Research and development project whole-process collaborative technology service system based on digital twinning

Abstract

This invention discloses a collaborative technology service system for the entire R&D project process based on digital twins, belonging to the field of digital twin technology. This invention constructs a digital twin model covering the entire R&D process, realizing attribute binding and bidirectional mapping between data from participating teams and virtual R&D entities. This enables automatic identification and quantification of data differences across teams, effectively solving the problem of difficulty in detecting data inconsistencies of the same R&D object across different teams in traditional methods. The digital twin model defines a collaborative process including serial, parallel, conditional judgment, and cyclic processing, and quantifies the execution order of each sub-task through a parallel task priority calculation formula, improving the execution efficiency and resource utilization of multi-team collaboration. Furthermore, by generating a parameter difference matrix and blank area description through multi-source data fusion processing, combined with an error type prediction model, potential risks can be proactively identified in the early stages of R&D, preventing problems from being concentrated in the integration phase.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, specifically to a collaborative technical service system for the entire process of R&D projects based on digital twins. Background Technology

[0002] As the complexity of integrated circuit and complex hardware product development continues to increase, R&D projects typically involve collaboration among multiple teams, including those responsible for architecture design, front-end design, verification, back-end implementation, and testing. During the R&D process, each team generates a large amount of design parameter data, verification record data, configuration change data, and external demand data from the market and supply chain. This data is scattered across different R&D tools and management systems, with varying data formats, time bases, and measurement standards.

[0003] In existing technologies, collaborative management of R&D projects mainly relies on project management platforms for task allocation and progress tracking. Data interaction between teams is primarily based on manual export and import, lacking a unified data alignment and integration mechanism. Different teams often have discrepancies in their understanding and data description of the same R&D object. For example, the architecture team and the verification team may have inconsistent descriptions of configuration parameters for the same IP block, and coverage data in different verification environments cannot be directly compared. These discrepancies are difficult to detect in the early stages of R&D and often only become apparent during the integration phase, leading to a significant increase in rework costs.

[0004] At the application level of digital twin technology, most existing solutions focus on digital twin modeling for a single stage in the R&D process. For example, they may only create a digital twin model for the functional verification process of a chip, or only simulate and extrapolate the manufacturing process. There is a lack of a unified digital twin framework covering the entire R&D process. The digital twin models for each stage are independent of each other, making cross-stage data linkage and collaborative simulation impossible. This results in the evaluation of R&D solutions being limited to a local scope, preventing a quantitative comparison and optimization of the collaborative effectiveness of different solutions from a global perspective.

[0005] In terms of R&D risk management, existing technologies mostly adopt a post-event analysis approach, that is, tracing back logs to locate the cause after a problem occurs, lacking the ability to proactively identify potential risks based on multi-source data fusion analysis. There is a lack of cross-comparison mechanisms between the verification activities and change records of different teams, and existing technologies cannot automatically identify and issue warnings about whether there are blind spots in the verification activities performed by different teams on the same object within the same time period.

[0006] In terms of R&D decision support, existing solutions lack a closed-loop feedback mechanism between simulation results and actual execution data. The parameters and evaluation rules of the simulation model are usually not updated after initial establishment, causing the deviation between the model's predictions and actual performance to accumulate and increase over time, thus continuously reducing the reference value of the simulation. Summary of the Invention

[0007] To address the aforementioned technical issues, this technical solution provides a collaborative technical service system for the entire R&D project process based on digital twins, resolving the problems mentioned in the background section.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A collaborative technical service system for the entire R&D project process based on digital twins, including: Data Acquisition Module: Collects multi-source data generated by participating teams throughout the entire R&D project process. The multi-source data includes design parameter data, verification record data, configuration change data, and external requirement data. The collected multi-source data is processed for timestamp alignment and unit unification. Module Establishment: Based on the processed multi-source data, construct a digital twin model of the R&D project, bind the data of each participating team to the corresponding virtual R&D entities in the digital twin model, and establish a mapping relationship between data and model; Define the module: Define the collaborative process of R&D tasks in the digital twin model, decompose the collaborative process into multiple sub-tasks, determine the serial, parallel, conditional judgment and loop processing execution logic between each sub-task, and define the execution order of each sub-task. Generation module: It integrates and processes the data of each participating team in the digital twin model, compares the data differences of the same R&D object in different teams, generates difference analysis results, and identifies potential risks in the R&D process based on the difference analysis results; Evaluation module: Based on the digital twin model, the R&D plan is simulated and deduced in multiple dimensions to obtain the performance index data of each plan under different working conditions, and the collaborative effectiveness of each plan is evaluated and an evaluation ranking is generated. Update module: Compare the simulation results with the actual R&D execution data, and calibrate and update the parameters and evaluation rules of the digital twin model based on the comparison analysis results to form a decision feedback closed loop.

[0009] Preferably, the collection of multi-source data generated by various participating teams throughout the entire R&D project process specifically includes: The unified data interface accesses the R&D design tools output data of each participating team. The R&D design tools output data includes the configuration parameter set of each IP block, the external connection modification description of each IP block, and the verification activity record of each IP block. The configuration parameter set includes register configuration values, clock frequency parameters, power consumption parameters, and signal integrity parameters. Process data is extracted from the R&D management systems of each participating team. The process data includes personnel and task identifiers in task assignment records, time nodes and completion percentages in progress update data, and change content and approval status in change approval records. Access external demand data sources, including user scenario descriptions and fault feedback from market feedback data, product specifications and delivery times from customer order data, and material availability and delivery information from supply chain status data. The above-mentioned multi-source data is subjected to timestamp alignment processing. With the unified time base of the R&D project as a reference, the timestamps of data from different sources are converted into a unified format and sorted according to the time series. To unify the units of data with different dimensions, different physical quantities such as temperature, voltage, and frequency are converted into a unified measurement system. This is specifically achieved through the following dimensional conversion formula: ; in, This represents the normalized value after dimension unification. This represents the numerical value of the physical quantity in the original data. This represents the minimum value of the physical quantity across all data sources. This represents the maximum value of the physical quantity across all data sources.

