Design and manufacturing integrated multi-agent collaborative optimization system and method

Abstract

This application provides a multi-agent collaborative optimization system and method integrating design and manufacturing, comprising: after receiving an initial constraint set in the processing stage, the agent calculates the increase in processing difficulty caused by changes in parameter values ​​using a constraint propagation mechanism, generates processing data, and continues to propagate it to the agent in the cost stage to obtain an extended constraint set containing the processing data; after receiving the complete constraint set, each agent in each stage performs real-time perception and detects conflicts; if the parameter optimization direction is opposite to the processing difficulty direction and conflicts with the cost direction, conflict points on the geometric features are marked to obtain a conflict set containing the conflict points; real-time perception and conflict detection are repeated on the updated constraint set; if there are no new conflict points, the final feasible solution space is output; if there are residual conflicts, the conflict negotiation mechanism is returned to continue iterating until a conflict-free feasible solution space is obtained.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a multi-agent collaborative optimization system and method that integrates design and manufacturing. Background Technology

[0002] The integration of industrial design and manufacturing has become a core development direction for modern manufacturing to improve efficiency, reduce costs, and ensure quality. Throughout the entire product lifecycle, from concept to finished product, decisions made during the design phase directly impact subsequent processing, assembly, and performance. Only by achieving seamless collaboration between design and manufacturing can we truly adapt to rapidly evolving market demands and increasingly complex trends in personalized customization.

[0003] While many current methods attempt to connect the design and manufacturing processes, they often remain at a loose level of information transmission, resulting in constraints failing to flow effectively and respond promptly between different stages. Constraints such as design, process, cost, and time inherently exhibit characteristics of mutual restraint and even direct conflict.

[0004] Therefore, how to establish a system in a multi-agent collaborative environment that allows various constraints to be automatically propagated across stages, perceived in real time, and finds a feasible solution space acceptable to all parties in the event of a conflict through intelligent means has become a core problem that urgently needs to be solved in integrated design and manufacturing, such as integrated industrial design and manufacturing, for multi-agent collaborative optimization systems and methods. Summary of the Invention

[0005] This invention provides a multi-agent collaborative optimization system and method integrating design and manufacturing, mainly comprising:

[0006] Obtain the geometric feature constraints and strength requirements generated in the design phase, extract parameter values ​​and material data from them, and transmit the parameter values ​​and material data to the processing phase agent and the cost phase agent through the constraint propagation mechanism to generate the initial constraint set received by the agent in each phase.

[0007] Based on the geometric feature constraints and strength requirements, the parameter values ​​and the material data are separated to form a set of propagation constraints that includes geometric constraint correlation and strength constraint correlation;

[0008] The propagation constraint set is transmitted to the processing stage agent, which determines the processing route based on the geometric constraint portion in the initial constraint set and outputs the processing process constraints.

[0009] The propagation constraint set is transmitted to the cost stage agent, which calculates the cost value corresponding to each processing technology constraint based on the strength constraint part in the initial constraint set and the material data, and outputs the cost constraint set.

[0010] In the processing stage agent, the incremental processing difficulty caused by the change in parameter value is calculated through the constraint propagation mechanism, processing data is generated and transmitted to the cost stage agent, forming an extended constraint set containing the processing data;

[0011] In the cost stage agent, the cost increment caused by the change of parameter value is calculated through the constraint propagation mechanism, cost data is generated and transmitted back to the design stage agent and the processing stage agent to form a complete constraint set containing the cost data.

[0012] For the complete set of constraints, real-time perception and conflict detection are performed in the agents at each stage, and conflict points on geometric features are marked to form a conflict set containing the conflict points;

[0013] The conflict points in the conflict set are handled by a conflict negotiation mechanism to generate a set of suggestions after multi-party negotiation. The feasible parameter range is iteratively calculated by combining a multi-objective optimization mechanism to update the geometric feature constraints and propagate them to agents at all stages.

[0014] This invention provides a multi-agent collaborative optimization system and system integrating design and manufacturing, mainly comprising:

[0015] The constraint acquisition and transmission module is used to acquire the geometric feature constraints and strength requirements generated in the design phase, extract parameter values ​​and material data from them, and use a constraint propagation mechanism to transmit the parameter values ​​and material data to the agents in the processing and cost phases to obtain the initial constraint set received in each phase.

[0016] The processing constraint extension module is used to calculate the increase in processing difficulty caused by changes in parameter values ​​after the agent receives the initial constraint set during the processing stage, using the constraint propagation mechanism, generating processing data and continuing to propagate it to the agent in the cost stage, thus obtaining an extended constraint set containing the processing data.

[0017] The cost constraint extension module is used to calculate the cost increment caused by the change of parameter values ​​after the agent receives the extended constraint set in the cost stage, using the constraint propagation mechanism to generate cost data and propagate the cost data back to the agents in the design and processing stages to obtain a complete constraint set containing cost data.

[0018] The conflict detection and marking module is used to perceive and detect conflicts in real time after the intelligent agent receives the complete set of constraints at each stage. If the direction of parameter optimization is opposite to the direction of processing difficulty and conflicts with the direction of cost, the conflict points on the geometric features are marked to obtain a conflict set containing the conflict points.

[0019] The conflict negotiation and optimization module is used to propose adjustment suggestions from agents at each stage for each conflict point in the conflict set using a conflict negotiation mechanism, resulting in a set of suggestions after multi-party negotiation. The set of suggestions is then input into a multi-objective optimization mechanism, and feasible parameter ranges are obtained through iterative calculation. The initial geometric feature constraints are updated according to the feasible parameter ranges and propagated to agents at all stages.

[0020] The iterative convergence module is used to repeatedly perceive and detect conflicts in real time on the updated constraint set. If there are no new conflict points, it outputs the final feasible solution space. If there are residual conflicts, it returns to the conflict negotiation mechanism to continue iterating until a conflict-free feasible solution space is obtained.

[0021] The present invention provides a computer-readable storage medium comprising a stored program, wherein, when the program is executed, any one of the methods described above is performed by a processor.

[0022] This invention provides an electronic device, comprising: a processor; and a memory for storing instructions executable by the processor.

[0023] The processor is configured to read the executable instructions from the memory and execute the instructions to implement any one of the methods described above.

[0024] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0025] This invention discloses an integrated intelligent design and manufacturing cost control method for complex parts oriented towards multi-stage collaborative optimization. By constructing an intelligent agent collaborative framework across three stages—design, manufacturing, and cost—it achieves cross-stage constraint propagation of geometric feature constraints, strength requirements, and material data. Parameter values ​​and material data extracted in the design stage are sequentially transmitted to the manufacturing and cost stages via a constraint propagation mechanism. In the manufacturing stage, the intelligent agent calculates the increase in manufacturing difficulty caused by parameter changes and generates manufacturing data for further forward propagation. In the cost stage, the intelligent agent calculates the cost increment and propagates the cost data back to the previous two stages, forming a closed-loop set containing complete constraints. Based on this set, the intelligent agents at each stage perceive and detect conflicts in real time. When the direction of parameter optimization contradicts the direction of manufacturing difficulty and cost, the conflict point on the geometric features is accurately marked. For the conflict point, each intelligent agent proposes adjustment suggestions through a conflict negotiation mechanism, and iteratively solves the feasible parameter range through a multi-objective optimization mechanism, thereby updating the initial constraints and propagating them throughout all stages. This process is iterated until all conflicts are eliminated, outputting a conflict-free feasible solution space. This invention effectively solves the problems of increased processing difficulty and cost control caused by frequent design changes, and achieves dynamic balance and global optimization among design, processing and cost, significantly improving the collaborative efficiency and economy of complex parts throughout their entire life cycle. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a multi-agent collaborative optimization method for integrated design and manufacturing according to the present invention.

[0027] Figure 2 This is a schematic diagram of the structure of a multi-agent collaborative optimization system that integrates design and manufacturing according to the present invention.

[0028] Figure 3 This is a schematic diagram of the DAG (Directed Acyclic Graph) model of the constraint propagation mechanism of this invention.

[0029] Figure 4 This is a convergence curve of the conflict negotiation iteration in this invention.

[0030] Figure 5 This is a multi-objective optimization Pareto front distribution map for the present invention.

[0031] Figure 6 This is a heat map showing the correlation between the three-stage constraint parameters of this invention.

[0032] Figure 7 This is a surface diagram illustrating the three-dimensional trade-offs between cost, quality, and efficiency in this invention.

[0033] Figure 8 This is a topology diagram of the DAG constraint propagation network of the present invention.

[0034] Figure 9 This is a timing diagram of the three-stage intelligent agent interaction of the present invention.

[0035] Figure 10 This is a spatial distribution map of the collision point detection and marking in this invention.

[0036] Figure 11 This is a schematic diagram illustrating the industrial manufacturing application scenario of the present invention. Detailed Implementation

[0037] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0038] Terminology Definition

[0039] 1. Constraint propagation mechanism: This refers to a mechanism based on directed acyclic graphs (DAGs) or rule-based reasoning to achieve automatic parameter association, state updates, and cross-stage data transmission in the design, manufacturing, and cost stages. Its core is to ensure that the impact of parameter changes can be perceived by relevant intelligent agents in real time.

[0040] 2. High-difficulty factors: These refer to key influencing factors that exceed preset thresholds during the processing (such as tool wear rate, machine tool vibration amplitude, etc.), resulting in limited processing technology, reduced efficiency, or significantly increased costs.

[0041] 3. Conflict point: refers to the specific point where the direction of parameter optimization for the same geometric feature is opposite to the direction of processing difficulty and contradicts the direction of cost. It needs to be adjusted and resolved through a negotiation mechanism.

