A digital engineering full life cycle management method based on a unified data source

Abstract

The present invention belongs to the technical field of intelligent management of digital engineering. The present invention discloses a digital engineering full-life cycle management method based on a unified data source. The method includes: creating a project on a platform and determining project requirements; dividing tasks according to each requirement item in the project requirements and allocating each requirement item in the project requirements to the corresponding task; evaluating the priority of each task according to the requirement item; designing a model for each task according to the priority and obtaining corresponding model data; feeding back the requirement item and model data of each task to the platform; interconnecting tasks with adjacent priorities through work package association in the platform. The present invention realizes a high degree of coordination between project requirements and tasks, realizes digital engineering full-life cycle management across tools and processes, solves problems such as data separation and difficult information transmission, improves project management efficiency and reduces project management difficulty at the same time.

Description

Technical Field

[0001] The present invention relates to the technical field of digital engineering intelligent management. More specifically, the present invention relates to a digital engineering full-life cycle management method based on a unified data source. Background Art

[0002] Currently, in the implementation of digital engineering, the traditional product full-life cycle management method is adopted, which only manages the task distribution and the final result files, and cannot track and manage the process data of the model; the task management is disjointed from the model construction, and different tools are used and managed separately, resulting in data separation and island phenomena, that is, the data between task management and model construction cannot be automatically shared or interconnected, increasing the management complexity and the difficulty of information transmission; making the data need to be manually imported and exported, which is not only time-consuming and laborious, but also difficult to ensure the timeliness, accuracy, integrity and traceability of the data; thus causing the management personnel to be unable to accurately master the process data and difficult to understand the actual task progress, and further affecting the control of project risks.

[0003] Of course, there are also intelligent data interconnection methods. For example, the Chinese patent with the publication number CN118154119A discloses a full-life cycle intelligent management system for water conservancy projects based on digital twin; including: S1: a data acquisition module, which collects various index data of the water conservancy project in real time through sensors and transmits the data into the system; S2: a digital twin modeling module, which associates the actual water conservancy project with its digital model to establish a digital twin model; S3: a monitoring and diagnosis module, which based on the digital twin model, conducts real-time monitoring and fault diagnosis on the operation state of the water conservancy project; S4: a prediction and early warning module, which based on historical data and model analysis, predicts the future hydrological situation and the project state and gives early warnings in advance; solving the problems of difficult data acquisition, lack of real-time monitoring, slow response and lack of prediction ability in the full-life cycle intelligent management of water conservancy projects.

[0004] The above technology realizes intelligent data interconnection through a digital twin model, but the digital twin focuses on the virtual representation of specific entities and cannot effectively integrate the data collected from different tools and stages, still resulting in data islands; and the digital twin model usually focuses on the real-time operation state and cannot cover the life cycle management of the entire digital engineering, with weak integration and coordination ability; in addition, the digital twin model is slightly insufficient in the coordination and management of multiple tasks and faces the complexity of collaboration.

[0005] In view of this, the present invention proposes a digital engineering full-life cycle management method based on a unified data source to solve the above problems. Summary of the Invention

[0006] To overcome the above defects of the prior art and to achieve the above object, the present invention provides the following technical solution: A digital engineering full life cycle management method based on a unified data source, comprising:

[0007] S1: Create a project on the platform and determine the project requirements;

[0008] S2: Divide tasks according to each requirement item in the project requirements, and allocate each requirement item in the project requirements to the corresponding task;

[0009] S3: Evaluate the priority of each task according to the requirement item;

[0010] S4: Design a model for each task according to the priority and obtain the corresponding model data; Feed back the requirement item and model data of each task to the platform;

[0011] S5: Through the work package association in the platform, interconnect tasks with adjacent priorities, and transfer requirement items and model data.

[0012] Further, the project includes a project overview and a project plan; the project requirements are the functions and features required by the project;

[0013] The platform is a digital engineering full life cycle management platform; The platform defines API standards and API requirements. The API standard is the rules and agreements for defining and standardizing the interaction of application program interfaces; Each model construction tool provides the corresponding API according to the API requirements defined by the platform; Integrate the APIs of each model construction tool into the platform by programming and synchronize user identity authentication.