[0010] Preferably, the construction of a digital twin model of the R&D project based on the processed multi-source data specifically includes: Based on the design documents, CAD drawings and layout data submitted by each participating team, construct a three-dimensional geometric model of each physical entity in the R&D project. The three-dimensional geometric model includes the physical structure of each IP block, the connection relationship between each module and the geometric features of each interface. The configuration parameter data of each participating team is mapped to the attribute fields of the corresponding 3D geometric model, the verification record data is associated with the verification status field of the corresponding 3D geometric model, and the configuration change data is recorded as the version change log of the corresponding 3D geometric model, thus establishing a two-way mapping relationship between data and model. Version comparison of the 3D geometric models of the same research object in different teams was performed. Parameter difference fields and structural difference fields were extracted between the versions. The parameter difference fields were recorded as numerical deviations, and the structural difference fields were recorded as topology change descriptions. The numerical deviations were quantified using the following parameter deviation calculation formula: ; in, Indicates the first The team and the first Overall parameter deviation value between teams Indicates the first The weighting coefficients of each parameter. Indicates the first The first team The values ​​of the parameters, In the j-th team, the first... The values ​​of the parameters, Indicates the total number of parameters involved in the comparison; The model parameter calibration rules are trained based on data from historical R&D projects. The model parameter calibration rules include parameter deviation tolerance threshold and structural change impact range determination logic. When the deviation between newly accessed data and existing model data exceeds the parameter deviation tolerance threshold, the model parameter adjustment process is automatically triggered, and the scope of related models that need to be updated synchronously is determined according to the structural change impact range determination logic.

[0011] Preferably, the collaborative process for defining R&D tasks in the digital twin model specifically includes: Obtain dependency data between subtasks in the R&D task. The dependency data includes the identifier of the preceding task, the identifier of the following task, and the dependency type. The dependency type is classified into serial dependency, parallel dependency, conditional branch dependency, and loop iteration dependency. For each subtask, the input data requirements for that subtask are extracted from the corresponding participating team data. The input data requirements include the required data type, data precision requirements, and data timeliness requirements. At the same time, the output data format for that subtask is defined, which includes data field names, data field types, and data encoding rules. Based on the dependency type and execution constraints of each subtask, the subtasks are arranged into an ordered task execution sequence. The task execution sequence includes parallel task groups that can be executed simultaneously and condition judgment nodes that need to determine the direction based on the condition judgment results. The execution constraints include resource consumption limits, time window limits, and data dependency integrity requirements. Configure exception handling rule data for each task execution node. The exception handling rule data includes task termination triggering conditions, task continuation triggering conditions, parameter adjustment range, and execution speed reduction. When the real-time data during task execution meets any exception handling rule, intervene in task execution according to the corresponding rule. The execution priority of each parallel subtask in the parallel task group is determined by the following parallel task priority calculation formula: ; in, Indicates the first The combined priority score of each parallel subtask. Indicates the first The estimated completion time of each parallel subtask. Indicates the first The amount of resources required for each parallel subtask This represents the total amount of currently available resources. Indicates the first The urgency level of each parallel subtask. , , These are the weights for the time factor, resource factor, and urgency factor, respectively, and they satisfy the following conditions: .

[0012] Preferably, the data fusion processing of each participating team in the digital twin model specifically includes: Multiple instances of the same research and development object are selected, and the configuration parameter sets of each instance in its respective participating team are obtained. The configuration parameter sets of each instance are compared field by field, and the fields with differences are recorded as difference fields. The difference values, difference directions, and difference magnitudes of the difference fields are used to form a parameter difference matrix. The parameter difference matrix is ​​constructed using the following formula: ; in, Represents the parameter difference matrix. Indicates the first The first instance and the first The difference value of each instance on the corresponding parameter field, Indicates the total number of instances. Indicates the total number of parameter fields; Obtain the verification activity records of each participating team, align the verification activity records of each team with the unified timeline of the R&D project, identify the verification activities performed by different teams on the same R&D object within the same time period, cross-compare the types, coverage and verification results of the verification activities, and generate a verification activity difference report. Coverage data from multiple instances is integrated, and set operations are performed on the coverage data of each instance. The IP block functional regions not covered by any instance are identified through set difference operations. These identified functional regions are recorded as blank areas, and a blank area description is generated, including the blank area's location, functional description, and impact assessment. The blank areas are determined using the following set difference formula: ; in, Represents the set of blank regions. Represents the complete set of functional areas of an IP block. Indicates the first The set of verification coverage areas for each instance. Indicates the total number of instances; Based on error type data and corresponding triggering scenario data recorded in historical R&D projects, an error type prediction model is trained. The trained error type prediction model is then used to analyze the current fused data, outputting the potential error types present in the current data, as well as the triggering scenario description for each error type. The triggering scenario description includes the combination of data conditions required to trigger the error type.

[0013] Preferably, the identification of potential risks in the R&D process based on the results of the difference analysis specifically includes: Match the predicted error type with the tasks currently being performed by each participating team, identify the target teams and target task nodes affected by the error type, and record the matching results as risk impact scope data; For each error type and its corresponding triggering scenario, similar error type and triggering scenario combinations are retrieved from the historical collaboration solution library, and corresponding historical collaboration processing suggestions are extracted. The historical collaboration processing suggestions are then adaptively adjusted based on the actual constraints of the current R&D project to generate collaboration processing suggestions corresponding to the current error type. The collaboration processing suggestions include the parameter fields that need to be adjusted and the adjustment range, the types of verification activities that need to be added and the verification coverage, and the participating teams and coordination content that need to be coordinated. The generated collaborative processing suggestions are prioritized according to the magnitude of their impact and urgency in the risk impact data. This prioritization is performed using the following risk priority calculation formula: ; in, This indicates the overall priority ranking score of the risk items. The risk impact range score is indicated, with a value ranging from 1 to 10. This indicates the urgency of the risk, with a value ranging from 1 to 10. This represents a risk probability score, with a value ranging from 0 to 1. , , These are the influence range weight factor, urgency weight factor, and probability weight factor, respectively, and they satisfy the following conditions: ; The priority-ordered collaborative processing suggestions are integrated into a collaborative solution document, which includes the execution order of each collaborative processing suggestion, the responsible team for execution, and a description of the expected results. The collaborative solution document is distributed to all relevant participating teams. After receiving it, each participating team feeds back the execution results to the digital twin model. The execution results include the actual adjusted parameter values, the actual additional verification activity records, and the actual coordination execution status.

