Engine production line process parameter debugging method, device and equipment based on digital twinning and medium

Abstract

The invention discloses an engine production line process parameter debugging method and device based on digital twinning, equipment and a medium. The method comprises the following steps: constructing a whole-process digital twin system synchronously mapped with a physical engine production line in real time; when a fault is detected in the test run process, fault features are extracted from the test run data, and key process parameters of a preorder manufacturing link associated with the fault features are reversely positioned from the production line whole process data based on a preset correlation matrix; on the basis of the positioned key process parameters, generating a new process parameter combination, and performing quality prediction on an adjustment scheme containing the new process parameter combination by using a pre-trained quality prediction model; and if the prediction result is qualified, inputting the new process parameter combination and the historical production process data into the digital twin system for virtual simulation operation so as to verify the validity of the process adjustment scheme. According to the invention, accurate tracing and efficient and reliable process parameter debugging of engine test run faults are realized.

Description

Technical Field

[0001] This application relates to the field of process parameter debugging technology, specifically to a method, apparatus, equipment, and medium for debugging process parameters of an engine production line based on digital twins. Background Technology

[0002] In engine manufacturing, process planning is a core element in ensuring product quality and production efficiency. The trial production process of a new process typically follows a sequence of component machining, automated assembly, manual assembly, and engine testing. However, when abnormal noises, vibrations, or performance failures occur during the final testing phase, tracing the root cause is often extremely difficult.

[0003] Currently, factories generally rely on manual methods to troubleshoot problems. Process engineers need to review multiple scattered paper or electronic documents, such as machining records, assembly work orders, equipment operation logs, and trial reports, and then rely on their personal experience to analyze and make judgments. This method has significant shortcomings: First, the data at each stage is isolated from each other. Spindle vibration and tool wear data from machining equipment, robotic arm motion parameters and tightening curves from automated assembly lines, operator information and tightening torques at manual assembly stations, and vibration spectrum and sound pressure characteristics of test benches are stored in different control systems and databases. There is a lack of effective correlation between these data, making it impossible to form a complete data chain from raw materials to finished products.

[0004] Second, traceability is inefficient and its accuracy is difficult to guarantee. Manual investigation requires cross-departmental coordination and multi-system queries, which consumes a lot of time. The investigation results heavily rely on the personal experience of engineers, and different people may reach different conclusions, leading to repeated problems and prolonging the process debugging cycle.

[0005] Third, the analytical methods are one-sided and simplistic. Existing methods can usually only analyze static data from a single isolated process, failing to comprehensively consider the combined effects of various factors such as changes in temperature and humidity at the production site, real-time fluctuations in equipment status, and differences in personnel operation. For complex quality problems caused by the combined effects of multiple processes and factors, it is often difficult to accurately pinpoint the problem.

[0006] Fourth, there is a lack of process reproducibility and virtual verification capabilities. When a failure occurs, it is impossible to recreate the actual production scenario at the time of the problem. Process engineers can only speculate based on discrete data records, making it difficult to detect instantaneous anomalies or interactive defects. Furthermore, during the process trial operation phase, there is a lack of effective means to systematically assess and optimize equipment operational stability and process parameter matching before implementation, hindering the early prevention of quality risks.

[0007] Digital twin technology has provided a new avenue for the digital transformation of the manufacturing industry in recent years. Some technical solutions have already applied digital twins to shop floor production scheduling, assembly process visualization, or single-part machining process monitoring. However, most of these solutions only focus on a specific stage of the entire production process, such as optimizing the machining sequence, simulating assembly animation, or predicting the machining quality of simple parts like valves. They do not address the complex structures and long process chains of products like engines, failing to build a comprehensive digital twin environment that spans machining, automated assembly, manual assembly, and even commissioning. In particular, they lack a systematic method that can automatically reverse-engineer and locate potential problem parameters in preceding manufacturing stages based on commissioning fault characteristics, and verify the effectiveness of process parameter adjustments through virtual simulation.

[0008] Therefore, the engine manufacturing industry urgently needs a new technological approach to solve the problems of difficulty in tracing test failures and low efficiency in process debugging, and to achieve intelligent diagnosis and process optimization based on full-process data association.

[0009] It should be noted that the above statements are only used to provide background information related to this application and do not necessarily constitute prior art. Summary of the Invention

[0010] To address the aforementioned technical problems, embodiments of this application provide a method, apparatus, equipment, and medium for debugging process parameters of an engine production line based on digital twins.

[0011] The first aspect of this application provides a method for debugging process parameters of an engine production line based on digital twins, the method comprising: Construct a full-process digital twin system that is synchronously mapped in real time to the physical engine production line, wherein the movement of the digital twin system is driven by real-time collected full-process production line data; When a fault is detected during the trial run, fault features are extracted from the trial run data, and based on a preset correlation matrix, key process parameters of the preceding manufacturing process associated with the fault features are located in reverse from the full production line data. Based on the identified key process parameters, a new combination of process parameters is generated, and a pre-trained quality prediction model is used to predict the quality of the adjustment scheme containing the new combination of process parameters. If the prediction results are satisfactory, the new combination of process parameters and historical production process data will be input into the digital twin system for virtual simulation to verify the effectiveness of the process adjustment plan.

