Calibration method and system of mobile calibration digital platform

Abstract

The invention relates to the technical field of engineering drawing verification, in particular to a verification method and system of a mobile verification digital platform, and the method comprises the steps: obtaining project approval data, including obtaining project information, configuring professional types and setting personnel roles; drawing a drawing based on the obtained project approval data, and performing manual lightweight operation on the drawn drawing file; and performing review operation according to the drawn drawing file, constructing a prediction model to optimize a review operation process, performing prediction model training based on the constructed prediction model, and performing data storage on the reviewed drawing file. According to the method, the situation of information transmission lag in traditional offline verification or non-real-time online verification can be changed in a real-time interaction mode, so that the verification process is smoother and more efficient, and all parties can communicate and cooperate in time.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering drawing review, and in particular to a review method and system for a mobile review digital platform. Background Art

[0002] In today's engineering design field, as the scale of construction projects continues to expand and the complexity increases, the review of drawings in the design process is becoming more and more critical. As the core guiding document of the engineering project, the accuracy and quality of the design drawings are directly related to the success or failure of the project. Whether it is a large-scale building complex, infrastructure construction, or industrial plant design, it needs to go through a strict review process to ensure that the design meets the specification requirements, functional requirements and safety standards. However, the current traditional drawing review is highly dependent on CAD software and LAN environment. As a professional design tool, CAD software is powerful, but it needs to be installed on specific computer equipment, and its operation has certain requirements for hardware configuration. This requires reviewers to perform review work in an office equipped with corresponding CAD software and hardware conditions. The LAN limits the scope of work of the reviewers. Drawing files can only be accessed and processed when connected to the company's internal LAN. This reliance on a specific environment severely restricts the flexibility of the review work and cannot meet the needs of remote collaboration and mobile office that may appear in modern engineering design projects. Due to the reliance on the office environment and LAN, the review work can only be carried out during working hours and in the office. Once the reviewers leave the office, such as on a business trip, working from home, or when urgent review drawings are needed at the construction site, they cannot carry out work in time. This causes the review process to be frequently interrupted due to the absence of personnel, and the continuity of the review cannot be guaranteed. For some urgent projects or design changes that require rapid response, the time and site limitations of the existing review technology have become a bottleneck for project advancement, which may lead to project delays and increased costs. At this stage, a review method and system for a mobile review digital platform is needed. Summary of the invention

[0003] In order to solve the problems of lack of flexibility and cumbersome and complicated proofreading work in traditional proofreading work, the present invention provides a proofreading method and system for a mobile proofreading digital platform.

[0004] In a first aspect, the present invention provides a review method for a mobile review digital platform, which adopts the following technical solution: A review method for a mobile review digital platform, comprising: Obtain project establishment data, including obtaining project information, configuring professional types, and setting personnel roles; Draw drawings based on the acquired project approval data, and perform manual lightweight operations on the drawn drawing files; Perform review operations based on the drawn drawing files, including sending the lightweight drawing files to the review area for real-time data synchronization between the web end and the CAD end, and transmitting the annotation information to the CAD end in real time through the real-time communication channel established by the Websocket method; Constructing prediction models to optimize the review process, including constructing a multi-layer perceptron neural network to predict the types of problematic areas in the drawing files; Perform prediction model training based on the constructed prediction model, including updating the weights and biases of the prediction model using the back propagation algorithm; The reviewed drawing files are stored in a data format, including a combination of incremental backup and full backup.

[0005] Furthermore, the manual lightweight operation of the drawn drawing file includes traversing the graphic entities and data blocks in the drawing file using a breadth-first search algorithm, identifying key graphic elements and attribute information, simplifying and compressing the extracted information, then removing redundant data, and reorganizing and storing the processed file, wherein the graphic entity simplification rate formula is: , in, Expressed as the graphic entity simplification rate, Represents the number of graphic entities in the original file. Expressed as the number of optimized graphic entities.

[0006] Furthermore, the real-time communication channel established through the Websocket method includes integrating Websocket client code on the web side and the cad side respectively, establishing a connection through a specified server address, triggering corresponding processing logic when the connection is successful, a message is received, and the connection is closed, and the web side encapsulates the proofreading annotation information into a specific format and sends it to the server through the Websocket connection, which is then forwarded to all connected cad sides through the server.