[0014] Preferably, the multi-dimensional simulation and deduction of the R&D plan based on the digital twin model specifically includes: Load the R&D scheme parameters to be evaluated in the digital twin environment. The R&D scheme parameters include design parameter combinations, process parameter combinations, resource allocation parameters and verification strategy parameters. Map each parameter to the adjustable attribute fields of the corresponding virtual entity in the digital twin model. For each R&D scheme parameter combination, simulation is performed sequentially in the digital twin environment. During the simulation, real-time status data of each key node is collected. The key nodes include the functional verification nodes of each IP block, the interface verification nodes of each module, and the integration verification nodes of each subsystem. The performance index data of each key node after the simulation is completed is obtained. The performance index data includes functional coverage, timing margin, power consumption value, and signal integrity index. The performance index data of each key node for each R&D scheme parameter combination are summarized, and the offset of each key node relative to the reference process parameters is calculated. The offset is calculated using the following formula: ; in, Indicates the first Percentage of performance offset for each critical node Indicates the first Performance metrics of key nodes under the current simulation scheme Indicates the first under the reference process parameters Baseline performance metrics for each key node; By combining the weight coefficients of each key node in the overall R&D process, a collaborative efficiency evaluation function is constructed to quantitatively score the parameter combinations of each R&D scheme, generating a simulation result table containing the scores of each scheme and the performance indicators of each key node. The collaborative efficiency evaluation function is calculated using the following formula: ; in, This indicates the collaborative effectiveness evaluation score of the R&D plan. Indicates the first The weight coefficient of each key node in the overall R&D process. Indicates the first Performance offset of each critical node This indicates the maximum allowed performance offset threshold. Indicates the total number of critical nodes; The parameter combinations of each R&D scheme in the simulation results table are sorted in descending order according to the score of the collaborative effectiveness evaluation function. A predetermined number of schemes with the highest scores are selected as candidate optimal schemes, and the performance characteristics of each candidate optimal scheme in each target dimension are marked.

[0015] Preferably, comparing the simulation results with the actual R&D execution data specifically includes: Record the execution time, responsible person, data content on which the decision was based, specific actions taken to execute the decision, and actual effects after the decision is executed for each R&D decision. Compile the above information into decision trace data and store it persistently. The prediction results of the digital twin model for each decision node are compared item by item with the actual effects in the decision trace data. The deviation value between the prediction result and the actual result of each decision node is calculated using the following formula: ; in, This represents the root mean square deviation of the decision node. Indicates the digital twin model for the first The predicted value of each decision indicator, Indicating the first in the decision trace data The actual value of each decision indicator Indicates the total number of decision-making indicators; Decision nodes whose deviation values ​​exceed a preset threshold are marked as high deviation nodes; For each high-deviation node, extract the data content and specific actions that the node relied on when making the decision from the decision trace data, compare them with the input parameters and simulation logic of the node in the digital twin model, identify the data factors or model factors that cause the deviation, and record the identification results as root cause analysis data. Based on the root cause analysis data, the parameters or evaluation logic of the corresponding nodes in the digital twin model are modified. The modification includes adjusting the values ​​of the model parameters, modifying the judgment conditions of the evaluation logic, and supplementing the missing influencing factors.

[0016] Preferably, the step of calibrating and updating the parameters and evaluation rules of the digital twin model based on the comparative analysis results specifically includes: Records marked as successful in decision-making trace data are filtered out, and the data features and execution results corresponding to successful decisions are extracted. These are then added to the model training set as positive sample data, and the corresponding decision rules in the model are reinforced using the positive sample data. The records marked as decision failures in the decision trace data are filtered, the data features and execution results corresponding to the failed decisions are extracted, the reasons for failure are analyzed and the model parameters or rules that need to be corrected are determined, and the corresponding parameters or rules in the model are corrected and trained using the failure sample data. According to a preset time period, the digital twin model is fully refreshed using the latest collected data from the entire R&D project process. The intermediate state data of the model generated during the full data refresh process is compared with the model state data before the refresh to verify the effectiveness of the model update. The effectiveness of the model update is verified by the following model performance change ratio formula: ; in, This represents the rate of change in model performance. This represents the collaborative effectiveness evaluation score after the model update. This represents the collaborative effectiveness evaluation score before the model update. The model update is considered valid when the value is greater than zero. If the value is less than or equal to zero, the model update is deemed invalid and the system is rolled back to the state before the update. Establish version management data for digital twin models, recording the update content, update time, performance comparison data of the model before and after the update, and the person responsible for the update operation for each model update.

[0017] Preferably, the process of aligning timestamps and unifying units for the collected multi-source data further includes: For fields with missing values ​​in multi-source data, missing values ​​are filled using interpolation of nearby time points or the mean of similar objects, based on the historical data distribution characteristics of the field. The filled data is then marked as processed data. The nearby time point interpolation is performed using the following linear interpolation formula: ; in, Indicates a point in time Interpolation padding at the location, Indicates a point in time Known data values ​​at that location Indicates a point in time Known data values ​​at that location ; For fields with outliers in multi-source data, upper and lower thresholds are set based on the data fluctuation range of the field. Data points exceeding the upper and lower thresholds are marked as suspicious data. Suspicious data undergoes secondary verification. If the secondary verification still determines that the data is abnormal, it is removed and recorded as outlier removed data. The upper and lower thresholds are determined by the following formula: ; ; in, This represents the upper threshold for outlier detection. This represents the lower threshold for outlier detection. This represents the mean of the data in this field. This represents the standard deviation of the data in this field. This represents the anomaly determination coefficient, with a value ranging from 2 to 3; Processed data and anomaly-removed data are stored separately. A unified dimensional conversion is performed on the processed data to convert similar physical quantities from different sources into a unified unit of measurement. The converted data undergoes a consistency check to ensure that the numerical differences of the same physical quantity across different data sources are within acceptable limits. This consistency check is performed using the following consistency deviation formula: ; in, This indicates the percentage deviation in consistency between two data sources for the same physical quantity. This represents the value of the physical quantity in data source A. Indicates data source The value of this physical quantity, when Data is considered consistent when the value is less than the preset consistency tolerance threshold. When the data is greater than or equal to the preset consistency tolerance threshold, the data is determined to be inconsistent and the data tracing process is triggered.

[0018] Compared with existing technologies, this invention provides a collaborative technical service system for the entire process of R&D projects based on digital twins, which has the following beneficial effects: This invention eliminates data barriers between different teams and tools by performing timestamp alignment, dimensional unification, and intelligent processing of missing and outlier values ​​on multi-source data throughout the entire R&D process, providing a unified and reliable data foundation for subsequent analysis. Based on the processed multi-source data, a digital twin model covering the entire R&D process is constructed, enabling attribute binding and bidirectional mapping between data from participating teams and virtual R&D entities. This allows for the automatic identification and quantification of cross-team data differences, effectively solving the problem of difficulty in detecting data inconsistencies of the same R&D object across different teams in traditional methods. The digital twin model defines a collaborative process including serial, parallel, conditional, and cyclical processing, and quantifies the execution order of each sub-task through a parallel task priority calculation formula, improving the execution efficiency and resource utilization of multi-team collaboration. Furthermore, by generating a parameter difference matrix and blank area description through multi-source data fusion processing, combined with an error type prediction model, potential risks can be proactively identified in the early stages of R&D, preventing problems from being concentrated in the integration phase. Finally, through multi-dimensional simulation and collaborative performance evaluation functions, each R&D solution is quantitatively scored and ranked, achieving objectivity and quantifiability in solution selection. Meanwhile, by comparing the simulation results with the actual execution data to form a decision feedback loop, the digital twin model can be continuously calibrated and updated, avoiding the problem of model accuracy decaying over time. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the system module framework of the present invention; Figure 2 This is a schematic diagram of the method flow for S201-S205 in this invention; Figure 3 This is a schematic diagram of the method flow for S301-S304 in this invention; Figure 4 This is a schematic diagram of the method flow for S401-S404 in this invention. Detailed Implementation

