Nuclear power plant digitized human-machine interface intelligent verification and confirmation system and scheduling method

Abstract

The nuclear power plant digital human-computer interface intelligent verification and confirmation system and scheduling method relate to the technical field of human-computer interface verification, and the system comprises a nuclear power plant human-computer interface intelligent verification and confirmation overall intelligent framework structure, a subjective evaluation and analysis subsystem architecture and a scheduling method, a V&V reasoning intelligent subsystem function overall structure and a scheduling method, and a picture sub-element or function module image recognition and decomposition subsystem overall structure and a scheduling method. By constructing the nuclear power plant digital human-computer interface intelligent verification and confirmation system, intelligent verification and confirmation are realized by using digital technology, the effectiveness of the human-computer interface can be quickly and accurately evaluated and confirmed, the efficiency of verification and confirmation is greatly improved, and the statistical verification result is simplified, so that the method has higher practicability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of human-computer interface verification, and particularly relates to a nuclear power plant digital human-computer interface intelligent verification and confirmation system and a scheduling method. BACKGROUND

[0002] The human-computer interface of a nuclear power plant is an important place for an operator to interact with the power plant, and is a main carrier for the operator to obtain information, parameters and complete related tasks. Various parameters, regulations and alarms are presented on the human-computer interface, and the operator performs cognitive activities such as diagnosis, decision-making and execution according to the information. With the increasing maturity of computer technology, control and information technology, and the improvement of automation, nuclear power plants gradually adopt digital control systems instead of traditional analog control systems. The introduction and development of human-computer interfaces based on digital technology may improve the reliability of the system and the performance of personnel, but the introduction of digital technology changes the role of the operator and the situational environment, which may generate new human factors problems. Therefore, before the digital human-computer interface is truly used in a nuclear power plant, an effective method should be used to verify and confirm (V&V) the human-computer interface.

[0003] If the designer does not understand the adverse effects of the characteristics of the digital human-computer interface on the operator, although the designer may follow or consider human factors engineering (HFE) design guidelines in the design process of the human-computer interface, the designer may ignore some deeper safety problems or ignore the behavior characteristics of the operator in complex human-computer interaction, so it is difficult to ensure that the designed human-computer interface can meet the operation requirements of the power plant. Moreover, so far, China has not developed a mature method and tool to evaluate the effectiveness of the human-computer interface, so that the design is more suitable for the cognitive behavior characteristics of the operators in China, and in particular, the tool for intelligent verification and confirmation of the human-computer interface is still in the exploratory stage. Therefore, at the present stage, the verification and confirmation of the human-computer interface of the nuclear power plant still adopts manual methods, such as written evaluation method, static simulation simulation method and operator interview method. Obviously, the manual verification and confirmation is very time-consuming, and the statistical complexity of the verification results is low and the efficiency is low. SUMMARY

[0004] One of the purposes of the present application is to provide a nuclear power plant digital human-computer interface intelligent verification and confirmation system to improve the verification efficiency and reduce the statistical complexity of the verification results.

[0005] In order to solve the above technical problems, the present application adopts the following technical solutions: a nuclear power plant digitized human-machine interface intelligent verification and confirmation method, comprising an interface layer for function scheduling, picture import and export, picture intelligent recognition and decomposition, procedure import and export, data and parameter input and output, verification analysis; an intelligent agent system function layer comprising a plurality of systems for evaluation, analysis, recognition, decomposition, summarization and reasoning; a data layer comprising a V&V qualitative evaluation index database, a V&V standard database, a V&V knowledge base, a picture decomposition element library, a task picture and procedure library; and a hardware layer comprising a client, a switch, a router and a server.

[0006] Preferably, the intelligent agent system function layer comprises a V&V subjective evaluation and analysis subsystem, a picture sub-element and function module image recognition and decomposition subsystem, a V&V result export and analysis and summarization sub-module and a V&V reasoning intelligent subsystem.

[0007] More preferably, the V&V subjective evaluation and analysis subsystem comprises three verification modes: HSI task support verification, HFE design verification and comprehensive system verification.

