Graphical-based health management algorithm fusion method and system and medium

Abstract

The invention belongs to the technical field of equipment health management, and particularly relates to a graphic-based health management algorithm fusion method, equipment and a medium, which can meet the requirement of low-code development of an algorithm and improve the algorithm development efficiency and the verification test efficiency at the same time. According to the method, the PHM algorithm can be developed on the basis of a graphical mode, meanwhile, a development template is provided, developers can call existing basic algorithm library content and develop professional algorithms on the basis of understanding the development process, meanwhile, the functions of data access, algorithm verification testing, evaluation report generation and the like are provided, and the development efficiency is improved. The requirement for low-code development of the algorithm is met, and the algorithm development efficiency and the verification test efficiency are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of equipment health management, and specifically relates to a graphics-based health management algorithm fusion method, system and medium. Background Art

[0002] With the rapid development of high-end equipment such as aerospace, aviation, high-speed rail, energy, and large facilities in my country, the country has put forward higher requirements for the safety, economy, and intelligence of high-end equipment. In order to meet the above needs of high-end equipment operation and security, health management technology is gradually gaining attention. Health management technology (hereinafter referred to as PHM technology) is an advanced management technology based on big data, which realizes the intelligence and autonomy of equipment status through real-time monitoring, diagnosis, and prediction of equipment through intelligent processing algorithms. It is the core technology to improve the safety and economy of equipment.

[0003] With the rapid development of the PHM field, the research and development of health management-related algorithms has gradually become the key to the application of PHM systems. However, based on traditional text-based and symbolic programming languages, it is difficult for industry experts to develop PHM-related algorithms for target devices. In addition, the compilation and debugging of algorithms are abstract and difficult to understand, which increases the difficulty of use for industry experts. Existing technologies are difficult to meet the development needs of industry experts for PHM algorithms for target devices. Summary of the invention

[0004] In view of the above problems, the present invention proposes a graphical health management algorithm fusion method, system and medium, which can meet the needs of low-code algorithm development, while improving the algorithm development efficiency and verification and testing efficiency.

[0005] To achieve the above purpose, the technical solution of the present invention is as follows:

[0006] The present invention provides a graphical health management algorithm fusion method, comprising the following steps:

[0007] Establishing a task goal and selecting a template for solving the target task, wherein the task template is an implementation goal, including fault diagnosis, state detection analysis, numerical calculation or statistical analysis;

[0008] According to the development step prompts, the PHM algorithm process is built by dragging graphical elements;

[0009] In the canvas, the algorithm model is formed by connecting the algorithms to establish the input and output relationship of the algorithm. There cannot be any algorithm modules independent of the main model in the canvas, and there cannot be multiple models at the same time.

[0010] Configure the parameters of the algorithm function in table form and by writing a Json file;

[0011] After the setup is complete, select the data source, evaluation method, and evaluation report;

[0012] Call the algorithm execution engine to perform algorithm testing and verification and output the evaluation results.

[0013] Among them, when establishing the task objectives, the following related method contents are further included:

[0014] Fault diagnosis includes: preprocessing, feature extraction, fault diagnosis and fault prediction;

[0015] State detection analysis includes: data preprocessing, feature extraction, feature analysis and state recognition;

[0016] Numerical computing includes: function operation, preprocessing and signal processing;

[0017] Statistical analysis includes: data acquisition, data preprocessing, data description analysis, data statistical modeling analysis, indicator evaluation and data synthesis.

[0018] The algorithm function is based on an existing basic function library or is edited and developed, supporting two modes: online editing mode and basic function library combined with source code development.

[0019] The algorithm parameter configuration implements directly passing input parameters to the function and obtaining output parameters, as well as adjusting the position parameters and default parameters of the function.

[0020] Among them, the data source configuration method includes csv or .txt format files, the evaluation method adopts the evaluation index content of the algorithm, including accuracy, precision, recall rate and F1 score, and the evaluation report content outputs charts.

[0021] The algorithm verification is controlled by a distributed scheduling algorithm engine, which supports the synchronous execution of multiple algorithm models, performs indicator evaluation on the algorithm during the execution process, and finally outputs the evaluation results.

[0022] The specific steps of graphical algorithm configuration include:

[0023] Organize algorithm model requirements and clarify model-related components;

[0024] Check whether there is a related model function in the model library. If so, select and drag it from the basic library to the canvas. If not, create a new basic function and edit the source code.

