A product development whole-process-oriented digital management method and system

Abstract

The application discloses a kind of product-oriented digital management method and system for whole process of research and development, it is related to product digital management technical field, the digital management method includes: extracting technical content data and adding metadata label to build correlation graph atlas;Degree of completion is calculated layer by layer, and weight is dynamically adjusted;Based on process characteristics and degree of completion, the reliability is dynamically predicted by prediction model;When the predicted reliability is lower than the threshold value, trigger early warning and locate risk source, adjust weight and qualified threshold value in reverse;Output degree of completion evaluation and reliability prediction result.The weight and threshold value of the application are dynamically adjusted by the reliability prediction result to drive the degree of completion evaluation in reverse, so that the evaluation system can automatically focus on high-risk links, forming a self-adaptive closed-loop optimization mechanism of "the greater the risk, the more stringent the evaluation", effectively reducing the research and development decision cost, and improving the reliability of product research and development.

Description

Technical Field

[0001] This invention relates to the field of product digital management technology, specifically to a digital management method and system for the entire product development process. Background Technology

[0002] Currently, in Product Lifecycle Management (PLM) systems, the assessment of R&D completion is often limited to simple statistics of task completion percentages, lacking quantitative assessment algorithms based on multi-dimensional data throughout the entire process. This results in highly subjective and coarse-grained assessments. Simultaneously, reliability assessments are typically conducted independently outside the system, lacking dynamic correlation with R&D process data and failing to effectively utilize massive amounts of design, simulation, and testing data for ongoing prediction. Furthermore, existing completion management and reliability management are isolated, lacking inherent logical connections and bidirectional communication mechanisms. The analysis and utilization of R&D process data primarily remains at the level of querying, retrieval, and report statistics, lacking dedicated analytical algorithms for assessment and prediction. This results in R&D decisions lacking accurate data support, hindering the achievement of closed-loop optimization of the R&D process. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a digital management method and system for the entire product development process. By setting up a model for prediction and evaluation, and dynamically adjusting the evaluation weights based on the data from the development process, an adaptive closed-loop optimization mechanism is formed, which effectively reduces the cost of development decisions and improves the reliability of product development.

[0004] This invention provides a digital management method for the entire product development process, the management method comprising: S11: Extract technical content data from each stage of the product development process in the product lifecycle management system, add process evaluation metadata tags to the technical content data, and construct a development process association graph based on the logical relationship between the technical content data. S12: Based on the technical content data, calculate the completion degree of each technical content node, the completion degree of each process stage, and the overall completion degree of product development layer by layer. Node weights are assigned according to the criticality of each technical content node in the corresponding process stage, and stage weights are assigned according to the importance of each process stage in the overall development. The node weights and stage weights are dynamically adjusted according to the dynamic changes in the development process and the feedback of subsequent reliability prediction results. S13: Based on the process characteristic data in the technical content data, as well as the completion degree of each process stage and the overall completion degree of product development, the prediction reliability of the product under the current development progress is dynamically predicted through the prediction model. S14: Compare the predicted reliability with a preset reliability threshold. When the predicted reliability is lower than the preset reliability threshold, trigger a risk warning and locate the source of risk. Adjust the node weight and the stage weight in reverse according to the risk warning and the source of risk, and increase the completion qualification threshold of the technical content node in the corresponding process stage. S15: Output the product development completion assessment results and reliability prediction results through the visualization interface of the product lifecycle management system.

[0005] Furthermore, the step of calculating the completion rate of each technical content node, the completion rate of each process stage, and the overall product development completion rate based on the technical content data includes: For each technical content node, the node completion rate is calculated by weighted summation based on the progress completion index, quality achievement index, and resource consumption index of that technical content node. For each process stage, the completion degree of all technical content nodes belonging to that process stage is weighted and summed according to the node weights to calculate the stage completion degree of that process stage. The overall product development completion rate is calculated by weighting and summing the completion rates of all stages in the entire product development process according to the stage weights.

[0006] Furthermore, for each technical content node, the node completion degree is calculated by weighted summation based on the progress completion index, quality achievement index, and resource consumption index of that technical content node, including: The progress completion index of the technical content node is calculated based on the ratio of the actual progress to the planned progress of the task recorded in the product lifecycle management system and in combination with the task duration sensitivity coefficient. The quality achievement index is calculated by comprehensively considering the deliverable review pass rate, defect density, and change frequency recorded in the product lifecycle management system. The resource consumption index is calculated based on the deviation rate between the actual resource consumption and the budgeted resources recorded in the product lifecycle management system.

[0007] Furthermore, for each process stage, the completion degree of all technical content nodes belonging to that process stage is weighted and summed according to the node weights to calculate the stage completion degree of that process stage, including: The process assesses the criticality of metadata tag reading and the level of technical risk. The node weight is determined by comprehensively considering the criticality of the technical content node in its respective process stage, the proportion of the workload corresponding to the technical content node to the total workload of its respective process stage, and the technical risk level corresponding to the technical content node.

[0008] Furthermore, before performing dynamic reliability prediction, the method includes a step of extracting the input features of the prediction model: Completeness features are extracted from the completion of each process stage and the overall completion of product development, and the rate of change of completion is calculated based on timestamp data. The rate of change of completion is then incorporated into the completion features. The total number of product components, the number of newly added parts, and the proportion of common parts are extracted from the bill of materials data in the technical content data as design quality characteristics. The frequency of design changes, the scope of impact of changes, and the response time of changes are extracted from the change record data in the technical content data as change characteristics. The simulation verification pass rate and test problem closure rate are extracted from the simulation reports and test reports in the technical content data as verification features; Retrieve historical reliability data of similar products at the same completion level as the current product from the historical project data in the technical content data, as historical reference features; The extracted completion features, design quality features, change features, verification features, and historical reference features are combined into an input feature set and input into the prediction model.

[0009] Furthermore, the prediction model is a machine learning regression model trained in advance using complete historical project data from a historical project database, which includes historical R&D process data and historical product final reliability verification data.

[0010] Furthermore, the management method also includes: New technology content data generated during product development is added to the training set to incrementally train the prediction model and update the model parameters. After the product completes the final reliability verification test, the actual reliability verification data will be sent back to the historical project database as labeled data for periodic retraining of the prediction model.