[0012] In some embodiments of this application, the construction of a full-process digital twin system that is synchronously mapped in real time to the physical engine production line includes: Throughout the entire process of engine production, including incoming materials, machining, automated assembly, manual assembly, and testing, real-time data on the production environment, production process, and product quality inspection results are collected at each stage. Each engine unit and key component is assigned a unique identifier, and the collected data is associated with and stored with the unique identifier. Based on the collected data, a digital twin system containing a fixed model, a motion model, and a behavior rule model is constructed in virtual space. The real-time collected data is then injected into the model through a data-driven engine to drive the motion model to perform synchronous mapping.

[0013] In some embodiments of this application, the unique identifier is an RFID electronic tag; the production process data includes: RFID electronic tags corresponding to engine parts during the incoming material process; machining center number, tool ID and wear status, cutting parameters, and spindle operating parameters during the machining process; robotic arm ID, robotic arm motion parameters, and gluing parameters during the automatic assembly process; workstation number, operator ID, and power tool parameters during the manual assembly process; and test bench number, operating status parameters, and fault codes during the test run process.

[0014] In some embodiments of this application, the correlation matrix is ​​constructed based on historical production data and is used to map the fault weights between specific trial run fault characteristics and process parameters of multiple preceding manufacturing stages, and can be dynamically iteratively updated according to newly added production and fault cases. In the correlation matrix, for each test failure feature, at least one key parameter from the main component change points, the processing change points, the automatic assembly process change points, and the manual assembly process change points are recorded and associated, and each key parameter is assigned a failure weight; wherein, the failure weight is determined through the accumulation of historical failure cases and production experience; when using the correlation matrix for reverse localization, the system automatically selects at least one key process parameter with the highest weight from the main component change points, the processing change points, the automatic assembly process change points, and the manual assembly process change points according to the level of the failure weight; The construction and updating of the correlation matrix and the training of the quality prediction model are all based on the same sample feature set, which includes: test run features, changes in major components, changes in the processing process, changes in the automatic assembly process, and changes in the manual assembly process.

[0015] In some embodiments of this application, the step of reversibly locating key process parameters of preceding manufacturing stages associated with the fault characteristics from the entire production line data based on a preset correlation matrix includes: Based on the fault characteristics, at least one key process parameter with the highest fault weight in the material processing, machining, automatic assembly, and manual assembly stages is automatically selected, and a reference value of the parameter in historical normal production is provided.

[0016] In some embodiments of this application, the sample feature set and the corresponding final product quality label are associated and stored in a quality data retention library for constructing the correlation matrix and training the quality prediction model.

[0017] In some embodiments of this application, the quality prediction model is a random forest model and is trained in the following manner: A training sample set is constructed by using the features in the sample feature set as input features and whether the product is ultimately abnormal as a binary classification label. Multiple sample subsets are generated from the training sample set through multiple random samplings with replacement. For each subset of samples, a decision tree is trained by randomly selecting some input features and splitting the nodes. The prediction results of all decision trees are integrated, and the final output of the random forest model is obtained through a voting mechanism.

[0018] A second aspect of this application provides a digital twin-based engine production line process parameter debugging device, the device comprising: A construction module is used to build a full-process digital twin system that is synchronously mapped with the physical engine production line in real time. The state of the virtual entity in the digital twin system is updated by real-time collected full-process production line data. The positioning module is used to extract fault features from the test run data when a fault is detected during the test run, and to reverse locate the key process parameters of the preceding manufacturing process associated with the fault features from the full process data of the production line based on a preset correlation matrix. The prediction module is used to generate new combinations of process parameters based on the located key process parameters, and to use a pre-trained quality prediction model to predict the quality of the adjustment scheme containing the new combination of process parameters. The verification module is used to input the new combination of process parameters and historical production process data into the digital twin system for virtual simulation operation when the prediction results are qualified, so as to verify the effectiveness of the process adjustment plan.

[0019] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the digital twin-based engine production line process parameter debugging method described in this application.

[0020] The fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the engine production line process parameter debugging method based on digital twins described in this application.

[0021] In this application, a full-process digital twin system is constructed that is synchronously mapped in real time to the physical engine production line. The state of the virtual entity in the digital twin system is driven and updated by real-time collected full-process production line data. This breaks down the barriers between data systems in each link and forms a data system that covers the entire production process and is synchronized between the virtual and real systems. This provides a unified and dynamic data foundation for subsequent intelligent traceability and simulation verification, making traceability and analysis no longer based on isolated and static fragmented information.

[0022] When a fault is detected during the trial run, fault features are extracted from the trial run data. Based on a preset correlation matrix, key process parameters of the preceding manufacturing stages associated with the fault features are located in reverse from the entire production line data. This transforms the time-consuming process of relying on manual experience and cross-departmental troubleshooting into an automated and precise analysis based on data and model rules. The system considers the complex correlation between fault features and multiple stages and factors, enabling it to quickly pinpoint the most likely root cause, significantly improving the efficiency and accuracy of problem localization and shortening the process debugging cycle.