[0007] Furthermore, the construction of a multilayer perceptron neural network predicts the area with problems, including constructing an input layer, multiple hidden layers and an output layer, wherein the input layer receives preprocessed historical review data features, the hidden layer performs a nonlinear transformation on the input data through an activation function, and the output layer outputs the type of the predicted problem area.

[0008] Furthermore, the hidden layer includes a plurality of neuron nodes, and the input of the hidden layer neuron is the weighted sum of the output of the input layer neuron plus the bias, and the input formula of the neuron is: , in, Represented as the weight from the input layer to the first hidden layer, Denotes the bias of the first hidden layer, Represented as the input value of the jth neuron in the first hidden layer, Represented as the output value of the i-th neuron in the input layer.

[0009] Furthermore, the use of the back-propagation algorithm to update the weights and biases of the prediction model includes calculating the difference between the predicted output and the true label according to the cross-entropy loss function, using the cross-entropy loss function to derive the gradient of the loss function with respect to the input of the output layer neurons, and using the gradient descent algorithm to update the weights and biases according to the calculated gradient.

[0010] Furthermore, the storage is performed by combining incremental backup and full backup, including storing drawing files in a distributed file system, storing and backing up project information and review records in a relational database, and then performing regular data storage by combining incremental backup and full backup, wherein the storage strategy formula is: , in, Expressed as the total backup data volume, Indicates the total amount of data when the first full backup is performed. Indicates the amount of data for each incremental backup.

[0011] In a second aspect, a proofreading system of a mobile proofreading digital platform includes: The data acquisition module is configured to acquire project establishment data, including acquiring project information, configuring professional types, and setting personnel roles; The drawing drawing module is configured to draw drawings based on the acquired project establishment data and perform manual lightweight operations on the drawn drawing files; The synchronization module is configured to perform review operations according to the drawn drawing files, including sending the drawing files after the lightweight operation to the review area for real-time data synchronization between the web end and the CAD end, and transmitting the annotation information to the CAD end in real time through the real-time communication channel established by the Websocket method; The model module is configured to construct a prediction model to optimize the review operation process, including constructing a multi-layer perceptron neural network to predict the type of areas with problems in the drawing file; The training module is configured to perform prediction model training based on the constructed prediction model, including updating the weights and biases of the prediction model using a back propagation algorithm; The storage module is configured to store the data of the drawing files that have been reviewed, including storage in a combination of incremental backup and full backup.

[0012] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for the review method of a mobile review digital platform.

[0013] In a fourth aspect, the present invention provides a terminal device including a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; and the computer-readable storage medium is used to store a plurality of instructions, wherein the instructions are suitable for being loaded by the processor and executing the review method of the mobile review digital platform.

[0014] In summary, the present invention has the following beneficial technical effects: 1. The present invention uses the Websocket method to establish a real-time communication channel to achieve real-time data synchronization between the web end and the CAD end. This real-time interactive mode changes the situation of information transmission lag in traditional offline review or non-real-time online review, making the review process smoother and more efficient. All parties can communicate and collaborate in a timely manner, reducing design errors and repeated modifications caused by poor information flow, and further improving the flexibility and response speed of the review work.

[0015] 2. The present invention optimizes the drawing files through manual lightweight operations, traverses the graphic entities and data blocks in the drawings through the breadth-first search algorithm, identifies key graphic elements and attribute information, removes redundant data, and significantly reduces the data volume of the drawing files. This greatly improves the loading and browsing speed of drawings on the web, and reviewers can open and view drawings faster, saving waiting time, simplifying the drawing operation links in the review process, and improving the overall review efficiency.

[0016] 3. The present invention predicts the types of areas with problems by constructing a multi-layer perceptron neural network prediction model, thereby providing auxiliary decision-making basis for reviewers. During the review process, the model can predict the types of areas in the drawings that may have problems based on the characteristics of historical review data, such as building structure problems, water supply and drainage problems, etc. Reviewers can conduct targeted key reviews based on these prediction results, reduce the workload of blind inspections, and improve review efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the overall process of a proofreading method of a mobile proofreading digital platform according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0019] Example 1 Reference Figure 1, a review method of a mobile review digital platform of this embodiment includes: Obtain project establishment data, including obtaining project information, configuring professional types, and setting personnel roles; Draw drawings based on the acquired project approval data, and perform manual lightweight operations on the drawn drawing files; Perform review operations based on the drawn drawing files, including sending the lightweight drawing files to the review area for real-time data synchronization between the web end and the CAD end, and transmitting the annotation information to the CAD end in real time through the real-time communication channel established by the Websocket method; Constructing prediction models to optimize the review process, including constructing a multi-layer perceptron neural network to predict the types of problematic areas in the drawing files; Perform prediction model training based on the constructed prediction model, including updating the weights and biases of the prediction model using the back propagation algorithm; The reviewed drawing files are stored in a data format, including a combination of incremental backup and full backup.