[0014] Further, the method for dividing tasks according to each requirement item in the project requirements includes:

[0015] Set different digital tags for different project names and mark them as project tags; Set different digital tags for different requirement items in the project requirements and mark them as item tags; Preset a task set, and each task set includes the tasks corresponding to each requirement item, and the task set corresponds to the requirement item one by one; Set different digital tags for different task sets and mark them as set tags; Among them, the project tag, the item tag and the set tag are all different;

[0016] Take the project tag and an item tag as a set of analysis data, input each set of analysis data into the trained task prediction model, and predict the set tag corresponding to each set of analysis data; Obtain the task set corresponding to each requirement item according to the set tag.

[0017] Further, the training process of the task prediction model includes:

[0018] Pre-collect the analysis data of group b, set corresponding set labels for all the analysis data of group b, where b is an integer greater than 1, and convert the analysis data and the corresponding set labels into a corresponding set of feature vectors;

[0019] Take each set of feature vectors as the input of the task prediction model. The task prediction model outputs a set of predicted set labels corresponding to each set of analysis data, and takes the actual set label corresponding to each set of analysis data as the prediction target. The actual set label is the set label preset corresponding to the analysis data; take minimizing the sum of prediction errors of all analysis data as the training target; train the task prediction model until the sum of prediction errors reaches convergence and then stop training; the task prediction model is a deep neural network model.

[0020] Further, the steps for evaluating the priority of each task include:

[0021] Evaluate the priority of the project requirements of the project to obtain the requirement priority of each project requirement; evaluate the priority of the requirement items corresponding to each project requirement to obtain the item priority of each requirement item in the corresponding project requirement; evaluate the priority of the tasks of each requirement item to obtain the task priority of each task in the corresponding requirement item; multiply the requirement priority, item priority, and task priority corresponding to each task in sequence to obtain the priority of each task.

[0022] Further, the method for evaluating the priority of the project requirements of the project includes:

[0023] Adopt the trained word vector conversion model to convert the project requirements of the project into corresponding word vectors in sequence; use the similarity algorithm to calculate the semantic similarity between every two word vectors; preset a similarity threshold, compare each semantic similarity with the similarity threshold, mark the semantic similarities with values greater than or equal to the similarity threshold as the same-kind similarities, and do not mark the semantic similarities with values less than the similarity threshold; classify each project requirement according to the classification rules in sequence to obtain a total of P requirement sets, and there are no identical project requirements in each requirement set, where P is an integer greater than 0;

[0024] Generate a sorting table for each requirement set, and sort the project requirements in each requirement set according to the sorting rules; count the number of project requirements in each sorting table, and mark the largest number of project requirements as the maximum number; take the maximum number as the priority of the project requirement ranked first in each sorting table, and the priorities of the remaining project requirements in each sorting table are obtained in sequence according to the positive order of the corresponding sorting table and the preset decreasing amplitude;

[0025] The method for priority evaluation of the requirement items corresponding to each project requirement and the tasks corresponding to each requirement item is the same as the method for priority evaluation of the project requirements of the project.

[0026] Furthermore, the training process of the word vector conversion model includes:

[0027] Pre-collect multiple project requirements; use the WordPiece algorithm to construct a word table containing the vocabulary in the project requirement domain; the vocabulary in the project requirement domain is a vocabulary set summarized from professional words and terms related to project requirements; convert the project requirements into an ID sequence through the word table, and add positional encoding as inputs; the inputs are encoded by the BERT encoder for deep Transformer encoding to obtain hidden sentence representations, and the hidden sentence representations are the word vectors; randomly select some words to mask and predict the corresponding original words as the unsupervised pre-training task objective; use cross-entropy to calculate the loss function, and optimize and modify the parameters of the word vector conversion model through backpropagation to complete the pre-training of the word vector conversion model; pre-collect the word vectors corresponding to the project requirements as annotations to construct a classification task; add a classification output layer on the basis of the pre-trained word vector conversion model, with the determination rate as the task objective; fix the BERT encoder and optimize and fine-tune the parameters of the output layer.

