Standardized process digital twin system and method for aviation equipment quality optimization

Abstract

The invention provides a standardized process digital twinning system and method for aviation equipment quality optimization, and relates to the technical field of digital twinning and information physical system auxiliary aviation equipment, and the system comprises an extraction module L1, a support module L2, an application verification module L3, an analysis module L4 and an optimization application module L5. By constructing the standardized process digital twin system for aviation equipment quality optimization, data-driven closed-loop optimization is realized, and a data closed loop from standard application to standard revision is constructed, so that the standard is based on aviation equipment development data, and the scientificity and timeliness of the aviation equipment standard are improved; through feedforward simulation verification and continuous feedback optimization based on operation data, standard defects can be found and repaired, and a solid foundation is provided for the whole life cycle of aviation equipment; by means of practical application and verification of the aviation equipment welding standard and feedback and optimization of the application process of the aviation equipment welding standard, the aviation equipment welding standard defects are successfully repaired.

Description

Standardized process digital twin system and method for optimizing the quality of aviation equipment Technical Field

[0001] This invention relates to the field of digital twin and cyber-physical system-assisted aviation equipment technology, specifically to a standardized process digital twin system and method for optimizing the quality of aviation equipment. Background Technology

[0002] As a typical complex engineering system, aerospace equipment systems rely heavily on a vast and sophisticated set of standards throughout their entire lifecycle, from design and manufacturing to operation, maintenance, and decommissioning. These standards cover various aspects, including materials, processes, design, testing, safety, and reliability. However, existing standards and standardization processes suffer from inherent flaws: static text carriers and semantic gaps. Traditional standards primarily exist in unstructured or semi-structured text formats such as PDF and Word, relying on manual interpretation by engineers. This leads to ambiguity in understanding, low application efficiency, and difficulty in direct integration and automated execution by industrial software such as Computer-Aided Design (CAD), Engineering Simulation (CAE), and Manufacturing Execution System (MES), thus creating a "semantic gap."

[0003] The disconnect between application and feedback: Valuable data on the application effects, compliance data, deviations, and quality issues caused by standards in actual engineering projects are often scattered across different systems, such as Product Lifecycle Management (PLM) and Maintenance, Repair, and Operations (MRO) systems, making them difficult to collect, correlate, and analyze systematically. This results in a lack of data-driven basis for standard revision work, relying heavily on expert experience and qualitative judgment.

[0004] The standard revision process suffers from severe lag in the optimization cycle: Aerospace technologies, such as additive manufacturing and new composite materials, are developing rapidly, while the revision process follows a traditional committee model, which is lengthy and often takes years. This "static" and discrete update model prevents standards from keeping pace with technological advancements and even hinders technological innovation.

[0005] The tacit and loss of knowledge: A large amount of tacit knowledge, such as expert discussions, identification and judgment criteria, and experimental data, is not effectively recorded and managed in the traditional model and is easily lost over time and with personnel changes.

[0006] In recent years, although there have been attempts to use knowledge graphs or ontology to structurally represent standards, most have remained at the level of static knowledge representation of standards and single-point compliance checks, failing to form a closed-loop intelligent system that covers the entire lifecycle of standards—from formulation to application to feedback and revision—and possesses dynamic learning and autonomous optimization capabilities. Therefore, this field urgently needs a new technological paradigm that treats the standard ontology and its lifecycle as a monitorable, simulateable, and optimizable "living" system to address the aforementioned challenges and fundamentally improve the quality of aviation equipment. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention aims to provide a standardized process digital twin system for optimizing the quality of aviation equipment. Through formal integration modeling and application programming interface support, standards can be seamlessly integrated and automatically executed by industrial software. Using data-driven closed-loop optimization, a data loop is constructed from standard application to standard revision, ensuring that standard optimization no longer relies solely on expert experience but is based on massive amounts of real-world aviation equipment development data. This significantly improves the scientific rigor and timeliness of aviation equipment standards, reducing the standardization cycle from months or years to weeks. Through feedforward simulation verification and continuous feedback optimization based on operational data, the system can proactively identify and correct design flaws in aviation equipment that may result from incomplete standards, thus providing a more solid foundation for the entire lifecycle of complex aviation equipment.

[0008] Specifically, on the one hand, this invention provides a standardized process digital twin system for optimizing the quality of aviation equipment, comprising: an extraction module L1, a support module L2, an application verification module L3, an analysis module L4, and an optimization application module L5; the extraction module L1 extracts aviation equipment standard knowledge based on aviation equipment standard documents and establishes a twin SKC, the twin SKC including: an ontology library. Graph Examples Rule base and model library Among them, the map examples The entity node set V in the data includes constraint variables. and confidence level C conf Constraint variables Used to identify the constraint strength of the standard clause corresponding to the node; confidence level C conf Used to quantify the credibility of the standard knowledge corresponding to nodes; supports module L2, used to traverse the rule base and model base in the twin SKC and encapsulate them into a set of application interfaces. Application verification module L3 is used for the application interface set based on support module L2. Design schemes within the standard window of an aerospace equipment manufacturing unit in a virtual engineering environment. Perform conformity verification and obtain verification results; Analysis module L4 includes: Deviation vector submodule, which, based on the verification results of application verification module L3, obtains the deviation vector between the virtual prediction result and the conformity rule requirement value, or between the virtual prediction and physical reality. The structured causal model submodule constructs a bias vector-based model. A structured causal model with target variables, using data from the aerospace equipment development process. The physical variables in the model are candidate causal variables. These variables are used to evaluate the observed data of aviation equipment quality performance indicators, generating a causal weight matrix. The targeted update submodule performs confidence decay based on causal weights, according to the causal weight matrix. Causal contribution weight in Using the state decay function to analyze the spectral instance Corresponding node confidence level C conf Perform targeted updates; correct the parameter submodule, and adjust the causal weight matrix. Contribution weights of each variable in Constraint variables passed to the twin SKC Perform collision analysis to generate prediction errors between the twin SKC and physical reality. Revision indicators with causal effects; optimization application module L5 is used to analyze the revision indicators output by module L4, and the Markov Decision Process (MDP) model is used for identification and judgment. By executing the iterative optimization of the twin SKC, the convergence of the virtual engineering environment and the standardization process of aviation equipment quality optimization is achieved within the preset deviation threshold.

[0009] Preferably, the twin SKC extracted from module L1 includes: an ontology library. Graph Examples Rule base and model library Ontology The descriptive logic is used to define concepts and axiomatic constraints for the physical domains covered in the standard document; (Graphical examples) The instantiation data of standard terms is stored in the form of RDF triples, set as a directed graph G=(V,E), where V is the set of entity nodes and E is the set of relation edges; rule base. Using a formal language of first-order logic, the performance requirements and constraints in standard documents are transformed into machine-executable rules. This formal language uses mathematical formulas composed of symbols to describe objects and the relationships between them; model library. Store the mathematical model associated with the standard terms.

[0010] Preferably, the entity node set V in the extraction module L1 contains key control variables, including process parameters, design parameters, performance indicators, and testing indicators, and a specific subset of nodes in the entity node set V. As a dynamic state attribute set A state Including constraint variables and confidence level C conf Two key state variables.

[0011] Preferably, the twin SKC obtained by the extraction module L1 is verified in the application verification module L3; the existence of problems with the aerospace equipment standard ontology is checked; after the preliminary verification of virtual compliance and physical feasibility, a design scheme that conforms to the standard window provided by the information system of each stage of the equipment development life cycle is received. ; Trigger the application verification module L3, which updates the application interface set by calling the support module L2. Perform simulation and verification calculations.