[0020] Specifically, a review method of a mobile review digital platform includes the following steps: like Figure 1 As shown, S1, obtaining project establishment data, including obtaining project information, configuring professional types, and setting personnel roles; Responsible for the management of the entire project life cycle, including project creation, information entry (such as project name, number, start and end time, project introduction, etc.) and project status tracking (such as in progress, completed, suspended, etc.), configure professional types, covering various professions such as architecture, structure, water, heating, electricity, etc., manually select the corresponding profession according to the characteristics of the drawings, ensure that the drawings can accurately match the review standards and personnel of the corresponding profession in the review process, set personnel roles, clarify the permissions and responsibilities of different roles such as professional leaders, reviewers, approvers, proofreaders, designers, etc., adopt a role-based authority allocation model (RBAC), and grant corresponding operation permissions according to user roles. For example, designers can create, edit and submit drawings for review; reviewers can annotate and review operations; professional leaders are responsible for project progress control and coordination and communication, etc.

[0021] S2. Draw drawings based on the acquired project approval data, and perform manual lightweight operations on the drawn drawing files; Based on the acquired project approval data and personnel roles, the designer uses a variety of drawing software to draw drawing files in the design area and generate dwg files. Then, the dwg file structure is parsed, and the breadth-first search algorithm is used to traverse the graphic entities and data blocks in the drawing files to identify key graphic elements (such as contour lines, center lines, etc.) and attribute information (such as layers, colors, line types, etc.). The extracted information is simplified and compressed, and the line segment merging algorithm is used to remove redundant data, and adjacent and collinear line segments are merged into one line segment to reduce the number of line segments. Then, the layer structure is optimized, empty layers and duplicate layer definitions are deleted, and layer data redundancy is reduced. Suppose the number of graphic entities in the original file is , after optimization, the number of graphic entities is , then the graphic entity simplification rate formula is: , in, Expressed as the graphic entity simplification rate, Represents the number of graphic entities in the original file. Expressed as the number of optimized graphic entities.

[0022] S3, performing review operations according to the drawn drawing files, including sending the drawing files after lightweight operations to the review area for real-time data synchronization between the web end and the CAD end, and transmitting the annotation information to the CAD end in real time through the real-time communication channel established by the Websocket method; After the drawings are completed and lightweight, the designer can send the drawings to the review area. The system automatically triggers the review notification mechanism and uses message queues (such as Kafka, RabbitMQ, etc.) to push review notifications to the corresponding reviewers. The notification content includes project details, drawing name, review priority, review deadline, etc. The reviewer logs in to the web-side digital platform, views the list of tasks to be reviewed in the review area, selects the review process initiated by the designer, and then opens the drawings on the web side for review operations, realizing real-time data synchronization between the web side and the CAD side. During the review process, when the reviewer annotates the drawings on the web side (such as selecting problem areas, adding text comments, marking dimensions, etc.), the real-time communication channel established through Websocket technology, first, create a Websocket server instance in Node.js, set the connection timeout and the maximum number of connections, integrate the Websocket client code on the web side and the CAD side, and use JavaScript on the web side to implement the Websocket client connection. The code example is as follows: socket.onmessage = function (event) { const data = JSON.parse(event.data); } / / Process the received data, such as updating the annotation display on the CAD side On the CAD side, integrate similar Websocket client code according to the CAD software used and the development environment it supports. For example, if AutoCAD is used for secondary development, you can use .NET language (such as C#) to implement Websocket connection. Code example: ClientWebSocket socket = new ClientWebSocket(); await socket.ConnectAsync(new Uri("ws: / / localhost:8080"),CancellationToken.None); Console.WriteLine("Websocket connection established"); / / Receive and process data pushed by the server awaitReceiveMessages(socket); string message = Encoding.UTF8.GetString(buffer, 0, result.Count); / / Process the received data, such as updating the annotation display in CAD When the client (web or cad) successfully connects to the Websocket server, onopen (in JavaScript) or the related logic after ConnectAsync succeeds (in C#) will be triggered. In these event handling functions, some initialization operations can be performed, such as sending a connection success message to the server, initializing the interface display, etc. If the connection fails, the corresponding error handling function will be triggered. In JavaScript, the connection error can be handled by the onerror event handling function. In C#, the connection failure can be handled by capturing the exception that the ConnectAsync method may throw. On the web side, when the reviewer creates a comment, the comment information (including the comment location, content, type, etc.) is encapsulated into a suitable data format (such as JSON), and then sent to the server through the Websocket connection. After the server receives the comment information sent by the web side, it traverses all connected cad clients and transmits the comment information (including the comment location coordinates, comment content, comment type, etc.) to the cad side in real time. Let the web side comment information be , the cad side annotation information is , the data synchronization formula is: , ensuring that designers can see review comments in real time on the CAD side.