[0020] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0021] Example 1 Please refer to Figure 1 As shown, the collaborative technical service system for the entire R&D project process based on digital twins includes: Data Acquisition Module: Collects multi-source data generated by participating teams throughout the entire R&D project process. The multi-source data includes design parameter data, verification record data, configuration change data, and external requirement data. The collected multi-source data is processed for timestamp alignment and unit unification. Module Establishment: Based on the processed multi-source data, construct a digital twin model of the R&D project, bind the data of each participating team to the corresponding virtual R&D entities in the digital twin model, and establish a mapping relationship between data and model; Define the module: Define the collaborative process of R&D tasks in the digital twin model, decompose the collaborative process into multiple sub-tasks, determine the serial, parallel, conditional judgment and loop processing execution logic between each sub-task, and define the execution order of each sub-task. Generation module: It integrates and processes the data of each participating team in the digital twin model, compares the data differences of the same R&D object in different teams, generates difference analysis results, and identifies potential risks in the R&D process based on the difference analysis results; Evaluation module: Based on the digital twin model, the R&D plan is simulated and deduced in multiple dimensions to obtain the performance index data of each plan under different working conditions, and the collaborative effectiveness of each plan is evaluated and an evaluation ranking is generated. Update module: Compare the simulation results with the actual R&D execution data, and calibrate and update the parameters and evaluation rules of the digital twin model based on the comparison analysis results to form a decision feedback closed loop.

[0022] As will be understood by those skilled in the art, this invention eliminates data barriers between different teams and tools by performing timestamp alignment, dimensional unification, and intelligent processing of missing and outlier values ​​on multi-source data throughout the entire R&D process, providing a unified and reliable data foundation for subsequent analysis. Based on the processed multi-source data, a digital twin model covering the entire R&D process is constructed, enabling attribute binding and bidirectional mapping between data from each participating team and virtual R&D entities. This allows for the automatic identification and quantification of cross-team data differences, effectively solving the problem of difficulty in detecting data inconsistencies of the same R&D object across different teams in traditional methods. The digital twin model defines a collaborative process including serial, parallel, conditional judgment, and cyclic processing, and quantifies the execution order of each sub-task through a parallel task priority calculation formula, improving the execution efficiency and resource utilization of multi-team collaboration. Furthermore, by generating a parameter difference matrix and blank area description through multi-source data fusion processing, combined with an error type prediction model, potential risks can be proactively identified in the early stages of R&D, preventing problems from being concentrated in the integration phase. Finally, by using multi-dimensional simulation and collaborative performance evaluation functions to quantitatively score and rank each R&D solution, objectivity and quantifiability of solution selection are achieved. Meanwhile, by comparing the simulation results with the actual execution data to form a decision feedback loop, the digital twin model can be continuously calibrated and updated, avoiding the problem of model accuracy decaying over time.

[0023] Please refer to Figure 2 As shown, the data collected includes multi-source data generated by various participating teams throughout the entire R&D project process, specifically: S201. Access the R&D design tools output data of each participating team through a unified data interface. The R&D design tools output data includes the configuration parameter set of each IP block, the external connection modification description of each IP block, and the verification activity record of each IP block. The configuration parameter set includes register configuration values, clock frequency parameters, power consumption parameters, and signal integrity parameters. S202. Extract process data from the R&D management system of each participating team. The process data includes personnel and task identifiers in the task assignment record, time nodes and completion percentages in the progress update data, and change content and approval status in the change approval record. S203. Access external demand data sources, including user scenario descriptions and fault feedback in market feedback data, product specifications and delivery time in customer order data, and material availability status and delivery information in supply chain status data. S204. Perform timestamp alignment processing on the above multi-source data. Using the unified time base of the R&D project as a reference, convert the timestamps of data from different sources into a unified format and sort them according to the time series. S205. Perform unit unification processing on data with different dimensions, converting different physical quantities such as temperature, voltage, and frequency into a unified measurement system. Specifically, this is done using the following unit unification conversion formula: ; in, This represents the normalized value after dimension unification. This represents the numerical value of the physical quantity in the original data. This represents the minimum value of the physical quantity across all data sources. This represents the maximum value of the physical quantity across all data sources.

[0024] Please refer to Figure 3 As shown, a digital twin model of the R&D project is constructed based on the processed multi-source data, specifically including: S301. Based on the design documents, CAD drawings and layout data submitted by each participating team, construct a three-dimensional geometric model of each physical entity in the R&D project. The three-dimensional geometric model includes the physical structure of each IP block, the connection relationship between each module and the geometric features of each interface. S302. Map the configuration parameter data of each participating team to the attribute fields of the corresponding 3D geometric model, associate the verification record data with the verification status field of the corresponding 3D geometric model, record the configuration change data as the version change log of the corresponding 3D geometric model, and establish a two-way mapping relationship between data and model. S303. Compare the versions of the 3D geometric models of the same research object in different teams, extract the parameter difference fields and structural difference fields between the versions, record the parameter difference fields as numerical deviations, and record the structural difference fields as topology change descriptions. The numerical deviations are quantified using the following parameter deviation calculation formula: ; in, Indicates the first The team and the first Overall parameter deviation value between teams Indicates the first The weighting coefficients of each parameter. Indicates the first The first team The values ​​of the parameters, In the j-th team, the first... The values ​​of the parameters, Indicates the total number of parameters involved in the comparison; S304. Model parameter calibration rules based on data from historical R&D projects. The model parameter calibration rules include parameter deviation tolerance threshold and structural change impact range determination logic. When the deviation between newly accessed data and existing model data exceeds the parameter deviation tolerance threshold, the model parameter adjustment process is automatically triggered, and the scope of related models that need to be updated synchronously is determined according to the structural change impact range determination logic.