[0042] 4. Feasible solution space: refers to the range of all parameter combinations that satisfy the design strength requirements, processing feasibility constraints and cost control objectives. It is the set of conflict-free parameter intervals that are finally output after iterative optimization.

[0043] like Figure 1 As shown, the multi-agent cooperative optimization method of the present invention includes the following steps: Step S101: Obtain the geometric feature constraints and strength requirements generated in the design stage, extract parameter values ​​and material data from them, and use a constraint propagation mechanism to transmit the parameter values ​​and material data to the agents in the processing stage and the cost stage, thereby obtaining the initial constraint set received by each stage; Step S102: After receiving the initial constraint set, the agent in the processing stage uses the constraint propagation mechanism to calculate the increase in processing difficulty caused by the change in parameter values, generate processing data, and continue to propagate it to the agent in the cost stage; Step S103: After receiving the extended constraint set, the agent in the cost stage calculates the cost increment and propagates it in reverse; Step S104: The agents in each stage perform real-time perception and detect conflicts; Step S105: For conflict points, a conflict negotiation mechanism is adopted, and the agents in each stage propose adjustment suggestions and input them into the multi-objective optimization mechanism; Step S106: Repeatedly perceive and detect the updated constraint set, and if there are no new conflict points, output the final feasible solution space.

[0044] like Figure 2 As shown, the multi-agent collaborative optimization system of the present invention includes: a constraint acquisition and transmission module, used to acquire geometric feature constraints and strength requirements generated in the design stage, and transmit parameter values ​​and material data to agents in the processing and cost stages using a constraint propagation mechanism; a processing constraint extension module, used to calculate the processing difficulty increment and generate processing data after the agent in the processing stage receives the initial constraint set; a cost constraint extension module, used to calculate the cost increment and propagate the cost data back to the agents in the design and processing stages; a conflict detection and marking module, used for agents in each stage to perceive and detect conflicts in real time; a conflict negotiation and optimization module, used for agents in each stage to propose adjustment suggestions for conflict points using a conflict negotiation mechanism; an iterative convergence module, used to repeatedly perceive and detect conflicts in real time on the updated constraint set; and an emergency handling module, used to deal with abnormal scenarios such as data transmission interruption and agent failure.

[0045] Specifically, this embodiment of a design-manufacturing integrated multi-agent collaborative optimization system and method may include:

[0046] Step S101: Obtain the geometric feature constraints and strength requirements generated in the design phase, extract parameter values ​​and material data from them, and use a constraint propagation mechanism to transmit the parameter values ​​and material data to the agents in the processing and cost phases to obtain the initial constraint set received by each phase.

[0047] It should be understood that, in this application, integrated design and manufacturing can specifically refer to the integration of industrial design and manufacturing, or it can refer to any suitable technical field, and the technical solutions of this application can be used in these suitable technical fields.

[0048] Obtain the geometric feature constraints and strength requirements documents output from the design phase. Perform parameter extraction on these documents, separating parameter values ​​and material data. Utilize a constraint propagation mechanism to process the parameter values ​​and material data, generating a propagated constraint set containing geometric and strength constraint associations. Transmit this propagated constraint set to the manufacturing stage agent, which receives it and obtains an initial constraint set containing both geometric and strength constraints. The manufacturing stage agent determines the manufacturing process route based on the geometric constraints in the initial constraint set and outputs the manufacturing process constraints. Transmit the propagated constraint set to the cost stage agent, which receives it and obtains an initial constraint set containing both geometric and strength constraints. The cost stage agent calculates the cost value corresponding to each manufacturing process constraint based on the strength constraints and material data in the initial constraint set and outputs the cost constraint set.

[0049] Specifically, the constraint propagation mechanism employs a parameter association model based on a directed acyclic graph (DAG). In this model, geometric parameters (such as hole diameter and wall thickness) serve as parent nodes, while machining process parameters (such as cutting step distance) and cost factors (such as material unit price and unit labor cost) serve as child nodes. When the value of a parent node changes abruptly, the propagation mechanism automatically triggers the state update of downstream nodes based on a preset logic function (such as a linear proportional function or a step response function) between nodes. For example, when the hole diameter parameter decreases to exceed the minimum tool diameter threshold, the logic function automatically sends a "process limitation" signal to the machining agent and simultaneously sends a cost increment reminder of "need to replace expensive tool" to the cost agent.

[0050] like Figure 3As shown, the constraint propagation mechanism of this invention adopts a parameter association model based on a directed acyclic graph. In this model, geometric parameters serve as parent nodes, including design-stage output parameters such as aperture parameters, wall thickness parameters, surface accuracy, and material strength; machining process parameters and cost factors serve as child nodes, including cutting step distance, tool type, machining time, material unit price, and unit energy consumption. Parent and child nodes are associated through preset logical functions, including linear proportional functions and step response functions. When the value of a parent node changes abruptly, the propagation mechanism automatically triggers the state update of downstream nodes according to the logical functions between nodes, achieving synchronous propagation of constraint information across stages.

[0051] like Figure 8 As shown, this invention illustrates a constraint propagation network topology based on a directed acyclic graph (DAG). The diagram is divided into three levels from top to bottom: design stage, manufacturing stage, and cost stage. The design stage includes five elliptical parameter nodes: aperture, wall thickness, surface accuracy, material strength, and load threshold. The manufacturing stage includes five rectangular process nodes: cutting step distance, tool type, machining time, machine tool vibration, and tool wear. The cost stage includes five diamond-shaped process nodes: material unit price, energy cost, labor cost, equipment depreciation, and scrap loss. Solid arrows represent forward constraint propagation paths, and dashed arrows represent reverse cost feedback paths. This topology clearly demonstrates the cross-stage parameter relationships and constraint propagation mechanism.

[0052] like Figure 6 As shown, this invention illustrates the distribution of constraint correlation strength among three types of parameters: design stage, machining stage, and cost stage. The heatmap uses grayscale gradients to represent correlation strength, with darker colors indicating stronger correlations. Design stage parameters include hole diameter, wall thickness, surface accuracy, material strength, and load threshold; machining stage parameters include cutting step distance, tool type, machining time, machine tool vibration, and tool wear; cost stage parameters include material unit price, energy consumption cost, labor cost, equipment depreciation, and scrap loss. The heatmap reveals a moderate correlation between design and machining stage parameters, a strong correlation between machining and cost stage parameters, and a weak indirect correlation between design and cost stage parameters. This correlation matrix provides a quantitative basis for the constraint propagation mechanism.

[0053] In a typical application scenario (such as machining of mechanical parts), the core node list of the DAG model is as follows:

[0054] - Parent node: aperture, wall thickness, surface accuracy, material yield strength, load threshold (all are output parameters in the design phase);

[0055] - Sub-nodes: Cutting step distance, tool type, machining time, material unit price, unit energy consumption cost (corresponding to machining process parameters and cost factors, respectively). Examples of logical functions between nodes: 1. Linear proportional function: When the hole diameter ≥ 5mm, cutting step distance = 0.8 × hole diameter; when the hole diameter < 5mm, cutting step distance = 0.5 × hole diameter; 2. Stepped response function: When the wall thickness exceeds 10mm, machining time = basic machining time × (1 + 0.3 × (wall thickness - 10) / 10), simultaneously triggering the activation of the "additional heat treatment cost" item in the cost factor.

[0056] For example, in the field of mechanical parts design, the first step is to obtain output files from the design software, such as geometric constraint files generated from CAD models and strength requirement documents generated from finite element analysis. These files contain information such as dimensional tolerances and stress thresholds for the parts. Automated scripts read the file contents to ensure data integrity and lay the foundation for subsequent processing.

[0057] Specifically, the parameter extraction operation involves using a parsing tool to scan the file and separate numerical values ​​such as length 10mm and thickness 5mm, while simultaneously extracting material data such as the elastic modulus and yield strength of steel. This separation helps to process different types of information independently and avoids confusion.

[0058] In one embodiment, the constraint propagation mechanism is a rule-based reasoning system that associates parameter values ​​with material data through logical chains, such as linking geometric dimensions with strength requirements, to form association rules such as "if the length exceeds a threshold, the material thickness needs to be adjusted," thereby generating a set of propagation constraints. This set includes geometric constraints such as shape symmetry and strength constraints such as maximum load-bearing capacity.

[0059] Specifically, the propagation process is derived step by step from the initial parameters. For example, assuming the part is a beam structure with geometric constraints requiring the beam to be 2m long and 0.5m wide, and the material to be aluminum alloy with a strength of 200MPa, the propagation mechanism will deduce that if the beam length increases by 10%, the strength needs to be increased accordingly to maintain stability, forming an associated chain to ensure that the constraints are transmitted consistently at different stages.

[0060] For example, this set of propagation constraints can be sent to the machining stage agent via an API interface. The latter is an AI-based process planning module that receives and parses the set, extracts geometric constraints such as surface accuracy requirements, and determines the process route, such as choosing CNC milling instead of casting to meet dimensional accuracy requirements.

[0061] Interface and Data Format Specifications: 1. API Interface: RESTful API version 3.0 is adopted, supporting integration with CAD software such as SolidWorks and UG, as well as the ANSYS finite element analysis tool. The interface address format is "http: / / [server IP]:[port] / agent / data / transmit"; 2. Data Format: JSON format is uniformly adopted. Core fields include: { "Parameter ID": "Unique identifier (e.g., G001-aperture)", "Value": "Specific parameter value (e.g., 8mm)", "Tolerance": "Allowable fluctuation range (e.g., ±0.05mm)", "Material Number": "Corresponding material standard number (e.g., GB / T 3077-2015)", "Strength Threshold": "Minimum strength value required by design (e.g., 350MPa)", "Data Type": "Geometric parameter / material parameter / process parameter / cost parameter"}; 3. Data Transmission Protocol: TCP / IP protocol is adopted, with a data transmission rate ≥1Mbps, a packet loss rate ≤0.3%, and a timeout retransmission time of 3s.