[0012] Preferably, the design scheme of the verified standard window is used. Applied to aerospace equipment manufacturing units, a dynamic analysis closed loop based on a structured causal model (SCM) is constructed. Specifically, this includes: obtaining the execution deviation vector, constructing and inferring the execution causal model, analyzing the confidence decay of execution based on causal weights, and performing mismatch analysis and strategy generation of execution standard attributes; and constructing a system with the output of the application verification module L3 as the target variable and process data as the basis. A cause-effect graph with physical variables as candidate causes; outputs the identified causes and their corresponding prediction errors. Physical variables with causal influence; used for continuous monitoring of application effects and causal attribution, enabling data aggregation and analysis.

[0013] Preferably, the dynamic analysis closed loop based on the Structured Causal Model (SCM) in the analysis module L4 specifically involves: obtaining the execution deviation vector and calculating the deviation vector based on the verification results transmitted by the application verification module L3. Perform causal model construction and inference, constructing a bias vector. A structured causal model for the target variable, with The physical variables in the model are candidate causal variables. The observed data are then estimated to generate a causal weight matrix. Perform confidence decay based on causal weights, according to the causal weight matrix. Causal contribution weight in Using the state decay function to apply to the corresponding nodes in the graph confidence level C conf Perform targeted updates; execute standard attribute mismatch analysis and strategy generation, and convert the causal weight matrix... Contribution weights of each variable in Standard constraint variables passed to module L2 Perform collision analysis to generate differentiated revision metrics.

[0014] Preferably, the state decay function in the analysis module L4 is specifically as follows: ; ;in, Output the state decay function; Input to the state decay function; α is the attenuation factor; α is the sensitivity coefficient; This is a preset deviation threshold; Weights are assigned to causal factors. This is the deviation vector.

[0015] Preferably, the optimization application module L5 uses the revision indicators output by the analysis module L4 to optimize the twin SKC ontology and knowledge graph; specifically, it optimizes the ontology library T, graph instance G, and model library M in the twin SKC; the optimization application module L5 has a built-in reinforcement learning-based recognition and judgment engine to output the optimized twin SKC', completing the closed loop from digital world optimization to physical world performance improvement.

[0016] On the other hand, this invention provides a standardized process digital twin method for optimizing the quality of aerospace equipment, which includes the following steps: S1: Invoking the extraction module L1 and the optimization application module L5 to digitally develop welding standards; Based on the aerospace equipment welding standard documents, extracting aerospace equipment welding standard knowledge, and converting the aerospace equipment welding standard documents into a structured welding standard twin SKC, wherein the welding standard twin SKC is represented as a quadruple, specifically: ;in, For welding standard twins; For welding standard body library; Examples of standard welding diagrams; A library of welding standard rules; For the welding standard model library; S2: Call the support module L2 and application verification module L3 to realize virtual design and verification based on welding standards; Traverse the rule base R and model library in the welding standard twin SKC output by step S1. It is automatically encapsulated into a series of application interface APIs for use by information systems at all stages of the entire lifecycle of aerospace equipment development, and outputs a set of application interfaces. Verify the welding standard twin SKC output in step S1; after successful verification, receive the design scheme of the welding standard window. S3: Perform simulation prediction; call analysis module L4 to collect and analyze physical world application data; perform physical world application effect monitoring and causal attribution, and realize data collection and analysis; use the welding standard window design scheme verified through step S2. The process involves: S1) Data-driven causal attribution of the prediction deviation between virtual prediction and physical reality; S2) Constructing and solving the structured causal model (SCM) to explain the source of the prediction deviation error and obtaining the causal graph; S3) Calling the optimization application module L5 and the extraction module L1 to realize the data-driven evolution and closure of welding standards; S4) Automatically optimizing and updating the welding standard twin; S5) Optimizing and updating the welding standard twin (SKC) based on the causal graph identified in step S3, and continuously applying it to the physical world.

[0017] Preferably, the design scheme of the welding standard window in step S2 is as follows: ;in, Design scheme for welding standard windows; This refers to the standard welding temperature in the standard welding process. The standard welding pressure in the standard welding process; The holding time in the standard welding process; calling the application interface set. Design scheme for welding standard window The performance is simulated and predicted in a virtual environment to obtain the prediction results.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention improves the accuracy and ease of use of aviation equipment standards. Through formal integration modeling and application programming interface (API) support, it eliminates the semantic ambiguity of the standards, enabling the standards to be seamlessly integrated and automatically executed by industrial software, freeing engineers from tedious research and manual comparison.

[0019] (2) This invention realizes data-driven closed-loop optimization. By constructing a data closed loop from standard application to standard revision, the optimization of the standard no longer relies solely on expert experience, but is based on real-world aviation equipment development data, which significantly improves the scientificity and timeliness of aviation equipment standards, and shortens the aviation equipment standard cycle from several months or years to several weeks.

[0020] (3) This invention enhances the aviation equipment quality optimization system. Through feedforward simulation verification and continuous feedback optimization based on operational data, the system can detect and repair aviation equipment design defects that may be caused by imperfect standards in advance, thus providing a more solid foundation for the entire life cycle of complex aviation equipment.

[0021] (4) This invention makes the implicit knowledge of aviation equipment standardization, such as identification and judgment logic and data basis, explicit and model-based, and solidifies it in the aviation equipment quality optimization system, thus forming a valuable standard knowledge asset that is traceable and inheritable. Attached Figure Description

[0022] Figure 1 is a flowchart of the standardized process digital twin system for optimizing the quality of aviation equipment according to the present invention; Figure 2 is a schematic diagram of the interaction between the present invention and the information system at each stage of the entire life cycle of aviation equipment development; Figure 3 is a flowchart of the specific implementation of the method of the present invention; Figure 4 is a comparison diagram of the shear strength of ten batches before and after the standard evolution of the embodiment of the present invention; Figure 5 is a schematic diagram of the logical relationship of the five-layer closed-loop system architecture driven by the dual-loop system proposed in the present invention; Figure 6 is a flowchart of the standardized process digital twin method for optimizing the quality of aviation equipment according to the present invention. Detailed Implementation

[0023] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0024] This invention proposes a standardized process digital twin system for optimizing the quality of aerospace equipment, as shown in Figure 1. It includes an extraction module L1, a support module L2, an application verification module L3, an analysis module L4, and an optimization application module L5. Figure 5 shows a schematic diagram of the five-layer closed-loop optimization logic relationship driven by the dual-loop system proposed in this invention. The figure illustrates a complete, dynamically optimized closed-loop optimization system centered on the twin SKC, with the clockwise "feedforward application loop" representing the extraction module L1, support module L2, and application verification module L3, and the counter-clockwise "feedback optimization loop" representing the analysis module L4 and optimization application module L5.

[0025] The application object in this embodiment of the invention is an automated welding standard unit for aerospace equipment composite materials. The actuator of this unit is an industrial robot capable of performing resistance welding. The task of the application scenario is to perform standard operations for welding aerospace equipment composite materials according to a formulated thermoplastic composite material welding standard, abbreviated as XXX-2025.

[0026] The extraction module L1 extracts aviation equipment standard knowledge based on aviation equipment standard documents and establishes a twin SKC. The twin SKC includes: an ontology library. Graph Examples Rule base and model library Among them, the map examples The entity node set V in the data includes constraint variables. and confidence level C conf Constraint variables Used to identify the constraint strength of the standard clause corresponding to the node; confidence level Cconf Used to quantify the credibility of the standard knowledge corresponding to the nodes; extracts the twin SKC in module L1, specifically including: ontology library. Graph Examples Rule base and model library Ontology The descriptive logic is used to define concepts and axiomatic constraints for the physical domains covered in the standard document; (Graphical examples) The instantiation data of standard terms is stored in the form of RDF triples, set as a directed graph G=(V,E), where V is the set of entity nodes and E is the set of relation edges; rule base. Using a formal language of first-order logic, the performance requirements and constraints in standard documents are transformed into machine-executable rules. This formal language uses mathematical formulas composed of symbols to describe objects and the relationships between them; model library. Store the mathematical model associated with the standard terms.