[0028] Furthermore, the classification rules include:

[0029] Mark the project requirement currently being classified as the current requirement;

[0030] Analyze the semantic similarity between the current requirement and the project requirements in all requirement sets;

[0031] If there is a semantic similarity marked as a similar class similarity, add the current requirement to the requirement set corresponding to the project requirement corresponding to the similar class similarity;

[0032] If there is no semantic similarity marked as a similar class similarity, add the current requirement to a new requirement set;

[0033] If there are multiple semantic similarities marked as similar class similarities, mark the requirement sets where the project requirements corresponding to each similar class similarity are located as similar class sets, count the number of project requirements in each similar class set, mark the similar class set with the largest number of project requirements as the largest set, add the current requirement to the largest set, and add the project requirements in each similar class set not marked as the largest set to the largest set.

[0034] Furthermore, the sorting rules include:

[0035] Set different digital tags for each item requirement in each requirement set, and mark them as requirement tags. Input the requirement tags corresponding to each requirement set into the corresponding priority analysis model respectively, predict the requirement tag with the highest priority in each requirement set, and mark it as the maximum tag. The priority analysis model corresponds to the requirement set one by one. The training process of the priority analysis model is the same as that of the task prediction model, and both are deep neural network models. Generate a sorting table for each requirement set, rank the item requirement corresponding to each maximum tag at the first place in the corresponding sorting table, mark the item requirement ranked at the last place in each sorting table as the end requirement, add the item requirement with the highest similarity of the same type to the end requirement to the corresponding sorting table, and add all item requirements to the corresponding sorting table in turn.

[0036] Further, sort all tasks in descending order of priority, and design a model for each task according to the positive order of priority, in combination with each task and the requirement items corresponding to each task; the model data are various data and files generated or used during the modeling process; the work package association is to connect different work packages in project management; the work package is the task in the project, and through the work package association, the tasks with adjacent priorities are connected; in the adjacent tasks, the output of the task with higher priority is used as the input of the task with lower priority.

[0037] The technical effects and advantages of a digital engineering full life cycle management method based on a unified data source according to the present invention:

[0038] Utilize a unified data source and API standard to achieve seamless integration of various model construction tools, and apply a word vector conversion model and a similarity algorithm for requirement classification and sorting to formulate a multi-level priority evaluation mechanism, so as to achieve a high degree of coordination between project requirements and tasks; at the same time, connect tasks in an orderly manner through work package association, so as to realize digital engineering full life cycle management across tools and processes, solve problems such as data separation and difficult information transmission, and improve project management efficiency while reducing project management difficulty. Brief Description of the Drawings

[0039] Figure 1 It is a flowchart of a digital engineering full life cycle management method based on a unified data source according to Embodiment 1 of the present invention;

[0040] Figure 2 It is a schematic diagram of the working principle of the digital engineering full life cycle management platform according to Embodiment 1 of the present invention. Detailed Embodiment

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0042] Embodiment 1

[0043] Please refer to Figure 1 As shown, a digital engineering full-life cycle management method based on a unified data source in this embodiment includes:

[0044] S1: Create a project on the platform and determine the project requirements.

[0045] The project includes a project overview and a project plan; the project overview includes, for example, project basic information (such as project name, project description, project type, etc.), project goals, project scope, etc.; the project plan includes, for example, a project time plan (such as project cycle, time nodes of each task in the project), project resource allocation (such as project team, project budget, project cost, etc.), project requirements (such as technical requirements, functional requirements, etc.); the project requirements are the functions and features that the project needs to achieve, and the project requirements include, for example, providing a user feedback function, giving an early warning within 5 seconds, etc.