[0062] Specifically, during the machining phase, the intelligent agent evaluates feasible paths based on geometric constraints. For example, for complex curved surfaces, it prioritizes five-axis machining and outputs process constraints such as cutting speed and tool type to optimize production efficiency.

[0063] In one embodiment, the propagation constraint set is simultaneously transmitted to the cost stage agent, which also receives and forms an initial constraint set. Then, it focuses on strength constraints such as fatigue limits and material data such as unit prices, calculates the cost of each process, for example, the total material consumption and energy cost of CNC machining is 500 yuan / piece, while that of casting is 300 yuan / piece, and outputs the cost constraint set such as the threshold of the lowest cost path.

[0064] Specifically, this calculation process first estimates the amount of material used, then weights the labor time, for example, parts with high strength requirements need additional heat treatment, increasing costs by 20%, and finally forms a set of data to guide decision-making.

[0065] For example, in actual automotive parts production, this mechanism ensures that design constraints are propagated to the processing and cost stages, achieving overall optimization, reducing scrap rates, and controlling budgets.

[0066] In step S102, after the agent receives the initial constraint set during the processing stage, it uses the constraint propagation mechanism to calculate the increase in processing difficulty caused by the change in parameter values, generates processing data, and continues to propagate it to the agent in the cost stage to obtain an extended constraint set containing the processing data.

[0067] Initial constraint data is acquired from the agent in the processing stage. A constraint propagation mechanism is used to decompose parameter changes item by item, identifying key factors affecting processing difficulty and obtaining a set of factors influencing processing difficulty. For this set, each key factor is analyzed using preset rules. If a factor exceeds a preset threshold, it is marked as a high-difficulty factor, resulting in a list of high-difficulty factors. Based on this list, information processing tools are used to classify and organize the processing data, generating a subset containing high-difficulty factors to determine the distribution of processing difficulty increments. Using this distribution and the constraint propagation mechanism, the subset of processing data is passed to the agent in the cost stage, resulting in an extended constraint set. For this extended constraint set, an information matching method is used to associate the processing difficulty increments with the constraints of the cost stage, determining the scope of cost impact and obtaining cost impact data. Based on this cost impact data, data is filtered using preset logical rules. If outliers are found, they are labeled, resulting in a labeled subset of cost constraints. Using this labeled subset of cost constraints, information integration tools are used to generate the final processing and cost correlation data, determining the cross-stage constraint impact results.

[0068] In one embodiment, the preset rule is a threshold judgment rule based on the classification of processing difficulty level, specifically including: tool wear rate ≥ 0.05 mm / min, machine tool vibration amplitude ≥ 2 micrometers, processing time growth rate ≥ 30%, and meeting any one of the conditions is marked as a high difficulty factor.

[0069] In one embodiment, when obtaining the initial constraint data in the intelligent agent during the processing stage, it is first necessary to extract relevant parameters from the geometric feature constraints and strength requirements transmitted from the design stage.

[0070] For example, in the automotive parts manufacturing business, initial constraint data might include a surface radius of 20 mm and a material strength requirement of 500 MPa. This data is stored in a file format such as XML for easy reading by the intelligent agent. Next, a constraint propagation mechanism is used to decompose the parameter variations item by item. This constraint propagation mechanism is a graph theory-based algorithm framework that treats parameters as nodes and constraints as edges, analyzing the variations item by item through depth-first search.

[0071] For example, when the surface radius changes from 20 mm to 15 mm, the mechanism decomposes it into factors such as changes in the machining tool path and increases in cutting force, thereby identifying the key factors affecting the change in machining difficulty, such as tool wear rate and machine tool stability, and obtaining a set of factors affecting machining difficulty. This set is presented in list form, with each factor accompanied by a description of the magnitude of change to support subsequent evaluation.

[0072] For example, when analyzing each key factor in a set of factors affecting machining difficulty using preset difficulty assessment rules, these rules can be based on a fuzzy logic rule set from an expert system. A threshold could be set such that a tool wear rate exceeding 0.05 mm / min is considered high difficulty. The analysis process involves inputting each factor into a rule engine, which calculates membership degrees using membership functions. For instance, for machine tool stability, if the vibration amplitude exceeds a preset 2-micron threshold, it is marked as a high-difficulty factor, ultimately resulting in a list of high-difficulty factors. This list helps identify potential machining bottlenecks and effectively avoids production delays in business operations.

[0073] Threshold setting basis: 1. Tool wear rate threshold (0.05 mm / min): derived from the service life standard (≥5000 pieces) of commonly used carbide tools (model: WC-Co 6%). When the wear rate exceeds this threshold, the tool service life will be shortened to less than 1000 pieces, resulting in a surge in processing costs; 2. Machine tool vibration amplitude threshold (2 micrometers): referring to the vibration allowable value of precision machining equipment in GB / T 17421.2-2021 "General Rules for Machine Tool Inspection". Exceeding this threshold will cause the part dimensional tolerance to exceed the tolerance (≥±0.02 mm); 3. Machining difficulty level classification: [0,0.3] (low difficulty), [0.3,0.7] (medium difficulty), (0.7-1] (high difficulty). The classification is based on the growth rate of machining time (low difficulty ≤10%, 10% < medium difficulty ≤30%, high difficulty >30%).

[0074] Specifically, when classifying and organizing the processing data based on the list of high-difficulty factors, the information processing tool can be a data mining software module, such as using a clustering algorithm to group the data and generate a subset of processing data containing high-difficulty factors.

[0075] For example, in the machining of aero-engine blades, high-difficulty factors such as the machining of complex curved surfaces are classified into subsets, thereby determining the distribution of machining difficulty increments. This distribution is represented by a histogram, such as the proportion of difficulty increasing from medium to high, which provides a quantitative basis for cross-stage transfer.

[0076] In one embodiment, by distributing the processing difficulty increment and combining it with the constraint propagation mechanism, when a subset of processing data is passed to the cost stage agent, the propagation mechanism expands the constraint edges, incorporating the difficulty increment as a new node, thus obtaining an expanded constraint set.

[0077] For example, after transmission, the cost agent receives a set containing additional difficulty coefficients, such as a 20% increase in processing time, which ensures the accuracy of cost calculation.

[0078] For example, when using an information matching method to associate the incremental processing difficulty with the constraints of the cost stage for an extended constraint set, the information matching method is based on semantic similarity calculation. For example, cosine similarity is used to match the incremental difficulty with the cost formula to determine the scope of cost impact and obtain cost impact data. For example, if the increase in material waste rate leads to a 15% increase in cost, the cause of this data is the increase in processing difficulty, and the consequence is an overall budget overrun. In business, this can guide the optimization design to reduce costs.

[0079] Specifically, when filtering data based on cost impact data using preset logical rules, the logical rules can be a set of if-then statements. If there are outliers in the cost impact data, such as a sudden 50% increase, they are marked to obtain a subset of cost constraints.

[0080] For example, labeling items as "high risk" can help with risk management.

[0081] In one embodiment, when using the labeled cost constraint subset to generate the final processing and cost correlation data through an information integration tool, which is a database fusion module, the information integration tool integrates the subset into a report to determine the cross-stage constraint impact results.

[0082] For example, in the manufacturing of electronic device casings, results showed that the increase in difficulty led to a 10% increase in costs, which enabled efficient resource allocation and technological iteration in the business.

[0083] Step S103: After receiving the extended constraint set in the cost stage, the agent uses the constraint propagation mechanism to calculate the cost increment caused by the change in parameter values, generates cost data, and propagates the cost data back to the agents in the design and processing stages to obtain a complete constraint set containing cost data.

[0084] Obtain the extended constraint set and initial parameter values. For the constraints in the extended constraint set, use a constraint propagation mechanism to deduce the dependencies between parameters layer by layer, obtaining the current parameter value variation range. Based on the current parameter value variation range, calculate the direct cost increment and cumulative cost increment caused by each parameter adjustment, generating a cost data table. Classify the cost data table according to the constraint correspondence between the design and processing stages, obtaining a design stage cost subset and a processing stage cost subset. Inject the design stage cost subset into the design stage agent constraint update module to update the cost-related constraint items in the design stage constraint set. Inject the processing stage cost subset into the processing stage agent constraint update module to update the cost-related constraint items in the processing stage constraint set. Merge the updated design stage constraint set, processing stage constraint set, and cost stage constraint set, and after a consistency check, form a complete constraint set containing cost data.

[0085] For example, in obtaining the extended constraint set and initial parameter values, it can be understood that the extended constraint data is first extracted from the cost-stage agent. This data includes design parameters such as material thickness, processing parameters such as cutting speed, and initial parameter values ​​such as the default material cost coefficient of 0.5. This approach ensures the integrity of the underlying data for subsequent derivations, thus providing reliable input for parameter dependency analysis.

[0086] In one possible implementation, a constraint propagation mechanism is used to deduce the dependencies between parameters layer by layer for the constraints in the extended constraint set.

