CN115729536A - Generalized equipment fault prediction and health management modeling system - Google Patents

Abstract

The invention provides a generalized equipment failure prediction and health management modeling system, which comprises: the system comprises a PHM project management module, a PHM data management module, a visualized PHM modeling module and a visualized PHM model training module, wherein the PHM project management module is used for completing the creation of a PHM project, the PHM data management module is used for completing the import and management of various types of sensor data of equipment, the visualized PHM modeling module is used for expressing PHM model elements by using different types of primitives based on the visualized modeling technology of the graph theory to realize the open modeling of models such as data processing, fault diagnosis, prediction, health assessment and the like, and the visualized PHM model training module is used for completing the multi-training, verification assessment and optimization of the models aiming at machine learning algorithms. Therefore, the invention provides a visual general PHM business model modeling system integrating multiple functions of design, development, verification evaluation, iterative optimization and the like, which is used for meeting the PHM development requirements and mainly solving the problems of general development support of PHM business models and rapid development iteration of business models.

Description

Generalized equipment fault prediction and health management modeling system

Technical Field

The invention relates to a fault Prediction and Health Management (PHM) technology in equipment maintenance support, which is a universal, rapid development and iterative optimization system of a PHM service model.

Background

In the field of equipment maintenance and security, the technical focus of fault Prediction and Health Management (PHM) is to utilize advanced sensor integration and predict, diagnose, monitor and manage the status of equipment with the help of various algorithms and intelligent models. In recent years, rapid development of fault diagnosis and health management has led to a shift of maintenance and safeguard modes from state monitoring to state management, which is an innovative scheme of testing and maintenance diagnosis, and is a comprehensive fault detection, isolation, prediction and health management technology. PHM technology was introduced not to directly eliminate system failures, but to understand and predict when failures are likely to occur; or a series of maintenance activities are triggered when the failure of the original material is not reached, thereby realizing the purposes of autonomous guarantee and reducing the equipment use and guarantee cost.

PHM design, development and verification are a system project, and key technologies such as PHM graphical modeling, inference engine, integrated database, PHM model networking collaborative design and verification evaluation need to be solved, so that a design and development platform integrating multiple functions such as design, development, verification and evaluation is needed to meet the PHM development requirement. At present, foreign PHM system technology has a plurality of engineering application examples, but China is relatively weak in this respect, and research work in this respect is urgently needed to be carried out. The PHM system design, development and face certificate platform can be used for PHM design, auxiliary development, simulation, evaluation and release, and the maturation of PHM software is realized through repeated iteration. The PHM platform software design and development technology is promoted to be comprehensively mastered through breakthrough of visual hierarchical modeling, diagnosis and prediction algorithm model base design and PHM system verification and evaluation technology of the PHM system, and the method has important significance for improving the PHM technical level in China.

The PHM system visual design and interaction technology adopts a visual mode to complete the modeling of the whole PHM system, on a visual software platform, users can utilize various primitives provided by the system to conveniently complete a PHM design model diagram through mouse operation aiming at different objects, a PHM hierarchical model index distribution is established based on an index compromise coefficient method, the PHM hierarchical model design level is analyzed and evaluated, a PHM test scheme is designed and generated, and the tests are organized according to a certain sequence to form a model development and operation module with grade, system level PHM hierarchy and visualization. Therefore, a new PHM system can be constructed, the development process of the PHM system is greatly simplified, the development efficiency is improved, and the development time is saved.

The visual general PHM design software is a platform for assisting the development of a PHM system, and the specific reality, design and development factors and visual technology of the PHM system development are organically combined in the platform, so that the visualization of a modeling process, the visualization of a model operation process and the visualization of data are respectively realized.