[0023] S4. Construct a prediction model to optimize the review process, including constructing a multi-layer perceptron neural network to predict the type of problematic areas in the drawing file; Collect historical review records, including drawing review data of different types of projects (such as residential buildings, commercial complexes, industrial plants, etc.) and different professions (architecture, structure, water, heating, electricity, etc.). The collected data types include reviewers’ annotation information (such as annotation location, annotation content, problem type, etc.), basic attributes of drawings (such as project name, drawing number, design stage, drawing format, etc.), relevant parameters of the project (such as building scale, number of floors, structural type, etc.) and review results (such as passed, passed after modification, failed, etc.). Clean the collected duplicate data, supplement and repair the missing data through mean filling, and finally use TF-IDF (term frequency-inverse document frequency) method to convert text into vector representation to complete data preprocessing.

[0024] For the multi-layer perceptron neural network, the input layer receives the pre-processed review-related data features (such as various attributes of drawings, historical review information, etc.). Let the input vector be , where n is the number of features. Each neuron in the input layer directly passes the input data to the hidden layer. The hidden layer contains multiple neuron nodes. Suppose there are neurons in the hidden layer. The input of the hidden layer neuron is the weighted sum of the output of the input layer neurons plus the bias. The formula is: , in, Represented as the weight from the input layer to the first hidden layer, Denotes the bias of the first hidden layer, Represented as the input value of the jth neuron in the first hidden layer, It is represented as the output value of the i-th neuron in the input layer, n is represented as the number of neurons in the input layer, and then through the ReLU function , perform nonlinear transformation on the input and obtain the output of the hidden layer neurons: , in, It is represented as the output value of the jth neuron in the first hidden layer. For multiple hidden layers, the calculation method of subsequent hidden layers is similar. Taking the second hidden layer as an example, its input is the weighted sum of the outputs of the neurons in the previous hidden layer plus the bias. , in, is the weight from the first hidden layer to the second hidden layer, is the bias of the second hidden layer, is the output value of the jth neuron in the second hidden layer, and m is the number of neurons in the first layer. Then the output of the neurons in the second hidden layer is obtained through the activation function. And so on, until the last hidden layer is calculated.

[0025] Then from the hidden layer to the output layer, the softmax function is used for multi-classification output. Assume that the output layer has k categories and the output vector , the input of the output layer neuron is: , in, Represents the input value of the i-th neuron in the output layer, It is represented as the weight connecting the jth neuron in the last hidden layer and the ith neuron in the output layer. Represented as the bias of the i-th neuron in the output layer, It is represented as the output value of the jth neuron in the last hidden layer after being processed by the activation function, and then multi-classification output is performed through the softmax function. This allows the output of the model to be interpreted as a probability distribution, which meets the requirements of multi-classification problems and facilitates subsequent classification decisions based on the probability size, such as selecting the category with the largest probability as the prediction result. In the drawing review, based on the output category probability, it can be determined which type the problem area in the drawing is most likely to belong to, assisting reviewers in making decisions and further inspections.