[0046] The project is created on the platform by members within the project team, and the platform is a digital engineering full-life cycle management platform; the platform defines API (Application Programming Interface) standards and API requirements. The API standard is the rules and agreements that define and standardize the interaction of application programming interfaces, including data format, protocol, security, etc.; the API requirements include, for example, information transmission on how to perform user identity authentication, how to transmit model data, etc.; each model construction tool (such as CAD, CAE, MATLAB, etc.) provides corresponding APIs according to the API requirements defined by the platform, and the APIs provided by each model construction tool will allow the platform to perform operations such as authentication and management; the APIs of each model construction tool are integrated into the platform in a programming manner, and user identity authentication is synchronized; synchronizing user identity authentication means that when a user logs in to the platform, the platform will send the user's authentication token to all integrated model construction tools through the API, so that the user will be automatically logged in to all model construction tools without having to log in again.

[0047] It should be noted that the purpose of integrating the APIs of various model building tools into the platform is to centrally manage the model data of all model building tools and the project-related data in the platform, achieve unified data sources, avoid the situation of data islands, and the platform can cover the entire life cycle management of digital engineering, improving the integration and overall planning ability; the purpose of synchronizing user identity authentication is to reduce the time required for users to switch between different model building tools, improve work efficiency, enhance the user experience, and centralized identity authentication enables security policies to be uniformly managed in one place, which helps to improve the consistency and maintainability of security control.

[0048] S2: Divide tasks according to each requirement item in the project requirements, and assign each requirement item to the corresponding task to achieve collaborative development of project requirements.

[0049] For example, if the project requirement is to provide a user feedback function, the corresponding requirement items are feedback collection, feedback storage, feedback classification and processing, feedback reply, feedback report generation, etc.; for example, if the requirement item is feedback collection, the corresponding tasks are designing a feedback form, implementing the front-end submission function, implementing the back-end receiving function, etc., if the requirement item is feedback storage, the corresponding tasks are database design, data interface development, storage mechanism implementation, etc., if the requirement item is feedback classification and processing, the corresponding tasks are establishing classification rules, developing processing processes, etc.

[0050] The methods for dividing tasks according to each requirement item in the project requirements include:

[0051] Set different digital tags for different project names and mark them as project tags; set different digital tags for different requirement items in the project requirements and mark them as item tags; preset a task set, each task set includes the tasks corresponding to each requirement item, the task set corresponds to the requirement item one by one, and the task set is preset by those skilled in the art according to the requirement items in the actual situation; set different digital tags for different task sets and mark them as set tags; among them, the project tags, item tags, and set tags are all different;

[0052] Take the project tag and an item tag as a set of analysis data, input each set of analysis data into the trained task prediction model, and predict the set tag corresponding to each set of analysis data; obtain the task set corresponding to each requirement item according to the set tag.

[0053] The training process of the task prediction model includes:

[0054] Pre-collect the analysis data of group b, set corresponding set labels for all the analysis data of group b, where b is an integer greater than 1, and convert the analysis data and the corresponding set labels into a corresponding set of feature vectors; the set labels corresponding to the analysis data are collected by those skilled in the art during the operation of historical project division. For the b groups of analysis data, corresponding set labels are set according to the actual situation; corresponding set labels are sequentially set for the b groups of analysis data;

[0055] Take each set of feature vectors as the input of the task prediction model. The task prediction model outputs a set of predicted set labels corresponding to each set of analysis data, and takes the actual set label corresponding to each set of analysis data as the prediction target. The actual set label is the pre-set set label corresponding to the analysis data; take minimizing the sum of the prediction errors of all the analysis data as the training target; wherein, the calculation formula of the prediction error is η K =(β K -ε K ) 2 , where η K is the prediction error, K is the group number of the feature vectors corresponding to the analysis data, β K is the predicted set label corresponding to the Kth group of analysis data, and ε K is the actual set label corresponding to the Kth group of analysis data; train the task prediction model until the sum of the prediction errors reaches convergence and then stop training.

[0056] The above task prediction model is specifically a deep neural network model; it includes an input layer, a hidden layer, and an output layer; each hidden layer includes multiple neurons, and there are connections between each neuron and the neurons in the next layer. The connections contain weights that determine the importance and influence of data transmission in the neural network; an activation function is applied to each neuron between the hidden layer and the output layer. The activation function introduces non-linearity and allows the network to learn more complex patterns and features.