[0008] More preferably, the HSI task support verification, the HFE design verification and the comprehensive system verification are subjected to performance evaluation through a performance evaluation subsystem, the HSI task support verification performance evaluation involves task analysis range definition, detailed description of the actions that should be taken by personnel, detailed definition of alarms, information control, team cooperation and communication, the HFE design verification performance evaluation involves HSI input, HSI design concept, HFE design guide, HSI detailed design and integration, degraded I&C and HIS conditions, HIS testing and evaluation, and the comprehensive system verification performance evaluation involves power plant performance, individual task performance, state awareness, cognitive workload, interaction reliability, team, test bench and power plant personnel.

[0009] More preferably, the V&V reasoning intelligent subsystem comprises a presentation layer for using an efficient and feasible presentation mode or format for different types of features during feature extraction of imported pictures; a reasoning layer for searching and matching corresponding paradigms based on the knowledge base according to the type, feature vector and format of the extracted features, to automatically generate performance analysis conclusions and measures; a knowledge base for pre-established feature specifications and standards of different functional elements in all pictures under different verification types and performance requirements; and a result and suggestion data layer for saving measure information and rule information after picture evaluation under different types.

[0010] More preferably, in the V&V reasoning intelligent subsystem, verification parameter value input or import is adopted for comprehensive system verification and HSI task support verification, and then verification conclusions are obtained through intelligent matching of standard modes.

[0011] More preferably, the picture sub-element and the function module image recognition and decomposition subsystem comprise: a function classification layer for separating different function modules of the picture to facilitate performance analysis of the same function module, and the classifier adopts a convolutional neural network method; a target detection module for completing picture recognition, feature analysis, and picture preprocessing of the same function module; a feature extraction and representation module for completing picture feature extraction and representation in a certain manner to facilitate matching or comparison with a feature library; a feature library for saving standard information of picture features, symbols, vectors, procedures, semantics, and patterns under different verification types; and a result and suggestion data layer for saving measure information and rule information after picture evaluation under different types.

[0012] More preferably, the scheduling method of the V&V subjective evaluation and analysis subsystem comprises the following steps:

[0013] (1) selecting a task to be verified;

[0014] (2) determining a verification category;

[0015] (3) importing a picture corresponding to the task according to the task to be verified, and selecting relatively important pictures for verification according to needs and verification methods;

[0016] (4) selecting an evaluation method;

[0017] (5) according to the verification category, calling out a corresponding index system table from a database, if it is subjective evaluation, letting personnel select a corresponding level for each index in the form of an online table, statistically analyzing each index level, outputting a result, and going to step (7); if the "online score evaluation" method is selected, letting personnel fill in a single index evaluation value corresponding to each index level in an online manner, taking an average value of each index input value, and then calculating to obtain a quantitative evaluation result, and going to step (6);

[0018] (6) outputting the quantitative evaluation result, and backstepping and analyzing the input value of the evaluation process index to obtain an index with a relatively low input value;

[0019] (7) matching corresponding design measures and suggestions in the database according to the index with a poor level obtained by the "subjective evaluation" method or the index with a relatively low input value obtained by the "online score evaluation" method, and outputting the suggestions;

[0020] (8) completing verification.

[0021] More preferably, the scheduling method of the V&V reasoning intelligent subsystem comprises the following steps:

[0022] (1) According to the user request, select a certain task, if it is HFE design verification, import a picture under a certain task, and then execute step (2); if it is integrated system verification or HSI task support verification, import or input parameter level value, and execute step (3);

[0023] (2) Select the picture to be verified and confirmed in turn, identify and decompose the picture, type classification, knowledge representation and feature extraction, and group features or patterns according to verification categories;

[0024] (3) Based on different verification types, select one or more reasoning methods according to functional features;

[0025] (4) Perform matching degree and similarity analysis according to the extracted features or patterns or the corresponding criteria or standards of the knowledge base;

[0026] (5) According to the knowledge base standard and similarity matching result, query the corresponding rule set from the result and suggestion data layer, and output the deficiencies and suggestions obtained by reasoning;

[0027] (6) If there are still tasks to be verified, jump to step (1), otherwise, the program ends.