[0023] Based on the identified key process parameters, new combinations of process parameters are generated. A pre-trained quality prediction model is then used to predict the quality of adjustment schemes incorporating these new parameter combinations. Before implementing the new process parameters in actual production or simulation verification, a scenario-based simulation using a data model is conducted to predict their effects. This effectively avoids ineffective process trials, transforming the traditional lagging cycle of production-testing-problem discovery-readjustment into a proactive iteration of prediction-optimization-re-verification. This significantly improves the scientific rigor and first-time success rate of process debugging, enabling low-cost, risk-free, and visualized comprehensive verification of the effectiveness of process adjustment schemes.

[0024] If the prediction results are satisfactory, the new combination of process parameters and historical production process data are input into the digital twin system for virtual simulation to verify the effectiveness of the process adjustment plan. In this way, not only is the problem scenario restored, but a platform for verifying the solution is also provided, enabling process personnel to intuitively confirm whether the adjustment has eliminated the cause of the failure. This provides sufficient confidence before the process documents are solidified and put into actual production, achieving a leap from speculative adjustment to verification optimization.

[0025] In summary, this application, through the organic combination of the above four steps, forms a complete closed loop for process parameter debugging, from data integration to intelligent diagnosis, then to predictive optimization, and finally to virtual verification. This enables rapid, accurate, and intelligent tracing of engine test failures and efficient and reliable process parameter debugging.

[0026] The above description is merely an overview of the technical solutions of the embodiments of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0027] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic diagram illustrating the steps of a digital twin-based engine production line process parameter debugging method provided in some embodiments of this application; Figure 2 This is a schematic diagram of a data transmission process provided in some embodiments of this application; Figure 3 This is a schematic diagram illustrating an example of a correlation matrix provided in some embodiments of this application; Figure 4 This is a schematic diagram of a decision tree construction provided in some embodiments of this application; Figure 5 This is a schematic diagram of the workflow of a method for debugging process parameters of an engine production line based on digital twins, according to another embodiment of this application. Figure 6 A schematic diagram of the structure of a digital twin-based engine production line process parameter debugging device provided in some embodiments of this application; Figure 7 This is a schematic diagram of the structure of an electronic device provided in some embodiments of this application. Detailed Implementation

[0028] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which embodiments of this application pertain; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of this application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings of this application are intended to cover non-exclusive inclusion.

[0030] In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0031] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0032] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0033] In related technologies, during the process planning and trial production stages of engine manufacturing, tracing the root cause of a fault discovered during the final test run is extremely difficult. This is mainly because: data from machining, automated assembly, manual assembly, and test run stages are scattered and isolated, making it difficult to form a complete traceable data chain; manual troubleshooting is inefficient and heavily reliant on personal experience; comprehensive analysis of multiple stages and factors is impossible; and there is a particular lack of effective means to "reproduce the scenario" of the production process when the fault occurred and to conduct "virtual verification" before applying new process parameters. Existing digital twin technology solutions mostly focus on optimizing or monitoring a single stage, lacking a solution that is comprehensive across the entire manufacturing process for complex products like engines, and thus cannot achieve a complete closed-loop solution.

[0034] To address the aforementioned technical problems, this application provides a method for debugging process parameters of an engine production line based on digital twins. To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0035] Figure 1 This is a schematic diagram illustrating the steps of a digital twin-based engine production line process parameter debugging method provided in some embodiments of this application, such as... Figure 1As shown, it includes: S1. Construct a full-process digital twin system that is synchronously mapped with the physical engine production line in real time, wherein the state of the virtual entity in the digital twin system is updated by real-time collected full-process production line data.

[0036] In one specific implementation, this step applies to an engine production line that includes five sequential processes: incoming materials, machining, automated assembly, manual assembly, and test run. First, full-element data collection and association are performed. This means that in the five major processes of engine production—incoming materials, machining, automated assembly, manual assembly, and test run—production environment data, production process data, and product quality inspection results data are collected in real time at each stage. Each engine unit and key component is assigned a unique identifier, and the collected data is associated and stored with these unique identifiers to form a traceable and complete data chain. The production environment data mainly refers to the temperature and humidity of the production workshop; this type of data is used to track changes in the physical properties of the model in the digital twin system. In this way, the technical barriers of isolated data storage and lack of interconnectivity in traditional manufacturing models are fundamentally broken down, laying a unified data foundation for subsequent accurate traceability and collaborative analysis.