[0011] Furthermore, the method of dynamically predicting the reliability of the product under the current R&D progress using a prediction model based on the process characteristic data in the technical content data, the completion degree of each process stage, and the overall completion degree of product development includes: When the predicted reliability is lower than a preset reliability threshold or the predicted reliability shows a continuous downward trend, a risk warning is triggered, and the contribution of each input feature to the predicted reliability is calculated using an interpretability analysis method. Based on the contribution of each input feature, the process stage and technical content node that contributes the most to reliability risk are located. Risk warning tasks are automatically generated in the product lifecycle management system and pushed to the corresponding responsible persons.

[0012] Furthermore, the step of comparing the predicted reliability with a preset reliability threshold, triggering a risk warning and locating the risk source when the predicted reliability is lower than the preset reliability threshold, and adjusting the node weight and stage weight in reverse according to the risk warning and the risk source, and increasing the completion qualification threshold of the technical content node within the corresponding process stage includes: The predicted reliability is compared with a preset reliability threshold, and when the predicted reliability is lower than the preset reliability threshold, it is identified as a high-risk level. The stage weight of the process stage identified as high-risk level will be increased in the overall completion calculation, and the completion qualification threshold of the technical content nodes within the high-risk process stage will be increased. The adjusted weights and thresholds are applied to the next completed metric evaluation, forming a loop iteration that continues until product development is completed.

[0013] This invention also provides a digital management system for the entire product development process, the digital management system comprising: Data acquisition and storage module: Extracts technical content data from each stage of the product development process in the product lifecycle management system, adds process evaluation metadata tags to the technical content data, and constructs a development process association graph based on the logical relationship between the technical content data; Complete the quantification and evaluation module: Based on the technical content data, calculate the completion degree of each technical content node, the completion degree of each process stage, and the overall completion degree of product development layer by layer. Among them, the node weight is assigned according to the criticality of each technical content node in the corresponding process stage, the stage weight is assigned according to the importance of each process stage in the overall development, and the node weight and the stage weight are dynamically adjusted according to the dynamic changes in the development process and the feedback of subsequent reliability prediction results. Reliability dynamic prediction module: Based on the process characteristic data in the technical content data, as well as the completion degree of each process stage and the overall completion degree of product development, the prediction model dynamically predicts the predicted reliability of the product under the current development progress. Collaborative scheduling module: compares the predicted reliability with a preset reliability threshold. When the predicted reliability is lower than the preset reliability threshold, a risk warning is triggered and the source of risk is located. Based on the risk warning and the source of risk, the node weight and the stage weight are adjusted in reverse, and the completion qualification threshold of the technical content node in the corresponding process stage is increased. Visualization module: Outputs the completion assessment results and reliability prediction results of product development through the visualization interface of the product lifecycle management system.

[0014] This application constructs a dual-algorithm collaborative architecture for quantitative assessment of R&D completion and dynamic prediction of process reliability. It establishes a two-way data transmission and iterative feedback mechanism between the two algorithms. The reliability prediction results drive the dynamic adjustment of the weight and threshold of the completion assessment, enabling the assessment system to automatically focus on high-risk links and form an adaptive closed-loop optimization mechanism of "the greater the risk, the stricter the assessment". This effectively reduces the cost of R&D decision-making and provides a valuable time window for design optimization. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a digital management method for the entire product development process in an embodiment of the present invention. Figure 2 This is a schematic diagram of the technical node completion calculation and analysis process for the entire product development process in this embodiment of the invention; Figure 3 This is a schematic diagram of a digital management system for the entire product development process in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: Please refer to Figure 1This invention provides a digital management method for the entire product development process, which is executed in a product lifecycle management system. The digital management method includes: Step S11: Extract technical content data from each stage of the product development process in the product lifecycle management system, add process evaluation metadata tags to the technical content data, and construct a development process association graph based on the logical relationship between the technical content data.

[0019] Specifically, the technical content data encompasses multi-source heterogeneous data across the entire lifecycle, from requirements analysis, solution design, detailed design, simulation verification, prototype manufacturing to testing and mass production implementation. This includes, but is not limited to, design documents, bills of materials, simulation reports, test records, process parameters, and change logs. Since raw data is often scattered and stored in various formats, making direct use difficult, this embodiment uses process evaluation metadata tags for structured processing. These tags can include dimensions such as task type, stage affiliation, technical difficulty level, criticality coefficient, quality review score, and problem closure status, thereby transforming unstructured business data into computational elements that can be recognized by the algorithm.

[0020] Building upon this foundation, we further explored the inherent logic between data to construct a correlation graph of the R&D process. This graph not only records the static reference relationships between data but also establishes causal chains such as "design scheme - simulation verification - test results" and impact chains such as "design change - affected materials - related processes." This approach breaks down the data silos between functional modules in traditional systems, providing a complete and traceable data foundation for subsequent quantitative assessment and risk tracing, ensuring full coverage and relevance of the assessment objects.

[0021] Step S12: Based on the technical content data, calculate the completion degree of each technical content node, the completion degree of each process stage, and the overall completion degree of product development layer by layer. The node weight is assigned according to the criticality of each technical content node in the corresponding process stage, and the stage weight is assigned according to the importance of each process stage in the overall development. The node weight and stage weight are dynamically adjusted according to the dynamic changes in the development process and the feedback of subsequent reliability prediction results.

[0022] In this process, completion calculation is not a simple percentage-based statistic, but rather employs a bottom-up, multi-layered aggregation logic. First, a basic assessment is conducted on the most granular technical content nodes. Then, these nodes are aggregated according to their weights to determine the completion rate of each process stage. Finally, the overall product development completion rate is obtained by aggregating the weights of each stage. The initial weight allocation can be determined according to preset rules, but more importantly, it possesses dynamic adjustment capabilities. This dynamic adjustment mechanism has two driving forces: first, the natural evolution of the development process itself; for example, as the project progresses from the design to the verification stage, the weight given to quality considerations is usually higher than that of schedule. Second, reliability prediction feedback from subsequent steps S13 and S14; when the system identifies reliability risks in specific stages, it automatically intervenes in the assessment system, tilting resources and attention towards high-risk areas. In this way, completion assessment is no longer a static "accounting" process, but a "live" indicator that responds to the project's health status, effectively avoiding the phenomenon of "inflated progress but numerous hidden dangers" in traditional management.