[0025] Connect basic functions and establish their input and output relationships;

[0026] Determine whether the connected function needs to be modified, and edit the source code if necessary;

[0027] Configure the algorithm function parameters;

[0028] Determine whether the entire algorithm is configured. If it is completed, select the data source, add evaluation methods, evaluation chart reports, etc., and end the configuration process.

[0029] The present invention also provides a graphical health management algorithm fusion system for implementing the method of the present invention, the system comprising:

[0030] A graphical interface for users to build algorithm processes through drag-and-drop operations;

[0031] Algorithm library, which stores a variety of basic algorithm modules for users to choose from;

[0032] Parameter configuration tool, which provides the function of configuring the input parameters and output parameters of the algorithm function;

[0033] Data source management module, responsible for selecting and accessing external data sets;

[0034] Evaluation module, used to evaluate the performance of the algorithm model and output the evaluation results;

[0035] Distributed scheduling algorithm engine, used to control the distributed execution and result evaluation of the algorithm.

[0036] The system further comprises one or more algorithm execution engines, which are scheduled by a distributed scheduling algorithm engine module to achieve parallel execution of the algorithm.

[0037] The present invention also provides a computer-readable medium having a computer program stored thereon, and when the computer program is executed, the computer is enabled to execute the method described in the present invention.

[0038] The beneficial effects of the present invention are as follows:

[0039] 1. The method of the present invention can develop PHM algorithms based on a graphical method, and provide development templates at the same time, so that developers can call the existing basic algorithm library content and develop professional algorithms based on understanding the development process. At the same time, it provides functions such as data access, algorithm verification testing, and evaluation report generation, which meets the needs of low-code algorithm development and improves the efficiency of algorithm development and verification testing.

[0040] 2. Through the graphical interface and drag-and-drop operation, the present invention allows users to build complex algorithm models without writing code, greatly reducing the technical threshold for algorithm development and improving development efficiency; it supports online editing mode and basic function library to meet the development needs of personnel at different levels, and both beginners and professional developers can get started quickly; users can flexibly select or create basic functions according to actual needs, and connect algorithms, edit content and configure parameters, making the entire system highly flexible.

[0041] 3. The present invention supports multiple data source formats (such as .csv, .txt, etc.) and multiple evaluation methods (such as accuracy, precision, recall, F1 score, etc.), which can adapt to different application scenarios and changes in demand; provides detailed parameter configuration options, including the management of input parameters and output parameters, and configuration in the form of tables or JSON files, ensuring the consistency and accuracy of algorithm execution; uses confusion matrix as an evaluation tool, which can intuitively show the relationship between the model prediction results and the true category, helping developers to better understand the performance of the model, so as to optimize the algorithm in a targeted manner.

[0042] 4. The method of the present invention adopts a graphical interface. It is easy to understand and communicate, which helps team members to quickly reach a consensus and accelerate the project progress. In addition, the construction of the basic function library and algorithm model library is conducive to the accumulation and reuse of existing algorithm resources, reducing duplication of work and improving overall R&D efficiency.

[0043] 5. The present invention can run multiple rules and algorithms simultaneously through a distributed scheduling algorithm engine, realize parallel processing of tasks, greatly shorten the time cost of algorithm verification, and improve the response speed and processing capacity of the system.

[0044] 6. The present invention forms a complete closed loop from demand collation, model construction to parameter configuration, and finally to evaluation report generation, ensuring quality control at each link and ultimately outputting a high-quality algorithm model. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flowchart of a graphical health management algorithm fusion method according to an embodiment of the present invention.

[0046] Figure 2 It is a graphical algorithm configuration flow chart of an embodiment of the present invention.

[0047] Figure 3 It is a schematic diagram of a graphical algorithm configuration interface according to an embodiment of the present invention.

[0048] Figure 4 It is a schematic diagram of algorithm parameter configuration according to an embodiment of the present invention.

[0049] Figure 5 It is a flowchart of the algorithm engine execution of an embodiment of the present invention.

[0050] Figure 6 Schematic diagram of an evaluation method according to an embodiment of the present invention.

[0051] Figure 7 Schematic diagram of evaluation results of an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0053] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0054] The present invention provides a graphical health management algorithm fusion method. Figure 1 This is a flowchart of a graphical health management algorithm fusion method according to an embodiment of the present invention, comprising the following steps:

[0055] 1) Establish a task goal and select a template to solve the target task;

[0056] Specifically, the task template is an implementation goal, such as fault diagnosis, state detection analysis, numerical calculation or statistical analysis, etc. Through different algorithm templates, different algorithm classifications can be associated.