[0037] Figure 2 A schematic diagram of a data transmission process is given, such as... Figure 2 As shown, the blue lines represent the production process data (corresponding to...). Figure 2 The process information (including the transmission of information) is as follows. Specifically, during the incoming material process, each engine component is affixed with an RFID electronic tag as a unique identifier. This tag information includes the supplier, delivery time, etc., and is manually scanned and entered into the Manufacturing Execution System (MES). Figure 2 The MES system is used in the production line. Subsequently, at every stage of the component's flow through the production line, its RFID tag serves as the core anchor point for data association, ensuring that all process data can be accurately attributed to a specific engine unit. The machining center automatically records and uploads data to the MES system, including: machining center number, tool ID and wear level, cutting parameters, and spindle operating parameters, such as spindle speed and vibration frequency. This high-frequency, multi-dimensional process data reflects the microscopic state of the equipment and the execution of the process in real time, forming the core input for the digital twin system to perform dynamic mapping and real-time monitoring. The transmitted data also includes: robotic arm ID, robotic arm motion parameters, and adhesive application parameters during automated assembly; workstation number, operator ID, and power tool parameters during manual assembly; and test bench number, operating status parameters, and fault codes during the test run, including operating frequency, speed, load, vibration spectrum, exhaust parameters, and abnormal noise characteristics. It can be understood that the operating status parameters collected during the test run, especially the vibration spectrum and acoustic characteristics, are the direct basis for identifying engine performance faults and extracting fault feature vectors.

[0038] Figure 2 The black lines in the middle indicate the quality inspection result data (corresponding to...) Figure 2 The transmission of test result information mainly includes product quality information after the completion of each of the four processes: machining, automated assembly, manual assembly, and trial run. Figure 2 The information displayed includes machining quality, assembly quality, and test run results. During machining, data such as the clamping force of parts, and the flatness and roughness measured after milling the end faces of parts, are recorded using point cloud technology. Automatic assembly processes include data such as the gaps and sealing of assembled parts. Manual assembly processes include data such as torque checks and checking for missing bolts. Test runs include data such as checking for leaks.

[0039] Furthermore, based on the data collected above, a digital twin system comprising a fixed model, a motion model, and a behavior rule model is constructed in virtual space. The fixed model includes static geometric models such as factory buildings, production line layout fences, transport lines, and machine tool housings, primarily used to check for interference during movement in the virtual environment. The motion model includes movable models such as machine tool spindles, robots, and engine components. The behavior rule model defines the motion logic of each physical entity, such as the robotic arm and spindle, as well as the process flow, such as assembly sequence and interaction rules.

[0040] The data-driven engine injects real-time or near real-time production line data, especially production process data, from the MES system into the digital twin system. This data drives the motion model to perform synchronous movement or displacement. Simultaneously, environmental data can be used to update the physical properties, such as thermodynamic performance, of relevant entities in the behavior rule model. This allows the state of virtual entities in the digital twin system to be updated based on real-time collected production line data, achieving a precise mapping with the physical production line. During the trial production phase of a new process, this system can be used to monitor equipment operating status and conduct preliminary assessments of process adaptability.

[0041] Step S1 breaks down the barriers between data systems in each stage, forming a virtual-real synchronized data system covering the entire production process. This provides a unified and dynamic data foundation for subsequent intelligent traceability and simulation verification, making traceability and analysis no longer based on isolated, static fragmented information.

[0042] S2. When a fault is detected during the trial run, fault features are extracted from the trial run data, and based on a preset correlation matrix, key process parameters of the preceding manufacturing process associated with the fault features are located in reverse from the full production line data.

[0043] In one specific implementation, the correlation matrix is ​​constructed based on historical production data and is used to map the fault weights between specific trial run fault characteristics and process parameters of multiple preceding manufacturing stages. It can also be dynamically updated based on newly added production and fault cases.

[0044] Specifically, during system initialization or operation, this correlation matrix is ​​jointly constructed and maintained by engineers and operators based on actual production fault location cases. The system automatically or manually updates the fault weights in the matrix monthly based on newly added successful or failed debugging cases to ensure its accuracy and real-time performance.

[0045] Figure 3 This illustrates a correlation matrix in structured tabular form, such as... Figure 3 As shown, each row represents a complete historical production or debugging case, i.e., one test. Its fields include: serial number, test characteristics, whether it is abnormal, major component 1 change point (fault weight), machining process change point (fault weight), automatic assembly process change point (fault weight), and manual assembly process change point (fault weight). The "machining process change point (fault weight)" field can include specific parameters such as "spindle speed (weight X%)", "tool wear (weight Y%)", and "workpiece clamping force (weight Z%)" and their corresponding weight values. The core idea of ​​this matrix is ​​to treat each complete process test, such as from the introduction of a new process to the completion of the test run, as a "sample," recording all "change points" that may affect quality throughout the entire manufacturing process and their corresponding "fault weights". Here, "change points" refer to process parameters or state quantities, such as spindle speed, tool wear, and workstation number, whose parameter values ​​may deviate from standard or empirical values ​​in a specific process, and are therefore considered potential quality-influencing factors. The "fault weight" is a quantitative indicator initially assigned by domain experts through analysis of a large number of historical samples, combining statistical data analysis such as the co-occurrence frequency of parameters and faults, and the significance of parameter deviations. This weight is then validated and iterated through the system to characterize the strength of the correlation between a point of change and a specific fault. A key design feature is that for the same point of change, such as "spindle speed," the corresponding fault weight value (X%) can differ in different case record rows. This reflects the dynamic and context-dependent nature of fault diagnosis and weight allocation: in different cases, due to differences in specific process backgrounds, abnormal manifestations, or expert judgments involved in the analysis, even the same factor may have different weights of responsibility for the final fault. The system automatically or manually updates the fault weights in the matrix monthly based on newly added successful or failed debugging cases to ensure accuracy and real-time performance.