[0023] Step S13: Based on the process characteristic data in the technical content data, as well as the completion rate of each process stage and the overall completion rate of product development, dynamically predict the reliability of the product under the current development progress through the prediction model.

[0024] This step transforms reliability assessment from "post-event verification" to "process-as-you-go." The predictive model's input not only includes the completed quantification results and their trends from step S12, but also deeply integrates process characteristics inherent in the technical content data, such as design complexity, change stability, and verification adequacy. These characteristics collectively constitute a multi-dimensional vector space reflecting the product's current quality maturity. The predictive model, through reasoning from this real-time data, outputs the predicted reliability at the current point in time, which can be expressed as mean time between failures, failure rate, or success probability. Because the input data is updated in real-time with the R&D process, the prediction results also evolve dynamically, enabling the R&D team to obtain quantitative expectations about the final product quality early in the development process, thus gaining a valuable time window for design optimization.

[0025] Step S14: Compare the predicted reliability with the preset reliability threshold. When the predicted reliability is lower than the preset reliability threshold, trigger a risk warning and locate the source of risk. Adjust the node weight and stage weight in reverse according to the risk warning and the source of risk, and increase the completion qualification threshold of the technical content node in the corresponding process stage.

[0026] This step forms the core feedback loop of the dual-algorithm collaborative closed loop. When the prediction results show that the potential risk exceeds the acceptable range, the system not only issues an alarm, but more importantly, automatically executes corrective actions. On the one hand, it uses attribution analysis to locate the specific process stage or technical node that caused the risk; on the other hand, it injects this risk information back into the evaluation system of step S12. Specifically, this manifests as increasing the weight of high-risk stages, increasing their influence in the overall evaluation, and raising the qualification threshold for nodes within that stage, forcing the team to complete the relevant work to a higher standard in order to be considered "qualified".

[0027] Step S15: Output the product development completion assessment results and reliability prediction results through the visual interface of the product lifecycle management system.

[0028] Finally, the evaluation and prediction data generated through the aforementioned closed-loop iteration are presented on the user interface in the form of intuitive charts, dashboards, or trend curves. This includes not only current completion and reliability levels but also risk warnings, weight adjustment records, and historical trend comparisons—all supporting decision-making information. By transforming abstract algorithmic results into visualized management insights, managers can gain a real-time overview of R&D, quickly identify bottlenecks and risks, and thus make more scientific and objective R&D decisions.

[0029] Specifically, in this embodiment, based on technical content data, the completion degree of each technical content node, the completion degree of each process stage, and the overall product development completion degree are calculated layer by layer. This process constructs a three-level quantitative mapping system from micro-tasks to macro-projects, aiming to solve the problem of coarse granularity in traditional evaluations. Specifically, this calculation process is not a simple linear accumulation, but is implemented through a multi-layer weighted aggregation model.

[0030] For details, please refer to Figure 2 Step S12 includes: S121: For each technical content node, the node completion degree is calculated by weighted summation based on the node's progress completion indicators, quality achievement indicators, and resource consumption indicators. In specific implementation, the node completion degree... The calculation formula is as follows: ; in, The schedule completion metric, in its physical sense, represents the fulfillment of tasks within the time dimension. This metric is not solely determined by the ratio of actual progress to planned progress; it is also adjusted using a task duration sensitivity coefficient. The duration sensitivity coefficient reflects the degree to which a task delay affects the overall project's critical path, typically ranging from 0.8 to 1.2. For example, for a core technology development task on the critical path, its sensitivity coefficient could be set to 1.2, meaning that even a small delay will significantly impact the project's critical path. The sensitivity coefficient is amplified in the middle, thus attracting the system's attention; while for non-critical auxiliary document organization tasks, the sensitivity coefficient can be set to 0.8 to smooth out the fluctuations in its impact on the overall evaluation.

[0031] This quality achievement metric characterizes the technological maturity of deliverables. It integrates three dimensions: deliverable review pass rate, defect density, and change frequency. The review pass rate reflects the efficiency of passing on the first attempt; defect density (such as defects per thousand lines of code or errors per page of documentation) directly quantifies the intrinsic quality of the technical content; and change frequency indirectly reflects the stability of the solution.

[0032] This is a resource consumption indicator, calculated based on the deviation rate between actual resource consumption and budgeted resources. The smaller the deviation rate, the higher the resource consumption. A higher value indicates a greater degree of control over the research and development process.

[0033] , , These are the weighting coefficients for the schedule, quality, and resource dimensions, respectively, and they satisfy... This multi-dimensional node evaluation method transforms abstract R&D activities into computable data objects, avoiding the evaluation distortion caused by a single progress perspective.

[0034] S122: For each process stage, the completion rates of all technical content nodes belonging to that stage are weighted and summed according to their node weights to calculate the stage completion rate. In this hierarchy, the node weights... Determining the weighting of nodes is crucial, as it determines the differences in the contribution of different tasks to the stage objectives. Specifically, during process evaluation, the system reads the criticality markers and technical risk levels from the metadata tags. The node weights are then determined comprehensively based on the criticality markers of the technical content nodes within their respective process stages, the proportion of the workload corresponding to each technical content node to the total workload of that process stage, and the technical risk level of the corresponding technical content node. In the detailed design stage, a module design task involving core security functions, although its workload may only account for 20%, will automatically be assigned a high node weight (e.g., 0.5) due to its "high" criticality marker and "high" technical risk level, making it the dominant factor determining the completion rate of that stage. Conversely, the concurrent general support design task, despite accounting for 30% of the workload, may have its weight compressed to 0.1 due to its lower criticality and risk level. This metadata tag-based weighting mechanism ensures that the evaluation results accurately reflect the progress in overcoming key R&D priorities and technical challenges.

[0035] S123: The overall product development completion rate is calculated by weighting and summing the completion rates of all stages in the entire product development process according to their respective stage weights.

[0036] Stage weights This reflects the strategic importance of each stage in the entire product lifecycle. For example, in the development of complex equipment, the design and verification / testing stages... This phase is typically higher than the standard engineering release phase because the former defines the product's DNA, while the latter confirms the product's bottom line.