[0057] Among them, fault diagnosis includes methods such as associative preprocessing, feature extraction, fault diagnosis, and fault prediction; state detection analysis includes data preprocessing, feature extraction, feature analysis, and state recognition; numerical calculation includes functional classifications such as associative function operation, preprocessing, and signal processing, and can respectively realize content analysis of signals such as vibration, pressure, and temperature; statistical analysis includes algorithmic contents such as associative data acquisition, data preprocessing, data description analysis, data statistical modeling analysis, indicator evaluation, and data synthesis;

[0058] 2) Build the PHM algorithm according to the development step prompts, and build the algorithm process based on graphical elements by dragging and dropping;

[0059] In this embodiment, the algorithm process is developed in a graphical mode, and the algorithm is organized by dragging the basic algorithm module in the canvas. During the development process, according to the specific task, the algorithm configuration process can be viewed and selected from the algorithm library.

[0060] 3) During the construction process, you can connect algorithms, edit algorithm content, and configure algorithm parameters, including:

[0061] In the canvas, an algorithm model is formed by connecting algorithms, and the input and output relationship of the algorithm is established, so as to determine the input source and output target of the algorithm, and reflect the complex execution relationship of different algorithm functions; the canvas cannot have algorithm modules (there can be only one module in the canvas) that are independent of the main model (the model refers to the algorithm composed of multiple modules), nor can there be multiple models at the same time.

[0062] The algorithm can be based on the existing basic function library, or it can be edited and developed. The basic function library is the function library content that has been established in the system. During the editing process, the algorithm source code can be edited through the online editing mode; through the basic function library + source code development, it can meet the development needs of personnel at different levels;

[0063] The configuration of algorithm parameters is to configure the input parameters and output parameters of the algorithm function, which mainly includes two modes: one is to configure in table form, and the other is to configure by writing Json files.

[0064] Specifically, parameter configuration can achieve two functions: first, it can directly pass input parameters to the function and obtain output parameters without the user having to make a separate function call within the function; second, the positional parameters and default parameters of the function are adjustable parameters, which need to be exposed to the front-end user for dynamic adjustment by the user;

[0065] The table format manages input parameters and output parameters. For input parameters, the main attributes include parameter Chinese name, English name, parameter type (list, string, etc.), whether it is required, interaction method (text box, selection box), optional values, default values, etc.; output parameter attributes include parameter Chinese name, English name, parameter type, interaction method, parameter description, etc.

[0066] The Json file provides the definition of input parameters and output parameters in the Json mode. The definition content must conform to the Json mode. After the configuration is completed, it is saved in the parameter configuration file.

[0067] 4) After the setup is completed, you can select data sources, evaluation methods, and evaluation reports;

[0068] After the configuration of the main algorithm model is completed, you can choose to configure the data source, evaluation method, evaluation result report, etc. Among them, the data source is to select the existing data set in the system, and the data source configuration method can choose .csv or txt format file; the evaluation method can use the evaluation indicators of the algorithm, including accuracy, precision, recall rate, F1 score, etc.; the evaluation report content can be output as a chart, which can be output after configuration.

[0069] 5) Call the algorithm execution engine to perform algorithm testing and verification and output the evaluation results.

[0070] After configuring the algorithm model and verification scheme, the algorithm can be executed. Before execution, the graphical algorithm model is automatically converted into an algorithm model. During the execution process, it is executed through the algorithm engine (an algorithm engine based on distributed computing, which can run multiple rules and algorithms at the same time), extracting data from the data source, calculating the algorithm rules, and evaluating the algorithm indicators. Finally, the evaluation results are output.

[0071] The graphical algorithm configuration process includes:

[0072] 1) Organize the algorithm model requirements and clarify the relevant components of the model;

[0073] 2) According to the model content, find out whether there is a related model function in the model library. If so, select the basic library algorithm function, drag the model from the basic library to the canvas, and go to step 4); if not, go to 3);

[0074] 3) Create a new basic function and edit the source code, then drag the model from the basic library to the canvas;

[0075] 4) Connect the basic functions and establish their input and output relationships;

[0076] 5) Determine whether the connected function needs to be modified. If it needs to be modified, edit the source code and then proceed to step 6); if it does not need to be modified, directly proceed to step 6);

[0077] 6) Configure the algorithm function parameters, including input parameters and output parameters;

[0078] 7) Determine whether the entire algorithm is configured, if so, proceed to step 8), otherwise, proceed to step 2);

[0079] 8) Select the data source for the entire algorithm model, add the evaluation method, evaluation chart report and other contents, and end the entire configuration process. Figure 2 A graphical algorithm configuration flow chart is given in .