[0027] The entity node set V in the extraction module L1 contains key control variables, including process parameters, design parameters, performance indicators, and testing indicators. A specific subset of nodes in entity node set V is also included. As a dynamic state attribute set A state Including constraint variables and confidence level C conf Two key state variables and constraint variables Used to identify the constraint strength of the standard clause corresponding to the node; confidence level C conf Used to quantify the credibility of the knowledge carried by a node.

[0028] The extraction module L1 extracts aviation equipment standard knowledge from aviation equipment standard documents and transforms these documents into a structured, machine-executable twin (SKC). This process takes an unstructured aviation equipment standard document as input, which includes text descriptions of aviation equipment process windows, graphs showing the relationship between material properties and temperature, and tables of geometric requirements for standard welded joints.

[0029] The extraction module L1 uses its internal multimodal natural language processing model to perform deep analysis on the aviation equipment standard documents, automatically extracting the entity, relation, rule, and model information contained within. The final output is an initialized twin SKC, which serves as the sole source of standard knowledge and the object of operation for all subsequent steps. In this embodiment, the twin SKC is formally represented as a quadruple, specifically including: an ontology library. Graph Examples Rule base and model library In this embodiment of the invention, the mathematical model structure of the twin SKC is a quadruple SKC=<T,G,R,M> The format visually illustrates the four core components—ontology TBox, knowledge graph ABox, rule base, and model base—as well as the relationships within the quadruples.

[0030] A complete aviation equipment standard may contain hundreds of clauses; therefore, a complete Standard Knowledge Base (SKC) is a vast and complex aviation equipment standards knowledge base. For clarity, we will focus on highlighting the core content closely related to the subsequent steps of this embodiment, but this does not represent the entirety of the Standard Knowledge Base (SKC) content, including: Ontology. Descriptive logic is used to define concepts and axiomatic constraints for the physical domains covered by the standard. This is used to define general concept classes and logical relationships between concepts in the aerospace equipment field, providing a unified semantic template for standard knowledge. In this embodiment, the concept of "standard welding process" is defined, denoted as WeldingProcess, and the axiomatic constraint WeldingProcess is applied. The constraint `hasParameter.ProcessParameter` defines a relationship where `hasParameter` indicates the "has parameter" attribute, connecting a process to its parameters; `ProcessParameter` represents the concept of "process parameters." This constraint clarifies that any standard welding process must possess the attribute of "process parameters."

[0031] Atlas Examples The standard's specific factual data is stored in the form of a directed graph G=(V,E), where V is the set of entity nodes and E is the set of relation edges. In this embodiment, the system first constructs a node representing the standard ontology, "XXX-2025," and a node representing the process object, "resistance welding process," in the entity node set V, and connects them through edges representing "setting" logic. Based on this, according to the text description of the process window in the standard document, the system instantiates three specific parameter nodes in set V: "standard welding temperature T," "standard welding pressure P," and "holding time t," and attaches them to the "resistance welding process" node through relation edges that "have parameters." These three parameter nodes are identified by the system as a specific subset of nodes representing key control variables. In the initial state of this embodiment, the system sets the constraint variables of these three parameter nodes. Initialized as "mandatory", confidence level C conf Initialized to 1.0.

[0032] rule base Using a formal language of first-order logic, performance requirements and constraints in standards are transformed into machine-executable rules. This formal language uses precise mathematics or formulas composed of symbols to describe objects and the relationships between them. For example, a rule regarding minimum performance requirements might be: Where x is the physical object of the welded standard joint; σ is the shear strength of the welded standard joint; This is the minimum permissible shear strength specified in the standard. To select any physical object for welding standard joints; This is a logical judgment predicate for welding standard joints, used to determine whether the variable welding standard joint physical object x belongs to the welding standard joint category; This is a logical predicate with strength, used to describe the physical object x of a standard welded joint having a strength value σ.

[0033] Model library The mathematical model associated with the standard terms is stored. In the steps of this embodiment, a key model is the internal solution process of a process parameter-performance index prediction model that incorporates physical mechanisms, expressed as follows: ;in, is the shear strength predicted by the model; f is the model function; T is the standard welding temperature; p is the standard welding pressure; t is the holding time. Through the above process, a complete, logically consistent twin SKC containing an adaptive model for the aerospace equipment field is formed.

[0034] The support module L2 is used to traverse the rule base and model base in the twin SKC and encapsulate them into a set of application interfaces. The L2 support module traverses the rule base and model library in the twin SKC, encapsulating them into a set of application interfaces F for use by information systems at all stages of the equipment development lifecycle. API ;F API It achieves synchronized output of standard knowledge and dynamic states, for any application interface f in the set. i Its output O i It is set as a structure containing metadata: O i ={Value i ,Attribute i}, where Value i The predicted value is calculated from the model library or the standard threshold is obtained by querying the rule library; Attribute i The key control variable node corresponding to this knowledge point The current dynamic state attribute includes at least the current constraint variable and the current confidence level.

[0035] In this embodiment, the module focuses on generating two core interfaces: one is the strength prediction interface predict_strength(T,p,t), and the other is the strength compliance check interface check_strength_requirement(σ).

[0036] When the information system calls the strength prediction interface at any stage of the equipment development lifecycle, this interface internally calls the process-performance prediction model f in the model library M to calculate and return the predicted shear strength value. Simultaneously, this interface returns the current process parameter nodes, including the dynamic status attributes of the welding standard temperature T, welding standard pressure p, and holding time t, informing the caller that these parameters are "mandatory" constraints in the current standard version, and that all have a confidence level of 1.0. When the strength compliance check interface is called, this interface uses the minimum performance constraint rules in the rule library R to make a judgment and returns a Boolean compliance result. Simultaneously, this interface returns the performance index node, which is the dynamic status attribute of the shear strength σ, explicitly indicating that the current confidence level of this index is 1.0.

[0037] Figure 2 illustrates the interaction between the system of this invention and information systems at each stage of the entire lifecycle of aerospace equipment development. It shows the interaction with the entire lifecycle of aerospace equipment development, including design, manufacturing, and operation and maintenance. The figure depicts the digital twin system of this invention, with the standard and standardized digital twin system DT-S&SP Platform as the core. Through its application programming interface (API) gateway, it conducts bidirectional data interaction with the computer-aided design (CAD) / product lifecycle management (PLM) system in the design stage, the manufacturing execution system (MES) / Internet of Things (IoT) system in the manufacturing stage, and the maintenance, repair, and operation (MRO) / fault prediction and health management (PHM) system in the operation and maintenance stage.

[0038] Application verification module L3 is used for the application interface set based on support module L2. Design schemes within the standard window of an aerospace equipment manufacturing unit in a virtual engineering environment. Perform conformity verification and obtain verification results; verify the twin SKC obtained from extraction module L1 in verification module L3; check whether there are any problems with the aerospace equipment standard ontology; after passing the preliminary verification of virtual conformity and physical feasibility, receive a design scheme that conforms to the standard window provided by the information system of each stage of the equipment development life cycle. ; Trigger the application verification module L3, which updates the application interface set by calling the support module L2. Perform simulation and verification calculations. Then, verify the design scheme of the standard window. Applied to aerospace equipment manufacturing units, a dynamic analysis closed loop based on a structured causal model (SCM) is constructed. Specifically, this includes: obtaining the execution deviation vector, constructing and inferring the execution causal model, analyzing the confidence decay of execution based on causal weights, and performing mismatch analysis and strategy generation of execution standard attributes; and constructing a system with the output of the application verification module L3 as the target variable and process data as the basis. A cause-effect graph with physical variables as candidate causes; outputs the identified causes and their corresponding prediction errors. Physical variables with causal influence; used for continuous monitoring of application effects and causal attribution, enabling data aggregation and analysis.