[0046] In the correlation matrix, for each test failure feature, at least one key parameter from the main component change points, the processing change points, the automatic assembly process change points, and the manual assembly process change points are recorded and associated, and each key parameter is assigned a failure weight; wherein, the failure weight is determined through the accumulation of historical failure cases and production experience; when using the correlation matrix for reverse localization, the system automatically selects at least one key process parameter with the highest weight from the main component change points, the processing change points, the automatic assembly process change points, and the manual assembly process change points according to the level of the failure weight.

[0047] For example, if an abnormal noise in a specific frequency band, such as 2000Hz, is detected during a test run, it can be identified as test run feature A1. The system extracts the dominant frequency of this abnormal noise from the test run data as a fault feature. The system then queries the correlation matrix based on this fault feature. In a specific matrix example, refer to... Figure 3 For the record “Serial No. 1: From Faulty Engine A1 to Abnormality (Y)”, the system will automatically filter out the features with the highest weight in each link according to the fault weight recorded in the table: “spindle speed” (weight 40%) from the “process change point (fault weight)” field, “robotic arm ID” (weight 20%) from the “automatic assembly process change point (fault weight)” field, and “workstation number” (weight 10%) from the “manual assembly process change point (fault weight)” field, as the most suspicious key process parameters.

[0048] Furthermore, based on a preset correlation matrix, the key process parameters of the preceding manufacturing stages associated with the fault characteristics are reverse-located from the full-process data of the production line. This includes: automatically selecting at least one key process parameter with the highest fault weight among the material extraction, machining, automatic assembly, and manual assembly stages according to the fault characteristics, and providing a reference value of the parameter in historical normal production.

[0049] After filtering out the aforementioned suspicious parameters, the system will further search the correlation matrix for records where the "test drive feature" is A1 but the "abnormality" is N, indicating normality. For example... Figure 3 The system will calculate the average or normal range of the corresponding parameters such as spindle speed, robot arm ID, and workstation number in these normal cases, and provide this "average" or "range" as a reference value as a benchmark for adjusting process parameters. Then, combined with other information in the quality data retention database and the above reference value, a new combination of process parameters will be determined.

[0050] The construction and updating of the correlation matrix and the training of the quality prediction model are all based on the same sample feature set, which includes: test run features, changes in major components, changes in the processing process, changes in the automatic assembly process, and changes in the manual assembly process.

[0051] Understandably, the S2 step transforms the time-consuming process of relying on human experience and cross-departmental investigation into an automated and precise analysis based on data and model rules. The system considers the complex correlation between fault characteristics and multiple links and factors, and can quickly identify the most likely root cause, significantly improving the efficiency and accuracy of problem localization and shortening the process debugging cycle.

[0052] S3. Based on the identified key process parameters, generate new combinations of process parameters, and use a pre-trained quality prediction model to predict the quality of the adjustment scheme containing the new combination of process parameters.

[0053] In one specific implementation, the sample feature set and the corresponding final product quality label are associated and stored in a quality data retention library for constructing the correlation matrix and training the quality prediction model. Based on the key process parameters located in step S2 and their historical reference values, a new combination of process parameters is determined and generated. Subsequently, the pre-trained quality prediction model is used to predict the adjustment scheme. The input features of the quality prediction model here include: trial run features, changes in major components, changes in the processing procedure, changes in the automated assembly process, and changes in the manual assembly process.

[0054] The quality prediction model is a random forest model and is trained in the following way: a training sample set is constructed using features from the sample feature set as input features and whether the product is ultimately abnormal as a binary classification label; multiple sample subsets are generated from the training sample set through multiple random sampling with replacement; for each sample subset, a decision tree is trained by randomly selecting some input features and splitting nodes; the prediction results of all decision trees are integrated, and the final output of the random forest model is obtained through a voting mechanism.

