A general health management platform for typical devices based on a multi-model environment

Abstract

This invention discloses a general health management platform for typical equipment based on a multi-model environment. As a fully digital platform, it adopts a B / S architecture and consists of a multi-signal flow graph and the AMEsim multi-model environment, a model database, and functional modules for fault injection, fault diagnosis, fault discovery, and status assessment. The two model environments provide typical equipment with different modeling paradigms. The model database stores comprehensive test information and provides data interaction interfaces. The functional modules provide multiple types of health management function nodes. This invention is the core of the health management platform software, achieving the integration of two model environments, the standardization of model paradigms, and the unification of data interfaces. It provides multiple types of health management services, fully utilizes model test information, and is applicable to health management tasks for various types of typical equipment. Furthermore, the platform has advantages such as strong portability and multi-user parallel operation, and can meet a wide range of health management needs.

Description

Technical Field

[0001] This invention relates to typical equipment health management, belonging to the field of system health management, and particularly to a general health management platform for typical equipment based on a multi-model environment. Background Technology

[0002] System Health Management (SHM) technology for typical equipment reduces the time and cost of maintenance through efficient and cost-effective diagnostic and predictive activities. It uses status information of subsystems and components to provide actionable information, thereby enabling intelligent decision-making to improve performance, safety, reliability and maintainability.

[0003] Among numerous system health management technologies, model-based methods have become a crucial technical means for typical equipment health management because they overcome the knowledge bottleneck problem of expert rules and are not limited by the data imbalance problem of data-driven methods. Important examples in the field of model-based typical equipment health management systems include NASA's Livingstone system, a fault diagnosis and recovery engine based on a qualitative model; and the HUGIN Bayesian inference model developed by HUGIN EXPERT. These two models are widely used in model-based health management, but due to their high degree of qualitative analysis, complex modeling theory, and generally limited integration of health management functions, their application in China is relatively limited.

[0004] Domestic research on model-based health management technologies for typical equipment started relatively late, and there has been little focus on the development of general-purpose systems. The main problems include:

[0005] 1) Limited modeling methods. Existing platforms often employ only a single modeling method, making it difficult to comprehensively describe the information of typical equipment objects.

[0006] 2) Missing model information. Model information often only pertains to signals or physical parameters, and the information is not fully matched.

[0007] 3) Low integration of health management services. Existing platforms often only focus on fault reasoning and engine construction, and the integration of fault prediction or status assessment services for typical equipment systems is not high.

[0008] To address the aforementioned issues, this invention provides a general health management platform for typical devices based on a multi-model environment. It integrates two model environments, multi-signal flow graph and AMEsim, consolidating model information and providing various health management services. This improves the utilization rate of model information and enhances the scalability and versatility of the model-based health management platform for typical devices. Summary of the Invention

[0009] The purpose of this invention is to solve the problems of traditional typical equipment health management platform model paradigm being single, lacking or not fully utilizing model information, and low integration of health management services. It proposes a typical equipment general health management platform based on a multi-model environment, which integrates two model environments, multi-signal flow graph and AMEsim, unifies data interfaces, makes full use of model information, and provides multiple types of health management services. The platform is suitable for multi-user, multi-project parallel scenarios and has certain universality, soundness and portability.

[0010] The health management platform is designed and developed using a B / S architecture with Django as the platform framework. Users can complete interactive operations solely through the web interface. Front-end action events trigger HTTP requests to access the server and retrieve backend business logic. As a typical, general-purpose health management platform for devices, it features rapid response, flexible interfaces, user-friendliness, clearly defined functions, high algorithm integration, and good encapsulation. Users only need to focus on business events, and the process executes automatically.

[0011] The health management platform's multi-model environment includes a multi-signal flow graph modeling environment and an AMEsim model environment. The two environments are not ranked. The platform provides a visual multi-signal flow graph modeling environment to construct hierarchical signal flow models for typical devices, and a visual AMEsim model environment to view, edit, and perform simulation analysis of energy flow physical models for typical devices. The two model environments are independent and do not interfere with each other.

[0012] The health management platform's model information database unifies multi-model information, analyzes fault propagation paths and signal cross-linking relationships in multi-signal flow graphs, matches multi-signal flow graph measurement points with corresponding physical parameters in the AMEsim model, records test attributes and initial parameter data, and monitors dynamic simulation events of the physical model, updating physical simulation variable values ​​in real time to provide information sources and data support for health management services. The database uses MySQL or SQLite, and its engine is integrated with the Django framework as the model layer.