The PHM system of the large equipment is divided into member level and regional level station system level three-level management according to job function levels. The member-level PHM system realizes the fault detection and the state monitoring of the PHM system, does not perform prediction and fault isolation, and submits fault and state reports to corresponding area levels through corresponding data buses for diagnosis. On the basis of PHM data such as fault reports, state information and the like of member systems, the regional PHM system realizes fusion diagnosis of PHM data information of the regional I by using model-based reasoning and rule-based reasoning algorithms in an expert system, forms a regional PHM fault state report and transmits the regional PHM fault state report to a system-level PHM system for fusion diagnosis and decision making. The prior art still has the following problems:

the development generalization degree of the domestic PHM algorithm model is low, and the adaptability of the same algorithm model in other environments is poor; the PHM software design and development software tools are various, interfaces are not uniform, and integration is difficult; in the aspect of domestic PHM model development, a mature software development support tool is lacked, cases are deficient, the flow is not standard, and the standard is not unified, so that the problem of low overall development efficiency is caused.

Disclosure of Invention

In order to solve the above technical problem, the present invention provides a generalized equipment failure prediction and health management modeling system, including:

the PHM project management module is used for completing the establishment of PHM projects, supporting the structural design of equipment with multi-level system level, regional level and component level, configuring and managing related parameters of the equipment, forming a sensor parameter matrix of the equipment and associating the sensor parameter matrix with PHM data;

the PHM data management module is responsible for completing the importing and management of various types of sensor data of equipment and supporting a user-defined data format, and equipment state data provides input data for visual PHM modeling and visual PHM model training and is necessary data for fault prediction and monitoring management of the equipment;

the visual PHM modeling module is based on a visual modeling technology of a graph theory, utilizes different types of primitives to express PHM model elements, realizes open modeling of models such as data processing, fault diagnosis, prediction, health assessment and the like, provides modeling support for different objects and various types of PHM functional models, greatly improves the efficiency and the generalization support of model development, provides basic algorithm primitives for visual PHM modeling, provides basic index verification algorithm primitives for PHM model verification indexes, and uses the results trained by a visual PHM model training module in the PHM modeling module for use and operation;

the visual PHM model training module is used for finishing multiple times of training and optimization of the model aiming at a machine learning algorithm, providing basic parameter adjusting parameters in various algorithm models, supporting training data and test data classification of data in the PHM data management module, providing output data for the training model by the training data, helping the algorithm to finish and optimize and evaluating a training result in a test operator, wherein an evaluation index is provided by a PHM model verification index. And (4) the trained algorithm result of the visual PHM model training module is quoted, loaded and operated in the visual PHM modeling module to complete the issuing work of the algorithm.

The PHM model verification index management module is used for providing an evaluation and verification operator for the PHM in the PHM model verification index after the model development is finished, and the evaluation result can be quickly obtained by connecting the PHM model and the PHM model so as to help a user finish quick debugging on PHM model training and optimization;

the verification indexes include a success detection rate (PODS), a false alarm rate (POFA), an Accuracy (Accuracy), a receiver operating characteristic curve, a detection threshold (DetectionThreshold), an overall confidence (overallconfiguration), a Stability (Stability), a duty sensitivity (dutysensivity), a noise sensitivity (noissesensitivity), and the like. Specific measures of predictive capability include Accuracy (Accuracy), precision (Precision), timeliness of prediction to failure time (TTF), confidence of prediction (Confidence), similarity (Similarity), sensitivity (Sensitivity).

And the generalized algorithm library provides a generalized primitive algorithm required by modeling for the visualized PHM modeling module and the visualized PHM model training module, and provides functions of adding and deleting, inquiring, modifying, upgrading and expanding the algorithm.

According to another aspect of the present invention, a generalized equipment failure prediction and health management modeling system is provided, comprising the following steps:

step 1: the PHM engineering management module creates a PHM engineering, creates specific equipment and a multi-level 'system level-regional level-component level' structure of the equipment in the engineering, describes health key information in the structure, binds the equipment and the PHM model after establishing the PHM model, and carries out PHM work aiming at the specific equipment;

step 2: the PHM data management module is used for importing sensor data of single equipment, supporting the time sequence data import of multiple sensors and single sensors, enabling one file to be a data set, and defining the acquisition characteristics and data characteristics of the sensors after the import;