[0080] Specifically, in Figure 3A schematic diagram of the graphical algorithm configuration interface is given in Figure 3 The following shows a method of using support vector machine for classification, including data source (loading sklearn dataset), support vector machine method (C support vector machine), dataset segmentation method (sklearn dataset segmentation), model training method (sklearn model training), model prediction method (sklearn model prediction), model analysis result saving (joblib model saving), model evaluation (multi-classification indicator evaluation), evaluation result report (drawing confusion matrix), etc. Figure 4 The algorithm parameter configuration diagram is given in . The algorithm parameter configuration provides configuration of the input parameters and output parameters of the function. The figure shows the input parameters of the data set segmentation, including feature matrix, label, feature name list, training set ratio, label column name, label deletion method, etc. The output parameters are training set data, training set label, and test set data.

[0081] The algorithm verification execution process is as follows Figure 5 As shown, the following steps are included:

[0082] 1) Algorithm verification is controlled by a distributed scheduling algorithm engine, which controls the data source access data content and calls the algorithm execution module for distributed execution (multiple algorithm models can be executed simultaneously). At the same time, the algorithm output result module can evaluate and assess the algorithm results.

[0083] 2) The data source access module can synchronously access multiple data sources according to the configuration of the algorithm model. The data source can be the data source of one algorithm model or multiple algorithm models. The data source access is used to extract data and pass the data to the algorithm.

[0084] 3) The algorithm execution module may include multiple algorithm execution engines, which are scheduled by the distributed scheduling algorithm engine module. The main execution process is as follows:

[0085] (1) For an algorithm model, traverse and search for the first function of the function module (multiple functions may be found depending on the data source);

[0086] (2) Function scheduling and execution are performed through the execution engine, and multiple functions can be executed in parallel;

[0087] (3) Determine whether the function has been executed. If yes, proceed to step (4); otherwise, wait for step (3) to be executed.

[0088] (4) Obtain the execution result of the function and mark the function as executed;

[0089] (5) Find the next function execution node (multiple functions) through the function output relationship. If the function execution has been completed, go to step 6); if the function execution has not been completed, go to step (7);

[0090] (6) Determine whether all functions have been executed. If not, execute step (5); if yes, output the execution result;

[0091] (7) If the function execution condition is met (the judgment condition is that the input parameters all have data and can be executed), then go to step (2); otherwise, wait in step (7).

[0092] 4) The algorithm result output module can be used to evaluate and assess the algorithm, and can output the algorithm evaluation results. It can perform a comprehensive evaluation (such as accuracy, precision, recall, etc.) based on the algorithm results and based on the evaluation module, and can perform a performance evaluation (such as confusion matrix) based on the algorithm results and based on the algorithm evaluation module.

[0093] Figure 6 Schematic diagram of the evaluation method of the embodiment of the present invention. The evaluation method provides evaluation indicators of the algorithm model, including accuracy, precision, recall rate, and F1 value. Figure 7 : is a schematic diagram of the evaluation results of an embodiment of the present invention. The evaluation results use a confusion matrix to perform performance analysis. The confusion matrix is ​​one of the important tools for evaluating model performance in classification algorithms. It shows the relationship between the results predicted by the model and the actual categories. By recording the number of true positives (TruePositive), false positives (FalsePositive), true negatives (TrueNegative) and false negatives (FalseNegative) for each category, it helps us understand which categories the model performs better on and which categories are easily misclassified. Figure 7 The confusion matrix in has the following meanings:

[0094] x-axis (Predicted label): represents the category predicted by the model.

[0095] y-axis (True label): represents the true category.

[0096] Diagonal: The diagonal elements of the matrix represent the number of samples correctly classified by the model. The higher the diagonal elements, the higher the prediction accuracy of the model for that class.

[0097] Off-diagonal: Off-diagonal elements represent the number of samples that are misclassified, which can be used to analyze common mistakes made by the model.

[0098] Accuracy: Accuracy is the ratio of the number of samples correctly classified by the model to the total number of samples. Accuracy represents the overall classification ability of the model. The higher the value, the better the accuracy of the model. The current accuracy is 100%.

[0099] Misclass (misclassification rate): The misclassification rate is the ratio of the number of samples misclassified by the model to the total number of samples, which is complementary to Accuracy. The misclassification rate reflects the proportion of model errors. The lower the value, the less misclassification the model has. The misclassification rate is 0%.

