production resource to move from the current position to the starting point of the vector object.

[0017] Preferably, in order to reflect the process priority of different regions, the weight of the multi-objective function is dynamically adjusted according to the level of the vector object: for the multi-laser lap joint object, the weight of the benefit index is set to the highest; for the fine profile object, the weight of the benefit index is preferentially considered; and for the core entity object, the cost index for improving the processing efficiency is preferentially considered.

[0018] Preferably, the step of dynamic scheduling is a continuous iterative loop process, and the steps are: at the beginning of the scheduling period, the dynamic collaborative matching matrix is generated; for the current idle laser, the corresponding optimal matching vector object is found from the matrix; the processing task of the vector object is assigned to the corresponding laser; when the laser completes the assigned task, the state of the laser becomes idle, and the processing state of the part is updated, and the system immediately triggers the scheduling period until all the vector objects of the current layer are processed.

[0019] Preferably, the method for establishing the space-time process potential field map comprises: performing finite element meshing on the three-dimensional model and performing thermal-structural coupling simulation to obtain theoretical distribution data of temperature field, stress field and phase change history of the part at any space-time node in the entire processing process under ideal process conditions; extracting case data from a historical database, the part material type, the preset quality level and the geometric complexity classification recorded in the case data are completely consistent with the current manufacturing task, and the case data includes complete sensor process records and final quality detection results; fusing and calibrating the theoretical distribution data and the case data to generate the space-time process potential field map; and the potential field map is used to define the ideal process state that should be reached by each spatial position of the part at the processing time in a four-dimensional data structure.

[0020] Preferably, the step of calling the correction strategy specifically comprises: when the real-time process state deviates from the ideal state defined by the spatiotemporal process potential field map, the system selects and executes a correction action from a pre-set strategy library, the correction action includes: adjusting the power, scanning speed and spot size of the currently working laser in real time without changing the scanning path; re-planning the scanning order and scanning strategy of the subsequent vector objects to be processed in the affected area; temporarily changing the task assignment to idle lasers to prioritize key compensation tasks generated due to state deviation.

[0021] Preferably, the step of closed-loop management and control includes: before executing the correction strategy, the system estimates the impact of the strategy on the total processing time and potential material loss; compares the estimated impact with the total budget and total working hours initially set for the manufacturing task; only when the estimated total cost and total time after correction are still within the allowed floating range, the correction strategy is executed; if it exceeds the range, the system will mark the deviation as a high-risk event and prompt manual intervention.

[0022] Compared with the prior art, the application has the following beneficial effects:

[0023] 1. The application constructs a hierarchical process topology set containing core entities, fine contours and overlapping areas through a vector feature engine, so that the data not only carries geometric information, but also predefines independent process rules and targets. This fundamentally overcomes the defects in the background art of relying only on discrete and static geometric information for decision-making, and improves the dimension of data processing from a single geometric path to a deep definition of manufacturing intent and physical process, providing a high-quality and high-dimensional decision basis for subsequent dynamic scheduling and closed-loop control.

[0024] 2. The application establishes a spatiotemporal process potential field map as an ideal state benchmark, and constructs a dynamic collaborative matching matrix with risk, benefit and cost as indicators, transforming resource allocation from passive and rigid task assignment in the background art to proactive risk assessment and strategy generation. It can quantitatively assess local heat accumulation and other risks before task assignment, and improve quality control from a post-detection remediation mode to an in-process intervention and pre-prevention level, significantly enhancing the process robustness and yield of complex component manufacturing.

[0025] 3. The present application integrates the above-mentioned contents to construct a full-process automatic control method of "data definition (topology set) → state perception (sensing data) → ideal state comparison (potential field map) → optimization decision (matching matrix) → closed-loop execution (correction strategy)". The method realizes the optimal collaborative deployment of multiple laser resources in the time and space dimensions under the constraints of cost and cycle through continuous iterative scheduling cycles and hierarchical correction strategies. This enables the present application to have self-adaptive ability to dynamic and nonlinear manufacturing environment, and fundamentally solves the defect that efficiency and quality are difficult to balance in the background technology. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A multi-laser SLM layer vector data multi-source collaborative dynamic allocation method structure schematic diagram of the present application;

[0027] Figure 2 A multi-laser SLM layer vector data multi-source collaborative dynamic allocation method flowchart of the present application;