[0037] Specifically, in this embodiment, the node weights and stage weights are dynamically adjusted based on the dynamic changes in the R&D process and the feedback from subsequent reliability prediction results. This dynamic adjustment mechanism includes three specific triggering scenarios, reflecting the adaptive capability of the technical solution to the objective laws of R&D. The first is a stage-based progressive adjustment, where the weight coefficients of the quality dimension increase as the R&D process progresses from the early to the later stages. Gradually increase the weighting coefficient of the progress dimension. The corresponding reduction. For example, in the requirements analysis phase, The initial value can be 0.4. The value is 0.35; when entering the prototype trial production stage, Automatically decays to 0.25. The weight has been increased to 0.5. This is because any quality issues in the later stages of R&D can lead to huge rework costs. The system automatically strengthens quality constraints through algorithms, which aligns with engineering practice. The second type is risk-driven adjustment. When the subsequent reliability prediction model outputs a risk warning pointing to a specific process stage, the stage weight in the overall completion calculation is adjusted. Automatically increase.

[0038] For example, if the predictive model detects a high risk in the "simulation verification phase," the system will [impose restrictions on that phase]. The weight was temporarily increased from 0.15 to 0.25, forcing management resources and evaluation focus to concentrate on this stage. The third type is change-triggered adjustment, where the node weight of the affected technical content nodes is adjusted when design changes occur during R&D. Automatic recalculation. Changes often mean increased uncertainty. The system ensures that the revised work receives more rigorous acceptance by increasing the weight threshold of relevant nodes. Through these three dynamic adjustment mechanisms, completion assessment is no longer a rigid administrative statistical tool, but has evolved into an intelligent data processing system that can respond to the project's health status in real time and automatically balance the relationship between progress and quality, effectively supporting the refined management and control of the R&D process.

[0039] Specifically, in this embodiment, before performing dynamic reliability prediction, a step of extracting the input features of the prediction model is included. This step serves as a bridge connecting R&D process data and reliability prediction results, and its core lies in transforming unstructured R&D behavior into a quantitative vector that can characterize the potential quality risks of the product. Specifically, this feature extraction process is not a simple data transfer, but rather a construction of a multi-dimensional feature space containing five dimensions based on a deep understanding of the mechanisms affecting R&D reliability.

[0040] First, completion features are extracted from the completion levels of each process stage and the overall product development completion level. The rate of change of completion level is then calculated based on timestamp data and incorporated into the completion feature. If the completion level of a certain stage fluctuates drastically within a short period or suddenly increases after a long period of stagnation, it often indicates a loss of control over the development pace or the existence of rushed work. This behavior pattern is significantly positively correlated with the later reliability failure rate. Therefore, the system reads the timestamp sequence of task status changes and uses a sliding window algorithm to calculate the increment of completion level per unit time. This increment is then used as a key temporal feature input to the model, characterizing the stability of the development pace, thereby capturing process risks that static values ​​cannot reflect.

[0041] Secondly, the total number of product components, the number of newly added parts, and the percentage of common parts are extracted from the bill of materials data in the technical content data as design quality characteristics. This set of characteristics directly maps the inherent complexity and maturity of the design from the product structure dimension. Among them, the total number of components reflects the integration scale of the system; the larger the scale, the more interaction interfaces and the more complex the failure paths. The number of newly added parts represents the proportion of novel elements in the design that have not been fully verified; the more new parts, the higher the risk of introducing unknown failure modes. The percentage of common parts represents the degree of design reuse; a high proportion of common parts means that a mature solution with historical verification has been adopted, which helps to improve the overall reliability baseline. Through the combination of these three indicators, the predictive model can make a quantitative judgment on the "inherent constitution" of the product in the early stage of design, without having to wait until the physical testing stage to discover problems.

[0042] Secondly, design change frequency, scope of impact, and response time are extracted from the change records in the technical content data as change characteristics. Design changes are a direct manifestation of uncertainty in the R&D process.

[0043] Furthermore, excessively high change frequency indicates that the technical solution has not yet converged and is in a state of repeated trial and error; a wide range of changes (such as involving multiple subsystems or cross-level BOM modifications) means that the chain reaction triggered by the changes is difficult to fully control and is prone to introducing derivative defects; excessively long change response time may indicate that the team has insufficient understanding of the problem or difficulty in resource coordination, resulting in a prolonged risk exposure window. This embodiment analyzes the log data in the change management module to transform these qualitative descriptions into quantitative indicators, enabling the model to keenly perceive the degree of "shock" in the R&D process and thus predict its negative impact on the final reliability.

[0044] Furthermore, simulation verification pass rate and test problem closure rate are extracted from simulation and test reports in the technical content data as verification features. These two indicators evaluate the effectiveness of defect elimination from both theoretical and physical verification perspectives. The simulation verification pass rate reflects the initial correctness of the design scheme in the virtual environment; a low pass rate indicates insufficient design margin or large deviations in modeling assumptions. The test problem closure rate measures the thoroughness of the resolution of actual problems; unclosed problems are often the source of potential problems left to the client. In particular, the system defines "problem closure" as not only fixing the fault phenomenon but also passing regression testing and updating the relevant knowledge base, rather than merely implementing temporary workarounds. This strict feature definition ensures the authenticity of the input data and avoids false high-reliability predictions.

[0045] Finally, historical reliability data of similar products at the same completion level as the current project are retrieved from the historical project data in the technical content data, serving as historical reference features. Considering that the process data of a single project may have sparsity or noise, introducing historical reference features can provide a benchmark anchor for current predictions. The system calculates similarity based on metadata tags such as product type, technical indicators, and application scenarios, filters out historical projects with high similarity, and extracts their reliability performance at the same completion node as the current project. For example, when the current project is in the detailed design stage and the completion level is 60%, the system will retrieve the defect density or review score of similar products in history at 60% completion. This feature construction method based on "homogeneous alignment" effectively solves the problem of prediction inaccuracy caused by insufficient data in the early stages of new projects, enhancing the model's generalization ability and robustness.