[0055] Specifically, the training and prediction process of this random forest model is as follows, and its logic can be found in [reference needed]. Figure 4 The diagram shown illustrates the construction of a decision tree. Figure 4 This visually demonstrates the process from the correlation matrix to the construction of multiple decision trees. For example... Figure 4As shown, the original correlation matrix contains multiple historical samples, each with rich feature dimensions such as test drive characteristics, changes in each stage, and their weights. First, using the key parameters selected from the correlation matrix as the core, a sample feature set is determined, including test drive characteristics, component changes, processing changes, automated assembly changes, and manual assembly changes. The sample labels are set as binary categorical variables, i.e., "product abnormal" or "product normal." Simultaneously, 70% of the samples are randomly selected from the quality information database as the training set, and 30% as the test set, to ensure consistent sample distribution between the training and test sets and avoid model overfitting. Then, the random forest model is trained, as follows... Figure 4 As shown, based on the Bootstrap bootstrapping sampling method, multiple different sample subsets are randomly selected from the training set. A decision tree is trained independently for each subset. Figure 4 As shown in "Decision Tree 1", "Decision Tree 2" to "Decision Tree n", each decision tree is based on a randomly selected subset of samples and some features randomly selected during the tree construction process. During the node splitting process of each decision tree, some features are randomly selected from the feature set, and the optimal features, such as the Gini index or the feature with the highest information gain, are selected for node splitting to ensure diversity among decision trees. This process is repeated to generate multiple independent decision trees, which together form a random forest model. Finally, the new process parameters adjusted by the process engineers, such as the optimized spindle speed and robotic arm motion parameters, are substituted into the trained random forest model. Each decision tree in the model outputs a prediction result: product normal or product abnormal. The overall prediction result of the model is determined using a majority voting mechanism. If the prediction result is "product normal", the process enters the digital twin scenario simulation and restoration stage (step S4); if the prediction result is "product abnormal", the process returns to the process parameter adjustment stage (step S2), prompting for re-optimization of parameters and re-prediction until a satisfactory prediction result is output. Figure 5 This is a schematic diagram of the workflow for a method of debugging process parameters on an engine production line based on digital twins, and the above process also reflects... Figure 5 The core component of "product prediction using the random forest algorithm" in the "historical data route (blue solid line)".

[0056] The S3 step effectively avoids ineffective process trials, transforming the traditional lagging cycle of production-testing-problem finding-readjustment into a forward-looking iteration of prediction-optimization-re-verification, significantly improving the scientific nature and first-time success rate of process debugging.

[0057] S4. If the prediction result is satisfactory, the new combination of process parameters and historical production process data are input into the digital twin system for virtual simulation to verify the effectiveness of the process adjustment plan.

[0058] In one specific implementation, the "historical production process data" used for verification is... Figure 5 "Process trial operation process (corresponding to)" Figure 5 When the new process is first implemented (as indicated by the black dashed line in the diagram), the digital twin system uses real-time data routing (corresponding to the five processes of incoming materials, machining, automated assembly, manual assembly, and trial run on the actual production line) to monitor the process. Figure 5 The complete equipment operation and status data (represented by the solid black line in the image) is recorded and stored in the motion data repository. This data corresponds strictly to any abnormal test results.

[0059] In another specific implementation, such as Figure 5 As shown, this step corresponds specifically to the final verification stage of the historical data route in the diagram. Its input is the new combination of process parameters generated after the prediction in step S3 has passed and the process engineers have completed the improvements. The verification process is completed in the digital twin system, and the specific steps are as follows: Step 1: Preparing Verification Data and Scenario. The system first retrieves complete historical production process data from the motion data repository, strictly corresponding to the current trial run fault. This data originates from the initial process trial run and covers raw operational data such as machining center number, tool wear, cutting parameters, and spindle vibration frequency recorded during machining; robotic arm motion and adhesive application parameters recorded during automated assembly; workstation number, operator operations, and power tool parameters recorded during manual assembly; and fault codes and operating frequencies recorded during the trial run. Simultaneously, relevant quality inspection results are obtained from the quality data repository, such as component morphology and fit clearances recorded in point cloud data. These data collectively constitute a snapshot of the original production site state at the time of the fault.

[0060] Step 2: Reconstruct and load the virtual environment in the digital twin system. Based on the historical data mentioned above, the system reconstructs a high-fidelity virtual production scenario in the digital twin system. This virtual environment consists of a fixed model, a motion model, and a firmware model that defines its behavioral logic. The fixed model is used to construct the static environment such as workshop walls and equipment enclosures; the motion model is used to simulate the activities of machine tool spindles, industrial robots, and engine components. During reconstruction, the system injects historical data into these models to accurately reproduce the state and movement trajectory of each piece of equipment when the fault occurred.

[0061] Step 3: Execute the virtual simulation. In the reconstructed virtual environment, the system loads and replaces the original process parameters with a new combination of process parameters. These new parameters may include optimized cutting parameters, robotic arm motion trajectories, and power tool operating parameters. Then, the system drives the digital twin environment to reproduce the entire process from machining to assembly from scratch based on the new parameters. This process is executed by the reconstructed virtual environment module and is monitored in real time by monitoring equipment. Engineers can flexibly adjust the viewing perspective in this virtual environment, slow down or pause the simulation process, and observe changes in key indicator readings such as machining shape and position anomalies and assembly parameter anomalies through monitoring equipment, thereby intuitively analyzing the execution effect of the new process.

[0062] Step 4: Analyze the results and determine validity. The core purpose of virtual simulation is to verify whether any anomalies will occur in the entire production process under the new process parameters in the virtual environment. The system closely monitors the entire simulation process. If all processes can be completed smoothly and completely according to the new parameters, and the monitoring equipment does not report any preset anomalies throughout the process, such as no alarms for abnormal machining shapes or assembly parameters, then the process improvement plan is deemed qualified, and its effectiveness is verified. This successful verification result will be used as the output process parameter reference value and ultimately used to form a fixed process document that can be put into mass production.