[0100] The present invention provides a graphical health management algorithm fusion system for implementing the method of the present invention, including:

[0101] A graphical interface for users to build algorithm processes through drag-and-drop operations;

[0102] Algorithm library, which stores a variety of basic algorithm modules for users to choose from;

[0103] Parameter configuration tool, which provides the function of configuring the input parameters and output parameters of the algorithm function;

[0104] Data source management module, responsible for selecting and accessing external data sets;

[0105] Evaluation module, used to evaluate the performance of the algorithm model and output the evaluation results;

[0106] Distributed scheduling algorithm engine, used to control the distributed execution and result evaluation of the algorithm.

[0107] Furthermore, the system also includes one or more algorithm execution engines, which are scheduled by a distributed scheduling algorithm engine module to achieve parallel execution of the algorithm.

[0108] As another aspect, the present application also provides a computer-readable storage medium, which may be a computer-readable storage medium included in the device described in the above embodiment; or it may be a computer-readable storage medium that exists independently and is not assembled into the device. The computer-readable storage medium may be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD ROM, or any other form of storage medium known in the technical field. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the graphical health management algorithm fusion method described in the present application.

[0109] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below.

Claims

1. A graphical health management algorithm fusion method, characterized in that: The following steps are involved: Establishing a task goal and selecting a template for solving the target task, wherein the task template is an implementation goal, including fault diagnosis, state detection analysis, numerical calculation or statistical analysis; According to the development step prompts, the PHM algorithm process is built by dragging graphical elements; In the canvas, the algorithm model is formed by connecting the algorithms, and the input and output relationship of the algorithm is established. In addition, the canvas cannot have algorithm modules that are independent of the main model, and multiple models cannot exist at the same time. The algorithm function parameters are configured in table form and in Json file writing. After the setup is complete, select the data source, evaluation method, and evaluation report; Call the algorithm execution engine to perform algorithm testing and verification and output the evaluation results.

2. The method according to claim 1, characterized in that: When establishing the task objectives, further include the following related method contents: Fault diagnosis includes: preprocessing, feature extraction, fault diagnosis and fault prediction; State detection analysis includes: data preprocessing, feature extraction, feature analysis and state recognition; Numerical computing includes: function operation, preprocessing and signal processing; Statistical analysis includes: data acquisition, data preprocessing, data description analysis, data statistical modeling analysis, indicator evaluation and data synthesis.

3. The method according to claim 1, characterized in that The algorithm function is based on an existing basic function library or is edited and developed, supporting two modes: online editing mode and basic function library combined with source code development.

4. The method according to any one of claims 1 to 3, characterized in that: The algorithm parameter configuration implements directly passing input parameters to a function and obtaining output parameters, as well as adjusting positional parameters and default parameters of a function.

5. The method according to claim 4, characterized in that The data source configuration method includes csv or .txt format files, the evaluation method adopts the evaluation index content of the algorithm, including accuracy, precision, recall rate and F1 score, and the evaluation report content outputs charts.

6. The method according to claim 5, characterized in that The algorithm verification is controlled by a distributed scheduling algorithm engine, which supports the synchronous execution of multiple algorithm models, evaluates the algorithm indicators during the execution process, and finally outputs the evaluation results.

7. The method according to claim 1, characterized in that The specific steps of graphical algorithm configuration include: sorting out the algorithm model requirements and clarifying the relevant components of the model; Check whether there is a related model function in the model library. If so, select and drag it from the basic library to the canvas. If not, create a new basic function and edit the source code. Connect basic functions and establish their input and output relationships; Determine whether the connected function needs to be modified, and edit the source code if necessary; Configure the algorithm function parameters; Determine whether the entire algorithm is configured. If it is completed, select the data source, add evaluation methods, evaluation chart reports, etc., and end the configuration process.

8. A graphical health management algorithm fusion system, characterized in that: For implementing the method according to any one of claims 1 to 7, the system comprises: A graphical interface for users to build algorithm processes through drag-and-drop operations; Algorithm library, which stores a variety of basic algorithm modules for users to choose from; Parameter configuration tool, which provides the function of configuring the input parameters and output parameters of the algorithm function; Data source management module, responsible for selecting and accessing external data sets; Evaluation module, used to evaluate the performance of the algorithm model and output the evaluation results; Distributed scheduling algorithm engine, used to control the distributed execution and result evaluation of the algorithm.

9. The system according to claim 8, characterized in that The system further comprises one or more algorithm execution engines, which are scheduled by a distributed scheduling algorithm engine module to achieve parallel execution of algorithms.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed, the computer is caused to perform the method according to any one of claims 1 to 7.

Images