[0039] In this embodiment of the invention, the application verification module L3 receives a design scheme conforming to a standard window provided by an external process planning system. In this embodiment, the design includes a set of specific process parameters: standard welding temperature. Set to 380 degrees Celsius, standard welding pressure Set to 0.6MPa, holding time The time limit is set to 120 seconds. Application verification module L3 first calls the strength prediction interface updated by support module L2, inputting the aforementioned process parameters into the process-performance prediction model f in the twin SKC model library M. Simulation calculations are then performed in a virtual environment to obtain the predicted shear strength value of the aerospace equipment. The predicted value is 48.5 MPa. The system is marked as the theoretical performance benchmark Y theory Subsequently, the application verification module L3 calls the strength compliance check interface, using the minimum performance requirement rule set in the twin SKC rule base R, namely the minimum allowable value of shear strength. In this embodiment, the pressure is 45 MPa, which is related to the theoretical performance benchmark Y mentioned above. theory Perform conformity verification.

[0040] If the aviation equipment standard verification results If the result is a failure, it indicates a contradiction within the twin SKC. For example, the standard window set in the rule base R of the twin SKC, after being mapped through model f in its model base M, may not achieve the desired performance value for aerospace equipment shear strength. Unable to meet the minimum performance requirements set in its rule base R. At this point, an "Inherent Defect Report of Standard" is automatically generated, which enters the analysis module L4 to perform causal inference, and then triggers the optimization application module L5 to revise the defective standard.

[0041] If the aviation equipment standard verification results If the test passes, it indicates that the twin SKC is self-consistent and feasible under the current virtual understanding; the standard window design scheme The approved aerospace equipment for physical world manufacturing enters analysis module L4 for continuous monitoring and data collection on its application effects in the real physical world. In this embodiment, the aerospace equipment standard verification results... Passed.

[0042] Analysis module L4 includes: a deviation vector submodule, which obtains the deviation vector between the virtual prediction result and the compliance rule requirement value, or between the virtual prediction and physical reality, based on the verification results of application verification module L3. The structured causal model submodule is constructed using bias vectors. A structured causal model with target variables, using data from the aerospace equipment development process. The physical variables in the model are candidate causal variables. These variables are used to evaluate the observed data of aviation equipment quality performance indicators, generating a causal weight matrix. The targeted update submodule performs confidence decay based on causal weights, according to the causal weight matrix. Causal contribution weight in Using the state decay function to analyze the spectral instance Corresponding node confidence level Perform targeted updates; the parameter correction submodule will update the causal weight matrix. Contribution weights of each variable in Constraint variables passed to the twin SKC Perform collision analysis to generate a response to the prediction error. Revision indicators with causal effects.

[0043] This invention uses the analysis module L4 to monitor the effects of physical world applications and perform causal attribution, achieving data aggregation and analysis; and incorporates the design scheme of a standard window that is virtually verified through the aviation equipment standard knowledge application verification module L3. This system applies automated welding standard cells to the physical world and performs data-driven causal attribution on prediction discrepancies between virtual and physical realities. Its core task is to monitor the effectiveness of the standard in real-world applications through the L4 analysis module and use artificial intelligence algorithms to identify key modeling factors leading to incomplete ontology, rules, and prediction models in the twin SKC. Specifically, the system first designs the... The instructions are sent to the automated welding standard robot in the physical world for execution. During the welding standard process, the robot's sensors collect process data in real time at a frequency of 50Hz. phy This includes the real-time monitored fluctuation values ​​of the welding standard current. After the welding standard is completed, the actual shear strength is measured by destructive testing. The value is 46.2 MPa. Analysis module L4 receives the theoretical performance benchmark Y from application verification module L3. theory The predicted shear strength is 48.5 MPa, and the calculated virtual-to-real deviation vector E is obtained. dev The modulus is 2.3 MPa.

[0044] Next, the analysis module L4 constructs a system based on the deviation E. dev The system uses a structured causal model (SCM) for the target variable. phy The physical variables in the standard are considered as candidate causal variables, including key control variables already defined within the standard, temperature, pressure, time, and potential variables not defined outside the standard, such as current fluctuations. Using Do-calculus to estimate the observed data, the calculation results show that: real-time current fluctuations... The average fluctuation amplitude of 3.5A has a significant pure causal effect on the reduction of shear strength, with a corresponding causal contribution weight as high as 0.91, explaining most of the prediction error; while the contribution weight of existing parameters is relatively low.

[0045] Subsequently, the system performs confidence decay based on causal weights. In this embodiment, the undefined latent variable "current fluctuation" The causal contribution weight of "0.91" is extremely high, exceeding the preset latent variable dominance threshold of 0.8, indicating that the system's judgment bias is mainly dominated by this external variable. Therefore, the system automatically triggers the confidence decay suppression mechanism to maintain the set key control variables. confidence level C conf The system remains unchanged to protect the stability of existing standard knowledge. Meanwhile, because the cumulative contribution of input variables is high, the system determines that the performance indicator nodes themselves have no confidence risk and therefore do not decay.

[0046] Finally, the system performs standard attribute mismatch analysis and strategy generation. The system performs a collision analysis between the calculated causal contribution weights and the standard attribute states. The analysis reveals that the physical variable current fluctuation... The contribution weight of 0.91 is much higher than the preset high threshold of 0.8, and this variable has not yet been included in the set of key control variables of the current twin SKC. Based on the "complete rules for latent variables," the system generates a clear standard revision indicator: "to reduce current fluctuations." Incorporating key control variables The constraint variables are set to be mandatory. The analysis results and the causal attribution matrix are output to the optimization application module L5 as the basis for triggering the standard evolution.

[0047] Analysis module L4 is used for continuous monitoring and causal attribution of application effects, realizing data aggregation and analysis. This module constructs a dynamic analysis closed loop based on the Structured Causal Model (SCM). The specific execution logic includes: First, obtaining the execution deviation vector: Calculate the deviation vector based on the verification results transmitted by application verification module L3. If the verification result fails, calculate the theoretical performance benchmark. Value required by compliance rules Virtual deviation vector between If the verification result is satisfactory, further collect real-world performance data. and full-element process data Calculate real performance data Compared with theoretical performance benchmark The virtual and real deviation vectors between .

[0048] The second step involves constructing and inferring a causal model: constructing a model based on a deviation vector. A structured causal model for the target variable, using full-factor process data. The physical variables in the table are candidate causal variables, and the physical variables include the key control variables already set in the standard. And latent variables not specified outside the standard; use Do-calculus to estimate the observed data, calculate the pure causal effect of each physical variable on the bias, and generate a causal weight matrix. , of which elements Representative variable The third step is to calculate the causal contribution weights; based on the causal weight matrix... Causal contribution weight in Using the state decay function to apply to the corresponding nodes in the graph confidence level Perform targeted updates; set the state decay function as: ;in, Output the state decay function; Input to the state decay function; This is the attenuation factor, calculated based on the verification scenario.

[0049] If the scenario is a virtual verification failure, then the decay factor... β is a fixed penalty coefficient, with a default value of 0.5, representing a significant reduction in confidence due to theoretical contradictions. This coefficient can be adjusted according to the rigor requirements of the standard type.

[0050] If the scenario involves physical deviations, then the attenuation factor... for: ; where α is the sensitivity coefficient, with a default value of 1.0, which is used to adjust the response speed of the confidence level to deviations. This coefficient can be increased for critical development stages. The preset deviation threshold can be determined based on the required range of key control variables in the standard.