[0087] For example, suppose that in the design of a mechanical part, increasing the material thickness will lead to a longer processing time. In this case, the mechanism will start from the thickness parameter, propagate to the time parameter, and then to the cost parameter, and find that when the thickness increases from 10mm to 12mm, the time increase is 20%. This derivation process involves constructing a dependency graph, where each node represents a parameter and the edges represent the influence relationship. The range boundary is calculated through iterative algorithms such as forward propagation, thus clearly revealing the chain effect between parameters.

[0088] For example, based on the current range of parameter value changes, calculate the direct cost increment and cumulative cost increment caused by adjusting each parameter, and generate a cost data table.

[0089] Specifically, in the automotive parts processing business, if the cutting speed is adjusted from 100m / min to 120m / min, the direct increase might be an increase of 5 yuan per piece in energy consumption, while the cumulative increase would add up to 15 yuan per piece in the total cost. The data table would list columns such as parameters, adjustment values, direct increases, and cumulative values. This calculation is based on formulas such as Increment = Base Value × Rate of Change, but avoids complex numerical values. It uses business logic, such as multiplying the energy unit price by the time difference, to reason and ensure that the data in the table supports decision-making.

[0090] In one possible implementation, the cost data table is classified and divided according to the constraint correspondence between the design stage and the processing stage, resulting in a cost subset for the design stage and a cost subset for the processing stage.

[0091] For example, in the design of electronic product casings, the design phase subset may include the material costs associated with thickness adjustments, while the manufacturing phase subset focuses on the labor costs of cutting parameters. By classifying the data through matching rules such as keyword association, it is helpful to isolate the influence of different phases and avoid cross-phase confusion.

[0092] For example, the cost subset of the design phase can be injected into the constraint update module of the design phase agent to update the cost-related constraints in the constraint set of the design phase.

[0093] In one possible implementation, this is similar to a feedback loop.

[0094] For example, when the increased thickness of the cost subset results in a 10% overrun of the total cost, the module will adjust constraints such as reducing the upper limit of thickness from 12mm to 11mm, thereby optimizing design constraints and achieving the technical goal of cost control, such as reducing overall production costs.

[0095] In one possible implementation, a subset of processing stage costs is injected back into the processing stage agent constraint update module to update cost-related constraint terms in the processing stage constraint set.

[0096] For example, in precision instrument manufacturing, if a subset indicates that the speed adjustment increment is too high, the module will update constraints such as limiting the speed range to 110 m / min, which ensures that the manufacturing process is more economical.

[0097] For example, the merged and updated design stage constraint set, manufacturing stage constraint set, and cost stage constraint set are combined and, after a consistency check, form a complete constraint set containing cost data.

[0098] Specifically, consistency checks involve verifying that there are no conflicts between parameters, such as the compatibility between design thickness and processing speed. If there are no anomalies, they are integrated into a unified set, which can bring cross-stage optimization effects to the business, such as improving the overall efficiency of the product.

[0099] In step S104, after receiving the complete set of constraints, the agents at each stage perform real-time perception and detect conflicts. If the direction of parameter optimization is opposite to the direction of processing difficulty and conflicts with the direction of cost, the conflict points on the geometric features are marked to obtain a conflict set containing the conflict points.

[0100] Obtain the complete constraint set uploaded by the agents at each stage. The stage agents perform real-time parsing of the complete constraint set to obtain the current processing state set. Perform constraint conflict detection on the current processing state set to determine if the parameter optimization direction is opposite to the processing difficulty direction. If the parameter optimization direction is opposite to the processing difficulty direction, further determine if this opposite direction conflicts with the cost direction. If the parameter optimization direction, processing difficulty direction, and cost direction all conflict simultaneously, extract the geometric features corresponding to the conflict location. Mark all points satisfying the three-way conflict conditions on these geometric features to obtain a conflict point set. Associate the conflict point set with the corresponding geometric features and output it to the next stage agent to form a complete conflict set containing the conflict points.

[0101] like Figure 9As shown, this invention illustrates the message interaction sequence between the design agent, processing agent, and cost agent. The sequence diagram is arranged vertically in chronological order and mainly includes five stages: Stage 1 is the constraint propagation stage, including: the design agent sending geometric and strength constraint propagation messages to the processing and cost agents; and the cost agent feeding back cost data to the design and processing agents; Stage 2 is the conflict detection stage, including: conflict detection requests and conflict marking responses between the agents; Stage 3 is the negotiation and optimization stage, including: the exchange of negotiation suggestions; and final parameter confirmation. Solid arrows represent active requests, dashed arrows represent feedback responses, and dotted arrows represent negotiation messages.

[0102] like Figure 10 As shown in the figure, this invention demonstrates the detection and marking distribution of conflict points in the constraint parameter space. The horizontal axis represents the relative values ​​of design parameters, and the vertical axis represents the relative values ​​of processing parameters. The gray area within the dashed border represents the feasible solution space. Circular gray dots represent feasible parameter points located within the feasible solution space; X-shaped black dots represent detected conflict points, mainly distributed near the boundaries of the feasible solution space, including types such as cost overrun conflicts, processing difficulty conflicts, and quality-cost conflicts; square white dots represent conflict points that have been resolved after being processed by the conflict negotiation mechanism. The dotted line represents the constraint boundary. This distribution diagram visually demonstrates the spatial location and resolution status of conflict points.

[0103] In one embodiment, the conflict detection between the parameter optimization direction, the machining difficulty direction, and the cost direction is achieved through directional vectorization mapping: the strengthening or weakening trend of the parameter is defined as a positive or negative vector. When the design agent requires the parameter to change in the direction of reducing cost (such as reducing wall thickness), while the machining agent determines, based on preset machine tool load rules, that this change will lead to a decrease in machining stability (i.e., the machining difficulty vector is opposite to the optimization vector), and the cost agent determines that the additional labor cost caused by the increase in machining difficulty exceeds the material saving cost, the system automatically identifies a three-way conflict. This vectorization comparison does not require complex numerical fitting, but rather achieves rapid location of the conflict point by judging the "gain / loss" logical evaluation of each agent's change trend of the same parameter.

[0104] For example, in one possible implementation, obtaining the complete set of constraints uploaded by agents at each stage can be understood as a data aggregation process in a multi-agent collaborative system.

[0105] Specifically, such systems typically involve multiple agents across the design, manufacturing, and cost phases. Each agent is responsible for maintaining and uploading its local constraint set, which includes parameters such as material thickness, manufacturing precision, and budget constraints. Through a central coordination module, these constraint sets are collected and integrated in real time to form a unified and complete constraint set, ensuring the comprehensiveness of subsequent analyses.

[0106] For example, in the automotive parts manufacturing business, the intelligent agent may upload constraints including the geometric dimensions and strength requirements of the parts during the design phase, while the intelligent agent uploads constraints on the processing path and machine tool load during the manufacturing phase, and constraints on material and labor costs are added during the cost phase. When these constraints are aggregated, they form a set covering the entire production chain, providing a foundation for real-time analysis.

[0107] In one possible implementation, the process of obtaining the current processing state set by real-time parsing of the complete set of constraints by a staged intelligent agent is actually the decomposition of constraints and state mapping using a parsing engine.

[0108] Specifically, the parsing engine scans the variable dependencies in the constraint set, such as mapping dimensional parameters in the design constraints to machine tool settings in the machining state, thereby generating a state set that describes the current production progress, parameter values, and potential deviations.

[0109] For example, when manufacturing an engine block, the parser may identify the current machining status, including the deviation between the actual measured value and the design value of the block wall thickness, as well as the machine tool speed setting. If the constraint set specifies that the wall thickness must be between 2.5mm and 3.0mm, the parser will output a state set listing the state items with an actual wall thickness of 2.7mm, which helps with subsequent conflict detection.

[0110] For example, the step of performing constraint conflict detection on the current set of processing states and determining whether the direction of parameter optimization is opposite to the direction of processing difficulty can be achieved by comparing direction vectors.

[0111] Specifically, parameter optimization might aim to reduce size to lower material costs, while machining difficulty might aim to increase size to reduce machining risks. If the two are opposite—one decreasing and the other increasing—further judgment will be triggered. This detection ensures production efficiency in business operations. For example, in parts machining, if the optimization direction is to reduce the hole diameter to save material, but the machining difficulty direction is to increase the hole diameter to avoid tool breakage, this creates an opposition, and the system will mark it as a potential conflict point.

[0112] In one possible implementation, if the direction of parameter optimization is opposite to the direction of processing difficulty, the process of determining whether the opposite direction conflicts with the cost direction involves multi-dimensional vector analysis.

[0113] Specifically, the cost direction may be towards minimizing expenditures, and a conflict is identified if the opposite direction leads to additional processing steps that increase costs.

[0114] For example, in cylinder block machining, if additional milling is required in the opposite direction to adjust the dimensions, this will increase the negative impact of cost. The system determines whether a three-way conflict has occurred by calculating the cost increment, such as additional labor costs, thereby avoiding losses caused by blind optimization.

[0115] For example, if the direction of parameter optimization, the direction of processing difficulty, and the direction of cost all conflict simultaneously, the implementation of extracting the geometric features corresponding to the location of the conflict can be done using geometric modeling tools.

[0116] Specifically, the system extracts features such as surfaces or edges from the CAD model, and conflict parameters are mapped onto these features. For example, it identifies specific surface areas with wall thickness conflicts on parts. This can accurately locate the source of the problem in business operations and improve debugging efficiency.

[0117] In one possible implementation, the step of marking all points that satisfy the three-way conflict condition on the geometric feature to obtain the set of conflict points is accomplished through point cloud analysis.

[0118] Specifically, for the extracted geometric features, the system iterates through each point and checks whether the conflict conditions in three directions are met simultaneously. For example, adjusting the parameters of the point may lead to optimization failure, increased difficulty, and increased cost. Then, these points are marked to form a set.