and step 3: the visual PHM modeling module is used for creating a PHM model, dragging operators in the generalized algorithm library into the PHM model, and combining the operators into a complete DAG data stream to finish algorithm editing; after the editing is finished, the PHM model is released, and the released PHM model is bound with equipment; after binding, loading equipment sensor data and operating a PHM (physical fitness management) model to obtain state monitoring, fault diagnosis, fault prediction and health assessment tasks aiming at the equipment;

and 4, step 4: for the model needing to be trained, the visual PHM model training module drags a machine learning algorithm and a PHM model verification index, the verification index gives a training effect, multiple times of training and optimization of the model are completed, and finally a model training result is obtained, and the result can be related and quoted in the visual PHM modeling module;

and 5: when a user needs to expand the algorithm, a new algorithm is created in the general algorithm library, and the algorithm is converted into an operator through release and is used as a generalized operator in the visual PHM modeling module.

Has the advantages that:

the invention provides a universal PHM modeling function, and can support PHM modeling of slave components, subsystems, region levels and equipment levels; a standard interface mode is adopted, so that model exchange is facilitated; the health state of the life piece can be evaluated, and life knowledge is added to realize the prediction of the residual life; supporting the embedding of various artificial intelligence algorithms/models to form the comprehensive integration of the PHM model and the artificial intelligence model; according to the user requirements, a code compiling function can be selected and matched to generate an executable embedded running program; a visual model editing mode is adopted, and the interface is friendly; the test verification of the model is supported, and indexes such as fault detection rate, fault isolation rate, prediction coverage rate, prediction precision and the like can be given; the system comprises a PHM case library, a generalized algorithm library and a component library.

The invention provides a visual PHM modeling environment, follows the primitives of the PHM basic development process, has an open system architecture, enables a platform to meet the requirements of usability and expandability, and provides a convenient and simple modeling environment. The imaging means that a topological graph formed by the elements and the attributes thereof is used for representing a PHM model of the system, aiming at complex equipment and the application of a visualization technology, a user can conveniently know an equipment mechanism and an algorithm calculation process, the process control, management and monitoring visualization are restored through the process visualization of element combination, a clear mapping relation is established between people and a PHM system development task, the PHM system development process is also monitored, and the whole development process is more transparent.

The layering of the invention refers to establishing a model according to a hierarchical structure of an equipment system, specifically which hierarchical model is to be established, and is determined according to the purposes of testing and diagnostic analysis. If in a system level test it is necessary to isolate the occurring faults to the subsystem level, the topology should be composed of primitives, attributes, etc. representing the subsystems. By adopting a hierarchical modeling method, the components at the upper level can be formed by utilizing the information of the components at the lower level. Hierarchical modeling provides conditions for object-oriented model development. The modeling environment supports the encapsulation of objects, forming a high-level part, the newly created part containing all the information that makes up his substructure. Through hierarchical modeling, the composition relation of complex equipment can be quickly realized, hierarchical reasoning and fault diagnosis are realized in fault diagnosis, and higher fault diagnosis accuracy is achieved.

Drawings

FIG. 1: the universal equipment fault prediction and health management modeling software system is formed;

FIG. 2: modeling equipment configuration in a hierarchical mode;

FIG. 3: verifying indexes by the PHM model;

FIG. 4: training and optimizing a PHM model;

FIG. 5 is a schematic view of: and 4, modeling a system core module relation between the universal equipment fault prediction and the health management software.

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments, and all other embodiments obtained by a person skilled in the art based on the embodiments of the present invention belong to the protection scope of the present invention without creative efforts.

The invention provides a generalized equipment fault prediction and health management modeling system, which is a software platform of PHM service model generalization, rapid development and iteration technology for large-scale complex equipment, integrates functions of large-scale complex equipment hierarchical modeling, PHM service model visual modeling, PHM service model verification, algorithm rapid iteration optimization and general algorithm management, and provides a PHM service model meeting requirements for PHM application. As shown in fig. 1.