[0028] More preferably, the scheduling method of the picture sub-element and the functional module image recognition and decomposition subsystem comprises the following steps:

[0029] (1) Select the task or scene to be verified and confirmed, and select the picture to be verified according to the selected task or scene;

[0030] (2) Preprocess the image based on the background removal, grayscale transformation, image denoising, and binary processing functions in image processing;

[0031] (3) Extract features based on functional modules, and save the features as expected output features;

[0032] (4) Train the picture image based on the convolutional neural network, and construct a functional block recognition model;

[0033] (5) Analyze and extract the features of each functional block, compare and analyze the semantic features in a certain functional block of a certain picture based on the semantic library, and compare the differences with the semantic knowledge base;

[0034] (6) Analyze the features of each functional block based on the knowledge base, obtain the semantic analysis result based on the performance feature set, and go to step (15);

[0035] (7) Compare and analyze the gradient features in a certain functional block of a certain picture based on the gradient library, and compare the differences with the gradient knowledge base;

[0036] (8) Based on the knowledge base, the characteristics are analyzed, the results after gradient analysis are obtained based on the performance characteristic base, and step (15) is transferred;

[0037] (9) Based on the knowledge base, the knowledge characteristics in a function block in a certain picture are analyzed and compared, and the differences with the knowledge base are compared;

[0038] (10) Based on the knowledge base, the characteristics are analyzed, the results after geometric analysis are obtained based on the performance characteristic set, and step (15) is transferred;

[0039] (11) Based on the rule base, the rule characteristics in a function block in a certain picture are analyzed and compared, and the differences with the rule base are compared;

[0040] (12) Based on the rule base, the characteristics are analyzed, the results after rule analysis are obtained based on the performance characteristic set, and step (15) is transferred;

[0041] (13) Based on the mode base, the mode characteristics in a function block in a certain picture are analyzed and compared, and the differences with the mode base are compared;

[0042] (14) Based on the mode base, the characteristics are analyzed, the results after mode analysis are obtained based on the performance characteristic set;

[0043] (15) Based on the performance standard characteristic base, the deficiencies of each function module are output;

[0044] (16) Based on the suggestion characteristic base, the design improvement suggestions are output.

[0045] Compared with the prior art, the nuclear power plant digital human-machine interface intelligent verification and confirmation system is constructed, intelligent verification and confirmation is realized by using digital technology, the effectiveness of the human-machine interface can be quickly and accurately evaluated and confirmed, thereby the efficiency of verification and confirmation is greatly improved, the verification result statistics are simplified, and the application has higher practicability. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 It is a schematic diagram of the overall architecture of the application;

[0047] Figure 2 It is a schematic diagram of the architecture of the V&V subjective evaluation and analysis subsystem in the application;

[0048] Figure 3 It is a schematic diagram of the architecture of the V&V reasoning intelligent subsystem in the application;

[0049] Figure 4 It is a schematic diagram of the architecture of the picture sub-element and function module image recognition and decomposition subsystem in the application. DETAILED DESCRIPTION

[0050] For the convenience of those skilled in the art to understand, the present application is further illustrated below in conjunction with the embodiments and the accompanying drawings, the content mentioned in the embodiments is not a limitation on the present application.

[0051] The overall framework of the nuclear power plant digitized human-machine interface intelligent verification and confirmation system is task-oriented, realizes the import, export, selection and verification of the picture and procedure under the corresponding task, and the main function modules include verification process, each sub-function design, element library, knowledge base, standard library, etc. The specific intelligent function architecture and relationship are as shown in Figure 1 , which involves the fields and technologies of function implementation, algorithm, image recognition, knowledge base, inference engine, mathematical analysis model, checking and matching method, database, etc.

[0052] The nuclear power plant digitized human-machine interface intelligent verification and confirmation system mainly includes: an interface layer for function scheduling, picture import and export, picture intelligent recognition and decomposition, procedure import and export, data and parameter input and output, verification analysis; an intelligent agent system function layer including multiple systems for evaluation, analysis, recognition, decomposition, summarization and reasoning; a data layer including a V&V qualitative evaluation index database, a V&V standard database, a V&V knowledge base, a picture decomposition element library, a task picture and procedure library; and a hardware layer including a client, a switch, a router and a server.

[0053] In the above system, the intelligent agent system function layer includes a V&V subjective evaluation and analysis subsystem, a picture sub-element and function module image recognition and decomposition subsystem, a V&V result export and analysis and summarization submodule, and a V&V reasoning intelligent subsystem. The V&V subjective evaluation table is a verification system composed of an index system, an evaluation level, a suggestion and a measure based on human-machine interface confirmation and verification. Based on different verification types, HIS (Human-System Interface) task support verification, HFE (Human Factor Engineering) design verification, and comprehensive system verification, there are different index systems, so the index systems of different types of verification are different. At present, the verification standard mainly refers to the NUREG-0711 standard of the United States. In the face of different verification categories, verification requirements, and multiple verification tables with different functions, the scheduling, allocation and result analysis automation program and implementation framework of the V&V evaluation table need to be designed. Based on the V&V process and combined with modern technology, the function structure and implementation process of the V&V subjective evaluation table scheduling and analysis subsystem are as shown in Figure 2 .