[0119] For example, on the curved surface of an engine block, if optimizing the wall thickness at certain points increases the machining difficulty and raises the cost, the system will mark these points to form a set. This set, once associated, can guide engineers to modify the design in a targeted manner, resulting in lower iteration costs and higher production consistency in the business.

[0120] For example, the process of associating the set of conflict points with their corresponding geometric features and outputting them to the next stage agent to form a complete set of conflict points is achieved through data encapsulation and transmission mechanisms.

[0121] Specifically, after association, the set is packaged into structured data and sent to downstream agents, such as agents in the optimization stage, for further adjustments. This ensures the continuous transmission of conflict information and improves robustness throughout the manufacturing process.

[0122] Step S105: For each conflict point in the conflict set, an adjustment suggestion is proposed by agents at each stage using a conflict negotiation mechanism to obtain a suggestion set after multi-party negotiation. The suggestion set is then input into a multi-objective optimization mechanism, and a feasible parameter range is obtained through iterative calculation. The initial geometric feature constraints are updated according to the feasible parameter range and propagated to agents at all stages.

[0123] Through a conflict negotiation mechanism, specific information about each conflict point is obtained from the conflict set. Pre-defined negotiation rules are used to integrate the feedback from agents at each stage, resulting in a preliminary set of adjustment suggestions. Based on this preliminary set, the results of multi-party negotiations are summarized and organized. A multi-objective optimization mechanism is used to assign weights to the various components of the suggestion set, determining the optimized suggestion priority sequence. Through iterative calculation, key parameters are extracted from the optimized suggestion priority sequence. Multiple iterations are performed to adjust the dependencies between parameters, resulting in a feasible parameter range that meets the constraints. Based on the feasible parameter range and the initial definition of geometric features, the feature constraints are dynamically updated, obtaining an updated set of constraints. For the updated set of constraints, the information transmission needs between agents at each stage are analyzed. Constraints are distributed to the corresponding agents through a pre-defined propagation mechanism, determining whether all relevant nodes are covered. If all relevant nodes are covered, the current constraint distribution status is recorded, and feedback data after distribution is obtained. If not, the propagation path is readjusted to determine the final propagation coverage result. By analyzing the feedback data, we can determine how well the agents adapt to the constraints at each stage. The feedback data is then categorized and organized using an information processing step to obtain a status report for each agent after executing the constraints.

[0124] The preset propagation mechanism is a master-slave tree network propagation mechanism. The master node is the constraint acquisition and transmission module, and the slave nodes are the intelligent agents at each stage. The propagation path priority is: master node → design stage intelligent agent → processing stage intelligent agent → cost stage intelligent agent. The backup path is: master node → cost stage intelligent agent → processing stage intelligent agent → design stage intelligent agent. When the delay of the master path is >1s or the data packet loss rate is >3%, the system automatically switches to the backup path.

[0125] In one possible implementation, the preset negotiation rule is a weight allocation rule of "safety margin priority", with a strength safety margin ≥15% as the first priority, a processing difficulty threshold ≤0.7 as the second priority, and a cost increment ≤10% as the third priority, with weight ratios of 0.4, 0.3, and 0.3, respectively.

[0126] During the conflict negotiation process, the adjustment suggestions proposed by the agents at each stage include suggestion priority labels. When multiple parties cannot reach a consensus, the system automatically allocates weights based on the principle of "safety margin priority": that is, prioritizing ensuring that the strength requirements are not exceeded, and then seeking the lowest cost solution within the range of processing feasibility. Through this hierarchical negotiation decision-making, the system can avoid getting stuck in infinite iteration loops and ensure that it outputs a final feasible solution space that meets the practical requirements of engineering within a limited number of computation steps.

[0127] The specific quantitative standards for the "safety margin priority" principle are as follows: 1. Strength safety margin ≥ 15% (i.e., actual strength ≥ design strength × 1.15), this is the first priority constraint that cannot be broken; 2. Processing difficulty threshold ≤ 0.7 (difficulty levels are divided into 0-1, 0 for no difficulty, 1 for extreme difficulty), this is the second priority constraint; 3. Cost increment ≤ 10% (compared to the baseline cost), this is the third priority constraint. Weight allocation adopts the Analytic Hierarchy Process (AHP), with fixed weight percentages: design strength 0.4, processing feasibility 0.3, cost control 0.3. When the consensus among multiple parties is lower than 0.7, automatic weight reallocation is triggered to ensure that the first priority constraint is not violated.

[0128] like Figure 5 As shown, this invention illustrates the Pareto front distribution of design cost versus manufacturing quality in a multi-objective optimization process. The horizontal axis represents the relative value of design cost, and the vertical axis represents the relative value of manufacturing quality. Black dots represent the Pareto optimal solution set, which is undominant in the cost-quality tradeoff; gray squares represent feasible solutions located inside the Pareto front; light gray triangles represent the initial solution before optimization. The solid curve fits the Pareto front boundary, with the upper left region representing the direction of the ideal solution. The results show that the multi-objective optimization mechanism of this invention can effectively solve for the Pareto optimal solution set, providing decision-makers with feasible solutions for different cost-quality tradeoffs.

[0129] like Figure 7 As shown, this invention illustrates a surface illustrating the trade-off relationship between three objectives: design cost, manufacturing quality, and production efficiency. In the figure, the X-axis represents the relative value of design cost, the Y-axis represents the relative value of manufacturing quality, and the Z-axis represents the percentage of production efficiency. The surface uses a grayscale gradient to represent the efficiency distribution, with lighter colors indicating higher efficiency. The surface shape reveals that production efficiency reaches a high level when cost and quality requirements are low; however, as cost and quality requirements increase, production efficiency exhibits a non-linear decreasing trend. Star-shaped markers represent the optimal balance point between cost, quality, and efficiency. This three-dimensional trade-off surface provides intuitive visual decision support for multi-objective optimization, helping decision-makers find the best trade-off among the three objectives.

[0130] In one possible implementation, when the system initiates a conflict negotiation mechanism, it first extracts detailed information about each conflict point from the marked conflict set. For example, in the field of automotive parts processing, a conflict point might involve the bending radius parameter of the car body frame. The optimization direction requires a reduction in radius to improve aerodynamic performance, but this increases processing difficulty and costs. The negotiation rules are preset to an integration method based on a priority matrix. The specific process includes collecting feedback from design-stage agents, such as "reducing the radius can reduce the drag coefficient by 0.05," from manufacturing-stage agents, such as "reducing the radius requires additional precision molds, increasing processing time by 20%," and from cost-stage agents, such as "upgrading the mold will incur an additional expenditure of 150,000 yuan." Through these feedbacks, the system applies rules such as a weighted average method to integrate and calculate the consensus degree of each feedback. For example, the consensus degree threshold is set to 0.7. If it exceeds this threshold, a preliminary adjustment suggestion is formed, such as "adjusting the radius to the median value of 1.2 meters, balancing performance and cost."

[0131] For example, when summarizing the results of multi-party negotiations, the system organizes the initial set of suggestions, assuming the set contains three suggestions: performance-first, cost-first, and a balanced solution. A multi-objective optimization mechanism, such as the Pareto optimization algorithm, is employed. Its principle is to find the optimal frontier through non-dominated sorting. The specific analysis process involves first defining an objective function, such as maximizing aerodynamic efficiency for performance and minimizing additional expenditure for cost. Then, weights are assigned to the suggestions, such as performance weight 0.4, difficulty weight 0.3, and cost weight 0.3. A genetic algorithm iteratively generates subsets to determine the priority sequence, such as "balanced solution first, followed by performance-first."

[0132] Key algorithm parameters: 1. Pareto optimization algorithm: number of non-dominated sorting iterations = 100, crowding distance threshold = 0.1, number of optimal solutions selected = 5; 2. Genetic algorithm: population size = 50, crossover probability = 0.6, mutation probability = 0.05, termination condition is no better solution appearing for 10 consecutive generations; 3. Iterative calculation termination condition: conflict rate ≤ 0.1% (conflict rate = number of conflict points / total number of geometric feature points × 100%), and no new conflict points are generated for 3 consecutive iterations.

[0133] In one possible implementation, key parameters such as bending radius and material thickness are extracted from a priority sequence, and dependencies are handled using iterative calculation methods. For example, the initial radius of 1.5 meters and the thickness of 2 mm are interdependent. If the radius decreases, the thickness needs to be increased to maintain strength. The solution is approximated by iterative adjustments such as the Newton-Raphson method, and iterated multiple times until the parameter range, such as radius 1.1-1.3 meters and thickness 2.1-2.4 mm, meets all constraints.

[0134] For example, by combining the initial definition of geometric features such as the equation of the frame surface, the feature constraints are dynamically updated to obtain a new set such as "lower radius limit 1.1 meters, upper radius limit 1.3 meters".

[0135] In one possible implementation, the information transmission requirements are analyzed, and constraints are distributed through a propagation mechanism such as a tree network. If all nodes, such as design, manufacturing, and testing agents, are covered, the status is recorded and feedback data, such as "the manufacturing agent has a fitness rate of 95%", is obtained. Otherwise, the path is adjusted, such as by adding a backup link, to ensure coverage.

[0136] For example, by classifying and organizing feedback data, and using information processing steps such as clustering algorithms to group data, a status report can be obtained, such as "Designing intelligent agents is fully adaptable, but manufacturing intelligent agents requires fine-tuning."

[0137] In one possible implementation, this integration mechanism ensures the continuity of the processing flow. For example, in subsequent verification, the adjusted parameter range can be directly applied to prototype testing, reducing the number of iterations.