[0013] The health management service module of the health management platform includes several parts: testability analysis, fault injection, fault diagnosis, fault discovery, and status assessment. The platform provides specific health management services, fully utilizing model information to comprehensively describe the health status of typical equipment objects. Testability analysis extracts information from the multi-signal flow graph model, using a fault-testability matrix (D-matrix) as its representation, calculating indicators such as fault detection rate and fault isolation rate, and statistically analyzing fuzzy and redundant groups for user reference. Fault injection allows users to edit AMEsim model parameters and perform dynamic simulation, thereby achieving fully digital fault injection for typical equipment objects at the physical level, and provides a 3D model interface for visualizing the fault injection. Fault diagnosis integrates the multi-signal flow graph and the AMEsim model, matches measurement point parameter information, and uses an anomaly inference engine for typical equipment and a general fault diagnosis algorithm to determine the current fault status of the object, presenting the results in a list. The system provides diagnostic results for known fault modes; for fault mining and monitoring, users select associated measurement points, and the system extracts corresponding variable data from the physical model through database model matching information data tables, organizes the data structure, analyzes historical trends, and presents a health score chart of the current equipment's associated measurement points in the form of a graph, and provides configuration nodes for users to add unknown fault information; for status assessment, the platform automatically subscribes to historical fault information in the database, and scores typical equipment subsystems and systems from bottom to top based on the failure rate, and classifies each level into four levels: normal, good, suspected fault, and fault, displaying the probability and rating of fault occurrence for the current subsystem, sub-system, and system.

[0014] The advantages of this invention are:

[0015] 1) The general health management platform for typical devices based on a multi-model environment breaks through the limitations of the single modeling method of traditional health management platforms, solves the problem of missing or insufficient model description information, and enables users to have a more comprehensive understanding of the current health status of the object.

[0016] 2) This invention uses Django as the overall development framework, which has high platform stability, good portability, and is suitable for multi-user, multi-project parallel task scenarios.

[0017] 3) This invention provides a wealth of health management service modules, with built-in algorithms, automatic process execution, high user flexibility, and clear and visible results.

[0018] 4) This invention integrates multi-model information of typical equipment, with a standardized modeling paradigm, unified data interface, and information that can be easily decoupled and fully utilized. Attached Figure Description

[0019] Figure 1 The present invention provides an overall architecture diagram of a typical device general health management platform based on a multi-model environment;

[0020] Figure 2 A schematic diagram of the environment for modeling multi-signal flow graphs;

[0021] Figure 3 This is a schematic diagram of the AMEsim model environment;

[0022] Figure 4 Design diagrams for a database of typical equipment models;

[0023] Figure 5 Hierarchical diagram of functional modules for health management services. Detailed Implementation

[0024] The following detailed description, in conjunction with the accompanying drawings, illustrates the general health management platform for typical devices based on a multi-model environment and its implementation method provided by this invention.

[0025] 1. System Framework

[0026] This invention provides a general health management platform for typical devices based on a multi-model environment and its implementation method. The overall architecture of the health management platform is as follows: Figure 1 As shown, the health management development architecture is a B / S architecture. The platform is divided into a model layer, a data layer, a service layer, and an application layer. The AMEsim model environment in the model layer simulates system operation, and the multi-signal flow graph environment describes system behavior. The data layer receives and integrates the physical data from AMEsim and the test information from the multi-signal flow graph. The service layer encapsulates algorithm modules and handles business logic. The application layer is developed using Django as the framework, MySQL or SQLite as the database engine, and Python as the programming language. It provides user interaction and visual process configuration functions, supports parallel task scenarios, and has complete task views for user creation, permission allocation, and project creation.

[0027] 2. Multi-model environment

[0028] The health management platform's multi-model environment includes two model environments: a multi-signal flow graph model environment and an AMESim model environment. This section describes the specific implementation methods for the health management platform's multi-model environment.

[0029] 2.1 Multi-signal flow graph model environment

[0030] To address the limitations of the Livingstone model (due to its high degree of qualitative analysis and its applicability to reasoning about unexpected events) and the challenges of Bayesian models (due to their high modeling difficulty), a modeling approach based on multi-signal flow graph theory is proposed. The multi-signal flow graph model takes the signal flow of typical equipment objects as its research object and uses the FMECA table of typical equipment objects as modeling material. It identifies fault propagation paths, describes the dependencies between fault measurement points, encapsulates subsystems, and configures measurement point test information, thereby establishing an information flow model for typical equipment objects. Users can use multi-signal flow graph modeling templates (including modules, subsystems, tests, switches, AND / OR nodes, and connectors) to complete hierarchical modeling of typical equipment objects through drag-and-drop operations, fully describing their signal flow and fault measurement point relationships. A schematic diagram of the multi-signal flow graph modeling environment is shown below. Figure 2 .