[0054] As can be seen from Figure 2 , the picture performance manual evaluation system is based on three verification methods, and the evaluation range, evaluation index and measures and suggestions set after evaluation should be constructed respectively. For Figure 2The functions involved, in the running process, the scheduling method given is:

[0055] (1) Select the task to be verified;

[0056] (2) Determine the verification category;

[0057] (3) According to the task to be verified, import the corresponding screen of the task, according to the need and verification method, select the relatively important screen for verification;

[0058] (4) Select the evaluation method;

[0059] (5) According to the verification category, call out the corresponding index system table from the database, if it is subjective evaluation, let the personnel select the corresponding level of each index in the form of online table, according to the level of each index to carry out statistical analysis, output the result, go to step (7); If the "online score evaluation" method is selected, let the personnel fill in the single index evaluation value corresponding to each index level in an online manner, take the average value of each index input value, and then calculate according to the evaluation method to obtain the quantitative evaluation result, go to step (6);

[0060] (6) Output the quantitative evaluation result, and back-propagate and analyze the input value of the evaluation process index to obtain the index with lower input value;

[0061] (7) According to the index with poor level obtained by the "subjective evaluation" method, or the index with lower input value obtained by the "online score evaluation" method, match the corresponding design measures and suggestions in the database, and output the suggestions;

[0062] (8) Complete the verification.

[0063] For the V&V reasoning intelligent subsystem, it can realize intelligent verification and confirmation of the performance covered by the three verification methods based on the verification criteria for the input parameter value, the identified and decomposed screen, and generate conclusions and suggestions. The function architecture of the V&V reasoning intelligent subsystem is as shown in Figure 3 .

[0064] In Figure 3In the middle, because the objects of the integrated system verification and HSI task support verification are different, image recognition and segmentation are rarely involved, so the verification parameter value input or import method is adopted, and then the verification conclusion is obtained through intelligent matching of the standard mode. When the imported picture is extracted, the features of different types of functions are represented in an efficient and feasible manner or format. The reasoning layer extracts the features and, according to their types, feature vectors and formats, searches and matches their corresponding paradigms based on the knowledge base to automatically generate performance analysis conclusions and measures. The knowledge base is a pre-established feature specification and standard for different functional elements in all pictures under different verification types and performance requirements, and its representation method should be consistent with the format of the features extracted after image recognition or decomposition.

[0065] At the same time, the important functions of the V&V reasoning intelligent subsystem are embodied in the technologies and functions involved in the representation layer, the reasoning layer and the knowledge base. The implementation of these technologies and functions is described in a series of other patents of V&V. This patent mainly describes the overall architecture and function design and scheduling of V&V. In the Figure 3 In the middle, when the functions and technologies of each layer are developed, the overall structure of the V&V reasoning intelligent subsystem is responsible for overall coordination, operation and scheduling to ensure that the reasoning intelligent system completes the corresponding requirements. To realize Figure 3 In the middle, the coordination of each layer function, parameter transmission and calling of corresponding functions are provided with a scheduling method of the V&V reasoning intelligent subsystem, which is responsible for the overall coordination, response and calling of the overall scheduling function of the module function, which includes the following steps:

[0066] (1) According to the user's request, select a certain task. If it is HFE design verification, import the picture under a certain task, and then execute step (2); if it is integrated system verification or HSI task support verification, import or input the parameter level value, and execute step (3);

[0067] (2) Select the picture to be verified and confirmed in turn, recognize and decompose the picture, classify the type, represent the knowledge and extract the features, and group the features or patterns according to the verification category;

[0068] (3) Based on different verification types, select one or more reasoning methods according to the function features;

[0069] (4) The extracted features or patterns, or the matching degree according to the parameter situation and the corresponding criteria or standards of the knowledge base, are analyzed for similarity;

[0070] (5) According to the knowledge base standard and the similarity matching result, the corresponding rule set is queried from the result and suggestion data layer, and the deficiencies and suggestions obtained by reasoning are output;

[0071] (6) If there are still tasks to be verified, jump to step (1), otherwise, the program ends.