[0057] It should be noted that the purpose of task division according to each requirement item in the project requirements is to refine the large project requirements into smaller executable resources, which is conducive to better optimizing resource allocation and task coordination; and refining tasks helps to more clearly define the project goals and progress, facilitating real-time monitoring and adjustment by project managers; at the same time, promoting the correspondence between requirements and tasks, and corresponding the requirement items and specific tasks helps to ensure that the requirements are fully implemented and covered.

[0058] S3: Evaluate the priority of each task according to the requirement items.

[0059] The steps for evaluating the priority of each task include:

[0060] Evaluate the priorities of the project requirements of a project to obtain the requirement priorities of each project requirement; evaluate the priorities of the requirement entries corresponding to each project requirement to obtain the entry priorities of each requirement entry in the corresponding project requirement; evaluate the priorities of the tasks of each requirement entry to obtain the task priorities of each task in the corresponding requirement entry; multiply the requirement priority, entry priority, and task priority corresponding to each task in sequence to obtain the priority of each task.

[0061] Exemplarily, the project includes project requirement 1 and project requirement 2. Project requirement 1 includes requirement entry 1 and requirement entry 2. Project requirement 2 includes requirement entry 3 and requirement entry 4. Requirement entry 1 includes task 1 and task 2. Similarly, each requirement entry includes two tasks. The priority of project requirement 1 is 1, the priority of project requirement 2 is 2. The priority of requirement entry 1 in project requirement 1 is 1, the priority of requirement entry 2 in project requirement 1 is 2. The priority of requirement entry 3 in project requirement 2 is 1, the priority of requirement entry 4 in project requirement 2 is 2. The priority of task 1 in requirement entry 1 is 2, the priority of task 2 in requirement entry 1 is 1. Since task 1 belongs to requirement entry 1 and requirement entry 1 belongs to project requirement 1, the priority of task 1 should be 1×1×2 = 2.

[0062] The method for evaluating the priorities of the project requirements of a project includes:

[0063] Adopt a trained word vector conversion model to sequentially convert the project requirements of the project into corresponding word vectors; use a similarity algorithm to calculate the semantic similarity between every two word vectors. The similarity algorithm can be, for example, cosine similarity calculation, Mahalanobis distance calculation, Euclidean distance calculation, etc.; preset a similarity threshold, and compare each semantic similarity with the similarity threshold. Mark the semantic similarities with values greater than or equal to the similarity threshold as similar similarities of the same category, and do not mark the semantic similarities with values less than the similarity threshold; classify each project requirement in sequence according to the classification rules to obtain a total of P requirement sets, and there are no identical project requirements in each requirement set. P is an integer greater than 0; the similarity threshold is obtained by those skilled in the art during historical priority evaluation. Collect multiple requirement sets, obtain the semantic similarities corresponding to every two project requirements in each requirement set, take the smallest semantic similarity as the minimum similarity of each requirement set, and take the average of multiple minimum similarities as the similarity threshold;

[0064] For each set of requirements, a sorting table is generated, and the item requirements in each set of requirements are sorted according to the sorting rules; the number of item requirements in each sorting table is counted, and the largest number of item requirements is marked as the maximum number; the maximum number is used as the priority of the item requirement ranked at the front of each sorting table, and the priorities of the remaining item requirements in each sorting table are obtained in sequence according to the positive order of the corresponding sorting table and a preset decreasing range, and the decreasing range is set by those skilled in the art according to the actual situation.

[0065] The training process of the word vector conversion model includes:

[0066] Pre-collect multiple item requirements; use the WordPiece algorithm to construct a word table containing the vocabulary in the field of item requirements; the vocabulary in the field of item requirements is a vocabulary set summarized from professional vocabulary and terms related to item requirements; convert the item requirements into an ID sequence through the word table, and add position encoding as inputs; the inputs are deeply encoded by a BERT encoder through a Transformer to obtain a hidden sentence representation, and the hidden sentence representation is the word vector; randomly select some words to be masked and predict the corresponding original words as the objective of the unsupervised pre-training task; use cross-entropy to calculate the loss function, and backpropagate to optimize and modify the parameters of the word vector conversion model to complete the pre-training of the word vector conversion model; pre-collect the word vectors corresponding to the item requirements as annotations to construct a classification task; add a classification output layer on the basis of the pre-trained word vector conversion model, with the determination rate as the task objective; fix the BERT encoder and optimize and fine-tune the parameters of the output layer.