[0051] When the causal weight matrix When the displayed bias is mainly dominated by undefined latent variables, i.e., the latent variables have extremely high weights (which can be set to greater than 0.8 by default), the system automatically suppresses the bias of the defined key control variables. The confidence level is reduced to protect the stability of existing standard knowledge.

[0052] Meanwhile, if the cumulative causal contribution of all candidate input variables is lower than the preset explanatory power threshold, for example, 0.5, but the bias E dev If the value still exceeds the allowed range, the performance metric node used as output is determined to have a confidence risk, and its performance is attenuated by a decay factor of [value missing]. , where γ is the target attenuation coefficient, and the default value is 0.8.

[0053] The fourth step is to perform standard attribute mismatch analysis and strategy generation: This involves generating the causal weight matrix. Contribution weights of each variable in Standard constraint variables passed to module L2 Collision analysis is performed to generate differentiated revision indicators; the collision analysis rules specifically include: completion rules for latent variables: if the causal weight matrix This indicates that the contribution weight of a certain physical variable is higher than a preset high threshold, for example, a contribution weight greater than 0.8. This threshold can be adjusted according to specific design, process, and sensitivity requirements of testing, and the variable is not yet included in the key control variables. If the set contains a blind spot in the judgment criteria, then the variable will be included. The proposed revision is to "set it as a mandatory constraint variable"; the upgrade rules for already set variables are as follows: if... If the contribution weight of a certain variable is higher than a high threshold, for example, higher than 0.8, but its current constraint variable is "recommended" or "not set", then the control is deemed insufficient, and a revision suggestion to "elevate the constraint variable to mandatory" is generated; for the downgrade rules of already set variables: specifically, when the scenario is a virtual verification failure, if If the contribution weight of a certain variable is lower than a preset low threshold (e.g., less than 0.1), but its current constraint variable is "mandatory," then there is a risk of over-constraint. The system generates a revision suggestion, "It is recommended to consider reducing the constraint variable of this variable to balance development costs," for the optimization application module L5 to weigh when formulating a comprehensive revision strategy. The final output includes a causal weight matrix. The updated node confidence set and the analysis results of the above governance suggestion signals serve as the basis for triggering the standard evolution of the optimization application module L5.

[0054] The optimization module L5 uses the revision indicators output by the analysis module L4 for identification and judgment. It employs a Markov Decision Process (MDP) model for this purpose, and through iterative optimization of the twin SKC, achieves convergence of the virtual engineering environment and aerospace equipment quality optimization and standardization process within a preset deviation threshold. The optimization module L5 uses the revision indicators output by the analysis module L4 to optimize the twin SKC ontology and knowledge graph; specifically, it optimizes the ontology library T, graph instances G, and model library M within the twin SKC. The optimization module L5 incorporates a reinforcement learning-based identification and judgment engine that ultimately outputs an optimized twin SKC' to guide subsequent batches, completing a closed loop from digital world optimization to physical world performance improvement.

[0055] In this embodiment of the invention, the optimization application module L5 is used to automatically optimize and update the twin; based on the cause-effect graph and revision suggestions identified by the analysis module L4, the twin SKC is optimized and updated in an unmanned, data-driven manner, and finally, through continuous application in the physical world, the optimization benefits to the physical process are demonstrated in a closed loop.

[0056] In this embodiment, the analysis results of the unmodeled physical variable ΔI output by the analysis module L4 trigger the optimization application module L5; the specific implementation process is as follows: Since the analysis module L4 found that the current fluctuation variable ΔI has a significant causal impact on the welding standard quality but was not set in the initial twin SKC, the system first optimizes the ontology library T and the graph instance G of SKC. In the ontology library T, the concept of the current fluctuation variable ΔI is automatically promoted to a new, formal process parameter subclass ProcessParameter; in the graph instance G, a hasParameter relationship is added to the welding standard method WeldingProcess instance under the welding standard XXX-2025 for thermoplastic composite materials, and it is associated with the new "current fluctuation variable" parameter node. According to the revision suggestions of module L4, the system automatically sets the constraint variables in the dynamic state attribute set of the new node to "mandatory" and initializes the confidence level to 1.0. After expanding the semantic level of standard knowledge, the prediction model f in the twin SKC model library M is optimized. This process includes: twin SKC prediction model structure optimization: The system uses the unmodeled physical variable identified by the analysis module L4, namely the current fluctuation variable ΔI, as a new, physically meaningful input term, and automatically updates the function signature of the prediction model f, forming a new twin SKC prediction model f' with the following structure: Twin SKC prediction model parameter correction: using the standard physical dataset containing the current fluctuation variable ΔI collected by the analysis module L4. By retraining or mechanistically adjusting the prediction model f' of this new structure, a set of more accurate model parameters corrected by physical world data can be obtained.

[0057] In this embodiment, since the analysis module L4 only found one clear optimization direction, this time a standard update action is performed. Triggered state transition conditional probability value It is deterministic, meaning that after performing the update action, the probability of the standard knowledge core transitioning from the current state to the new state is 1. However, the core value of the identification and judgment process MDP framework constructed in this invention lies in handling more complex scenarios with multiple optimization options. For example, when the analysis module L4 simultaneously discovers three feasible but resource-conflicting revision suggestions, the identification and judgment process MDP framework will use its reward function R and long-term value optimization parameters to calculate the optimal standard revision strategy π* that maximizes long-term cumulative benefits, thereby assisting in strategic, rather than "greedy" single-step identification and judgment. Therefore, this deterministic transition is a specific execution instance under the strategic framework of this identification and judgment process MDP. At the same time, the complete information of this optimization, including the triggering reason, data evidence summary, version change content of the twin SKC, and timestamp, is packaged into a transaction and recorded on the system's distributed ledger, making the optimization process transparent, traceable, and tamper-proof. After the update and logging are completed, the "update" operation is automatically performed, marking the optimized twin SKC' as the new version of the XXX-2025 standard, for example, Rev.A, and making its application programming interface (API) support immediately effective for all downstream systems through the support module L2.

[0058] The final output of the optimized application module L5 is a formally updated, optimized, and new version of the twin SKC'. After the update, it will serve as the new and sole "digital regulation" to guide all subsequent batches of the automated welding standard unit. In the subsequent 10 batches, the process parameter settings of the welding standard robot are constrained by the more precise models and rules in the twin SKC'. Figure 4 shows a comparison of the shear strength of ten batches before and after the standard evolution in this embodiment of the invention. After statistical analysis of the quality of these 10 batches of samples, the results show that compared with the samples guided by the initial twin SKC in the previous round, the shear strength of the new aerospace equipment... The performance improved by 3.7%, and more importantly, its strength consistency, measured by standard deviation, decreased by 15%. This significant improvement in strength consistency signifies a substantial enhancement in the stability of the manufacturing process, which has extremely high engineering value in the field of aerospace equipment in this embodiment. Through the automatic optimization and updating of the twin, a direct and quantifiable improvement and control of the physical welding standard process quality was ultimately achieved, completing a closed-loop implementation from digital world optimization to physical world performance enhancement.

[0059] The optimization module L5 is used to analyze the results output by the analysis module L4. It uses the Markov Decision Process (MDP) model to identify and determine the optimal revision strategy and executes automatic iteration and updates of the twin SKC. This module has a built-in reinforcement learning-based identification and judgment engine. The specific execution logic includes: Action space mapping: It receives multiple governance suggestion signals output by the analysis module L4 and maps them to a set of candidate actions A in the Markov Decision Process (MDP) model. Each action corresponds to a modification operation on the node state attribute in the twin SKC graph instance G, including "increase constraint variables", "decrease constraint variables" and "add node".

[0060] Reward Function Construction: Construct a comprehensive reward function R, which is based on the causal weight matrix. The expected "quality gain" positive reward is calculated based on the causal contribution weights in the calculation, and is based on the constraint variables. The magnitude of the change is used to estimate the expected negative reward in order to quantify the overall value of different revision strategies.