[0026] S5. Perform prediction model training based on the constructed prediction model, including updating the weights and biases of the prediction model using a back propagation algorithm; The collected historical review data is divided into training set, validation set and test set according to a certain ratio, with the ratio of 70% training set, 20% validation set and 10% test set. The true label of the training sample is , using one-hot encoding, for example, for a three-class problem, if the sample belongs to the second category, then , the predicted output is , then the cross entropy loss function is: , Among them, k represents the number of categories, is represented as the true label, Expressed as the predicted output, the loss function is derived from the cross entropy loss function to obtain the input of the output layer neuron The gradient formula is: , in, is represented as the true label, It is expressed as the predicted output, and then the hidden layer gradient is calculated, and the gradient of the loss function to its weight and bias is calculated. The gradient of the weight and bias of each hidden layer is calculated from the back to the front, until the gradient of the weight and bias between the input layer and the first hidden layer is calculated. According to the calculated gradient, the gradient descent algorithm is used to update the weight. The weight update formula is: , in, and They are represented as the old and updated values ​​of the weights, respectively. Expressed as the learning rate, it controls the step size of each update. Expressed as loss function L versus weight The gradient of the bias is updated in a similar way to the weight. The bias update formula is: , in, and are respectively represented as the old and new values ​​of the bias, Expressed as loss function L versus weight The model can gradually learn the patterns and rules in the data by continuously updating the weights and biases according to the gradients calculated by back propagation, thereby improving the accuracy of prediction of the problem areas in the review. After each update of the weights and biases, the model will re-perform the forward propagation calculation and loss function evaluation, and repeat this process until the preset training stop condition is met. The above prediction model is used to predict the category of the problem area, and the problem area is classified into different types such as building structure problems, water supply and drainage problems, and electrical problems. For example, when reviewing a building drawing, it is predicted that a certain problem area belongs to insufficient structural strength (category A), unreasonable pipeline layout (category B) or incorrect wire laying (category C). For the above three types of problems, if the prediction result is insufficient structural strength, the output may be (1, 0, 0); if it is unreasonable pipeline layout, it is (0, 1, 0); if it is incorrect wire laying, it is (0, 0, 1), which helps to accurately locate the review work and improve the review speed.

[0027] S6. Store the reviewed drawing files, including using a combination of incremental backup and full backup.

[0028] The data after the review is completed is stored and backed up, where a full backup refers to a complete backup of all data, including all drawing files, various project-related information, and all records in the review process, etc., while an incremental backup only backs up data that has changed since the last backup. Compared with a full backup, an incremental backup has the advantages of occupying less storage space and faster backup speed. During the drawing review process, if only a few drawings have been modified or only some review records have been updated, the incremental backup will only back up these modified and updated data, rather than the entire data set. This embodiment uses a combination of incremental backup and full backup for storage, including using a distributed file system to store drawing files, and using a relational database to store and back up project information and review records, and then using a combination of incremental backup and full backup for regular data storage, where the storage strategy formula is: , in, Expressed as the total backup data volume, Indicates the total amount of data when the first full backup is performed. Indicates the amount of data for each incremental backup.

[0029] Example 2 The difference between this embodiment and embodiment 1 is that this embodiment provides a review system for a mobile review digital platform, including: The data acquisition module is configured to acquire project establishment data, including acquiring project information, configuring professional types, and setting personnel roles; The drawing drawing module is configured to draw drawings based on the acquired project establishment data and perform manual lightweight operations on the drawn drawing files; The synchronization module is configured to perform review operations according to the drawn drawing files, including sending the drawing files after the lightweight operation to the review area for real-time data synchronization between the web end and the CAD end, and transmitting the annotation information to the CAD end in real time through the real-time communication channel established by the Websocket method; The model module is configured to construct a prediction model to optimize the review operation process, including constructing a multi-layer perceptron neural network to predict the type of areas with problems in the drawing file; The training module is configured to perform prediction model training based on the constructed prediction model, including updating the weights and biases of the prediction model using a back propagation algorithm; The storage module is configured to store the data of the drawing files that have been reviewed, including storage in a combination of incremental backup and full backup.

[0030] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device. The invention discloses a review method for a mobile review digital platform.

[0031] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executed by the review method of a mobile review digital platform.