[0046] After extracting the five types of features mentioned above, the system combines the extracted completion features, design quality features, change features, verification features, and historical reference features into an input feature set, which is then input into the prediction model. This prediction model is a machine learning regression model pre-trained using complete historical project data from a historical project database. This complete historical project data includes historical R&D process data and historical product final reliability verification data. The "complete historical project data" emphasized here refers to a closed-loop dataset covering the entire process from project initiation to mass production, with a clear final reliability acceptance conclusion, as opposed to fragmented data containing only partial stage records or lacking final result feedback. Only by training on such complete samples with causal correspondence can the model accurately learn the nonlinear mapping relationship between process features and final reliability. In specific implementations, algorithms such as gradient boosting trees, random forests, or long short-term memory networks can be used to construct the regression model, with mean time between failures (MTBF), failure rate, or reliability confidence as the output target. In this way, this embodiment establishes a complete technical chain from data perception to intelligent reasoning, laying a solid algorithmic foundation for subsequent dynamic early warning and closed-loop feedback.

[0047] In this embodiment, the management method also includes a continuous learning mechanism for the model and risk interpretability attribution to ensure that the predictive model can adapt to the evolution of the R&D model and provide credible decision-making basis. Specifically, new technology content data generated during product R&D is added to the training set to incrementally train the predictive model and update its parameters. This mechanism mainly targets dynamic changes during the R&D process, addressing the performance degradation problem that occurs over time. In actual operation, the system monitors the data flow in the product lifecycle management system in real time. When new technical content data (such as newly added design change records, the latest simulation test reports, or phase review results) accumulates to a preset threshold, or according to a preset time cycle (e.g., weekly or at the end of each milestone node), the incremental training task is automatically triggered. Incremental training adopts an online learning or small-batch fine-tuning strategy, only using recently generated new samples to locally update the model parameters, rather than retraining the entire model. This approach retains the general patterns learned by the model from historical data while quickly capturing R&D model changes or newly introduced technical risk factors specific to the current project, ensuring the timeliness and sensitivity of the prediction results.

[0048] Furthermore, after the product completes its final reliability verification test, the actual reliability verification data is fed back as labeled data to the historical project database for periodic retraining of the predictive model. Unlike incremental training, periodic retraining aims to calibrate the model's long-term baseline. Because incremental training may carry the risk of error accumulation or overfitting to recent noise, it is essential to rely on physically validated "true value" data for periodic correction. In practice, only after a project has completed all R&D processes and obtained a clear final reliability acceptance conclusion (such as measured MTBF values, failure rate statistics, etc.) will the project's complete process data be marked as high-confidence samples and added to the historical project database. The system typically initiates full or large-scale retraining tasks during off-peak business periods (such as quarterly or semi-annual periods), using the updated complete historical dataset to re-optimize the model structure and hyperparameters.

[0049] At the model output level, to overcome the "black box" nature of artificial intelligence algorithms and enhance user trust and execution of early warning information, this embodiment introduces a risk interpretability attribution mechanism. Based on process characteristic data in the technical content data, as well as the completion rate of each process stage and the overall product development completion rate, the predictive model dynamically predicts the product's reliability at the current development progress. On this basis, a risk warning is triggered when the predicted reliability falls below a preset reliability threshold or when the predicted reliability shows a continuous downward trend. A dual triggering condition design is adopted: an absolute threshold trigger is used to intercept explicit high-risk states, preventing products from passing inspections with defects; while a downward trend trigger is used to identify hidden signs of deterioration. Even if the current reliability is above the acceptable level, if it shows a monotonically decreasing trend for multiple consecutive evaluation cycles, it also indicates a systemic deviation in the development process, requiring early intervention. This combined triggering strategy significantly reduces the false negative rate and improves the comprehensiveness of risk perception.

[0050] After an alert is triggered, algorithms such as SHAP (Shapley Additive exPlanations), LIME (Local Interpretable Model-agnostic Explanations), or tree-based feature importance ranking can be used to quantify the marginal contribution of each input feature (such as completion rate change, number of new parts, change frequency, etc.) to the current prediction result. For example, if SHAP analysis shows that the feature "change frequency in the detailed design phase" contributes -0.15 to the reliability score, and its absolute value ranks first among all features, it indicates that this feature is the main driving factor causing the current risk. Subsequently, based on the contribution ranking of each input feature, the process stage and technical content node that contributes the most to the reliability risk is located. Since the input features have established a mapping relationship with the process evaluation metadata tags during construction, the system can accurately trace back the abstract feature contribution to specific business entities. For example, the "change frequency" feature can be associated with the "electrical control module design" node under the "detailed design phase," thereby clearly pointing out the source of risk, rather than vaguely indicating "poor design quality."

[0051] Finally, risk warning tasks are automatically generated in the product lifecycle management system and pushed to the corresponding responsible persons. This step achieves a closed loop from data analysis to management action. The system backend calls the task management interface of the product lifecycle management system to automatically create warning task sheets containing risk details, attribution analysis, and suggested measures based on risk attribution results. Based on the responsibility attribution information in the metadata tags, the tasks are accurately pushed to the to-do lists of relevant engineers or project managers. For example, when it is found that the problem closure rate of a certain test node is too low, leading to increased risk, the system will automatically send a special rectification task to the person in charge of that test to "improve the problem closure rate," along with the current closure rate data and target requirements. This automated task distribution mechanism not only reduces the delay and distortion of information conveyed manually, but more importantly, it directly embeds algorithmic insights into the daily R&D workflow, transforming risk management from passively viewing reports to proactively executing tasks, effectively improving the implementation efficiency and user stickiness of digital management methods.

[0052] In this embodiment, the predicted reliability is compared with a preset reliability threshold. When the predicted reliability is lower than the preset reliability threshold, it is identified as a high-risk level. The identification of the high-risk level is not only for alarm purposes, but also for initiating the adaptive reconstruction of the evaluation system. When the system monitors the predicted reliability value output by the reliability dynamic prediction module in real time, once the value falls below the preset safety baseline (i.e., the preset reliability threshold), the system determines that the current R&D status is at a high-risk level.

[0053] Furthermore, the preset reliability threshold is not a fixed value; it can be configured differently based on product type, R&D stage, or historical benchmark data. For example, the threshold in the prototype trial production stage is usually higher than that in the scheme design stage, in order to reflect stricter quality requirements for the physical verification stage.

[0054] Furthermore, the weight of process stages identified as high-risk in the overall completion calculation is increased, and the completion threshold for technical content nodes within high-risk process stages is raised. This is the core execution action for achieving "risk-driven assessment" in this application.