[0031] 2.2 AMEsim Model Environment

[0032] Based on multi-signal flow graph theory, modeling for typical equipment objects involves a high level of abstraction, and the fault-measurement point matching relationship cannot fully describe the system state. Therefore, it is necessary to expand the modeling approach to more comprehensively describe the increasingly complex information of typical equipment objects. The AMEsim model uses bond graphs as its modeling principle to describe the energy flow propagation relationship of the object system. It adopts object-oriented physical principle modeling and leverages the advantages of the AMEsim model library—rich in content, with realistic and physically meaningful variables, and widely applicable interfaces—by integrating the AMEsim model environment into a health management platform. Users can access the internal information of the AMEsim model through a web interface to achieve dynamic simulation. A schematic diagram of the AMEsim model environment is shown below. Figure 3 AMEsim's Python API is used to assist the platform's view layer in processing business logic, while the model database serves as the data engine in the model layer.

[0033] By packaging the AMEsim secondary development script, namely the Python Script library, and adding the Python interpreter path, the AMEsim Python API is introduced into the Django backend of the health management platform, thereby executing the business logic of the AMEsim physical simulation model and completing the secondary development of the AMEsim model.

[0034] After a user imports a pre-built AMEsim physical simulation model of a typical device, the platform backend automatically reads the model information, breaks down the model components, organizes the parameter data structure, and populates the model data table as subsequent data support.

[0035] Users can easily view the composition of the current model components in a chart format using the "Model View" button, and an HTML page describing the model automatically generated by AMEsim is also embedded for users to read.

[0036] Users can select a component via a drop-down menu, and after clicking, they can view the component's preset parameter data and edit it.

[0037] Users can configure simulation information, such as running time, running step size, and simulation mode, through the "Model Simulation" page, and can execute dynamic simulation functions to provide a visual curve display of the variable values ​​of model components.

[0038] 3. Typical Equipment Model Database

[0039] The typical device model database integrates information from multiple models, involving data tables, linking AMEsim model components, parameters, and variable fields, matching multi-signal flow graph measurement points with AMEsim parameter information, using MySQL or SQLite as the database engine, providing data services through the Django framework's model layer, and dynamically listening for user operation events to update the typical device object model information. See the detailed database design diagram. Figure 4 The Django-related data tables record user permissions, project tasks, and other information, while the model-related data tables record the correspondence between AMESim components and measurement points, parameter variable values, diagnostic results, and other health information.

[0040] 4. Health management services

[0041] This section describes the specific operational steps for the health management service function modules of the health management platform. The health management service functions described below are based on the user starting the platform (using the `python manage.py runserver` command) and correctly establishing a multi-signal flow diagram of typical device objects and the AMEsim model. See the detailed hierarchical diagram of the health management service function modules. Figure 5 .

[0042] 4.1 Testability Analysis Function Module

[0043] After a user correctly creates a multi-signal flow graph for an object, they can perform testability analysis on the current multi-signal flow graph. The testability analysis module uses a fault-testability matrix (D matrix) to describe the test information of the current multi-signal flow graph. The D matrix calculates indicators such as subsystem and system-level fault detection rate and fault isolation rate, and provides graphical displays of the information.

[0044] 4.2 Fault Injection Function Module

[0045] After the user correctly creates the AMEsim physical simulation model of the object, it is imported into the health management platform. The platform automatically decouples the model's hierarchical structure, extracts the parameter variables of each component, and organizes the data structure. After selecting a model component, the user can view the corresponding component parameter information and change the state of the AMEsim physical model by editing the component parameters. After running dynamic simulation on the platform, the corresponding fully digital fault injection function is completed. The fault injection function module has a 3D model interface, allowing users to upload the object's 3D physical model, visualize the fault injection process, and highlight the currently injected faulty component.

[0046] 4.3 Fault Diagnosis Function Module

[0047] After all the typical equipment models are correctly established and the simulation is run on the health management platform, users can call the object anomaly inference engine through the "Fault Diagnosis" page to analyze its original AMEsim physical simulation data and convert it into 0-1 vectors, which are then injected into the D matrix of the multi-signal flow graph model. The built-in encapsulated fault diagnosis algorithm is automatically started to identify faults in the current equipment and the diagnosis results are displayed in a list format.