[0028] Figure 3 A dynamic collaborative matching matrix construction and decision diagram of a multi-laser SLM layer vector data multi-source collaborative dynamic allocation method of the present application. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] Embodiment one:

[0031] The present embodiment provides a specific application of the multi-laser SLM layer vector data multi-source collaborative dynamic allocation method. The technical solution of the present application is applied to a large-format SLM additive manufacturing equipment with four laser and galvanometer systems. The application scenario of the present embodiment is to print a large, thin-walled, and complex lattice structure-containing aerospace-grade titanium alloy rectifier case. The difficulty of this manufacturing task lies in that the thin-walled lattice structure is extremely sensitive to heat accumulation and thermal stress, and is prone to deformation or cracking, and the huge part size requires multiple lasers to work efficiently to ensure a reasonable production cycle, which puts high requirements on the intelligent level of task allocation.

[0032] Reference Figure 1 and Figure 2The method of the embodiment first generates a hierarchical process topology set through a vector feature engine according to production order information contained in a manufacturing task and a three-dimensional model; meanwhile, the method acquires real-time process state data through a multi-modal sensing system and establishes a space-time process potential field map for guiding the machining process through offline simulation and historical data analysis.

[0033] Further, when preprocessing a three-dimensional model of a large aerospace structure, the step of multi-dimensional feature processing performed by the vector feature engine is specifically as follows: first, geometric and topological analysis is performed on initial layer vector data to identify and mark vector clusters representing key shapes of the part; second, according to the analysis results and in combination with the “astronautical grade” quality level defined in the production order, the vector clusters are classified and constructed into vector objects of different levels, including core entity objects for achieving efficient filling, fine contour objects for ensuring dimensional accuracy and surface quality, and multi-laser overlapping objects for ensuring metallurgical bonding strength, thereby constituting the hierarchical process topology set.

[0034] In a specific implementation of the embodiment, the step of geometric and topological analysis performed by the vector feature engine is specifically as follows: a k-d tree data structure is constructed to spatially index endpoints of the vector data to quickly query neighboring vectors within a specified radius, thereby determining the spatial adjacency relationship. The step of classifying and constructing vector clusters into vector objects of different levels is implemented through a preset rule set, for example: when a vector cluster has a compact filling degree greater than a threshold A (for example, 80%) and is not adjacent to the outer boundary of the part, it is classified as a core entity object; when it constitutes a closed loop and has an aspect ratio greater than a threshold B, it is classified as a fine contour object.

[0035] The embodiment enables the system to execute completely different manufacturing philosophies within one layer for different geometric regions of the part by differentiating the definition of process objects. Instead of using single, compromised process parameters, it can assign completely different process objectives to entity parts requiring high efficiency and fine parts requiring high quality. This ability to decompose objectives based on regional functions is an innovative approach to fundamentally resolving the core contradiction between efficiency and quality of large complex parts.

[0036] The data implementation of the hierarchical process topology set also includes:

[0037] The hierarchical process topology set is constructed as a graph data structure, in which the vector objects are nodes and the relationships between the objects are edges connecting the nodes. The topology relationships include two types: the first type is a spatial adjacency relationship, in which an adjacency edge is established between the corresponding nodes when the distance between two vector objects in the physical space is less than a preset threshold; the second type is a process dependency relationship, in which a directed dependency edge is established between the corresponding nodes when the processing quality of an object is a prerequisite for the successful processing of another object. These topology relationships are calculated and stored when the vector feature engine processes the initial data. In the scenario of printing the aerospace structure, when the vector feature engine processes the data of a layer, it not only classifies the vectors into "core", "fine contour", and other objects, but also constructs the topology graph between them. For example, for a fine reinforcing rib structure, tight "spatial adjacency" edges are established between all the vector object nodes constituting the structure. More importantly, for a "multi-laser lap joint object", the system analyzes the thermal influence between it and the adjacent area objects, thereby establishing a "process dependency" edge, which may record that the lap joint processing needs to wait for the adjacent area to cool down to below 800K before the optimal metallurgical bonding can be obtained. This complete topology graph is an important input for subsequent scheduling decisions.

[0038] The digitalization of the relationship of one vector object relative to another vector object into a process topology graph enables the decision system to understand the structural connectivity of the part and the internal correlation of the process from a macro perspective, rather than facing individual isolated processing tasks. This deep understanding of the context enables the scheduling decision to change from pursuing local optimization to pursuing global optimization.