[0138] Step S106: Repeat real-time perception and conflict detection on the updated constraint set. If there are no new conflict points, output the final feasible solution space. If there are residual conflicts, return to the conflict negotiation mechanism to continue iterating until a conflict-free feasible solution space is obtained.

[0139] By continuously monitoring the constraint set, data is collected at each stage using pre-defined perception tools to obtain real-time status information. Based on this information, potential conflict points are initially screened to determine if abnormal data exists; if so, they are marked as potential conflict points. For each marked potential conflict point, its specific location and related parameters are obtained, and compared with a pre-defined threshold range to determine if it is an actual conflict point. If determined to be an actual conflict point, the conflict negotiation module classifies the relevant data to obtain a categorized conflict dataset. Based on the categorized conflict dataset, conflict points are sorted using pre-defined priority rules to obtain a sorted conflict processing sequence. Using the sorted conflict processing sequence, the corresponding negotiation strategy is called one by one to adjust the parameters, resulting in an adjusted temporary solution space. A cyclic verification mechanism is used to perform multiple rounds of checks on the adjusted temporary solution space to determine if a conflict-free state has been reached; if so, the final feasible solution space is output.

[0140] like Figure 4As shown in the figure, this invention demonstrates the trend of the number of conflicts as a function of the number of iterations during the iterative convergence process of conflict negotiation. The horizontal axis represents the number of iterations, ranging from 1 to 20 rounds; the vertical axis represents the number of conflicts. The three curves represent three typical scenarios: the solid dotted curve represents a simple constraint scenario, exhibiting rapid convergence characteristics, reaching convergence around the 5th round; the dashed square dotted curve represents a medium-complexity scenario, reaching convergence around the 10th round; and the dotted-triangle curve represents a complex constraint scenario, reaching convergence around the 15th round. The horizontal dashed line represents the convergence threshold (number of conflicts less than 1). The results show that the iterative convergence mechanism of this invention can effectively handle constraint scenarios of different complexities, achieving conflict resolution within a finite number of iterations.

[0141] For example, in a smart manufacturing system, continuous monitoring of the constraint set can be achieved by deploying a sensor network. These sensors collect data from equipment at various stages of the production line in real time, such as parameters like temperature, pressure, and speed, thereby generating real-time status information.

[0142] Specifically, this sensing tool can be a pre-built module based on the Internet of Things that scans the data stream once per second to ensure that any subtle changes are captured. For example, when a machine's rotation speed exceeds the normal range, the system will immediately record and generate a status report to support subsequent conflict detection.

[0143] In one possible implementation, during the initial screening based on the collected real-time status information, anomaly detection algorithms can be used to analyze the data. For example, the deviation of the current data from the historical average can be compared. If the deviation exceeds 10%, it is judged as abnormal data and marked as a potential conflict point. This screening process emphasizes rapid response to prevent small problems from escalating into major conflicts. For instance, on an automotive assembly line, if an abnormal temperature is detected during the welding stage, the system will mark it to prevent it from affecting the overall production rhythm.

[0144] For example, for a potential conflict point marked, its location, such as the specific coordinates of the welding station and parameters such as the temperature value of 120 degrees Celsius, is obtained. By comparing it with a preset threshold range, such as 100-110 degrees Celsius, if it exceeds the threshold, it is confirmed as an actual conflict point.

[0145] In one possible implementation, this comparison process involves multi-dimensional parameter checks to ensure accuracy, such as simultaneously verifying stress parameters to avoid misjudgments, thereby providing a reliable basis for conflict negotiation.

[0146] Specifically, after identifying the actual conflict points, the conflict negotiation module categorizes the data, classifying conflicts into resource-related conflicts such as equipment occupancy conflicts and parameter-related conflicts such as numerical deviation conflicts, resulting in a categorized dataset. This module is typically an integrated software framework that uses machine learning models for automatic classification. For example, in a supply chain system, it categorizes supplier delays and inventory shortages separately for targeted handling.

[0147] In one possible implementation, based on the categorized dataset, conflict points are ranked using a pre-defined priority rule, such as a scoring system based on the degree of impact. For example, conflicts affecting production downtime are ranked first, resulting in a ranking sequence. This rule can be a custom logic tree, ensuring that high-priority conflicts are processed first, thereby optimizing overall efficiency.

[0148] For example, by sorting the sequence, negotiation strategies such as parameter fine-tuning strategies are called one by one to make adjustments, resulting in a temporary solution space.

[0149] Specifically, this strategy might involve simulating temperature parameters to drop from 120 to 105, generating a set of temporary solutions that, in smart grid systems, can be used to balance load conflicts and avoid overload.

[0150] In one possible implementation, a cyclic verification mechanism is used for the temporary solution space, performing multiple rounds of checks, such as simulating the solution space three times, to check for conflicts. If a conflict is found, the final solution space is output.

[0151] For example, in traffic management systems, this can verify whether the adjusted traffic light parameters have eliminated congestion and achieved smooth traffic flow. This mechanism ensures stability through iteration, resulting in more reliable system performance.

[0152] This invention provides a multi-agent collaborative optimization system and system integrating design and manufacturing, mainly comprising:

[0153] The constraint acquisition and transmission module is used to acquire the geometric feature constraints and strength requirements generated in the design phase, extract parameter values ​​and material data from them, and use a constraint propagation mechanism to transmit the parameter values ​​and material data to the agents in the processing and cost phases to obtain the initial constraint set received in each phase.

[0154] The processing constraint extension module is used to calculate the increase in processing difficulty caused by changes in parameter values ​​after the agent receives the initial constraint set during the processing stage, using the constraint propagation mechanism, generating processing data and continuing to propagate it to the agent in the cost stage, thus obtaining an extended constraint set containing the processing data.

[0155] The cost constraint extension module is used to calculate the cost increment caused by the change of parameter values ​​after the agent receives the extended constraint set in the cost stage, using the constraint propagation mechanism to generate cost data and propagate the cost data back to the agents in the design and processing stages to obtain a complete constraint set containing cost data.

[0156] The conflict detection and marking module is used to perceive and detect conflicts in real time after the intelligent agent receives the complete set of constraints at each stage. If the direction of parameter optimization is opposite to the direction of processing difficulty and conflicts with the direction of cost, the conflict points on the geometric features are marked to obtain a conflict set containing the conflict points.

[0157] The conflict negotiation and optimization module is used to propose adjustment suggestions from agents at each stage for each conflict point in the conflict set using a conflict negotiation mechanism, resulting in a set of suggestions after multi-party negotiation. The set of suggestions is then input into a multi-objective optimization mechanism, and feasible parameter ranges are obtained through iterative calculation. The initial geometric feature constraints are updated according to the feasible parameter ranges and propagated to agents at all stages.

[0158] The iterative convergence module is used to repeatedly perceive and detect conflicts in real time on the updated constraint set. If there are no new conflict points, it outputs the final feasible solution space. If there are residual conflicts, it returns to the conflict negotiation mechanism to continue iterating until a conflict-free feasible solution space is obtained.

[0159] The emergency handling module is used to deal with abnormal scenarios such as data transmission interruption and agent failure. Specifically, it includes: 1. Data transmission fault tolerance: A local caching mechanism is adopted, with a caching time of ≥10 minutes. Cached data includes constraint sets for each stage, parameter adjustment records, and conflict detection results. When transmission is interrupted (judgment condition: 3 consecutive data transmission failures, single timeout = 3s), the cached data is automatically activated to continue the current step. After transmission is restored, the cached data is synchronized. 2. Agent failure fault tolerance: A redundant agent design is adopted, with one backup agent configured for each stage. When the primary agent fails (judgment condition: no response for 5 consecutive seconds), the backup agent starts and takes over the work within 2 seconds. The historical data of the failed agent is synchronized to the backup agent through a distributed database. 3. Data consistency fault tolerance: The CRC32 checksum algorithm is used to verify the transmitted data. When the verification fails, data retransmission is automatically triggered. The number of retransmissions is ≤3. If it still fails, a manual intervention reminder mechanism is activated (via system background pop-up window + email notification).

[0160] like Figure 11As shown, this invention demonstrates a typical application scenario of a multi-agent collaborative optimization system in industrial manufacturing. The upper part of the diagram contains three business units: the design center is responsible for CAD / CAE geometric modeling and strength analysis; the machining workshop is responsible for CNC machine tool process planning and quality inspection; and the cost control unit is responsible for ERP system cost accounting and resource optimization. The central part is the multi-agent collaborative optimization platform, integrating core functional modules such as constraint propagation, conflict detection, negotiation optimization, iterative convergence, and emergency handling. Bidirectional arrows indicate data interaction between each business unit and the collaborative optimization platform, including constraint data uploading and optimization suggestion distribution. The bottom shows the industrial product output after collaborative optimization. This scenario diagram clearly illustrates the application architecture of this invention in a real industrial manufacturing environment.

[0161] I. Specific Explanation of the Constraint Propagation Mechanism

[0162] The constraint propagation mechanism is a parameter association model based on a directed acyclic graph (DAG), and its specific implementation is as follows:

[0163] DAG structure construction:

[0164] • Geometric parameters from the design phase (such as aperture, wall thickness, surface accuracy, material strength, and load threshold) serve as parent nodes.

[0165] • Process parameters during the machining stage (such as cutting step distance, tool type, machining time, machine tool vibration, and tool wear) are used as sub-nodes.

[0166] • Cost factors in the cost stage (such as material unit price, energy consumption cost, labor cost, equipment depreciation, and scrap loss) are treated as sub-nodes.