[0048] 4.4 Fault Discovery Function Module

[0049] Users can freely select associated measurement points. The platform backend listens for measurement point selection events, extracts corresponding physical parameter variables from the database, organizes the simulated data structure, uses it as input for the built-in algorithm, analyzes historical trends, mines the intrinsic relationships between parameters, and visualizes them in the form of a health score curve. It also provides users with the option to configure nodes for updating multi-signal flow graph models for unknown faults, thereby completing fault discovery and model evolution services.

[0050] 4.5 Status Assessment Function Module

[0051] The platform automatically subscribes to historical fault information in the database. Based on the failure rate, it scores typical equipment subsystems and systems from bottom to top. Each level is further categorized into four levels: normal, good, suspected fault, and fault. The platform displays the probability and rating of failure for the current subsystem, sub-system, and system. Users can easily view the status assessment results through the "Status Assessment" page, providing information support for user decision-making.

Claims

1. A general health management platform for typical equipment based on a multi-model environment, characterized in that: The health management platform is a software system based on a B / S architecture, including two model environments for typical devices, a complete object model database, and multiple health management service function modules. The general health management platform provides multiple types of health management services for typical equipment objects, including test analysis, fault injection, fault diagnosis, fault discovery and status assessment. The health management service function modules complete specified tasks by subscribing to model database information. Fault injection allows users to edit AMEsim model parameter information and perform dynamic simulation, thereby realizing fully digital fault injection of typical equipment objects at the physical level, and provides a 3D model interface to provide a visual display of fault injection. After the user correctly establishes the AMEsim physical simulation model of the object, it is imported into the health management platform. The platform automatically decouples the model structure hierarchy, extracts the parameter variables of each component, organizes the data structure, and allows the user to view the corresponding component parameter information after selecting the model component. The user can also change the state of the AMEsim physical model by editing the component parameters. After running dynamic simulation on the platform, the corresponding all-digital fault injection function is completed. The fault discovery monitoring system allows users to select associated measurement points, extract corresponding variable data from the physical model through database model matching information data tables, organize the data structure, analyze historical trends, and present the current device's associated measurement point health score chart in the form of a graph. It also provides configuration nodes for users to add unknown fault information. The aforementioned general health management platform provides a complete object model database for multi-model environments. The database standardizes model paradigms, integrates multi-model test information for the same object, designs data tables, automatically subscribes to model test information and parameter data, and binds multi-signal flow graphs and AMEsim model information. The content covers the object's fault propagation relationships, test attributes, and physical measurement point parameters. The database also provides communication interfaces to provide data interaction services for health management service modules. Users can quickly match model information through the object model database, and the database dynamically monitors the editing and execution of model test cases, allowing users to focus only on data access operations. Step 1: Establish a multi-signal flow graph model based on the signal cross-linking relationships of typical devices, and configure module test attributes; Step 2: Establish an AMEsim physical simulation model based on the physical structure information of typical devices, incorporate it into the general health management platform model environment, configure simulation information, and complete dynamic simulation; Step 3: Users select health management function modules according to their needs, complete corresponding health management tasks, and use the results to assist in the evaluation and decision-making of typical devices.

2. The typical equipment general health management platform based on a multi-model environment according to claim 1, characterized in that: The general health management platform provides two model environments for the target object: a modeling environment based on multi-signal flow graph theory and a modeling environment based on AMEsim. The multi-signal flow graph modeling environment provides an interactive and visual modeling operation interface through a web front-end webpage. The AMEsim-based modeling environment integrates the AMEsim modeling environment into the general health management platform. Users can view and edit the physical simulation model information of typical equipment objects through the web front-end webpage, visualize the structural composition of model components, parameter and variable data, configure simulation parameters, perform dynamic simulations, and view the physical model component parameter data in real time. The general health management platform provides modeling and modeling environments at both the object information flow and physical flow levels through a front-end visual interface. Users can create and view models of typical equipment using only the web interface.

3. The typical equipment general health management platform based on a multi-model environment according to claim 1, characterized in that: Through various health management services, users can fully utilize multi-model information analysis of typical devices to analyze the health status of objects, thereby providing decision support. Moreover, the health management services provided by the platform have a certain degree of universality and are widely applicable in practical scenarios.

Images