[0039] Further, in the above step, the independent process rules of the vector objects are respectively set from multiple dimensions. Among them, the geometric constraint of the fine contour object specifies the lower limit of the minimum wall thickness (for example, 0.5 mm), the process parameter constraint of the fine contour object specifies a narrow preset range of scanning speed and laser power to maintain stable energy input, and the process parameter constraint of the core entity object sets a wider preset range to achieve efficiency.

[0040] The embodiment converts the abstract process specifications and quality requirements into machine-readable and executable rigid constraints by embedding independent process rules in the vector objects, so that the manufacturing intent can be transmitted from the design end to the execution end without loss, and ensures that all subsequent automatic decisions do not exceed the predefined process bottom line. In this way, the starting point of quality control is moved from passive monitoring in the manufacturing process to active definition in data preparation, greatly enhancing the stability and predictability of the process.

[0041] Further, the real-time process state data is generated by the multi-modal sensing system during the printing process. The visual sensing unit determines the morphology of the molten pool and geometric abnormalities such as scratches and shortages on the powder bed surface in real time through a high-speed camera. The thermal flow sensing unit determines the temperature of the molten pool, the cooling rate, the surrounding thermal field distribution and other dynamic conditions in real time through a pyrometer. These different modal data are fused to form a complete state description of the current manufacturing process.

[0042] The embodiment fuses multi-modal sensing data to construct a process perception system that is more robust and has richer information dimensions than a single sensor. It can cross-verify information from different physical dimensions, effectively filter noise or false positives of a single sensor, form a high-confidence comprehensive judgment of the molten pool state, and greatly improve the perception accuracy of the system for complex physical phenomena.

[0043] The step of fusing real-time process state data specifically includes a multi-modal data fusion and feature engineering process. The process first aligns the original pixel matrix from the visual sensing unit and the temperature matrix from the thermal flow sensing unit in milliseconds in space and time to form a unified multi-layer data graph. Then, the feature extraction module calculates a set of structured and low-dimensional feature vectors from the data graph in real time, including morphological features for characterizing the stability of the molten pool, thermodynamic features for characterizing the thermal cycle history, and abnormal features for characterizing the process environment. The final output feature vector is the real-time process state data used for quantitative calculation of matching degree. In the actual printing process of the machine case, when a laser is scanning, the high-speed camera and thermal imaging camera are recording the dynamic process of the molten pool synchronously. The feature engineering process in the embodiment runs in real time on the edge computing node, which quickly processes and converts hundreds of megabytes of raw video streams generated per second into a concise feature vector. This The feature vector containing rich physical connotations, rather than massive raw pixel data, is used to update the risk indicators related to the area in the dynamic cooperative matching matrix in real time, thereby realizing efficient real-time decision-making.

[0044] The embodiment successfully builds a bridge between a large number of disorganized original sensing signals and decision factors with low dimensions but extremely dense information by introducing multi-modal data fusion and feature engineering processes. It can convert complex and difficult physical phenomena into operation intelligence that machines can understand and apply to calculations. In this way, not only does the closed-loop control based on real-time feedback achieve calculation level, but also the subsequent scheduling decisions are greatly improved in accuracy and robustness due to the refinement and noise reduction at the data source.

[0045] Further, with reference to Figure 3 In the process of carrying out the printing task, for the construction method of the dynamic collaborative matching matrix, in a scheduling period, the system needs to select an optimal task for the No. 2 laser which is in an idle state from a plurality of vector objects to be processed. One of the candidate objects is a core entity object (A) located in the center area of the part, and the other is a fine contour object (B) located in the thin-walled lattice structure. The system calculates a quantitative matching degree for the two pairs of (2# laser-A object) and (2# laser-B object) according to a preset multi-objective function, respectively. In the calculation of the matching degree, the multi-objective function is quantified as follows:

[0046] The risk index is obtained by analyzing the real-time thermal field distribution map of the region where the object to be processed is located, extracting the average temperature, maximum temperature gradient and other characteristics of the region, and inputting these characteristics into a preset risk evaluation function to calculate a normalized risk score. Higher average temperature and temperature gradient correspond to higher risk score.

[0047] The benefit index is obtained by reading the static attributes carried by the object to be processed, especially its level and preset process sensitivity, and mapping them into a preset benefit lookup table to obtain a quantitative benefit score. Objects with higher level and higher process sensitivity correspond to higher benefit score.