[0167] • Nodes are linked through predefined logical functions to form a directed acyclic graph.

[0168] Node association logic function:

[0169] • Linear proportional function: For example, "when the hole diameter is ≥ 5mm, the cutting step distance = 0.8 × hole diameter; when the hole diameter is < 5mm, the cutting step distance = 0.5 × hole diameter". This function is determined based on actual machining experience to ensure a reasonable proportional relationship between machining efficiency and parameter changes.

[0170] • Stepped response function: For example, "when the wall thickness exceeds 10mm, the processing time = base processing time × (1 + 0.3 × (wall thickness - 10) / 10)". This function is determined based on the processing technology database. When the wall thickness increases beyond a threshold, the processing time increases in a stepped manner.

[0171] How the propagation mechanism works:

[0172] • Upon system startup, an initial DAG is constructed based on the geometric constraints and strength requirements established during the design phase.

[0173] When design parameters change, the system propagates the changes from parent nodes to child nodes through a Directed Acyclic Graph (DAG).

[0174] • Upon receiving a change, each node calculates the corresponding changes in the parameters of its child nodes based on a pre-defined logical function.

[0175] The propagation process employs a depth-first search algorithm to ensure that all dependencies are correctly calculated.

[0176] • During the propagation process, the system monitors the status changes between nodes in real time to ensure the accuracy and timeliness of the propagation.

[0177] Data transmission and processing:

[0178] • Data transmission is conducted using a unified JSON format, with core fields including parameter ID, value, tolerance, material number, strength threshold, and data type.

[0179] The transmission protocol uses TCP / IP to ensure data reliability and real-time performance.

[0180] • Packet loss rate is controlled below 0.3%, and timeout retransmission time is set to 3 seconds to ensure data transmission integrity.

[0181] II. Determination and Adjustment of Thresholds for High-Difficulty Factors

[0182] The determination of high-difficulty factors is based on the following thresholds, which were established through long-term engineering practice, industry standards, and experimental data:

[0183] Tool wear rate threshold:

[0184] • Threshold set at: 0.05 mm / min

[0185] • Basis for determination: Derived based on the service life standard (≥5000 pieces) of commonly used carbide cutting tools (model: WC-Co 6%). When the wear rate exceeds 0.05 mm / min, the tool life will be shortened to below 1000 pieces, leading to a surge in machining costs.

[0186] Adjustment method: In practical applications, fine-tuning can be made according to the tool type, material hardness, and machining accuracy requirements. For example, for high-precision machining, the threshold can be lowered to 0.03 mm / min; for roughing, it can be relaxed to 0.08 mm / min.

[0187] Machine tool vibration amplitude threshold:

[0188] Threshold set to: 2 micrometers

[0189] • Basis for determination: Refer to the permissible vibration values ​​for precision machining equipment in GB / T 17421.2-2021 "General Rules for Machine Tool Inspection". Exceeding this threshold will result in out-of-tolerance dimensional tolerances of the parts (≥±0.02mm).

[0190] Adjustment method: Adjust according to the machine tool type and machining accuracy requirements. For example, for high-precision CNC machine tools, the threshold can be set to 1.5 micrometers; for ordinary machine tools, it can be relaxed to 3 micrometers.

[0191] Processing difficulty level classification:

[0192] • Low difficulty: [0, 0.3] (processing time growth rate ≤ 10%)

[0193] Medium difficulty: [0.3, 0.7] (Processing time growth rate 10% (excluding 10%) - 30% (including 30%))

[0194] • High difficulty: (0.7-1] (processing time growth rate > 30%)

[0195] • Basis for classification: Based on actual processing experience, statistical analysis was used to determine the range of work hour growth rates corresponding to different difficulty levels. This range has been verified as effective in multiple actual projects.

[0196] III. Specific Logic and Implementation of Conflict Detection

[0197] Collision detection is achieved through direction vectorization mapping, and the specific logic is as follows:

[0198] Definition of parameter optimization direction:

[0199] • The optimization direction of parameters proposed by the intelligent agent during the design phase, such as "reducing wall thickness to reduce material costs", is represented by a vector as "negative direction" (reduction is negative, increase is positive).

[0200] Definition of processing difficulty direction:

[0201] • During the processing stage, the direction of difficulty change determined by the intelligent agent according to preset rules, such as "reducing the wall thickness will lead to an increase in processing difficulty", is represented by a vector as "positive" (increased difficulty is positive).

[0202] Definition of cost direction:

[0203] • The direction of cost changes calculated by the agent during the cost phase, such as "reducing wall thickness will lower material costs, but may increase processing difficulty, thereby increasing processing costs," the overall direction of cost changes depends on which impact is greater.

[0204] Conflict determination logic:

[0205] • Judgment criteria: The direction of parameter optimization is opposite to the direction of processing difficulty and conflicts with the direction of cost.

[0206] For example, if the parameter optimization direction is "reducing wall thickness" (negative), the processing difficulty direction is "increasing wall thickness" (positive), and the cost direction is "increasing cost" (positive), then a three-way conflict is formed.

[0207] The system automatically identifies conflict points by comparing vectors in various directions.

[0208] Conflict point marking implementation:

[0209] • Locate the specific position of the conflict point on the geometric features using geometric modeling tools.

[0210] • Use point cloud analysis technology to traverse each point on the geometric feature and check whether the three-way conflict condition is satisfied.

[0211] • Mark the points that meet the conditions to form a set of conflict points.

[0212] During point cloud analysis, the system automatically filters out boundary effects and measurement errors to ensure the accuracy of the markings.

[0213] IV. Specific Implementation of Multi-Objective Optimization Mechanism

[0214] The multi-objective optimization mechanism uses the Pareto optimization algorithm, and its specific implementation is as follows:

[0215] Objective function and weights:

[0216] • Objective 1: Design strength requirements (safety margin ≥ 15%), weight 0.4

[0217] • Objective 2: Processing feasibility (processing difficulty threshold ≤ 0.7), weight 0.3

[0218] • Objective 3: Cost control (cost increment ≤ 10%), weight 0.3

[0219] • Weight allocation basis: Based on engineering practice and industry standards, the weights are determined using the Analytic Hierarchy Process (AHP) to ensure that key indicators are given priority.

[0220] Pareto optimization algorithm process:

[0221] • Initialization: Based on the initial parameters from the design phase, generate 100 random parameter combinations to form an initial solution set.

[0222] • Non-dominated sorting: Sort the solution set according to the three objective function values ​​and find the non-dominated solutions (i.e., solutions that are not inferior to other solutions on all objective functions).

[0223] • Crowding Degree Calculation: Calculate the crowding degree of each non-dominated solution to ensure the diversity of the solution set and avoid over-concentration of solutions.

[0224] • Selection: Select elite solutions to advance to the next generation, ensuring the preservation of high-quality solutions.

[0225] • Crossover and mutation: Perform crossover and mutation operations on elite solutions to generate new solutions and maintain population diversity.

[0226] • Iteration: Repeat the above steps until the convergence condition is met.

[0227] Convergence condition:

[0228] • Conflict rate ≤ 0.1% (Conflict rate = number of conflict points / total number of geometric feature points × 100%)

[0229] No new conflict points were generated in three consecutive iterations.

[0230] • Maximum number of iterations: 100, to ensure the algorithm converges within a reasonable time.

[0231] Optimize result processing:

[0232] • Select a solution from the Pareto optimal solution set that satisfies the "safety margin priority" principle.

[0233] Prioritize ensuring design strength requirements, and then seek the lowest cost solution within the scope of fabrication feasibility.

[0234] • Output feasible parameter range, used to update geometric feature constraints

[0235] • Optimization results are presented through visualization tools to facilitate engineers' understanding and decision-making.

[0236] V. Specific methods for threshold adjustment

[0237] Safety margin threshold:

[0238] • Benchmark threshold: 15% (actual strength ≥ design strength × 1.15)

[0239] Adjustment method: For critical components or high-risk applications, the allowance can be increased to 20%; for non-critical components, it can be reduced to 10%.

[0240] • Adjustment criteria: Based on the importance of the component in the product and the severity of the consequences of failure, the adjustment is determined by the design engineer based on experience.

[0241] Processing difficulty threshold:

[0242] • Baseline threshold: 0.7 (Difficulty level 0-1, 0 is no difficulty, 1 is the most difficult)

[0243] Adjustment method: For high-precision machining, the threshold can be set to 0.6; for rough machining, it can be relaxed to 0.8.

[0244] • Adjustment criteria: Determined by the process engineer based on the machine tool's precision level, the complexity of the machining process, and the operator's skill level.

[0245] Cost increment threshold:

[0246] • Benchmark threshold: 10% (relative to benchmark cost)

[0247] Adjustment method: For cost-sensitive products, the threshold can be lowered to 5%; for high-end products, it can be relaxed to 15%.

[0248] • Adjustment basis: Determined by the cost control department based on the company's cost control strategy, product market positioning, and customer budget range.

[0249] VI. Specific Implementation of the Emergency Response Mechanism

[0250] Data transmission fault tolerance:

[0251] • Employs a local caching mechanism with a cache time of ≥10 minutes.

[0252] • The cached data includes constraint sets for each stage, parameter adjustment records, and conflict detection results.

[0253] • When transmission is interrupted (judgment condition: 3 consecutive data transmission failures, single timeout time = 3s), the buffered data is automatically used to continue the current step.

[0254] After transmission resumes, the system automatically synchronizes cached data to ensure data consistency.

[0255] Agent fault tolerance:

[0256] • A redundant agent design is adopted, with one backup agent configured for each stage.