[0025] Please refer to Figure 4 As shown, the collaborative process for R&D tasks is defined in the digital twin model, specifically including: S401. Obtain the dependency relationship data between each subtask in the R&D task. The dependency relationship data includes the identifier of the preceding task, the identifier of the following task, and the dependency type. Classify the dependency type into serial dependency, parallel dependency, conditional branch dependency, and loop iteration dependency. S402. For each subtask, extract the input data requirements for the subtask from the corresponding participating team data. The input data requirements include the required data type, data precision requirements, and data timeliness requirements. At the same time, define the output data format for the subtask. The output data format includes data field names, data field types, and data encoding rules. S403. Based on the dependency type and execution constraints of each subtask, arrange each subtask into an ordered task execution sequence. The task execution sequence includes parallel task groups that can be executed simultaneously and condition judgment nodes that need to determine the direction based on the condition judgment results. The execution constraints include resource consumption limits, time window limits, and data dependency integrity requirements. S404. Configure exception handling rule data for each task execution node. The exception handling rule data includes task termination triggering conditions, task continuation triggering conditions, parameter adjustment range, and execution speed reduction. When the real-time data during task execution meets any exception handling rule, intervene in task execution according to the corresponding rule. The execution priority of each parallel subtask in the parallel task group is determined by the following parallel task priority calculation formula: ; in, Indicates the first The combined priority score of each parallel subtask. Indicates the first The estimated completion time of each parallel subtask. Indicates the first The amount of resources required for each parallel subtask This represents the total amount of currently available resources. Indicates the first The urgency level of each parallel subtask. , , These are the weights for the time factor, resource factor, and urgency factor, respectively, and they satisfy the following conditions: .

[0026] The data from each participating team in the digital twin model were merged and processed, specifically including: Multiple instances of the same research and development object are selected, and the configuration parameter sets of each instance in its respective participating team are obtained. The configuration parameter sets of each instance are compared field by field, and the fields with differences are recorded as difference fields. The difference values, difference directions, and difference magnitudes of the difference fields are used to form a parameter difference matrix. The parameter difference matrix is ​​constructed using the following formula: ; in, Represents the parameter difference matrix. Indicates the first The first instance and the first The difference value of each instance on the corresponding parameter field, Indicates the total number of instances. Indicates the total number of parameter fields; Obtain the verification activity records of each participating team, align the verification activity records of each team with the unified timeline of the R&D project, identify the verification activities performed by different teams on the same R&D object within the same time period, cross-compare the types, coverage and verification results of the verification activities, and generate a verification activity difference report. Coverage data from multiple instances is integrated, and set operations are performed on the coverage data of each instance. The IP block functional regions not covered by any instance are identified through set difference operations. These identified functional regions are recorded as blank areas, and a blank area description is generated, including the blank area location, functional description, and impact assessment. The blank areas are determined using the following set difference formula: ; in, Represents the set of blank regions. Represents the complete set of functional areas of an IP block. Indicates the first The set of verification coverage areas for each instance. Indicates the total number of instances; Based on error type data and corresponding triggering scenario data recorded in historical R&D projects, an error type prediction model is trained. The trained error type prediction model is then used to analyze the current fused data, outputting the potential error types present in the current data, as well as the triggering scenario description for each error type. The triggering scenario description includes the combination of data conditions required to trigger the error type.

[0027] Based on the results of the difference analysis, potential risks in the R&D process were identified, specifically including: Match the predicted error type with the tasks currently being performed by each participating team, identify the target teams and target task nodes affected by the error type, and record the matching results as risk impact scope data; For each error type and its corresponding triggering scenario, similar error type and triggering scenario combinations are retrieved from the historical collaboration solution library, and corresponding historical collaboration processing suggestions are extracted. The historical collaboration processing suggestions are then adaptively adjusted based on the actual constraints of the current R&D project to generate collaboration processing suggestions for the current error type. The collaboration processing suggestions include the parameter fields that need to be adjusted and the adjustment range, the types of verification activities that need to be added and the verification coverage, and the participating teams and coordination content that need to be coordinated. The generated collaborative processing suggestions are prioritized according to the magnitude of their impact and urgency in the risk impact data. The prioritization is performed using the following risk priority calculation formula: ; in, This indicates the overall priority ranking score of the risk items. The risk impact range score is indicated, with a value ranging from 1 to 10. This indicates the urgency of the risk, with a value ranging from 1 to 10. This represents a risk probability score, with a value ranging from 0 to 1. , , These are the influence range weight factor, urgency weight factor, and probability weight factor, respectively, and they satisfy the following conditions: ; The priority-ordered collaborative processing suggestions are integrated into a collaborative solution document, which includes the execution order of each collaborative processing suggestion, the responsible team for execution, and a description of the expected results. The collaborative solution document is distributed to all relevant participating teams. After receiving it, each participating team feeds back the execution results to the digital twin model. The execution results include the actual adjusted parameter values, the actual additional verification activity records, and the actual coordination execution status.

[0028] Multi-dimensional simulation and deduction of the R&D plan based on the digital twin model, specifically including: Load the R&D scheme parameters to be evaluated in the digital twin environment. The R&D scheme parameters include the combination of design parameters, the combination of process parameters, the resource allocation parameters, and the verification strategy parameters. Map each parameter to the adjustable attribute fields of the corresponding virtual entity in the digital twin model. For each R&D scheme parameter combination, simulation is performed sequentially in the digital twin environment. During the simulation, real-time status data of each key node is collected. Key nodes include functional verification nodes of each IP block, interface verification nodes of each module, and integration verification nodes of each subsystem. Performance index data of each key node is obtained after the simulation is completed. Performance index data includes functional coverage, timing margin, power consumption and signal integrity index. For each R&D scheme parameter combination, the performance index data of each key node are summarized, and the offset of each key node relative to the reference process parameters is calculated. The offset is calculated using the following formula: ; in, Indicates the first Percentage of performance offset for each critical node Indicates the first Performance metrics of key nodes under the current simulation scheme Indicates the first under the reference process parameters Baseline performance metrics for each key node; By combining the weight coefficients of each key node in the overall R&D process, a collaborative efficiency evaluation function is constructed to quantitatively score the parameter combinations of each R&D scheme, generating a simulation result table containing the scores of each scheme and the performance indicators of each key node. The collaborative efficiency evaluation function is calculated using the following formula: ; in, This indicates the collaborative effectiveness evaluation score of the R&D plan. Indicates the first The weight coefficient of each key node in the overall R&D process. Indicates the first Performance offset of each critical node This indicates the maximum allowed performance offset threshold. Indicates the total number of critical nodes; The parameter combinations of each R&D scheme in the simulation results table are sorted in descending order according to the score of the collaborative effectiveness evaluation function. A predetermined number of schemes with the highest scores are selected as candidate optimal schemes, and the performance characteristics of each candidate optimal scheme in each target dimension are marked.