[0048] The cost index is directly calculated according to the current physical position of the idle production resource and the starting position of the object to be processed, and the Euclidean distance required for the laser galvanometer to jump is converted into a quantitative time cost score.

[0049] Finally, through operation logic, the above-mentioned benefit score is positively weighted, while the risk score and the cost score are negatively weighted, so as to calculate the final quantitative matching degree.

[0050] This embodiment converts the complex scheduling problem into a clear and quantifiable multi-objective optimization task, which discards the rigid allocation logic according to fixed rules or region division in the background technology, so that the system can dynamically and comprehensively consider the comprehensive situation of each potential decision in the three dimensions of risk, quality and cost, and then make intelligent resource allocation that meets the optimal benefit at present.

[0051] Further, to embody the process priority of different regions, the weight of the multi-objective function is dynamically adjusted according to the level of the vector object. The adjustment mechanism is realized by a preset weight lookup table, which establishes the mapping relationship between the level of the vector object and its corresponding weight coefficient vector. For example, when the system evaluates the fine contour object, the preset quality priority weight vector is called from the lookup table, which can be set to 0.7 for the benefit index weight, 0.15 for the risk index weight, and 0.15 for the cost index weight; when the core entity object is evaluated, the efficiency priority weight vector is called, which can be set to 0.2 for the benefit index weight, 0.2 for the risk index weight, and 0.6 for the cost index weight. When performing weighted operation, the system first obtains the level of the object to be evaluated, and retrieves the corresponding weight vector from the lookup table to complete the calculation.

[0052] The embodiment introduces a mechanism of dynamic weight adjustment, which tightly couples the manufacturing strategy at a high level and the mathematical optimization at a bottom level. In this way, the decision-making behavior of the system is no longer around a single target, but can automatically switch between different modes such as quality priority and efficiency priority according to the functional importance of the current processing region, thereby giving the system a high degree of decision-making flexibility.

[0053] Further, the steps involved in dynamic scheduling are a cyclic process in a continuous iteration state. In the scenario just mentioned, the system arranges the B object with the optimal matching degree to the No. 2 laser for processing according to the result calculated by the dynamic cooperative matching matrix. After several hundred milliseconds, the No. 4 laser in another region of the part completes its original task and its state changes to idle. At this moment, the system starts a new scheduling cycle: it first updates the actual processing state of the entire part, and then reconstructs a new dynamic cooperative matching matrix for all remaining objects to be processed and all idle lasers based on the new and real-time state, and finds the optimal task option for the No. 4 laser. This cycle of "complete task operation → update actual state → reconstruct matrix → find optimal scheme → assign new task" is continuously carried out at a high frequency until all vector objects at the current level are completed.

[0054] The embodiment uses the way of continuous iterative scheduling cycle to ensure that each allocation decision is made based on the latest and most accurate system state. The mechanism of frequent re-decision is an important embodiment of the dynamic nature of the application, which can respond to any subtle changes in the process, actively capture fleeting process opportunities, and quickly re-plan for unexpected exceptions, ensuring that the entire manufacturing process is always in the best state.

[0055] Further, before the printing task officially starts, the method for establishing the space-time process potential field map is as follows: first, finite element meshing is performed on the three-dimensional model of the aerospace structure, and thermal-structural coupling simulation is performed, so as to obtain the theoretical temperature field and stress field distribution of the part at any space-time node in the entire machining process under ideal process conditions; then, successful case data matching the material type, quality level and geometric complexity of the current part are extracted from the historical database; the theoretical simulation data and the real process data from the successful cases are fused and calibrated to generate the final space-time process potential field map.

[0056] Further, during the execution of the scheduled machining task, the specific operation of the step of calling the correction strategy is as follows: if there is a deviation between the real-time process state and the ideal state defined by the space-time process potential field map, the system will immediately select and execute the matching correction action from the pre-set strategy library. For example, if the local temperature exceeds the ideal value, the system will perform a real-time adjustment of the current laser power correction action; if the deviation is large, it may perform a temporary change in the task assignment of other idle lasers, prioritize key compensation tasks, and perform higher-level correction actions.

[0057] Further, the closed-loop management and control specifically involves the following steps: when the aforementioned correction strategy is executed, the system estimates the impact of the strategy on the total machining time and potential material loss, and then compares the estimated situation with the total budget and total working hours set by the manufacturing task initially. Only when the estimated total cost and total time after correction are within the allowed floating interval can the system execute the correction strategy. If it is out of this range, the system will mark the deviation as a high-risk event and immediately send an alarm to the operator to remind him to carry out manual intervention.