[0072] For the imported picture, before analyzing its performance, the different functional modules in the picture should be identified, feature extracted, classified, pattern or rule generated, etc. so that the function and procedure library of the module function decomposed from the picture can be compared for performance analysis. For the purpose of performance analysis, the overall structure of the picture sub-element and functional module image recognition and decomposition subsystem is as shown in Figure 4

[0073] In Figure 4 , the function classification layer is used to separate different functional modules of the picture to facilitate performance analysis of the same functional module. The classifier mainly uses the convolutional neural network method. The target detection module completes the picture recognition, feature analysis and picture preprocessing of the same functional module. The feature extraction and representation are used to complete the extraction and representation of the picture features in a certain way, which is convenient for matching or comparing with the feature library. The feature library mainly saves various picture features, symbols, vectors, procedures, semantics, patterns, etc. under different verification types. The representation method of the picture features in the standard library should be consistent with the features extracted from the evaluated picture. The establishment of the standard library requires a large number of standard element libraries, and the accuracy determines the accuracy of the picture performance intelligent analysis, so the establishment of the standard library is very important. The suggestion data layer mainly saves the measure information and rule information of the picture evaluation under different types, which provides background support for the conclusion output.

[0074] Figure 4 In, various functions are an organic whole, which work together and synchronously. When each layer function and technology is developed, the overall function of V&V picture intelligent recognition and analysis is responsible for overall coordination, operation and scheduling to ensure that the reasoning intelligent system completes the corresponding demand. Then, in order to realize the coordination of each layer function in Figure 4 “HFE design verification”, the parameter transmission and calling of the corresponding function, an overall scheduling method is needed to be responsible for the overall coordination and calling of each module function. The specific steps are as follows:

[0075] (1) Select the task or scene to be verified and confirmed, and select the picture to be verified according to the selected task or scene;

[0076] (2) Preprocess the image based on the background removal, grayscale transformation, image denoising and binaryzation processing in image processing;

[0077] (3) Extract features based on functional modules, and save the features as expected output features;

[0078] (4) Train the picture image based on the convolutional neural network to build a functional block recognition model;

[0079] (5) Analyze the characteristics of each function block and extract, compare and analyze the semantic characteristics in a certain function block in a certain picture based on the semantic library, and compare the differences with the semantic knowledge base;

[0080] (6) Analyze the characteristics of each function block based on the knowledge base, obtain the semantic analysis result based on the performance characteristic set, and go to step (15);

[0081] (7) Compare and analyze the gradient characteristics in a certain function block in a certain picture based on the gradient library, and compare the differences with the gradient knowledge base;

[0082] (8) Analyze its characteristics based on the knowledge base, obtain the gradient analysis result based on the performance characteristic library, and go to step (15);

[0083] (9) Compare and analyze the knowledge characteristics in a certain function block in a certain picture based on the knowledge base, and compare the differences with the knowledge base;

[0084] (10) Analyze its characteristics based on the knowledge base, obtain the geometric analysis result based on the performance characteristic set, and go to step (15);

[0085] (11) Compare and analyze the rule characteristics in a certain function block in a certain picture based on the rule library, and compare the differences with the rule library;

[0086] (12) Analyze its characteristics based on the rule library, obtain the rule analysis result based on the performance characteristic set, and go to step (15);

[0087] (13) Compare and analyze the mode characteristics in a certain function block in a certain picture based on the mode library, and compare the differences with the mode library;

[0088] (14) Analyze its characteristics based on the mode library, obtain the mode analysis result based on the performance characteristic set;

[0089] (15) Output the deficiencies of each function module based on the performance standard characteristic library;

[0090] (16) Output the design improvement suggestions based on the suggestion characteristic library.

[0091] The nuclear power plant digital human-machine interface intelligent verification and confirmation system and scheduling method provided by the application can achieve the following beneficial effects:

[0092] 1. Improve the efficiency of nuclear power plant human-machine interface verification and confirmation: the traditional manual verification and confirmation method is time-consuming and inefficient, and the application can quickly and accurately evaluate and confirm the effectiveness of the human-machine interface by constructing a nuclear power plant human-machine interface intelligent verification and confirmation function system, using digital technology and artificial intelligence algorithms, greatly improving the efficiency of verification and confirmation.