[0063] This step completes a full technical loop, from identifying problems during the initial process trial run, recording the process through real-time data mapping, then using historical data analysis to pinpoint the problems, predict and optimize, and finally verifying the results in a reconstructed virtual environment. This achieves a priori, low-cost, and high-fidelity verification of process parameter adjustment schemes, successfully shifting process debugging from the traditional model relying on physical trial and error to an intelligent optimization model based on digital simulation, greatly improving the efficiency of the debugging process and the reliability of the final results.

[0064] It is understood that the digital twin-based engine production line process parameter debugging method provided in this application, through the organic combination of the above four steps, forms a complete closed loop of process parameter debugging from data integration to intelligent diagnosis, then to predictive optimization, and finally to virtual verification, realizing rapid, accurate, and intelligent traceability of engine test failures and efficient and reliable process parameter debugging.

[0065] In some embodiments of this application, a device for adjusting process parameters of an engine production line based on digital twins is also provided, such as... Figure 6 As shown, the device includes: A construction module is used to build a full-process digital twin system that is synchronously mapped with the physical engine production line in real time. The state of the virtual entity in the digital twin system is updated by real-time collected full-process production line data. The positioning module is used to extract fault features from the test run data when a fault is detected during the test run, and to reverse locate the key process parameters of the preceding manufacturing process associated with the fault features from the full process data of the production line based on a preset correlation matrix. The prediction module is used to generate new combinations of process parameters based on the located key process parameters, and to use a pre-trained quality prediction model to predict the quality of the adjustment scheme containing the new combination of process parameters. The verification module is used to input the new combination of process parameters and historical production process data into the digital twin system for virtual simulation operation when the prediction results are qualified, so as to verify the effectiveness of the process adjustment plan.

[0066] The device, through the coordinated operation of its construction, positioning, prediction, and verification modules, forms a complete, closed-loop process debugging system encompassing data integration, intelligent diagnosis, predictive optimization, and virtual verification. It systematically solves the core problems of difficult fault tracing, long process debugging cycles, and low success rates during engine trial production, caused by data silos, reliance on manual labor, and a lack of effective analysis tools and verification methods. This achieves intelligent, precise, and efficient process parameter debugging.

[0067] In some embodiments of this application, an electronic device is also provided to execute the above-described method for debugging engine production line process parameters based on digital twins. Please refer to... Figure 7 This illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 7 As shown, the electronic device 9 includes: a processor 901, a memory 902, a bus 903, and a communication interface 904. The processor 901, the communication interface 904, and the memory 902 are connected via the bus 903. The memory 902 stores a computer program that can run on the processor 901. When the processor 901 runs the computer program, it executes the engine production line process parameter debugging method based on digital twin provided in any of the foregoing embodiments of this application.

[0068] The memory 902 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this device network element and at least one other network element is achieved through at least one communication interface 904 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0069] Bus 903 can be an ISA bus, PCI bus, or EISA bus, etc. Buses can be divided into address buses, data buses, control buses, etc. Memory 902 is used to store programs. After receiving execution instructions, processor 901 executes the program. The engine production line process parameter debugging method based on digital twins disclosed in any of the aforementioned embodiments of this application can be applied to processor 901, or implemented by processor 901.

[0070] The processor 901 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 901 or by instructions in software form. The processor 901 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 902. Processor 901 reads the information in memory 902 and, in conjunction with its hardware, completes the steps of the above method.

[0071] The electronic equipment provided in this application embodiment and the engine production line process parameter debugging method based on digital twin provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0072] This application also provides a computer-readable storage medium corresponding to the engine production line process parameter debugging method based on digital twins provided in the aforementioned embodiments. The medium stores a computer program (i.e., a program product). When the computer program is run by a processor, it executes the engine production line process parameter debugging method based on digital twins provided in any of the aforementioned embodiments.

[0073] It should be noted that computer-readable storage media may include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, optical disc or other optical and magnetic storage media, etc., which will not be elaborated here.

[0074] The computer-readable storage medium provided in this application embodiment and the engine production line process parameter debugging method based on digital twin provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application stored therein.

[0075] This application also provides a computer program product corresponding to the digital twin-based engine production line process parameter debugging method provided in the foregoing embodiments, including a computer program that is executed by a processor to implement the aforementioned digital twin-based engine production line process parameter debugging method.

[0076] The computer program product provided in this application embodiment is based on the same inventive concept as the engine production line process parameter debugging method based on digital twin provided in this application embodiment, and has the same beneficial effects as the method implemented by the computer program being executed by a processor.

[0077] It is understood that the description of the various embodiments above tends to emphasize the differences between the various embodiments, and the similarities or similarities can be referred to each other. For the sake of brevity, these will not be repeated here.