[0067] The classification rules include:

[0068] Mark the item requirement currently being classified as the current requirement;

[0069] Analyze the semantic similarity between the current requirement and the item requirements in all sets of requirements;

[0070] If there is a semantic similarity marked as the same-class similarity, add the current requirement to the set of requirements corresponding to the item requirements corresponding to the same-class similarity;

[0071] If there is no semantic similarity marked as the same-class similarity, add the current requirement to a new set of requirements;

[0072] If there are multiple semantic similarities marked as the same-class similarity, mark the set of requirements where the item requirements corresponding to each same-class similarity are located as the same-class set, count the number of item requirements in each same-class set, mark the same-class set with the largest number of item requirements as the largest set, add the current requirement to the largest set, and add the item requirements in each same-class set not marked as the largest set to the largest set.

[0073] The methods for priority assessment of the requirement items corresponding to each project requirement and the tasks corresponding to each requirement item are the same as those for priority assessment of the project requirements of the project.

[0074] The sorting rules include:

[0075] Each project requirement in each requirement set is set with a different digital label and marked as a requirement label. The requirement labels corresponding to each requirement set are respectively input into the corresponding priority analysis model to predict the requirement label with the highest priority in each requirement set, which is marked as the maximum label. The priority analysis models correspond one by one to the requirement sets. The training process of the priority analysis model is the same as that of the task prediction model, and both are deep neural network models. A sorting table is generated for each requirement set. The project requirement corresponding to each maximum label is ranked first in the corresponding sorting table. The project requirement ranked last in each sorting table is marked as the end requirement. The project requirement with the highest similarity of the same type to each end requirement is added to the corresponding sorting table, and all project requirements are sequentially added to the corresponding sorting tables.

[0076] It should be noted that the purpose of priority assessment of each task according to the requirement items is to help determine the execution order of the tasks, ensure the coherence and dependency relationships among the tasks, thereby improving the efficiency of task execution and enhancing the collaboration ability of the project team.

[0077] S4: Model design is carried out for each task according to the priority, and the corresponding model data is obtained; the requirement items and model data of each task are both fed back to the platform. The schematic diagram of the platform working principle is as Figure 2 shown;

[0078] The priorities of all tasks are sorted from high to low. The modelers in the project team carry out model design for each task according to the positive order of the priorities, and in combination with each task and the requirement items corresponding to each task. After the model design is completed, the requirement items and model data of each task are both fed back to the platform. The model data are various data and files generated or used during the modeling process. The model data includes, for example, design drawings, simulation results, parameter settings, data sets, etc. The model data is used to describe the attributes, behaviors of the model and its relationships with other systems or components.

[0079] It should be understood that the purpose of feeding back the requirement items and model data of each task to the platform is to ensure the timeliness, accuracy, integrity and traceability of the data, enabling project managers to accurately master the process data, timely understand the real task progress, thereby enhancing the ability to control project risks and reducing the project management difficulty.

[0080] S5: Through the work package association in the platform, tasks with adjacent priorities are interconnected to transfer requirement items and model data, realizing the automatic transfer of data.

[0081] Work package association in project management is to connect different work packages to ensure the smooth transfer of data and information between work packages; a work package is a task in a project, and through work package association, tasks with adjacent priorities are connected; in adjacent tasks, the output of the task with a higher priority is used as the input of the task with a lower priority, realizing the data interconnection between adjacent tasks and enhancing the coordination and management capabilities of multiple tasks.

[0082] This embodiment uses a unified data source and API standard to achieve seamless integration of various model construction tools, and applies a word vector conversion model and a similarity algorithm for requirement classification and ranking to establish a multi-level priority evaluation mechanism, realizing a high degree of coordination between project requirements and tasks; at the same time, tasks are connected in an orderly manner through work package association, thereby realizing the full life cycle management of digital engineering across tools and processes, solving problems such as data separation and difficult information transfer, improving project management efficiency while reducing the difficulty of project management.