[0032] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A review method for a mobile review digital platform, characterized in that: include: Obtain project establishment data, including obtaining project information, configuring professional types, and setting personnel roles; Draw drawings based on the acquired project approval data, and perform manual lightweight operations on the drawn drawing files; Perform review operations based on the drawn drawing files, including sending the lightweight drawing files to the review area for real-time data synchronization between the web end and the CAD end, and transmitting the annotation information to the CAD end in real time through the real-time communication channel established by the Websocket method; Constructing prediction models to optimize the review process, including constructing a multi-layer perceptron neural network to predict the types of problematic areas in the drawing files; Perform prediction model training based on the constructed prediction model, including updating the weights and biases of the prediction model using the back propagation algorithm; The reviewed drawing files are stored in a data format, including a combination of incremental backup and full backup.

2. The review method of a mobile review digital platform according to claim 1, characterized in that: The manual lightweight operation of the drawn drawing file includes traversing the graphic entities and data blocks in the drawing file using a breadth-first search algorithm, identifying key graphic elements and attribute information, simplifying and compressing the extracted information, then removing redundant data, and reorganizing and storing the processed files, wherein the graphic entity simplification rate formula is: ， in, Expressed as the graphic entity simplification rate, Represents the number of graphic entities in the original file. Expressed as the number of optimized graphic entities.

3. The review method of a mobile review digital platform according to claim 1, characterized in that: The real-time communication channel established by the Websocket method includes integrating Websocket client codes on the web side and the cad side respectively, establishing a connection through a specified server address, triggering corresponding processing logic when the connection is successful, receiving a message, and closing the connection respectively, and the web side encapsulates the proofreading and annotation information into a specific format and sends it to the server through the Websocket connection, and forwards it to all connected cad sides through the server.

4. The review method of a mobile review digital platform according to claim 1, characterized in that: The method of constructing a multilayer perceptron neural network to predict problematic areas includes constructing an input layer, multiple hidden layers, and an output layer, wherein the input layer receives preprocessed historical proofreading data features, the hidden layer performs a nonlinear transformation on the input data through an activation function, and the output layer outputs the type of the predicted problematic area.

5. The review method of a mobile review digital platform according to claim 1, characterized in that: The hidden layer includes multiple neuron nodes. The input of the hidden layer neuron is the weighted sum of the output of the input layer neuron plus the bias. The input formula of the neuron is: , in, Represented as the weight from the input layer to the first hidden layer, Denotes the bias of the first hidden layer, Represented as the input value of the jth neuron in the first hidden layer, Represented as the output value of the i-th neuron in the input layer.

6. The review method of a mobile review digital platform according to claim 1, characterized in that: The method uses the back propagation algorithm to update the weights and biases of the prediction model, including calculating the difference between the predicted output and the true label according to the cross entropy loss function, using the cross entropy loss function to derive the gradient of the loss function to the input of the output layer neuron, and using the gradient descent algorithm to update the weights and biases according to the calculated gradient.

7. The review method of a mobile review digital platform according to claim 1, characterized in that: The storage method of combining incremental backup and full backup includes storing drawing files in a distributed file system, storing and backing up project information and review records in a relational database, and then storing data regularly in a combination of incremental backup and full backup. The storage strategy formula is: , in, Expressed as the total backup data volume, Indicates the total amount of data when the first full backup is performed. Indicates the amount of data for each incremental backup.

8. A proofreading system for a mobile proofreading digital platform, executing the method according to claim 1, characterized in that: include: The data acquisition module is configured to acquire project establishment data, including acquiring project information, configuring professional types, and setting personnel roles; The drawing drawing module is configured to draw drawings based on the acquired project establishment data and perform manual lightweight operations on the drawn drawing files; The synchronization module is configured to perform review operations according to the drawn drawing files, including sending the drawing files after the lightweight operation to the review area for real-time data synchronization between the web end and the CAD end, and transmitting the annotation information to the CAD end in real time through the real-time communication channel established by the Websocket method; The model module is configured to construct a prediction model to optimize the review operation process, including constructing a multi-layer perceptron neural network to predict the type of areas with problems in the drawing file; The training module is configured to perform prediction model training based on the constructed prediction model, including updating the weights and biases of the prediction model using a back propagation algorithm; The storage module is configured to store the data of the drawing files that have been reviewed, including storage in a combination of incremental backup and full backup.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executed by a review method for a mobile review digital platform as claimed in claim 1.

10. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executed by a review method for a mobile review digital platform as described in claim 1.

Images