[0058] The embodiment integrates data definition of hierarchical process topology set, optimization decision of dynamic collaborative matching matrix and closed-loop execution based on space-time process potential field map, and constructs a full-process automatic control method. The method overcomes the defects caused by the staticity of the decision model and the unpredictability of the process risk in the background art, and provides a key technical support for realizing the unified optimization of quality, efficiency and cost benefit under the order business constraints in a dynamic and nonlinear manufacturing environment.

[0059] Embodiment two:

[0060] The embodiment provides another specific application of the multi-source collaborative dynamic allocation method of multi-laser SLM layer vector data. The technical scheme of the application is applied to the same large-format four-laser SLM additive manufacturing equipment independently developed by the embodiment, but the manufacturing task this time is a large and high-precision injection mold for the automobile manufacturing field. The mold contains a complex conformal cooling water channel, and has a high requirement for suppressing part warpage and ensuring the final size accuracy. The production order of the manufacturing task strictly defines the cost budget and the delivery cycle. The embodiment focuses on the specific implementation of the process reference establishment, closed-loop correction and production management of the application.

[0061] Further, in the embodiment, the method for establishing the space-time process potential field map is specifically: before the printing task starts, the preprocessing module of the system first performs finite element meshing on the three-dimensional model of the injection mold, and performs a detailed thermal-structural coupling simulation. The simulation aims to obtain the distribution data of the theoretical temperature field, stress field and phase change history of any space-time node in the mold during the entire processing process lasting for several days under ideal process conditions. At the same time, the system automatically retrieves and extracts the historical successful case data most matched with the current manufacturing task from the historical process database. These case data include complete sensor process records and final three-dimensional scanning quality inspection results. Finally, the system fuses and calibrates the distribution data obtained by the theoretical simulation with the process data from the real successful cases, for example, uses the real data to correct the prediction deviation of the cooling rate of the conformal water channel area in the simulation model, thereby generating a high-fidelity final space-time process potential field map defined by a four-dimensional data structure.

[0062] The embodiment introduces prior knowledge of process quality into the real-time control loop by establishing a spatiotemporal process potential field map. The control system in the background art lacks a real-time quantitative standard for judging the pros and cons of the current state when performing tasks. The potential field map described in the application provides a high-resolution quality benchmark for the manufacturing process, which is jointly backed by physical theory and empirical data. This fundamentally changes the manufacturing process from an open-loop instruction-driven mode that executes geometric paths to a closed-loop target optimization mode that actively seeks to approach the ideal state, providing an objective and accurate basis for all subsequent online corrections and closed-loop control.

[0063] Further, in the printing process of the mold, the step of calling the correction strategy specifically includes: when the equipment prints to a narrow conformal waterway partition wall of the mold, the multi-modal sensing system monitors the real-time process state of the region, and a significant deviation from the ideal state defined by the spatiotemporal process potential field map occurs, specifically manifested as an actual temperature that is 30% higher than an ideal temperature and a duration that exceeds 2 seconds. The system determines that this is a serious deviation that may lead to local overheating deformation. At this time, the system immediately selects and executes a cooperative level composite correction action from the preset strategy library. The correction action includes three parallel operations: first, real-time adjustment of the power of the No. 2 laser currently working in the region, reducing it by 15%; second, immediate re-planning of the scanning order of all subsequent vector objects to be processed in the affected area, changing from the original short-side filling strategy to a skip filling strategy to disperse heat; third, temporarily changing the task assignment of the idle No. 3 laser next to it, commanding it to perform a key compensation task, i.e., a low-power, large-spot preheating scan of the adjacent solid area of the overheated area, to actively guide heat conduction to the solid area, thereby quickly reducing the temperature of the partition wall.

[0064] The correction strategy library and selection mechanism constructed in the embodiment provide a hierarchical response system for the system to cope with process deviations. The correction means of traditional control systems are usually single and fixed. The application can call different levels of strategies from parameter fine-tuning to task re-planning, and even multi-resource collaborative intervention according to the severity and type of the deviation. This risk level matching capability with multiple correction means ensures that the intervention behavior of the system is always accurate, moderate and efficient, thereby giving the entire manufacturing process unprecedented resilience and robustness.