[0257] • When the primary agent fails (criteria: no response for 5 consecutive seconds), the backup agent starts up and takes over the work within 2 seconds.

[0258] Historical data from the faulty agent is synchronized to the backup agent via a distributed database to ensure operational continuity.

[0259] Data consistency and fault tolerance:

[0260] • The CRC32 checksum algorithm is used to verify the transmitted data.

[0261] • When verification fails, data retransmission will be automatically triggered, with a maximum of 3 retransmissions.

[0262] If the system still fails, a manual intervention and alert mechanism will be activated (via system background pop-ups and email notifications) to ensure system reliability.

[0263] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A multi-agent collaborative optimization method integrating design and manufacturing, characterized in that, include: Obtain the geometric feature constraints and strength requirements generated in the design phase, extract parameter values ​​and material data from them, and transmit the parameter values ​​and material data to the processing phase agent and the cost phase agent through the constraint propagation mechanism to generate the initial constraint set received by the agent in each phase. The constraint propagation mechanism adopts a parameter association model based on a directed acyclic graph. The parameter association model takes the geometric parameters in the design stage as parent nodes, the process parameters in the processing stage as child nodes, and the cost factors in the cost stage as grandchild nodes. The nodes are associated with each other through a preset logical function to form a directed acyclic graph. The preset logical function includes a linear proportional function and a step response function. When the value of the parent node changes, the constraint propagation mechanism propagates the change from the parent node to the child node through the directed acyclic graph. After each node receives the change, it calculates the corresponding change of the child node parameter according to the preset logic function, automatically triggering the state update of the downstream node, realizing the synchronous propagation of constraint information across stages, so that the impact caused by parameter changes can be perceived by the relevant intelligent agents in real time. Based on the geometric feature constraints and strength requirements, the parameter values ​​and material data are separated to form a propagation constraint set that includes geometric constraints and strength constraints; The propagation constraint set is transmitted to the processing stage agent, which determines the processing route based on the geometric constraint portion in the initial constraint set and outputs the processing process constraints. The propagation constraint set is transmitted to the cost stage agent, which calculates the cost value corresponding to each processing technology constraint based on the strength constraint part in the initial constraint set and the material data, and outputs the cost constraint set. In the processing stage agent, the incremental processing difficulty caused by the change in parameter value is calculated through the constraint propagation mechanism, processing data is generated and transmitted to the cost stage agent, forming an extended constraint set containing the processing data; In the cost stage agent, the cost increment caused by the change of parameter value is calculated through the constraint propagation mechanism, cost data is generated and transmitted back to the design stage agent and the processing stage agent to form a complete constraint set containing the cost data. For the complete set of constraints, real-time perception and conflict detection are performed in the agents at each stage, and conflict points on geometric features are marked to form a conflict set containing the conflict points.

2. The method according to claim 1, characterized in that, Also includes: The conflict points in the conflict set are handled by a conflict negotiation mechanism to generate a set of suggestions after multi-party negotiation. The feasible parameter range is iteratively calculated by combining a multi-objective optimization mechanism to update the geometric feature constraints and propagate them to agents at all stages.

3. The method according to claim 2, characterized in that, In the processing stage agent, the increment in processing difficulty caused by changes in parameter values ​​is calculated through the constraint propagation mechanism, processing data is generated and transmitted to the cost stage agent, forming an extended constraint set containing the processing data, including: For the initial constraint set in the intelligent agent during the processing stage, the parameter value changes are decomposed to determine the key factors affecting the processing difficulty, thus forming a set of factors influencing the processing difficulty. By analyzing the key factors affecting the processing difficulty set using preset rules, high-difficulty factors are marked and a list of high-difficulty factors is generated. Based on the list of high-difficulty factors, the processing data is categorized and organized to generate a subset of processing data containing high-difficulty factors, and the distribution of processing difficulty increments is determined. The processing data subset is transmitted to the cost stage agent through the incremental distribution of processing difficulty, forming the extended constraint set; For the extended constraint set, the incremental processing difficulty and cost constraints are correlated to generate cost impact data, which is then filtered, outliers are marked, and the final processing and cost correlation data is formed.

4. The method according to claim 2, characterized in that, The process involves handling conflict points in the conflict set through a conflict negotiation mechanism, generating a multi-party negotiated suggestion set, iteratively calculating the feasible parameter range using a multi-objective optimization mechanism, updating the geometric feature constraints, and propagating them to agents at all stages, including: For each conflict point in the conflict set, feedback from agents at each stage is integrated through preset negotiation rules to form a preliminary set of adjustment suggestions; Based on the preliminary set of adjustment suggestions, the results of multi-party consultations are summarized, and weights are allocated through the multi-objective optimization mechanism to determine the optimized suggestion priority sequence. Through iterative calculation, key parameters are extracted from the proposed priority sequence, and the parameter dependencies are adjusted cyclically to generate a feasible parameter range that meets the constraints. Based on the feasible parameter range, the geometric feature constraints are dynamically updated to form an updated set of constraints. For the updated set of constraints, the information transmission needs of the agents at each stage are analyzed, the constraints are distributed through a preset propagation mechanism, the distribution status is recorded and the feedback data is organized to generate the state information of each agent after executing the constraints.

5. The method according to claim 2, characterized in that, In the cost-stage agent, the cost increment caused by parameter value changes is calculated through the constraint propagation mechanism, cost data is generated, and it is transmitted back to the design-stage agent and the processing-stage agent, forming a complete constraint set containing the cost data, including: For the constraints in the extended constraint set, derive the parameter dependencies and determine the range of parameter value variations; Based on the range of parameter value changes, calculate the direct cost increment and cumulative cost increment, generate a cost data table, and classify it into a cost subset for the design stage and a cost subset for the processing stage. The cost subsets of the design phase and the cost subsets of the processing phase are injected into the corresponding agents, and the constraint set is updated to form the complete constraint set.

6. The method according to claim 2, characterized in that, For the complete set of constraints, real-time perception and conflict detection are performed in the agents at each stage, and conflict points on geometric features are marked to form a conflict set containing the conflict points, including: The complete set of constraints is analyzed by the intelligent agents at each stage to generate the current processing state set; For the current set of processing states, conflicts are detected between the direction of parameter optimization and the direction of processing difficulty and cost. Geometric feature points that meet the conflict conditions are marked to form the conflict set.

7. The method according to claim 2, characterized in that, The step of dynamically updating the geometric feature constraints based on the feasible parameter range to form an updated set of constraints includes: By continuously monitoring the updated set of constraints, data is collected at each stage to screen potential conflict points. For the potential conflict points, the actual conflict points are identified and classified, sorted by priority rules, and the parameters are adjusted one by one to generate a temporary solution space. The temporary solution space is verified by looping to determine whether a conflict-free state has been reached, and the final feasible solution space is output.

8. A multi-agent collaborative optimization system integrating design and manufacturing, characterized in that, The system includes: The constraint acquisition and transmission module is used to acquire the geometric feature constraints and strength requirements generated in the design phase, extract parameter values ​​and material data from them, and use a constraint propagation mechanism to transmit the parameter values ​​and material data to the agents in the processing and cost phases to obtain the initial constraint set received in each phase. The constraint propagation mechanism adopts a parameter association model based on a directed acyclic graph. The parameter association model takes the geometric parameters in the design stage as parent nodes, the process parameters in the processing stage as child nodes, and the cost factors in the cost stage as grandchild nodes. The nodes are associated with each other through a preset logical function to form a directed acyclic graph. The preset logical function includes a linear proportional function and a step response function. When the value of the parent node changes, the constraint propagation mechanism propagates the change from the parent node to the child node through the directed acyclic graph. After each node receives the change, it calculates the corresponding change of the child node parameter according to the preset logic function, automatically triggering the state update of the downstream node, realizing the synchronous propagation of constraint information across stages, so that the impact caused by parameter changes can be perceived by the relevant intelligent agents in real time. The processing constraint extension module is used to calculate the increase in processing difficulty caused by changes in parameter values ​​after the agent receives the initial constraint set during the processing stage, using the constraint propagation mechanism, generating processing data and continuing to propagate it to the agent in the cost stage, thus obtaining an extended constraint set containing the processing data. The cost constraint extension module is used to calculate the cost increment caused by the change of parameter values ​​after the agent receives the extended constraint set in the cost stage, using the constraint propagation mechanism to generate cost data and propagate the cost data back to the agents in the design and processing stages to obtain a complete constraint set containing cost data. The conflict detection and marking module is used to perceive and detect conflicts in real time after the intelligent agent receives the complete set of constraints at each stage. If the direction of parameter optimization is opposite to the direction of processing difficulty and conflicts with the direction of cost, the conflict points on the geometric features are marked to obtain a conflict set containing the conflict points. The conflict negotiation and optimization module is used to propose adjustment suggestions from agents at each stage for each conflict point in the conflict set using a conflict negotiation mechanism, resulting in a set of suggestions after multi-party negotiation. The set of suggestions is then input into a multi-objective optimization mechanism, and feasible parameter ranges are obtained through iterative calculation. The initial geometric feature constraints are updated according to the feasible parameter ranges and propagated to agents at all stages. The iterative convergence module is used to repeatedly perceive and detect conflicts in real time on the updated constraint set. If there are no new conflict points, it outputs the final feasible solution space. If there are residual conflicts, it returns to the conflict negotiation mechanism to continue iterating until a conflict-free feasible solution space is obtained.

9. A computer-readable storage medium, characterized in that, The storage medium includes a stored program, wherein, when the program is executed, the method described in any one of claims 1 to 7 is performed by a processor.

10. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1-7.

Images