[0093] 2. Improve the design quality of human-machine interface: Through the overall design idea and method provided by the present application, the designer can fully understand the influence of the characteristics of the digital human-machine interface on the operator, consider more deep safety issues and the behavior characteristics of the operator, and thus design a human-machine interface that is more in line with the operation requirements of the nuclear power plant, thereby improving the design quality.

[0094] 3. Reduce the risk of human factors: The introduction of digital technology may bring new human factors, and the intelligent verification and confirmation function system of the present application can evaluate the characteristics of the human-machine interface, find potential safety hazards and human factors, and provide corresponding improvement measures to reduce the risk of human factors and improve the safety and reliability of the nuclear power plant.

[0095] 4. Promote the development of intelligent human-machine interface verification and confirmation method: The intelligent verification and confirmation function system of the present application provides a new idea and method for the development of intelligent human-machine interface verification and confirmation method, which can promote the innovation and application of related technologies and make a positive contribution to the safety and efficiency improvement of the human-machine interface of the nuclear power plant.

[0096] In order to make the ordinary skilled in the art more convenient to understand the improvement of the present application over the prior art, some of the drawings and descriptions of the present application have been simplified, and the above examples are the preferred implementation of the present application, in addition to which the present application can be implemented in other ways, and any obvious substitutions within the scope of the present technical solution concept are within the protection scope of the present application.

Claims

1. A nuclear power plant digitized human-machine interface intelligent verification and confirmation system, characterized in that, The application comprises: an interface layer for function scheduling, picture import and export, picture intelligent identification and decomposition, procedure import and export, data and parameter input and output, verification analysis; an intelligent agent system function layer comprising a plurality of systems for evaluation, analysis, identification, decomposition, summarization, reasoning; a data layer comprising a V&V qualitative evaluation index database, a V&V standard database, a V&V knowledge base, a picture decomposition element library, a task picture and procedure library; a hardware layer comprising a client, a switch, a router, a server; the intelligent agent system function layer comprises a V&V subjective evaluation and analysis subsystem, a picture sub-element and functional module image identification and decomposition subsystem, a V&V result export and analysis and summarization sub-module, a V&V reasoning intelligent subsystem; the V&V reasoning intelligent subsystem comprises: a presentation layer for using an efficient and feasible representation method or format for different types of features during feature extraction of imported pictures; a reasoning layer for searching and matching the corresponding paradigm based on the knowledge base according to the type, feature vector and format of the extracted features, to automatically generate performance analysis conclusions and measures; a knowledge base for pre-established feature specifications and standards of different functional elements in all pictures under different verification types and performance requirements; a result and suggestion data layer for saving measure information and rule information after picture evaluation under different types.

2. The nuclear power plant digitized human-machine interface intelligent verification and confirmation system according to claim 1, characterized in that: The V&V subjective evaluation and analysis subsystem comprises three verification methods: HSI task support verification, HFE design verification and comprehensive system verification.

3. The nuclear power plant digitized human-machine interface intelligent verification and confirmation system according to claim 2, characterized in that: The HSI task support verification, HFE design verification and comprehensive system verification are subjected to performance evaluation through a performance evaluation subsystem, the HSI task support verification performance evaluation involves task analysis range definition, detailed description of the measures to be taken by personnel, detailed definition of alarms, information control, team cooperation and communication; the HFE design verification performance evaluation involves HSI input, HIS design concept, HFE design guide, HIS detailed design and integration, degraded I&C and HIS conditions, HIS testing and evaluation; the comprehensive system verification performance evaluation involves power plant performance, individual task performance, state awareness, cognitive workload, interaction reliability, team, test bench, power plant personnel.

4. The nuclear power plant digitized human-machine interface intelligent verification and confirmation system according to claim 1, characterized in that: In the V&V reasoning intelligent subsystem, for comprehensive system verification and HSI task support verification, verification parameter value input or import is adopted, and then intelligent matching of standard mode is performed to obtain verification conclusions.

5. The nuclear power plant digitized human-machine interface intelligent verification and confirmation system according to claim 3, characterized in that, The picture sub-element and functional module image identification and decomposition subsystem comprises: a function classification layer for separating different functional modules of a picture to facilitate performance analysis of the same functional module, and a classifier adopting a convolutional neural network method; a target detection module for completing picture identification, feature analysis and picture preprocessing of the same functional module; a feature extraction and representation module for completing picture feature extraction and representation in a certain way to facilitate matching or comparison with a feature library; a feature library for saving standard information of picture features, symbols, vectors, procedures, semantics and modes under different verification types; A result and suggestion data layer is used to save the measure information and rule information after the picture evaluation.