[0065] In the printing process of the mold, the step of calling the correction strategy specifically includes:

[0066] The method pre-establishes a structured correction strategy library, which is organized as a database containing multiple levels. The selection logic of the correction strategy is realized by a decision tree selection model pre-trained by historical data. The input of the model is a multi-dimensional deviation vector between the real-time process state and the ideal state defined by the spatiotemporal process potential field map, and the output is the correction strategy that should be called from the strategy library.

[0067] The multi-dimensional deviation vector for inputting the decision tree selection model includes at least the deviation value of real-time temperature and ideal temperature, the deviation value of real-time cooling rate and ideal cooling rate, and the deviation value of real-time molten pool area and ideal area identified by the visual system. When the mold is printed to a deep cavity thin wall structure, the system monitors the deviation between the real-time state of the area and the ideal state of the spatiotemporal process potential field map, forming a deviation vector containing multiple dimensional information, for example, the vector indicates that the temperature of the current area is 15℃ higher than the ideal value, and the cooling rate is 10% lower than the standard. The system immediately inputs the deviation vector into the pre-trained decision tree model. The model makes a series of judgments and branches along the nodes of the tree according to the specific values in the vector, the first node judges whether the temperature deviation is greater than 10℃, and the second node judges whether the molten pool shape is abnormal. Finally, the deviation vector falls on a leaf node, which clearly points to the correction strategy to be executed, that is, to call the second level of re-planning scanning strategy and apply the chessboard scanning mode to uniform the heat field of the area. The instruction is immediately issued to the execution unit.

[0068] By constructing a structured strategy library and a decision tree-based selection model, the logic of closed-loop correction is improved from simple rule judgment to a more intelligent and systematic level. The hierarchical strategy library enables the system to take measures of different granularity from parameter fine-tuning to multi-laser collaborative intervention according to the severity of the deviation. The application of the decision tree model makes the selection process not only efficient, but also clear and interpretable in decision path, which is crucial for the manufacturing of precision molds and other key components that require high reliability and traceability. The mechanism ensures that the system can match the most reasonable and efficient correction strategy for complex and variable real-time deviations.

[0069] Further, the closed-loop management and control step is embodied in a key decision-making step before the execution of the above-mentioned corrective action. After the system determines that the above-mentioned collaborative level corrective action needs to be executed, the built-in closed-loop management module will first make a quick cost and time impact estimation on the strategy. The estimation result shows that the execution of the set of corrective actions will increase the total processing time by about 15 minutes and the corresponding laser running cost. Subsequently, the module compares the estimated impact with the total budget and total working hours initially set for the mold manufacturing task. The comparison result shows that the additional 15 minutes is still within the 8-hour floating time range specified by the production order, and the additional cost does not exceed the 2% redundancy of the budget. Only when the judgment that "it is still within the allowed floating range" is true, the system finally confirms and autonomously executes the correction strategy. Otherwise, if the estimated impact exceeds the range, the system will mark the deviation as a high-risk decision event that needs human intervention, and push an alarm information to the operating engineer.

[0070] The closed-loop management step of the present embodiment builds a key business objective constraint layer on top of the technical control loop. It establishes a direct link between the optimization of the machine's physical process and the production management objectives of the factory. The automated system in the background art may make cost-irrelevant decisions to correct local technical problems due to the lack of such a mechanism. The present invention ensures that every autonomous correction behavior of the system is subject to economic rationality test, ensuring that technical autonomy always serves the macro business objectives. This management intelligence level is the core guarantee for the safe, reliable and trustworthy deployment of complex automated systems in industrial scenarios.

[0071] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A multi-laser SLM layer vector data multi-source collaborative dynamic allocation method, characterized in that, Comprise: According to the manufacturing task containing the three-dimensional model and the associated production order information, the three-dimensional model is sliced to generate initial layer vector data, multi-dimensional feature processing is performed through a vector feature engine to generate a hierarchical process topology set carrying independent process rules of vector objects; Fusion of the static attributes of each vector object in the hierarchical process topology set and the real-time process state data, quantitative calculation of the matching degree between each object to be processed and the idle production resources, and generation of a dynamic collaborative matching matrix; According to the optimal matching value in the dynamic collaborative matching matrix, the multi-laser production resources are dynamically scheduled; Through offline simulation and historical process data analysis, a space-time process potential field map is established, which is used to define the ideal process state of each spatial position of the part at the processing time, and the process strategy for correcting the state deviation is associated; Execute the assigned processing task, compare the real-time process state with the ideal state defined by the space-time process potential field map, call the correction strategy, and realize closed-loop management and control of the manufacturing task under the premise of meeting the cost budget and delivery cycle.