[0078] Those skilled in the art will understand that, in the methods described above in specific embodiments, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined based on its function and possible internal logic.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and not to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and they should all be covered within the scope of the claims and specification of this application. In particular, as long as there is no structural conflict, the various technical features mentioned in the embodiments can be combined in any way. This application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for debugging process parameters of an engine production line based on digital twins, characterized in that, The method includes: Construct a full-process digital twin system that is synchronously mapped in real time to the physical engine production line, wherein the state of the virtual entity in the digital twin system is updated by real-time collected full-process production line data; When a fault is detected during the trial run, fault features are extracted from the trial run data, and based on a preset correlation matrix, key process parameters of the preceding manufacturing process associated with the fault features are located in reverse from the full production line data. Based on the identified key process parameters, a new combination of process parameters is generated, and a pre-trained quality prediction model is used to predict the quality of the adjustment scheme containing the new combination of process parameters. If the prediction results are satisfactory, the new combination of process parameters and historical production process data will be input into the digital twin system for virtual simulation to verify the effectiveness of the process adjustment plan.

2. The method as described in claim 1, characterized in that, The aforementioned full-process digital twin system, which is constructed and synchronously mapped in real time with the physical engine production line, includes: Throughout the entire process of engine production, including incoming materials, machining, automated assembly, manual assembly, and testing, real-time data on the production environment, production process, and product quality inspection results are collected at each stage. Each engine unit and key component is assigned a unique identifier, and the collected data is associated with and stored with the unique identifier. Based on the collected data, a digital twin system containing a fixed model, a motion model, and a behavior rule model is constructed in virtual space. The real-time collected data is then injected into the model through a data-driven engine to drive the motion model to perform synchronous mapping.

3. The method as described in claim 2, characterized in that, The unique identifier is an RFID electronic tag; The production process data includes: RFID electronic tags corresponding to engine parts during the incoming material process; During machining, the machining center number, tool ID and wear status, cutting parameters, and spindle operating parameters are recorded; during automated assembly, the robotic arm ID, robotic arm motion parameters, and gluing parameters are recorded; during manual assembly, the workstation number, operator ID, and power tool parameters are recorded; and during commissioning, the test bench number, operating status parameters, and fault codes are recorded.

4. The method as described in claim 1, characterized in that, The correlation matrix is ​​constructed based on historical production data and is used to map the fault weights between specific trial run fault characteristics and process parameters of multiple preceding manufacturing stages. It can also be dynamically updated based on newly added production and fault cases. In the correlation matrix, for each test failure feature, at least one key parameter from the main component change points, the processing change points, the automatic assembly process change points, and the manual assembly process change points are recorded and associated, and each key parameter is assigned a failure weight; wherein, the failure weight is determined through the accumulation of historical failure cases and production experience; when using the correlation matrix for reverse localization, the system automatically selects at least one key process parameter with the highest weight from the main component change points, the processing change points, the automatic assembly process change points, and the manual assembly process change points according to the level of the failure weight; The construction and updating of the correlation matrix and the training of the quality prediction model are all based on the same sample feature set, which includes: test run features, changes in major components, changes in the processing process, changes in the automatic assembly process, and changes in the manual assembly process.

5. The method as described in claim 4, characterized in that, The method, based on a preset correlation matrix, reverse-engineers key process parameters of preceding manufacturing stages associated with the fault characteristics from the entire production line data, including: Based on the fault characteristics, at least one key process parameter with the highest fault weight in the material processing, machining, automatic assembly, and manual assembly stages is automatically selected, and a reference value of the parameter in historical normal production is provided.

6. The method as described in claim 4, characterized in that, The sample feature set and the corresponding final product quality label are associated and stored in the quality data retention library, which is used to construct the correlation matrix and train the quality prediction model.

7. The method as described in claim 6, characterized in that, The quality prediction model is a random forest model and is trained in the following way: A training sample set is constructed by using the features in the sample feature set as input features and whether the product is ultimately abnormal as a binary classification label. Multiple sample subsets are generated from the training sample set through multiple random samplings with replacement. For each subset of samples, a decision tree is trained by randomly selecting some input features and splitting the nodes. The prediction results of all decision trees are integrated, and the final output of the random forest model is obtained through a voting mechanism.

8. A process parameter debugging device for an engine production line based on digital twins, characterized in that, The device includes: A construction module is used to build a full-process digital twin system that is synchronously mapped with the physical engine production line in real time. The state of the virtual entity in the digital twin system is updated by real-time collected full-process production line data. The positioning module is used to extract fault features from the test run data when a fault is detected during the test run, and to reverse locate the key process parameters of the preceding manufacturing process associated with the fault features from the full process data of the production line based on a preset correlation matrix. The prediction module is used to generate new combinations of process parameters based on the located key process parameters, and to use a pre-trained quality prediction model to predict the quality of the adjustment scheme containing the new combination of process parameters. The verification module is used to input the new combination of process parameters and historical production process data into the digital twin system for virtual simulation operation when the prediction results are qualified, so as to verify the effectiveness of the process adjustment plan.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by a processor to implement the method as described in any one of claims 1-7.

Images