[0061] Optimal strategy solution: Solve the Markov decision process (MDP) model to find the optimal revision strategy π* that maximizes the long-term cumulative reward in the candidate action set A, thereby achieving an autonomous trade-off between quality and cost in complex scenarios with multiple potential adjustment options.

[0062] Evolutionary execution and state reset: Automatically execute update operations on the ontology T, graph instances G, and model library M according to the optimal strategy π*; specifically, for the revised node, the confidence C of its dynamic state attribute set is used. conf Resetting to version 1.0 signifies that the system has entered a new stable state.

[0063] Update and record: The evolved twin SKC is marked as the new version and updated. Key information of the identification and judgment process, including candidate strategies, reward evaluation values ​​and final decisions, is recorded in the distributed ledger to complete the closed loop.

[0064] The second aspect of this invention proposes a standardized process digital twin method for optimizing the quality of aerospace equipment. Figure 3 shows a flowchart of the specific implementation of this invention. Taking the aerospace composite material welding standard XXX-2025 as an example, the flowchart details the entire process from the input of the welding standard draft, through digital development, virtual design and verification, physical world data aggregation and analysis, to the final data-driven optimization of the welding standard, clearly mapping each step to the corresponding system module. It includes the following steps: Step S1: Calling the extraction module L1 and the optimization application module L5 to perform digital development of the standard; extracting aerospace equipment standard knowledge based on the aerospace equipment standard document, and converting the aerospace equipment standard document into a structured, machine-executable welding standard twin SKC. This process takes an unstructured aerospace equipment standard document as input, which contains text descriptions of the aerospace equipment process window, graphs showing the relationship between material properties and temperature, and tables of geometric requirements for welding standard joints. A multimodal natural language processing model is used to perform deep analysis on aerospace equipment standard documents, automatically extracting the entity, relation, rule, and model information contained within. The output is an initialized welding standard twin (SKC), which serves as the sole source of welding standard knowledge and the object of operation for all subsequent steps. The welding standard twin (SKC) is formally represented as a quadruple, specifically: ;in, For welding standard twins; For welding standard body library; Examples of standard welding diagrams; A library of welding standard rules; This is a standard welding model library.

[0065] Step S2: Invoke the support module L2 and application verification module L3 to implement standard-based virtual design and verification; traverse the welding standard rule library R and welding standard model library in the welding standard twin SKC output in step S1. It automatically encapsulates a series of application interface APIs for use by information systems at all stages of the equipment development lifecycle, and outputs a set of standard welding application interfaces for these APIs. Specifically: ;in, This is a collection of welding standard application interfaces, including application interfaces that provide support for welding standard knowledge. Supports the i-th application interface API in the set of standard welding application interfaces; Input fields are provided to support welding standard knowledge; The output domain is supported by welding standard knowledge; pre() supports preconditions for welding standard knowledge; post() supports postconditions for welding standard knowledge.

[0066] A virtual compliance and physical feasibility pre-verification is performed on the welding standard twin SKC output in step S1. This verifies whether there are inherent logical contradictions or physical infeasibility in the aerospace equipment standard body. Logical contradictions include conflicts between rules and models, while physical infeasibility includes situations where the set process window cannot produce products that meet performance requirements. This allows for the discovery of defects before the welding standard is applied to the physical world. After passing the virtual compliance and physical feasibility pre-verification, a design scheme provided by an external process planning system that conforms to the welding standard window is received. The design scheme includes a set of specific process parameters; the design scheme for the welding standard window is as follows: ;in, Design scheme for welding standard windows; The standard welding temperature in the standard welding process is 380°C in this embodiment. The standard welding pressure in the standard welding process is 0.6 MPa in this embodiment. The pressure holding time is the standard welding process; in this embodiment, the pressure holding time is 120 seconds.

[0067] Calling the welding standard application interface collection The application interface (API) in the design scheme for the welding standard window. The performance is simulated and predicted in a virtual environment to obtain the predicted shear strength value of the aerospace equipment. The pressure was 48.5 MPa; the process utilized the SKC quasi-knowledge model library, a welding standard twin. The core prediction model in Output predicted shear strength values ​​for aerospace equipment for: ;in, The predicted shear strength value for aerospace equipment; This is a welding standard prediction model.

[0068] The performance compliance check application interface API is invoked, and the minimum performance requirement rules set in the welding standard twin SKC rule base R are used to evaluate the predicted aerospace equipment shear strength value obtained in the previous step. Perform conformity verification to obtain a Boolean-type aviation equipment standard judgment parameter. for: ;in, These are standard judgment parameters for aviation equipment, and are Boolean data. The strength requirement check function is a mathematical encapsulation of the minimum performance requirement rules in the rule base R.

[0069] If the aviation equipment standard verification results If true, output the verification result of aviation equipment standards. If the result is "Pass", then the aviation equipment standard verification result will be output. Fail.

[0070] Step S3: Call the analysis module L4 to collect and analyze physical world application data; perform physical world application effect monitoring and causal attribution to achieve data collection and analysis; and implement the design scheme of the welding standard window that has passed the virtual verification in step S2. An automated welding standard unit applied to the physical world, and data-driven causal attribution of the prediction deviation between virtual prediction and physical reality.

[0071] The analysis focuses on the discrepancy between predicted and actual predictions. The core of this analysis is to construct and solve a structured causal model (SCM) to explain the sources of prediction error. This process can be formally described, and the prediction error can be calculated as follows: ;in, This is the prediction error between the current understanding of the welding standard twin SKC and the physical reality; To represent the actual shear strength of aviation equipment; In this embodiment, to predict the shear strength of the aerospace equipment, ,but .

[0072] Construct a system to predict bias As the target variable, with process data Non-destructive testing data All physical variables collected Reasons for being a candidate Cause-effect graph. Do-calculus is used to estimate the prediction bias for each candidate cause V. Pure causal effect The final output is one or more identified parameters related to the prediction error. Unmodeled physical variables with causal effects In this embodiment, calculations revealed that the system detected current fluctuation variables. The average fluctuation range reached 3.5A, with a significant causal effect of -2.1 MPa, explaining 91.3% of the prediction error. Therefore, the physical variables were not modeled. This output means that the welding standard prediction model in the initial welding standard twin SKC... The reason it's not precise enough is, in a key, physically significant way, due to the influence of current fluctuations. The impact.

[0073] Step S4: Invoke the optimization application module L5 and extraction module L1 to achieve data-driven standard evolution and closed-loop; perform automatic optimization and updating of the welding standard twin; based on the causal graph identified in step S3, perform unattended, data-driven optimization and updating of the welding standard twin SKC, and finally, through continuous application in the physical world, prove the optimization benefits to the physical process in a closed loop. In this embodiment, the unmodeled physical variables output in step S3 are obtained. The analysis results trigger step S4; its specific implementation process is as follows: Step S41: Optimize the welding standard twin SKC ontology and knowledge graph; the current fluctuation variable ΔI was not identified as a key process parameter in the initial welding standard twin SKC in step S1; optimize the ontology library T and knowledge graph instance G of the welding standard twin SKC. In the ontology library T, the concept of current fluctuation variable ΔI is automatically promoted to a new, formal process parameter subclass; in the knowledge graph instance G, it is associated with the new "current fluctuation variable" parameter.

[0074] Step S42: Optimize the structure of the welding standard twin SKC prediction model; using the current fluctuation variable ΔI as an input, automatically update the function signature of the prediction model f to form a new welding standard twin SKC prediction model. The structure is as follows: .

[0075] Step S43: Construct the identification and judgment process MDP framework, and use its reward function R and long-term value optimization parameters to calculate the optimal welding standard revision strategy π* that maximizes long-term cumulative benefits. After completing the update and recording, automatically execute the "update" operation to apply the optimized welding standard twin SKC' to all downstream systems through step S2.