[0083] Embodiment 2

[0084] This application also provides an electronic device, which may include one or more processors and one or more memories. Among them, computer-readable code is stored in the memory, and when the computer-readable code is run by one or more processors, it can execute a method for full life cycle management of digital engineering based on a unified data source as described above.

[0085] The method or system according to the embodiment of the present application can also be implemented by means of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as ROM or a hard disk, can store a method for full life cycle management of digital engineering based on a unified data source provided by the present application.

[0086] Furthermore, the electronic device may further include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components of the electronic device disclosed in the present invention can be omitted according to actual needs.

[0087] Embodiment 3

[0088] One embodiment of the present application discloses a computer-readable storage medium. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are run by a processor, a digital engineering full life cycle management method according to an embodiment of the present application described with reference to the above drawings can be executed. The storage medium includes but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory, etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0089] In addition, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be run by a processor to execute instructions corresponding to the method steps provided by the present application, such as: a digital engineering full life cycle management method based on a unified data source. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0090] The above is only the specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0091] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A digital engineering full life cycle management method based on a unified data source, characterized in that: include: S1: Create a project on the platform and determine project requirements; S2: Divide the tasks according to the requirements in the project requirements and assign each requirement item in the project requirements to the corresponding tasks; S3: Prioritize each task based on the requirements; S4: Design a model for each task according to the priority and obtain the corresponding model data; feedback the requirement items and model data of each task to the platform; S5: Through the work package association in the platform, tasks with adjacent priorities are interconnected to transfer requirement items and model data; The S5 also includes sorting the priorities of all tasks from large to small, and designing a model for each task according to the positive order of priority and in combination with each task and the requirement items corresponding to each task; the model data refers to various data and files generated or used in the modeling process; the work package association refers to connecting different work packages in project management; a work package is a task in a project, and tasks with adjacent priorities are connected through work package association; the output of a task with a larger priority among adjacent tasks is used as the input of a task with a smaller priority; The step of evaluating the priority of each task includes: evaluating the priority of the project requirements of the project to obtain the requirement priority of each project requirement; evaluating the priority of the requirement items corresponding to each project requirement to obtain the item priority of each requirement item in the corresponding project requirement; evaluating the priority of the task of each requirement item to obtain the task priority of each task in the corresponding requirement item; multiplying the requirement priority, item priority and task priority corresponding to each task in turn to obtain the priority of each task; The method for evaluating the priority of project requirements of a project includes: using a trained word vector conversion model to convert the project requirements of the project into corresponding word vectors in turn; using a similarity algorithm to calculate the semantic similarity between every two word vectors; presetting a similarity threshold, comparing each semantic similarity with the similarity threshold, marking the semantic similarity with a value greater than or equal to the similarity threshold as the same similarity, and not marking the semantic similarity with a value less than the similarity threshold; classifying each project requirement in turn according to the classification rule, obtaining a total of P requirement sets, wherein there is no identical project requirement in each requirement set, and P is an integer greater than 0; A sorting table is generated for each requirement set, and the project requirements in each requirement set are sorted according to the sorting rules; the number of project requirements in each sorting table is counted, and the number of project requirements with the largest value is marked as the maximum number; the maximum number is used as the priority of the project requirement at the front of each sorting table, and the priorities of the remaining project requirements in each sorting table are obtained in sequence according to the positive order of the corresponding sorting table and the preset decreasing range; The method of evaluating the priority of the requirement items corresponding to each project requirement and the tasks corresponding to each requirement item is consistent with the method of evaluating the priority of the project requirements.

2. A digital engineering full life cycle management method based on a unified data source according to claim 1, characterized in that: The project includes a project overview and a project plan; the project requirements are the functions and features that the project needs to achieve; The platform is a digital engineering full life cycle management platform; The platform defines API standards and API requirements. API standards are rules and conventions that define and regulate application program interface interactions. Each model building tool provides corresponding APIs based on the API requirements defined by the platform. Integrate the APIs of various model building tools into the platform programmatically and synchronize user authentication.