Referring to fig. 5, a generalized equipment failure prediction and health management modeling system of the present invention comprises:

the system comprises a PHM engineering management module (1), a PHM data management module (2), a visual PHM modeling module (3), a visual PHM model training module (4), a generalized algorithm library (5) and a PHM model verification index management module (6).

The PHM engineering management module (1) is used for creating engineering, defining a multi-level system level-area level-component level structure of the equipment, and binding the equipment with the model created by the visual PHM modeling module (3).

Equipment sensor data is imported and managed in the PHM data management module (2), and definition of file formats can be performed on different data of the file formats; and the imported equipment sensor data is used as data input of a visual PHM modeling module (3) and a visual PHM model training module (4). The equipment sensor data is associated with equipment in the project.

And the visual PHM modeling module (3) completes PHM model algorithm modeling and can be bound with equipment in the PHM engineering management module (1) after being released. And after loading the PHM data, operating the PHM model to obtain the operation result of the model and giving the state monitoring, fault diagnosis, fault prediction and health evaluation results of the equipment.

And the visual PHM model training module (4) is dragged into a machine learning algorithm and a PHM model verification index, a training effect is given by the verification index, multiple times of training and optimization of the model are completed, and a model training result is finally obtained and can be related and quoted in the visual PHM modeling module (3).

And the generalized algorithm library (5) is used for creating a new algorithm in the general algorithm library, converting the algorithm into an operator through release, and using the operator as a generalized operator in the visual PHM modeling module (3).

And (6) PHM model verification index management, namely, tightly attaching to the output result of the PHM algorithm model, providing a series of verification operators including fault accuracy, service life prediction accuracy, health assessment accuracy and the like, and calculating the verification result of the model in time.

Specifically, the PHM project management module is used for completing the establishment of a PHM project, supporting the structural design of equipment with multi-level system level, regional level and component level, configuring and managing related parameters of the equipment, forming a sensor parameter matrix of the equipment and associating the sensor parameter matrix with PHM data;

specifically, the PHM data management module is responsible for importing and managing various types of sensor data of equipment and supporting a user-defined data format, and equipment state data provides input data for visual PHM modeling and visual PHM model training and is necessary data for fault prediction and monitoring management of the equipment;

specifically, the visual PHM modeling module is based on a visual modeling technology of a graph theory, utilizes different types of primitives to express PHM model elements, realizes open modeling of models such as data processing, fault diagnosis, prediction, health assessment and the like, provides modeling support for different objects and various types of PHM functional models, provides basic algorithm primitives for visual PHM modeling, provides basic index verification algorithm primitives for the PHM model verification indexes, and uses and operates the results trained by a visual PHM model training module in the PHM modeling module;

specifically, the visual PHM model training module is used for finishing multiple times of training and optimization of the model aiming at a machine learning algorithm, providing basic parameter adjustment parameters in various algorithm models, supporting training data and test data classification of data in the PHM data management module, providing output data for the training model by the training data, helping the algorithm to finish and optimize, evaluating a training result in a test operator, and providing an evaluation index by a PHM model verification index. And (4) the trained algorithm result of the visual PHM model training module is quoted, loaded and operated in the visual PHM modeling module to complete the issuing work of the algorithm.

Specifically, the PHM model verification index management is used for providing an evaluation and verification operator for the PHM model after the model development is completed, and the evaluation operator is connected with the PHM model to quickly obtain an evaluation result and help a user to complete quick debugging on training and optimization of the PHM model;

the verification indexes include a success detection rate (PODS), a false alarm rate (POFA), an Accuracy (Accuracy), a receiver operating characteristic curve, a detection threshold (DetectionThreshold), an overall confidence (overallconfiguration), a Stability (Stability), a duty sensitivity (dutysensivity), a noise sensitivity (noissesensitivity), and the like. Specific measures of predictive capability include Accuracy (Accuracy), precision (Precision), timeliness of prediction to failure time (TTF), confidence of prediction (Confidence), similarity (Similarity), sensitivity (Sensitivity).