[0076] The final output is a formally updated, optimized, and new version of the welding standard twin, SKC'. This invention, through the automatic optimization and updating of the welding standard twin, ultimately achieves a direct and quantifiable improvement and control over the quality of physical welding standard processes, completing a closed-loop implementation from digital-world optimization to physical-world performance enhancement.

[0077] The digital twin system and method for welding standardization process proposed in this invention for optimizing the quality of aviation equipment enables closed-loop optimization of the welding standard, thereby achieving control over the physical welding standard process. This ensures that key performance indicators such as the shear strength σ of the welded standard joint consistently meet the standards, ultimately achieving the optimization of aviation equipment quality.

[0078] As shown in Figure 5, in a specific embodiment of the present invention, the extraction module L1 is a semantic modeling layer, the support module L2 is a service enabling layer, the application verification module L3 is a behavior verification layer, the analysis module L4 is a cognitive analysis layer, and the application optimization module L5 is an evolutionary governance layer.

[0079] The second embodiment of this invention applies to the digital assembly unit of a certain type of aircraft wing box. The scenario task is to verify the scientific validity of the tolerance allocation principles and benchmark thresholds specified in the standard document "Wing Box Section Assembly Tolerance Allocation and Coordination Specification" (hereinafter referred to as the "Tolerance Specification Standard") formulated internally by the enterprise, and to evolve and revise its clauses using the system of this invention. The core objective of this standard is to ensure that the final assembly clearance and step difference of the wing box meet the aerodynamic shape requirements. As shown in Figure 6, the specific implementation steps are as follows: Knowledge construction is performed using the extraction module L1, receiving the tolerance specification standard document, parsing it, and converting it into a formal twin SKC. During this process, the system converts the provisions in the standard text regarding "machined frame outline error" (corresponding to standard clause 3.1, specifying a tolerance zone of ±0.1mm) and "skin part edge manufacturing error" (corresponding to standard clause 4.2, specifying a tolerance zone of ±0.2mm) into entities and constraints in the knowledge graph. The system then converts the above clause objects into key control variable nodes in the knowledge graph. It initializes its dynamic state properties, sets its constraint variables to "mandatory", and initializes the confidence level to 1.0.

[0080] Meanwhile, the quasi-knowledge model base of the twin SKC Includes a Jacobi screw model describing the propagation of assembly errors. This model, as the inherent calculation engine of this standard, describes how manufacturing errors of the machined frame and skin are propagated and accumulated through the assembly kinematic chain, ultimately forming the mathematical relationship of the assembly's functional requirements, namely the skin seam gap.

[0081] Support updates are performed using the L2 support module; the Jacobi spinor model described above is encapsulated as a standardized application programming interface (API) for use by upstream digital pre-assembly systems to predict assembly quality under current standard terms.

[0082] Application verification module L3 was used for application and verification; Monte Carlo simulation was performed in a virtual environment. The system strictly adhered to the tolerance zones set in Clause 3.1 (±0.1mm) and Clause 4.2 (±0.2mm) of the tolerance specification standard, randomly generating 1000 sets of virtual parts and performing virtual assembly calculations. Verification results showed that, under strict adherence to the current standard clauses, the out-of-tolerance rate of the final assembly clearance was as high as 6.8%. This result indicates a defect in the clause settings of the tolerance specification standard itself; that is, the tolerance allocation principle specified in the standard cannot be adapted to ensure that the final assembly quality meets design requirements. Based on this, the system generated a "Standard Clause Defect Report" and triggered causal inference.

[0083] Data aggregation and analysis are performed using the L4 analysis module. A structured causal model is constructed based on virtual validation samples and a small number of actual trial assembly samples to analyze the causes of deviations. Since the received results are virtual validation failures, the system first calculates the virtual deviation vector E between the theoretical performance benchmark and the compliance rule requirement value. dev Next, the system used "machining frame outline error" and "skin part edge manufacturing error" as causal variables and employed Do-calculus to calculate the intervention effect. The calculation results showed that intervening in tightening the skin error (item 4.2) had a very high causal contribution weight to eliminating the deviation; while intervening in the machining frame error (item 3.1) had a weak causal effect.

[0084] Based on this weight analysis, module L4 performs confidence decay: due to the virtual verification failure scenario, the system significantly decays the confidence C of the "skin part edge manufacturing error" node using a fixed penalty coefficient β. conf Meanwhile, in the attribute mismatch analysis, the system generated differentiated revision suggestions: for high-weight skin errors, it suggested "tightening tolerances"; for low-weight machined frame errors where the original constraint was mandatory, based on the degradation rules, it generated an optimization signal that suggested "considering relaxing tolerances or reducing constraint variables to balance development costs".

[0085] The L5 optimization application module is used for collaborative evolution and governance; the identification and judgment process model is used to formulate the optimal standard revision strategy for standard clauses; the system sets discrete action sets for specific clauses: action a1: revise clause 4.2 to tighten the manufacturing error tolerance of the skin part edge to ±0.15mm.

[0086] Action a2: Revise Clause 3.1 to relax the tolerance of the machined frame outline to ±0.15mm.

[0087] Action a3: Keep the standard terms unchanged.

[0088] The identification and judgment process model integrates positive rewards (improved assembly quality) and negative rewards (increased machining difficulty). Through problem-solving, the system discovered a combined strategy {a1, a2}, which involves tightening skin tolerances to improve quality while simultaneously relaxing machining frame tolerances to balance machining time, yielding the greatest long-term cumulative return.

[0089] The system automatically executes this strategy, generating and updating the new tolerance specification standard. In the subsequent five batches of actual trial assembly, guided by the new standard, the measured assembly clearance deviation rate decreased from 6.8% to 1.2%, and due to the relaxation of tolerance requirements for time-consuming machined parts, the processing time per piece was reduced by approximately 3%. This demonstrates that the present invention, based on standardized process analysis, achieves a scientific revision of standard clauses, thereby improving equipment quality.

[0090] The beneficial effects of this invention are as follows: The embodiments of this invention improve the accuracy and usability of aviation equipment standards. Through formal integration modeling and application programming interface (API) support, it eliminates semantic ambiguity in standards, enabling seamless integration and automatic execution of standards by industrial software, freeing engineers from tedious research and manual comparison. It achieves data-driven closed-loop optimization. By constructing a data closed loop from standard application to standard revision, standard optimization no longer relies solely on expert experience but is based on real-world aviation equipment development data, significantly improving the scientific rigor and timeliness of aviation equipment standards, shortening the standardization cycle from months or years to weeks. The embodiments of this invention enhance the aviation equipment quality optimization system. Through feedforward simulation verification and continuous feedback optimization based on operational data, the system can proactively identify and repair design defects in aviation equipment that may result from imperfect standards, thus providing a more solid foundation for the entire lifecycle of complex aviation equipment. It makes the tacit knowledge of aviation equipment standardization, such as identification and judgment logic and data basis, explicit and model-based, and solidifies it within the aviation equipment quality optimization system, forming a traceable and inheritable valuable standard knowledge asset. The simulation of the standard welding process for aviation equipment and the digital assembly unit of the aircraft wing box in the examples realizes the practical application of standardized digital twin modeling.