3. A digital engineering full life cycle management method based on a unified data source according to claim 2, characterized in that: The method for dividing tasks according to each requirement item in the project requirements includes: Different digital labels are set for different project names and marked as project labels; different digital labels are set for different requirement items in the project requirements and marked as item labels; task sets are preset, each task set includes tasks corresponding to each requirement item, and task sets correspond to requirement items one by one; different digital labels are set for different task sets and marked as set labels; the project labels, item labels and set labels are all different; Take the project label and an item label as a set of analysis data, input each set of analysis data into the trained task prediction model, predict the set label corresponding to each set of analysis data; and obtain the task set corresponding to each requirement item according to the set label.

4. A digital engineering full life cycle management method based on a unified data source according to claim 3, characterized in that: The training process of the task prediction model includes: Collect b groups of analysis data in advance, set corresponding set labels for the b groups of analysis data, where b is an integer greater than 1, and convert the analysis data and the corresponding set labels into a corresponding set of feature vectors; Each group of feature vectors is used as the input of the task prediction model. The task prediction model takes a group of prediction set labels corresponding to each group of analysis data as output, and takes the actual set labels corresponding to each group of analysis data as prediction targets, where the actual set labels are pre-set set labels corresponding to the analysis data; minimizing the sum of prediction errors of all analysis data is used as the training target; the task prediction model is trained until the sum of prediction errors reaches convergence and the training is stopped; the task prediction model is a deep neural network model.

5. A digital engineering full life cycle management method based on a unified data source according to claim 4, characterized in that: The training process of the word vector conversion model includes: Collect multiple project requirements in advance; use the WordPiece algorithm to build a word list containing vocabulary in the field of project requirements; the vocabulary in the field of project requirements is a vocabulary collection that summarizes professional vocabulary and terminology related to project requirements; convert project requirements into ID sequences through the word list, and add position codes as inputs; the inputs are deep Transformer encoded through the BERT encoder to obtain latent sentence representations, which are word vectors; randomly select some word masks and predict the corresponding original words as the unsupervised pre-training task target; use cross entropy to calculate the loss function, and back-propagate to optimize and modify the parameters of the word vector conversion model to complete the pre-training of the word vector conversion model; collect the word vectors corresponding to the project requirements in advance as annotations and construct classification tasks; add a classification output layer based on the pre-trained word vector conversion model to determine the rate as the task target; fix the BERT encoder and optimize and fine-tune the output layer parameters.

6. A digital engineering full life cycle management method based on a unified data source according to claim 5, characterized in that: The classification rules include: Mark the project requirements currently being classified as current requirements; Analyze the semantic similarity between the current requirement and the project requirements in all requirement sets; If there is a semantic similarity that is marked as the same type of similarity, the current requirement is added to the requirement set corresponding to the project requirement corresponding to the same type of similarity; If there is no semantic similarity marked as similarity of the same kind, the current requirement is added to a new requirement set; If there are multiple semantic similarities marked as similarities of the same kind, then the requirement set corresponding to each project requirement of the same kind of similarity is marked as a similar set, the number of project requirements in each similar set is counted, the similar set with the largest number of project requirements is marked as the maximum set, the current requirement is added to the maximum set, and the project requirements in each similar set that is not marked as the maximum set are added to the maximum set.

7. A digital engineering full life cycle management method based on a unified data source according to claim 6, characterized in that: The sorting rules include: Set a different digital label for each project requirement in each requirement set and mark it as a requirement label. Input the requirement labels corresponding to each requirement set into the corresponding priority analysis model, predict the requirement label with the highest priority in each requirement set, and mark it as the maximum label. The priority analysis model corresponds to the requirement set one by one. The training process of the priority analysis model is consistent with the training process of the task prediction model, and both are deep neural network models. Generate a sorting table for each requirement set, rank the project requirement corresponding to each maximum label first in the corresponding sorting table, mark the project requirement ranked last in each sorting table as the last requirement, add the project requirement with the greatest similarity to each last requirement to the corresponding sorting table, and add all project requirements to the corresponding sorting table in sequence.

Images