2. The multi-laser SLM layer vector data multi-source collaborative dynamic allocation method according to claim 1, characterized in that, The step of multi-dimensional feature processing by the vector feature engine includes: geometric and topological analysis of the initial layer vector data, identification and marking of vector clusters representing key shapes of the part; according to the results of the geometric and topological analysis, and combining the quality level defined in the production order information, the vector clusters are classified and constructed into vector objects of different levels.

3. The method of claim 1, wherein, The independent process rules carried by the vector object cover multiple dimensions, including: geometric constraints for ensuring forming accuracy, which include wall thickness and hole diameter limits; and process parameter constraints for guiding energy input, which include preset ranges of scanning speed and laser power.

4. The multi-laser SLM layer vector data multi-source collaborative dynamic allocation method according to claim 1, characterized in that, The real-time process state data comes from a multi-modal sensing system, which includes: a visual sensing unit for identifying molten pool topography and powder bed geometric anomalies, and a thermal flow sensing unit for monitoring molten pool temperature and thermal field distribution.

5. The multi-laser SLM layer vector data multi-source collaborative dynamic allocation method according to claim 1, characterized in that, The construction method of the dynamic collaborative matching matrix is: for the pairing between the vector object to be processed and the idle production resource, the quantitative matching degree is calculated according to the multi-objective function; the multi-objective function weightedly combines the following indicators: risk indicators for evaluating the quantitative value of local heat accumulation, thermal stress concentration and process risk caused by assigning the production resource to the vector object; benefit indicators for evaluating the contribution to improving the metallurgical quality of the overlap area and ensuring the forming accuracy of the fine profile key quality objectives; Cost indicators for evaluating the time cost and energy cost required for the production resource to move from the current position to the starting point of the vector object.

6. The multi-laser SLM layer vector data multi-source collaborative dynamic allocation method according to claim 5, characterized in that, The weights of the multi-objective function are dynamically adjusted according to the level of the vector object: for multi-laser overlap objects, the weight of the benefit indicator is set to the highest; for fine profile objects, the weight of the benefit indicator is given priority; for core entity objects, the cost indicator for improving processing efficiency is given priority.

7. The method of claim 1, wherein, The dynamic scheduling step is a continuous iterative loop process, and the steps are: at the beginning of the scheduling period, the dynamic collaborative matching matrix is generated; for the current idle laser, the corresponding optimal matching vector object is found from the matrix; the processing task of the vector object is assigned to the corresponding laser; when the laser completes the assigned task, the laser state becomes idle, and the part processing state is updated, and the system triggers the scheduling period immediately until all vector objects of the current layer are processed.

8. The method of claim 1, wherein, The method for establishing the space-time process potential field diagram comprises: performing finite element meshing on the three-dimensional model and performing thermal-structural coupling simulation to obtain theoretical distribution data of temperature field, stress field and phase change history of the part at any space-time node in the entire processing process under ideal process conditions; extracting case data from a historical database, the part material type, the preset quality level and the geometric complexity classification recorded in the case data are completely consistent with the current manufacturing task, and the case data includes complete sensor process records and final quality detection results; fusing and calibrating the theoretical distribution data and the case data to generate the space-time process potential field diagram; and the space-time process potential field diagram is used to define the ideal process state that should be reached by each spatial position of the part at the processing time in a four-dimensional data structure.

9. The method of claim 1, wherein, The step of calling the correction strategy specifically comprises: when the real-time process state deviates from the ideal state defined by the space-time process potential field diagram, the system selects and executes a correction action from a preset strategy library, and the correction action comprises: adjusting the power, scanning speed and spot size of the currently working laser in real time without changing the scanning path; re-planning the scanning order and scanning strategy of the subsequent to-be-processed vector objects in the affected area; temporarily changing the task assignment of the idle laser to preferentially process the key compensation task generated due to the state deviation.

10. The method of claim 1, wherein, The step of closed-loop management and control comprises: before executing the correction strategy, the system estimates the influence of the correction strategy on the total processing time and potential material loss; comparing the estimated influence with the total budget and total working hours initially set for the manufacturing task; only when the estimated total cost and total time after correction are still within the allowed floating range, the correction strategy is executed; if it is out of range, the system will mark the deviation as a high-risk event and prompt manual intervention.

Images