[0055] Specifically, when a process stage is marked as high-risk, the system automatically modifies the parameter configuration of that stage in the overall evaluation model. On the one hand, increasing the stage weight means that the performance of that stage has amplified its impact on the overall completion rate, forcing managers to shift their attention from the overall average to specific weak links. On the other hand, raising the completion threshold directly increases the difficulty of "meeting the standard." For example, in the R&D project of a new energy vehicle electronic control system, if the prediction model shows that the prediction reliability of the "detailed design stage" is below the threshold and is identified as high-risk, the system can automatically increase the stage weight of that stage in the overall completion rate calculation from the initial 0.2 to 0.3, and at the same time raise the completion score of all technical content nodes within that stage from the default 80 points to 90 points. The physical meaning of this parameter adjustment is that, under high-risk conditions, the original 80 points may only represent "formal completion" and is insufficient to support subsequent reliability requirements. Therefore, the system uses an algorithm to force a higher delivery quality to be recognized as effective progress.

[0056] More importantly, the adjusted weights and thresholds are applied to the next completion metric assessment, forming a cyclical iteration that continues until product development is complete. This emphasizes that the technical solution of this application is a dynamic convergence process, rather than a single error correction operation. After completing a parameter adjustment, the system does not stop working but immediately re-executes the completion metric assessment using the updated weights and thresholds to obtain a corrected completion result. Subsequently, this corrected result, along with the latest process characteristic data, is input into the prediction model again to generate a new round of prediction reliability. If the new round of prediction results shows that the risk has been reduced to below the threshold, the system can gradually adjust the weights and thresholds back to normal levels in the next iteration; if the risk still exists or further deteriorates, the system will continue to maintain or further strengthen the high-pressure assessment strategy. This cyclical iterative mechanism runs through the entire R&D cycle until the product is finally completed and launched.

[0057] Through the aforementioned dual-algorithm collaborative closed-loop feedback mechanism, this application effectively solves the technical problem of "inflated completion rate but insufficient reliability" commonly found in existing technologies. In traditional management models, completion assessment and reliability assessment are often parallel processes. Teams may lower quality standards under schedule pressure to achieve a superficially high completion rate, leading to frequent quality issues later in the project. In this embodiment, the reliability prediction results directly and inversely reshape the evaluation scale of completion, establishing an adaptive technical logic of "the greater the risk, the stricter the assessment." When the system detects potential risks, it automatically tightens the acceptance criteria for that stage, ensuring that any attempt to sacrifice quality for schedule will not receive high marks in the system. This mechanism forces R&D resources and management efforts to proactively focus on high-risk stages, ensuring the authenticity of R&D progress and the robustness of product quality, achieving a shift from passive post-event remediation to proactive process immunity. Furthermore, as an optional implementation, in addition to triggering based on absolute thresholds, this closed-loop mechanism can also incorporate a downward trend trigger condition. That is, when the predicted reliability, although not falling below the threshold, shows a continuous downward trend, preventative adjustments to the weights and thresholds are initiated in advance, further enhancing the system's risk defense capabilities.

[0058] Example 2: like Figure 3 As shown, this embodiment provides a digital management system for the entire product development process. This system, as a software device, is embedded or integrated into a product lifecycle management system to execute the digital management methods described in the preceding embodiments. The digital management system includes a data acquisition and storage module 10, a metric evaluation module 20, a reliability dynamic prediction module 30, a collaborative scheduling module 40, and a visualization module 50.

[0059] The data acquisition and storage module 10 is used to extract technical content data from each stage of the product development process within the product lifecycle management system. It adds process evaluation metadata tags to the technical content data and constructs a development process relationship graph based on the logical relationships between the technical content data. Specifically, this module calls the underlying data interface of the product lifecycle management system to periodically or in real-time capture multi-source heterogeneous data such as design documents, bills of materials, and change records. Before the data is stored, the module's built-in tagging engine automatically parses the data content and adds process evaluation metadata tags according to preset rules, for example, marking a simulation report as "Verification Stage - High Criticality - Reviewed". Simultaneously, this module utilizes graph database technology to construct a development process relationship graph, transforming discrete data objects into a network structure with causal and influence chains, providing standardized data query and traceability services for upper-layer algorithms.

[0060] The quantification and evaluation module 20 is used to calculate the completion degree of each technical content node, each process stage, and the overall product development completion degree based on technical content data. It assigns node weights according to the criticality of each technical content node in its corresponding process stage, and stage weights according to the importance of each process stage in the overall development. The node and stage weights are dynamically adjusted based on the dynamic changes in the development process and feedback from subsequent reliability prediction results. Specifically, this module, as the core computing engine of the system, encapsulates a three-layer weighted aggregation algorithm model. It reads basic data and metadata tags from the data acquisition and storage module 10 and performs hierarchical calculations at the node, stage, and overall levels. More importantly, this module exposes a parameter configuration interface, enabling it to receive externally input weight adjustment and threshold update commands, thus breaking the limitations of static evaluation. For example, when receiving feedback signals from the collaborative scheduling module 40, this module automatically reloads the weight configuration file for the corresponding stage and applies the new parameters in the next calculation cycle, achieving dynamic reconstruction of the evaluation logic.

[0061] The reliability dynamic prediction module 30 is used to dynamically predict the reliability of the product under the current R&D progress based on process feature data in the technical content data, as well as the completion status of each process stage and the overall product development completion status. Specifically, this module integrates a pre-trained machine learning regression model inference service. On the one hand, it subscribes to the output results of the completion quantification evaluation module 20; on the other hand, it directly extracts process features such as change frequency and BOM complexity from the data acquisition and storage module 10, processes them through feature engineering, and inputs them into the model for inference. This module supports online hot updates and version management of the model, ensuring that the predictive capability continuously evolves with the accumulation of R&D data. In terms of physical deployment, this module can run as an independent microservice, interacting decoupled from other modules through RESTful APIs or message queues to ensure response performance in high-concurrency scenarios.