[0091] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A standardized process digital twin system for optimizing the quality of aviation equipment, characterized in that, include: The module consists of extraction module L1, support module L2, application verification module L3, analysis module L4, and application optimization module L5. Module L1 extracts aviation equipment standard knowledge based on aviation equipment standard documents and establishes a twin SKC. The twin SKC includes: ontology library. Graph Examples Rule base and model library Among them, the map examples The entity node set V in the data includes constraint variables. and confidence level C conf Constraint variables Used to identify the constraint strength of the standard clause corresponding to the node; confidence level C conf Used to quantify the credibility of the standard knowledge corresponding to nodes; supports module L2, used to traverse the rule base and model base in the twin SKC and encapsulate them into a set of application interfaces. Application verification module L3 is used for the application interface set based on support module L2. Design schemes within the standard window of an aerospace equipment manufacturing unit in a virtual engineering environment. Perform conformity verification and obtain verification results; Analysis module L4 includes: Deviation vector submodule, which, based on the verification results of application verification module L3, obtains the deviation vector between the virtual prediction result and the conformity rule requirement value, or between the virtual prediction and physical reality. The structured causal model submodule constructs a bias vector-based model. A structured causal model with target variables, using data from the aerospace equipment development process. The physical variables in the model are candidate causal variables. These variables are used to evaluate the observed data of aviation equipment quality performance indicators, generating a causal weight matrix. The targeted update submodule performs confidence decay based on causal weights, according to the causal weight matrix. Causal contribution weight in Using the state decay function to analyze the spectral instance Corresponding node confidence level C conf Perform targeted updates; correct the parameter submodule, and adjust the causal weight matrix. Contribution weights of each variable in Constraint variables passed to the twin SKC Perform collision analysis to generate prediction errors between the twin SKC and physical reality. Revision indicators with causal effects; optimization application module L5 is used to analyze the revision indicators output by module L4, and the Markov Decision Process (MDP) model is used for identification and judgment. By executing the iterative optimization of the twin SKC, the convergence of the virtual engineering environment and the standardization process of aviation equipment quality optimization is achieved within the preset deviation threshold.

2. The standardized process digital twin system for optimizing the quality of aviation equipment according to claim 1, characterized in that: The twin SKC extracted from module L1 includes: ontology library Graph Examples Rule base and model library Ontology The descriptive logic is used to define concepts and axiomatic constraints for the physical domains covered in the standard document; (Graphical examples) The instantiation data of standard terms is stored in the form of RDF triples, set as a directed graph G=(V,E), where V is the set of entity nodes and E is the set of relation edges; rule base. Using a formal language of first-order logic, the performance requirements and constraints in standard documents are transformed into machine-executable rules. This formal language uses mathematical formulas composed of symbols to describe objects and the relationships between them; model library. Store the mathematical model associated with the standard terms.

3. The standardized process digital twin system for optimizing the quality of aviation equipment according to claim 2, characterized in that: The entity node set V in the extraction module L1 contains key control variables, including process parameters, design parameters, performance indicators, and testing indicators. A specific subset of nodes in entity node set V is also included. As a dynamic state attribute set A state Including constraint variables and confidence level C conf Two key state variables.

4. The standardized process digital twin system for optimizing the quality of aviation equipment according to claim 1, characterized in that: The application verification module L3 verifies the twin SKC obtained from the extraction module L1; it checks whether there are any problems with the aerospace equipment standard ontology. After the preliminary verification of virtual compliance and physical feasibility, it receives a design scheme that conforms to the standard window provided by the information systems of each stage of the equipment development life cycle. ; Trigger the application verification module L3, which updates the application interface set by calling the support module L2. Perform simulation and verification calculations.

5. The standardized process digital twin system for optimizing the quality of aviation equipment according to claim 1, characterized in that: The design scheme of the verified standard window Applied to aerospace equipment manufacturing units, a dynamic analysis closed loop based on a structured causal model (SCM) is constructed. Specifically, this includes: obtaining the execution deviation vector, constructing and inferring the execution causal model, analyzing the confidence decay of execution based on causal weights, and performing mismatch analysis and strategy generation of execution standard attributes; and constructing a system with the output of the application verification module L3 as the target variable and process data as the basis. A cause-effect graph with physical variables as candidate causes; outputs the identified causes and their corresponding prediction errors. Physical variables with causal influence; used for continuous monitoring of application effects and causal attribution, enabling data aggregation and analysis.

6. The standardized process digital twin system for optimizing the quality of aviation equipment according to claim 1, characterized in that: The dynamic analysis closed loop based on the Structured Causal Model (SCM) in analysis module L4 specifically involves: obtaining the execution bias vector and calculating the bias vector based on the verification results transmitted by application verification module L3. Perform causal model construction and inference, constructing a bias vector. A structured causal model for the target variable, with The physical variables in the model are candidate causal variables. The observed data are then estimated to generate a causal weight matrix. Perform confidence decay based on causal weights, according to the causal weight matrix. Causal contribution weight in Using the state decay function to apply to the corresponding nodes in the graph confidence level C conf Perform targeted updates; Perform standard attribute mismatch analysis and strategy generation, and generate the causal weight matrix. Contribution weights of each variable in Standard constraint variables passed to module L2 Perform collision analysis to generate differentiated revision metrics.

7. The standardized process digital twin system for optimizing the quality of aviation equipment according to claim 6, characterized in that: The state decay function in analysis module L4 is as follows: ； ;in, Output the state decay function; Input to the state decay function; α is the attenuation factor; α is the sensitivity coefficient; This is a preset deviation threshold; Weights are assigned to causal factors. This is the deviation vector.

8. The standardized process digital twin system for optimizing the quality of aviation equipment according to claim 1, characterized in that: The optimization application module L5 uses the revision metrics output by the analysis module L4 to optimize the twin SKC ontology and knowledge graph; specifically, it optimizes the ontology library T, graph instances G, and model library M in the twin SKC; the optimization application module L5 has a built-in reinforcement learning-based recognition and judgment engine, which outputs the optimized twin SKC', completing the closed loop from digital world optimization to physical world performance improvement.

9. A standardized process digital twin method for optimizing the quality of aviation equipment, characterized in that, The standardization process is a welding standardization process, which includes the following steps: S1: Digital development of welding standards is performed by calling the extraction module L1 and the optimization application module L5; based on the aviation equipment welding standard documents, aviation equipment welding standard knowledge is extracted, and the aviation equipment welding standard documents are converted into a structured welding standard twin SKC. The welding standard twin SKC is represented as a quadruple, specifically: ;in, For welding standard twins; For welding standard body library; Examples of standard welding diagrams; A library of welding standard rules; For the welding standard model library; S2: Call the support module L2 and application verification module L3 to realize virtual design and verification based on welding standards; Traverse the rule base R and model library in the welding standard twin SKC output by step S1. It is automatically encapsulated into a series of application interface APIs for use by information systems at all stages of the entire lifecycle of aerospace equipment development, and outputs a set of application interfaces. Verify the welding standard twin SKC output in step S1; after successful verification, receive the design scheme of the welding standard window. S3: Perform simulation prediction; call analysis module L4 to collect and analyze physical world application data; perform physical world application effect monitoring and causal attribution, and realize data collection and analysis; use the welding standard window design scheme verified through step S2. The process involves: S1) Data-driven causal attribution of the prediction deviation between virtual prediction and physical reality; S2) Constructing and solving the structured causal model (SCM) to explain the source of the prediction deviation error and obtaining the causal graph; S3) Calling the optimization application module L5 and the extraction module L1 to realize the data-driven evolution and closure of welding standards; S4) Automatically optimizing and updating the welding standard twin; S5) Optimizing and updating the welding standard twin (SKC) based on the causal graph identified in step S3, and continuously applying it to the physical world.

10. The standardized process digital twin method for optimizing the quality of aviation equipment according to claim 9, characterized in that: The design scheme for the welding standard window in step S2 is as follows: ;in, Design scheme for welding standard windows; This refers to the standard welding temperature in the standard welding process. The standard welding pressure in the standard welding process; The holding time in the standard welding process; calling the application interface set. Design scheme for welding standard window The performance is simulated and predicted in a virtual environment to obtain the prediction results.

Images