[0029] The simulation results were compared with the actual R&D execution data, specifically including: Record the execution time, responsible person, data content on which the decision was based, specific actions taken to execute the decision, and actual effects after the decision is executed for each R&D decision. Compile the above information into decision trace data and store it persistently. The prediction results of the digital twin model for each decision node are compared with the actual effects in the decision trace data item by item. The deviation value between the prediction result and the actual result of each decision node is calculated using the following formula: ; in, This represents the root mean square deviation of the decision node. Indicates the digital twin model for the first The predicted value of each decision indicator, Indicating the first in the decision trace data The actual value of each decision indicator Indicates the total number of decision-making indicators; Decision nodes whose deviation values ​​exceed a preset threshold are marked as high deviation nodes; For each high-deviation node, extract the data content and specific actions that the node relied on when making the decision from the decision trace data, compare them with the input parameters and simulation logic of the node in the digital twin model, identify the data factors or model factors that cause the deviation, and record the identification results as root cause analysis data. Based on the root cause analysis data, the parameters or evaluation logic of the corresponding nodes in the digital twin model are modified. The modifications include adjusting the values ​​of the model parameters, modifying the judgment conditions of the evaluation logic, and supplementing the missing influencing factors.

[0030] Based on the comparative analysis results, the parameters and evaluation rules of the digital twin model are calibrated and updated, specifically including: Records marked as successful in decision-making trace data are filtered out, and the data features and execution results corresponding to successful decisions are extracted. These are then added to the model training set as positive sample data, and the corresponding decision rules in the model are reinforced using the positive sample data. The records marked as decision failures in the decision trace data are filtered, the data features and execution results corresponding to the failed decisions are extracted, the reasons for failure are analyzed and the model parameters or rules that need to be corrected are determined, and the corresponding parameters or rules in the model are corrected and trained using the failure sample data. According to the preset time period, the digital twin model is fully refreshed using the latest collected full-process data of the R&D project. The intermediate state data of the model generated during the full data refresh process is compared with the model state data before the refresh to verify the effectiveness of the model update. The effectiveness of the model update is verified by the following model performance change ratio formula: ; in, This represents the rate of change in model performance. This represents the collaborative effectiveness evaluation score after the model update. This represents the collaborative effectiveness evaluation score before the model update. The model update is considered valid when the value is greater than zero. If the value is less than or equal to zero, the model update is deemed invalid and the system is rolled back to the state before the update. Establish version management data for digital twin models, recording the update content, update time, performance comparison data of the model before and after the update, and the person responsible for the update operation for each model update.

[0031] The process of aligning timestamps and unifying units for multi-source data also includes: For fields with missing values ​​in multi-source data, missing values ​​are filled using interpolation of nearby time points or the mean of similar objects, based on the historical data distribution characteristics of the field. The filled data is then marked as processed data. Nearby time point interpolation is performed using the following linear interpolation formula: ; in, Indicates a point in time Interpolation padding at the location, Indicates a point in time Known data values ​​at that location Indicates a point in time Known data values ​​at that location ; For fields with outliers in multi-source data, upper and lower thresholds are set based on the data fluctuation range of the field. Data points exceeding the upper and lower thresholds are marked as suspicious data. Suspicious data undergoes secondary verification. If the secondary verification still determines that the data is abnormal, it is removed and recorded as outlier removed data. The upper and lower thresholds are determined by the following formula: ; ; in, This represents the upper threshold for outlier detection. This represents the lower threshold for outlier detection. This represents the mean of the data in this field. This represents the standard deviation of the data in this field. This represents the anomaly determination coefficient, with a value ranging from 2 to 3; Processed data and outlier-removed data are stored separately. A unified dimensional conversion is performed on the processed data to convert similar physical quantities from different sources into a unified unit of measurement. The converted data undergoes consistency verification to ensure that the numerical differences of the same physical quantity across different data sources are within acceptable limits. Consistency verification is performed using the following consistency deviation formula: ; in, This indicates the percentage deviation in consistency between two data sources for the same physical quantity. This represents the value of the physical quantity in data source A. Indicates data source The value of this physical quantity, when Data is considered consistent when the value is less than the preset consistency tolerance threshold. When the data is greater than or equal to the preset consistency tolerance threshold, the data is determined to be inconsistent and the data tracing process is triggered.

[0032] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A collaborative technical service system for the entire R&D project process based on digital twins, characterized in that: include: Data Acquisition Module: Collects multi-source data generated by participating teams throughout the entire R&D project process. The multi-source data includes design parameter data, verification record data, configuration change data, and external requirement data. The collected multi-source data is processed for timestamp alignment and unit unification. Module Establishment: Based on the processed multi-source data, construct a digital twin model of the R&D project, bind the data of each participating team to the corresponding virtual R&D entities in the digital twin model, and establish a mapping relationship between data and model; Define the module: Define the collaborative process of R&D tasks in the digital twin model, decompose the collaborative process into multiple sub-tasks, determine the serial, parallel, conditional judgment and loop processing execution logic between each sub-task, and define the execution order of each sub-task. Generation module: It integrates and processes the data of each participating team in the digital twin model, compares the data differences of the same R&D object in different teams, generates difference analysis results, and identifies potential risks in the R&D process based on the difference analysis results; Evaluation module: Based on the digital twin model, the R&D plan is simulated and deduced in multiple dimensions to obtain the performance index data of each plan under different working conditions, and the collaborative effectiveness of each plan is evaluated and an evaluation ranking is generated. Update module: Compare the simulation results with the actual R&D execution data, and calibrate and update the parameters and evaluation rules of the digital twin model based on the comparison analysis results to form a decision feedback closed loop.

2. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 1, characterized in that, The data collected from multiple sources by participating teams throughout the entire R&D project process specifically includes: The unified data interface accesses the R&D design tools output data of each participating team. The R&D design tools output data includes the configuration parameter set of each IP block, the external connection modification description of each IP block, and the verification activity record of each IP block. The configuration parameter set includes register configuration values, clock frequency parameters, power consumption parameters, and signal integrity parameters. Process data is extracted from the R&D management systems of each participating team. The process data includes personnel and task identifiers in task assignment records, time nodes and completion percentages in progress update data, and change content and approval status in change approval records. Access external demand data sources, including user scenario descriptions and fault feedback from market feedback data, product specifications and delivery times from customer order data, and material availability and delivery information from supply chain status data. The above-mentioned multi-source data is subjected to timestamp alignment processing. With the unified time base of the R&D project as a reference, the timestamps of data from different sources are converted into a unified format and sorted according to the time series. To unify the units of data with different dimensions, different physical quantities such as temperature, voltage, and frequency are converted into a unified measurement system. This is specifically achieved through the following dimensional conversion formula: ； in, This represents the normalized value after dimension unification. This represents the numerical value of the physical quantity in the original data. This represents the minimum value of the physical quantity across all data sources. This represents the maximum value of the physical quantity across all data sources.

3. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 2, characterized in that, The construction of a digital twin model for the R&D project based on the processed multi-source data specifically includes: Based on the design documents, CAD drawings and layout data submitted by each participating team, construct a three-dimensional geometric model of each physical entity in the R&D project. The three-dimensional geometric model includes the physical structure of each IP block, the connection relationship between each module and the geometric features of each interface. The configuration parameter data of each participating team is mapped to the attribute fields of the corresponding 3D geometric model, the verification record data is associated with the verification status field of the corresponding 3D geometric model, and the configuration change data is recorded as the version change log of the corresponding 3D geometric model, thus establishing a two-way mapping relationship between data and model. Version comparison of the 3D geometric models of the same research object in different teams was performed. Parameter difference fields and structural difference fields were extracted between the versions. The parameter difference fields were recorded as numerical deviations, and the structural difference fields were recorded as topology change descriptions. The numerical deviations were quantified using the following parameter deviation calculation formula: ； in, Indicates the first The team and the first Overall parameter deviation value between teams Indicates the first The weighting coefficients of each parameter, Indicates the first The first team The values ​​of the parameters, In the j-th team, the first... The values ​​of the parameters, Indicates the total number of parameters involved in the comparison; The model parameter calibration rules are trained based on data from historical R&D projects. The model parameter calibration rules include parameter deviation tolerance threshold and structural change impact range determination logic. When the deviation between newly accessed data and existing model data exceeds the parameter deviation tolerance threshold, the model parameter adjustment process is automatically triggered, and the scope of related models that need to be updated synchronously is determined according to the structural change impact range determination logic.

4. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 3, characterized in that, The collaborative process for defining R&D tasks in the digital twin model specifically includes: Obtain dependency data between subtasks in the R&D task. The dependency data includes the identifier of the preceding task, the identifier of the following task, and the dependency type. The dependency type is classified into serial dependency, parallel dependency, conditional branch dependency, and loop iteration dependency. For each subtask, the input data requirements for that subtask are extracted from the corresponding participating team data. The input data requirements include the required data type, data precision requirements, and data timeliness requirements. At the same time, the output data format for that subtask is defined, which includes data field names, data field types, and data encoding rules. Based on the dependency type and execution constraints of each subtask, the subtasks are arranged into an ordered task execution sequence. The task execution sequence includes parallel task groups that can be executed simultaneously and condition judgment nodes that need to determine the direction based on the condition judgment results. The execution constraints include resource consumption limits, time window limits, and data dependency integrity requirements. Configure exception handling rule data for each task execution node. The exception handling rule data includes task termination triggering conditions, task continuation triggering conditions, parameter adjustment range, and execution speed reduction. When the real-time data during task execution meets any exception handling rule, intervene in task execution according to the corresponding rule. The execution priority of each parallel subtask in the parallel task group is determined by the following parallel task priority calculation formula: ； in, Indicates the first The combined priority score of each parallel subtask. Indicates the first The estimated completion time of each parallel subtask. Indicates the first The amount of resources required for each parallel subtask This represents the total amount of currently available resources. Indicates the first The urgency level of each parallel subtask. , , These are the weights for the time factor, resource factor, and urgency factor, respectively, and they satisfy the following conditions: .

5. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 4, characterized in that, The data fusion processing of each participating team in the digital twin model specifically includes: Multiple instances of the same research and development object are selected, and the configuration parameter sets of each instance in its respective participating team are obtained. The configuration parameter sets of each instance are compared field by field, and the fields with differences are recorded as difference fields. The difference values, difference directions, and difference magnitudes of the difference fields are used to form a parameter difference matrix. The parameter difference matrix is ​​constructed using the following formula: ； in, Represents the parameter difference matrix. Indicates the first The first instance and the first The difference value of each instance on the corresponding parameter field, Indicates the total number of instances. Indicates the total number of parameter fields; Obtain the verification activity records of each participating team, align the verification activity records of each team with the unified timeline of the R&D project, identify the verification activities performed by different teams on the same R&D object within the same time period, cross-compare the types, coverage and verification results of the verification activities, and generate a verification activity difference report. Coverage data from multiple instances is integrated, and set operations are performed on the coverage data of each instance. The IP block functional regions not covered by any instance are identified through set difference operations. These identified functional regions are recorded as blank areas, and a blank area description is generated, including the blank area's location, functional description, and impact assessment. The blank areas are determined using the following set difference formula: ； in, Represents the set of blank regions. Represents the complete set of functional areas of an IP block. Indicates the first The set of verification coverage areas for each instance. Indicates the total number of instances; Based on error type data and corresponding triggering scenario data recorded in historical R&D projects, an error type prediction model is trained. The trained error type prediction model is then used to analyze the current fused data, outputting the potential error types present in the current data, as well as the triggering scenario description for each error type. The triggering scenario description includes the combination of data conditions required to trigger the error type.

6. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 5, characterized in that, The identification of potential risks in the R&D process based on the results of the difference analysis specifically includes: Match the predicted error type with the tasks currently being performed by each participating team, identify the target teams and target task nodes affected by the error type, and record the matching results as risk impact scope data; For each error type and its corresponding triggering scenario, similar error type and triggering scenario combinations are retrieved from the historical collaboration solution library, and corresponding historical collaboration processing suggestions are extracted. The historical collaboration processing suggestions are then adaptively adjusted based on the actual constraints of the current R&D project to generate collaboration processing suggestions corresponding to the current error type. The collaboration processing suggestions include the parameter fields that need to be adjusted and the adjustment range, the types of verification activities that need to be added and the verification coverage, and the participating teams and coordination content that need to be coordinated. The generated collaborative processing suggestions are prioritized according to the magnitude of their impact and urgency in the risk impact data. This prioritization is performed using the following risk priority calculation formula: ； in, This indicates the overall priority ranking score of the risk items. The risk impact range score is indicated, with a value ranging from 1 to 10. This indicates the urgency of the risk, with a value ranging from 1 to 10. This represents a risk probability score, with a value ranging from 0 to 1. , , These are the influence range weight factor, urgency weight factor, and probability weight factor, respectively, and they satisfy the following conditions: ; The priority-ordered collaborative processing suggestions are integrated into a collaborative solution document, which includes the execution order of each collaborative processing suggestion, the responsible team for execution, and a description of the expected results. The collaborative solution document is distributed to all relevant participating teams. After receiving it, each participating team feeds back the execution results to the digital twin model. The execution results include the actual adjusted parameter values, the actual additional verification activity records, and the actual coordination execution status.

7. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 6, characterized in that, The multi-dimensional simulation and deduction of the R&D plan based on the digital twin model specifically includes: Load the R&D scheme parameters to be evaluated in the digital twin environment. The R&D scheme parameters include design parameter combinations, process parameter combinations, resource allocation parameters and verification strategy parameters. Map each parameter to the adjustable attribute fields of the corresponding virtual entity in the digital twin model. For each R&D scheme parameter combination, simulation is performed sequentially in the digital twin environment. During the simulation, real-time status data of each key node is collected. The key nodes include the functional verification nodes of each IP block, the interface verification nodes of each module, and the integration verification nodes of each subsystem. The performance index data of each key node after the simulation is completed is obtained. The performance index data includes functional coverage, timing margin, power consumption value, and signal integrity index. The performance index data of each key node for each R&D scheme parameter combination are summarized, and the offset of each key node relative to the reference process parameters is calculated. The offset is calculated using the following formula: ； in, Indicates the first Percentage of performance offset for each critical node Indicates the first Performance metrics of key nodes under the current simulation scheme Indicates the first under the reference process parameters Baseline performance metrics for each key node; By combining the weight coefficients of each key node in the overall R&D process, a collaborative efficiency evaluation function is constructed to quantitatively score the parameter combinations of each R&D scheme, generating a simulation result table containing the scores of each scheme and the performance indicators of each key node. The collaborative efficiency evaluation function is calculated using the following formula: ； in, This indicates the collaborative effectiveness evaluation score of the R&D plan. Indicates the first The weight coefficient of each key node in the overall R&D process. Indicates the first Performance offset of each critical node This indicates the maximum allowed performance offset threshold. Indicates the total number of critical nodes; The parameter combinations of each R&D scheme in the simulation results table are sorted in descending order according to the score of the collaborative effectiveness evaluation function. A predetermined number of schemes with the highest scores are selected as candidate optimal schemes, and the performance characteristics of each candidate optimal scheme in each target dimension are marked.

8. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 7, characterized in that, The comparison of simulation results with actual R&D execution data specifically includes: Record the execution time, responsible person, data content on which the decision was based, specific actions taken to execute the decision, and actual effects after the decision is executed for each R&D decision. Compile the above information into decision trace data and store it persistently. The prediction results of the digital twin model for each decision node are compared item by item with the actual effects in the decision trace data. The deviation value between the prediction result and the actual result of each decision node is calculated using the following formula: ； in, This represents the root mean square deviation of the decision node. Indicates the digital twin model for the first The predicted value of each decision indicator, Indicating the first in the decision trace data The actual value of each decision indicator Indicates the total number of decision-making indicators; Decision nodes whose deviation values ​​exceed a preset threshold are marked as high deviation nodes; For each high-deviation node, extract the data content and specific actions that the node relied on when making the decision from the decision trace data, compare them with the input parameters and simulation logic of the node in the digital twin model, identify the data factors or model factors that cause the deviation, and record the identification results as root cause analysis data. Based on the root cause analysis data, the parameters or evaluation logic of the corresponding nodes in the digital twin model are modified. The modification includes adjusting the values ​​of the model parameters, modifying the judgment conditions of the evaluation logic, and supplementing the missing influencing factors.

9. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 8, characterized in that, The calibration and updating of the parameters and evaluation rules of the digital twin model based on the comparative analysis results specifically includes: Records marked as successful in decision-making trace data are filtered out, and the data features and execution results corresponding to successful decisions are extracted. These are then added to the model training set as positive sample data, and the corresponding decision rules in the model are reinforced using the positive sample data. The records marked as decision failures in the decision trace data are filtered, the data features and execution results corresponding to the failed decisions are extracted, the reasons for failure are analyzed and the model parameters or rules that need to be corrected are determined, and the corresponding parameters or rules in the model are corrected and trained using the failure sample data. According to a preset time period, the digital twin model is fully refreshed using the latest collected data from the entire R&D project process. The intermediate state data of the model generated during the full data refresh process is compared with the model state data before the refresh to verify the effectiveness of the model update. The effectiveness of the model update is verified by the following model performance change ratio formula: ； in, This indicates the rate of change in model performance. This represents the collaborative effectiveness evaluation score after the model update. This represents the collaborative effectiveness evaluation score before the model update. The model update is considered valid when the value is greater than zero. If the value is less than or equal to zero, the model update is deemed invalid and the system is rolled back to the state before the update. Establish version management data for digital twin models, recording the update content, update time, performance comparison data of the model before and after the update, and the person responsible for the update operation for each model update.

10. The collaborative technical service system for the entire R&D project process based on digital twins as described in claim 1, characterized in that, The process of aligning timestamps and unifying units for the collected multi-source data also includes: For fields with missing values ​​in multi-source data, missing values ​​are filled using interpolation of nearby time points or the mean of similar objects, based on the historical data distribution characteristics of the field. The filled data is then marked as processed data. The nearby time point interpolation is performed using the following linear interpolation formula: ； in, Indicates a point in time Interpolation padding at the location, Indicates a point in time Known data values ​​at that location Indicates a point in time Known data values ​​at that location ; For fields with outliers in multi-source data, upper and lower thresholds are set based on the data fluctuation range of the field. Data points exceeding the upper and lower thresholds are marked as suspicious data. Suspicious data undergoes secondary verification. If the secondary verification still determines that the data is abnormal, it is removed and recorded as outlier removed data. The upper and lower thresholds are determined by the following formula: ； ； in, This represents the upper threshold for outlier detection. This represents the lower threshold for outlier detection. This represents the mean of the data in this field. This represents the standard deviation of the data in this field. This represents the anomaly determination coefficient, with a value ranging from 2 to 3; Processed data and anomaly-removed data are stored separately. A unified dimensional conversion is performed on the processed data to convert similar physical quantities from different sources into a unified unit of measurement. The converted data undergoes a consistency check to ensure that the numerical differences of the same physical quantity across different data sources are within acceptable limits. This consistency check is performed using the following consistency deviation formula: ； in, This indicates the percentage deviation in consistency between two data sources for the same physical quantity. This represents the value of the physical quantity in data source A. Indicates data source The value of this physical quantity, when Data is considered consistent when the value is less than the preset consistency tolerance threshold. When the data is greater than or equal to the preset consistency tolerance threshold, the data is determined to be inconsistent and the data tracing process is triggered.

Images