[0062] The collaborative scheduling module 40 compares the predicted reliability with a preset reliability threshold. When the predicted reliability falls below the preset threshold, it triggers a risk warning and locates the source of the risk. Based on the risk warning and the source of the risk, it adjusts the node weights and stage weights in reverse, and increases the completion threshold of the technical content nodes within the corresponding process stage. Specifically, this module is the physical hub connecting the evaluation and prediction algorithm systems, undertaking the functions of bidirectional data transmission and control command generation. In terms of operation, the collaborative scheduling module 40 monitors the predicted reliability stream data output by the reliability dynamic prediction module 30 in real time. Once a risk condition is triggered, it not only generates a warning message, but more importantly, it automatically generates a control command package containing the target weight value and the new threshold. This command package is precisely sent to the parameter configuration interface of the quantification evaluation module 20 through the system's internal communication bus, forcibly driving the reshaping of the evaluation system. This hard connection or soft bus interaction mechanism between modules translates the abstract concept of "feedback adjustment" into concrete system-level data flow actions, ensuring the determinism and real-time performance of the dual-algorithm collaborative closed loop in engineering implementation, and avoiding the lag and errors caused by manual intervention.

[0063] The visualization module 50 is used to output the completion assessment results and reliability prediction results of product development through the visualization interface of the product lifecycle management system. Specifically, this module serves as the human-computer interaction front end of the system, subscribing to the status data of each backend module via WebSocket or polling mechanism, and rendering it as dynamic dashboards, trend curves, and risk heatmaps. For example, the interface can simultaneously display the current overall completion value and the predicted reliability confidence interval, and use color coding to intuitively identify the process stages marked as high-risk by the collaborative scheduling module 40. This module also supports drill-down interaction; users can click on a risk node to trace back to the underlying correlation graph and technical content data, thereby achieving a one-stop experience from macro-monitoring to micro-investigation.

[0064] Through the modular architecture described above, this application solidifies the methodological dual-algorithm collaborative logic into a concrete software system entity. The modules are tightly coupled through clearly defined data interfaces and control protocols. In particular, the physical feedback link established by the collaborative scheduling module 40 ensures that completion assessment and reliability prediction are no longer two isolated programs, but rather form an organic whole capable of automatically sensing risks and adjusting strategies. This device-level design not only provides a solid product protection safety net for the method claims but also guarantees the stability and reproducibility of the technical solution in practical engineering applications, effectively supporting improved transparency in R&D management and early risk identification.

[0065] Specifically, in this embodiment, during the early stages of requirements analysis and solution design at project initiation, the weighting coefficient α for the progress completion indicator is set to 0.40, the weighting coefficient β for the quality achievement indicator is set to 0.35, and the weighting coefficient γ for the resource consumption indicator is set to 0.25. This parameter combination reflects the objective reality that in the early stages of R&D, the focus on progress is slightly higher than on quality details. Simultaneously, based on the technical characteristics of each process stage of the electronic control system, detailed stage weights are preset. The weighting for each technical stage is as follows: scheme design stage (0.15), detailed design stage (0.20), simulation verification stage (0.25), prototype manufacturing stage (0.25), and testing verification stage (0.15). The default completion threshold for each technical node is set to 80 points. Based on this initial configuration, the system begins real-time data collection and evaluation throughout the entire R&D process.

[0066] When the project progressed to the middle of the detailed design phase, the reliability dynamic prediction module 30, based on feature extraction logic, detected an abnormal increase in the "design change frequency" feature value and an increase in the "number of new parts" exceeding the historical baseline by 20%, causing the model's current predicted reliability to drop rapidly from 0.92 to 0.78, below the preset reliability threshold of 0.85. At this point, the collaborative scheduling module 40 immediately triggered a risk warning and, through interpretability analysis, located the risk source as the "inverter control algorithm design" node and its associated "power module selection" node. Based on the dual-algorithm collaborative closed-loop feedback mechanism, the system automatically executed the following reverse adjustment actions: First, the stage weight of the detailed design phase identified as high-risk was adjusted in the overall completion calculation. The initial weight was increased from 0.20 to 0.30, making it the absolute focus of the current evaluation system; secondly, the completion threshold for all technical content nodes in this phase was raised from the default 80 points to 90 points, mandating higher delivery quality to be recognized as effective progress; finally, the adjusted weights and thresholds will be applied to the next completion quantification evaluation.

[0067] After adjusting the parameters mentioned above, the completion rate of the detailed design phase, recalculated by the system in the next evaluation cycle, was revised from 88% to 72%. Although the numerical completion rate showed a significant decrease, this truly reflected the actual health status of this stage after considering reliability risks, eliminating the quality hazards previously masked by simply pursuing progress. Based on the risk warning tasks automatically pushed by the system, the R&D team concentrated resources on a special review and optimization of the inverter control algorithm, and completed design convergence two weeks later. As the frequency of changes decreased and the verification pass rate increased, the predicted reliability gradually rebounded to above 0.88. The system then smoothly adjusted the stage weights and qualification thresholds back to normal levels in subsequent iterations, achieving adaptive recovery of the evaluation system.

[0068] On the visual interface, managers can not only view the real-time overall product development completion curve and predicted reliability confidence interval, but also clearly identify detailed design stages marked as high-risk through color coding. The risk attribution heatmap in the interface directly points to specific problem nodes and simultaneously displays the historical trajectory of weight adjustments and threshold change records. This visual interaction transforms abstract algorithmic logic into perceptible management insights, helping project managers quickly grasp the overall development landscape and bottlenecks, thereby effectively improving product development efficiency and reliability.

[0069] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0070] Furthermore, the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A digital management method for the entire product development process, characterized in that, The management method includes: S11: Extract technical content data from each stage of the product development process in the product lifecycle management system, add process evaluation metadata tags to the technical content data, and construct a development process association graph based on the logical relationship between the technical content data. S12: Based on the technical content data, calculate the completion degree of each technical content node, the completion degree of each process stage, and the overall completion degree of product development layer by layer. Node weights are assigned according to the criticality of each technical content node in the corresponding process stage, and stage weights are assigned according to the importance of each process stage in the overall development. The node weights and stage weights are dynamically adjusted according to the dynamic changes in the development process and the feedback of subsequent reliability prediction results. S13: Based on the process characteristic data in the technical content data, as well as the completion degree of each process stage and the overall completion degree of product development, the prediction reliability of the product under the current development progress is dynamically predicted through the prediction model. S14: Compare the predicted reliability with a preset reliability threshold. When the predicted reliability is lower than the preset reliability threshold, trigger a risk warning and locate the source of risk. Adjust the node weight and the stage weight in reverse according to the risk warning and the source of risk, and increase the completion qualification threshold of the technical content node in the corresponding process stage. S15: Output the product development completion assessment results and reliability prediction results through the visualization interface of the product lifecycle management system.