6. The nuclear power plant digitized human-machine interface intelligent verification and confirmation system according to claim 3, characterized in that, The scheduling method of the V&V subjective evaluation and analysis subsystem comprises the following steps: (1) selecting a task to be verified; (2) determining a verification category; (3) importing a picture corresponding to the task according to the task to be verified and selecting a relatively important picture for verification according to a need and a verification mode; (4) selecting an evaluation mode; (5) according to the verification category, calling out a corresponding index system table from a database, if it is subjective evaluation, allowing a person to select a corresponding level for each index in an online table form, performing statistical analysis according to the level of each index, outputting a result, and proceeding to step (7); if the "online score evaluation" mode is selected, allowing the person to fill in a single index evaluation value corresponding to the level of each index in an online manner, taking an average value of each index input value, and then performing calculation according to an evaluation method to obtain a quantitative evaluation result, and proceeding to step (6); (6) outputting the quantitative evaluation result, and backstepping and analyzing the input value of the evaluation process index to obtain an index with a relatively low input value; (7) according to the index with a poor level obtained by the "subjective evaluation" mode or the index with a relatively low input value obtained by the "online score evaluation" mode, matching a corresponding design measure and suggestion in the database, and outputting the suggestion; (8) completing verification.

7. The nuclear power plant digitized human-machine interface intelligent verification and confirmation system according to claim 4, characterized in that, The scheduling method of the V&V reasoning intelligent subsystem comprises the following steps: (1) according to a user request, selecting a task, if it is HFE design verification, importing a picture under a task, and then performing step (2); if it is comprehensive system verification or HSI task support verification, importing or inputting a parameter level value, and performing step (3); (2) selecting a picture to be verified and confirmed in turn, recognizing and decomposing the picture, classifying the type, representing knowledge, and extracting features or modes, and grouping features or modes according to a verification category; (3) based on different verification types, selecting one or more reasoning modes according to a functional feature; (4) performing matching degree and similarity analysis according to a corresponding criterion or standard in the knowledge base; (5) according to the knowledge base standard and the similarity matching result, querying a corresponding rule set from a result and suggestion data layer, and outputting an obtained deficiency and suggestion by reasoning; (6) if there is a task to be verified, jumping to step (1), otherwise, the program ends.

8. The nuclear power plant digitized human-machine interface intelligent verification and confirmation system according to claim 4, characterized in that, The scheduling method of the picture sub-element and functional module image recognition and decomposition subsystem comprises the following steps: (1) selecting a task or scene to be verified and confirmed, and selecting a picture to be verified according to the selected task or scene; (2) performing pre-processing on the image based on background removal, grayscale transformation, image denoising, and binarization processing in image processing; (3) extracting features based on functional modules, and saving the features as expected output features; (4) training the picture image based on a convolutional neural network, and constructing a functional block recognition model; (5) analyzing and extracting features of each functional block, comparing and analyzing semantic features in a certain functional block of a certain picture based on a semantic library, and comparing differences with a semantic knowledge base; (6) Analyze the feature of each function block based on the knowledge base, obtain the semantic analysis result based on the performance feature set, and go to step (15); (7) Compare and analyze the gradient feature in a function block of a picture based on the gradient base, and compare the difference with the gradient knowledge base; (8) Analyze the feature based on the knowledge base, obtain the result after gradient analysis based on the performance feature base, and go to step (15); (9) Compare and analyze the knowledge feature in a function block of a picture based on the knowledge base, and compare the difference with the knowledge base; (10) Analyze the feature based on the knowledge base, obtain the result after geometric analysis based on the performance feature set, and go to step (15); (11) Compare and analyze the rule feature in a function block of a picture based on the rule base, and compare the difference with the rule base; (12) Analyze the feature based on the rule base, obtain the result after rule analysis based on the performance feature set, and go to step (15); (13) Compare and analyze the mode feature in a function block of a picture based on the mode base, and compare the difference with the mode base; (14) Analyze the feature based on the mode base, obtain the result after mode analysis based on the performance feature set; (15) Output the deficiency of each function module based on the performance standard feature base; (16) Output the suggestion for design improvement based on the suggestion feature base.

Images