2. The digital management method according to claim 1, characterized in that, The calculation of the completion rate of each technical content node, the completion rate of each process stage, and the overall product development completion rate based on the technical content data includes: For each technical content node, the node completion rate is calculated by weighted summation based on the progress completion index, quality achievement index, and resource consumption index of that technical content node. For each process stage, the completion degree of all technical content nodes belonging to that process stage is weighted and summed according to the node weights to calculate the stage completion degree of that process stage. The overall product development completion rate is calculated by weighting and summing the completion rates of all stages in the entire product development process according to the stage weights.

3. The digital management method according to claim 2, characterized in that, For each technical content node, the node completion degree is calculated by weighted summation based on the node's progress completion indicators, quality achievement indicators, and resource consumption indicators. The progress completion index of the technical content node is calculated based on the ratio of the actual progress to the planned progress of the task recorded in the product lifecycle management system and in combination with the task duration sensitivity coefficient. The quality achievement index is calculated by comprehensively considering the deliverable review pass rate, defect density, and change frequency recorded in the product lifecycle management system. The resource consumption index is calculated based on the deviation rate between the actual resource consumption and the budgeted resources recorded in the product lifecycle management system.

4. The digital management method according to claim 2, characterized in that, For each process stage, the completion degree of all technical content nodes belonging to that process stage is weighted and summed according to the node weights to calculate the stage completion degree of that process stage, including: The process assesses the criticality of metadata tag reading and the level of technical risk. The node weight is determined by comprehensively considering the criticality of the technical content node in its respective process stage, the proportion of the workload corresponding to the technical content node to the total workload of its respective process stage, and the technical risk level corresponding to the technical content node.

5. The digital management method according to claim 1, characterized in that, Before performing dynamic reliability prediction, the method also includes a step of extracting the input features of the prediction model: Completeness features are extracted from the completion of each process stage and the overall completion of product development, and the rate of change of completion is calculated based on timestamp data. The rate of change of completion is then incorporated into the completion features. The total number of product components, the number of newly added parts, and the proportion of common parts are extracted from the bill of materials data in the technical content data as design quality characteristics. The frequency of design changes, the scope of impact of changes, and the response time of changes are extracted from the change record data in the technical content data as change characteristics. The simulation verification pass rate and test problem closure rate are extracted from the simulation reports and test reports in the technical content data as verification features; Retrieve historical reliability data of similar products at the same completion level as the current product from the historical project data in the technical content data, as historical reference features; The extracted completion features, design quality features, change features, verification features, and historical reference features are combined into an input feature set and input into the prediction model.

6. The digital management method according to claim 5, characterized in that, The prediction model is a machine learning regression model trained in advance using complete historical project data from a historical project database. The complete historical project data includes historical R&D process data and historical product final reliability verification data.

7. The digital management method according to claim 6, characterized in that, The management method also includes: New technology content data generated during product development is added to the training set to incrementally train the prediction model and update the model parameters. After the product completes the final reliability verification test, the actual reliability verification data will be sent back to the historical project database as labeled data for periodic retraining of the prediction model.

8. The digital management method according to claim 5, characterized in that, The method of dynamically predicting the reliability of the product under the current R&D progress using a prediction model, based on the process characteristic data in the technical content data, the completion degree of each process stage, and the overall completion degree of product development, includes: When the predicted reliability is lower than a preset reliability threshold or the predicted reliability shows a continuous downward trend, a risk warning is triggered, and the contribution of each input feature to the predicted reliability is calculated using an interpretability analysis method. Based on the contribution of each input feature, the process stage and technical content node that contributes the most to reliability risk are located. Risk warning tasks are automatically generated in the product lifecycle management system and pushed to the corresponding responsible persons.

9. The digital management method according to claim 1, characterized in that, The step of comparing the predicted reliability with a preset reliability threshold, triggering a risk warning and locating the risk source when the predicted reliability is lower than the preset reliability threshold, and adjusting the node weight and stage weight in reverse according to the risk warning and the risk source, and increasing the completion qualification threshold of the technical content node within the corresponding process stage includes: The predicted reliability is compared with a preset reliability threshold, and when the predicted reliability is lower than the preset reliability threshold, it is identified as a high-risk level. The stage weight of the process stage identified as high-risk level will be increased in the overall completion calculation, and the completion qualification threshold of the technical content nodes within the high-risk process stage will be increased. The adjusted weights and thresholds are applied to the next completed metric evaluation, forming a loop iteration that continues until product development is completed.

10. A digital management system for the entire product development process, characterized in that, The digital management system includes: Data acquisition and storage module: Extracts technical content data from each stage of the product development process in the product lifecycle management system, adds process evaluation metadata tags to the technical content data, and constructs a development process association graph based on the logical relationship between the technical content data; Complete the quantification and evaluation module: Based on the technical content data, calculate the completion degree of each technical content node, the completion degree of each process stage, and the overall completion degree of product development layer by layer. The node weight is assigned according to the criticality of each technical content node in the corresponding process stage, and the stage weight is assigned according to the importance of each process stage in the overall development. The node weight and the stage weight are dynamically adjusted according to the dynamic changes in the development process and the feedback of subsequent reliability prediction results. Reliability dynamic prediction module: Based on the process characteristic data in the technical content data, as well as the completion degree of each process stage and the overall completion degree of product development, the prediction model dynamically predicts the predicted reliability of the product under the current development progress. Collaborative scheduling module: compares the predicted reliability with a preset reliability threshold. When the predicted reliability is lower than the preset reliability threshold, a risk warning is triggered and the source of risk is located. Based on the risk warning and the source of risk, the node weight and the stage weight are adjusted in reverse, and the completion qualification threshold of the technical content node in the corresponding process stage is increased. Visualization module: Outputs the completion assessment results and reliability prediction results of product development through the visualization interface of the product lifecycle management system.

Images