Specifically, the generalized algorithm library provides a generalized primitive algorithm required by modeling for the visual PHM modeling module and the visual PHM model training module, and provides functions of adding and deleting, inquiring, modifying, upgrading and expanding the algorithm.

The system can realize hierarchical modeling of complex equipment, provides hierarchical structural modeling of equipment level, system level and finished product level according to the division of the equipment system and the hierarchical structure, realizes quick programming, development, test and iteration of PHM models with different types and functions, coordinates and distributes sensor parameters of different levels of the equipment, enhances FMECA information, can set diagnosis rules while programming fault modes, and establishes a fault signal matrix.

During specific implementation, a model is established according to a hierarchical structure of an equipment configuration, specifically which hierarchical model is to be established, and is determined according to the purpose of test and diagnostic analysis. If in a system level test it is necessary to isolate the occurring faults to the subsystem level, the topology should be composed of primitives, attributes, etc. representing the subsystems. By adopting a hierarchical modeling method, the components at the upper level can be formed by utilizing the information of the components at the lower level. Hierarchical modeling provides conditions for object-oriented model development. The modeling environment supports the encapsulation of objects, forming a high-level part, the newly created part containing all the information that makes up its sub-structure. According to the actual situation of the object, the modeling can adopt two modes of top-down and bottom-up, as shown in FIG. 2;

the PHM business model visual modeling module of the system can realize the visual modeling of the PHM business model, is used for defining the multi-dimensional visual PHM algorithm graphic element, the attribute of the PHM algorithm graphic element, the public attribute and the configuration rule of the PHM business model and controlling the configuration of the visual graphic element of the PHM business model by a user, and is convenient for the user to carry out the visual PHM business model modeling. The visual modeling adopts a data flow mode, various primitives are mainly connected through connecting lines to form a directed graph, in a canvas, primitive nodes are connected through directed data connecting lines to represent the flow direction of data, the whole data flow modeling is a directed connected graph for realizing the data, the data flows according to logic, and a PHM service is finally completed under the combination of different functional primitives. One data stream in the system is called a PHM model.

PHM visual modeling is to establish PHM classification primitive classes on the basis of a primitive library, wherein class members comprise names, parameters, realization functions and the like, and different classes are added to realize different PHM analysis primitives; and aiming at different analysis objects, different PHM analysis primitives are utilized, primitive parameters are edited, and after a realization function is selected, the primitives are connected to form a PHM visual analysis model diagram of the object.

The module firstly introduces the obtained measuring point/sensor layout optimization design scheme into a testability analysis case library in an XML file form, and stores the scheme as PHM auxiliary design knowledge; according to an OSA-CBM architecture, a PHM system development process is roughly divided into data acquisition, data preprocessing, feature extraction, fault diagnosis, fault prediction, health assessment and the like, and the functions are packaged in a primitive form.

Besides the capabilities of hierarchical equipment modeling and visual PHM business modeling, the most important thing is to verify and evaluate the PHM model of the observation object. In the measurement evaluation process of the PHM model, corresponding verification work is respectively carried out from two aspects of fault diagnosis capability and fault prediction capability, corresponding measurement evaluation indexes are respectively selected according to different selected verification algorithms to carry out PHM capability verification, finally, verification results are compared with specified values to carry out qualification judgment, verification conclusions are given, and a user carries out iterative optimization on the service model according to the conclusions.

Specifically, a typical algorithm is packaged according to a fault evaluation method of the complex equipment system, a PHM model verification index management module is constructed, and the algorithm can be evaluated by monitoring the state of the complex equipment system and estimating the health condition of the complex equipment system. And packaging the evaluation algorithm into a visual primitive, adding the corresponding primitive when the PHM business model is modeled, and operating the model to obtain an evaluation result. The verification index comprises the following contents:

success detection rate (PODS), false alarm rate (POFA), accuracy (Accuracy), recipient operational characteristic curve, detection threshold (DetectionThreshold), overall confidence (OverallConfidence), stability (Stability), duty sensitivity (dutysensivity), noise sensitivity (NoiseSensitivity), and the like. Specific measures of the prediction capability include Accuracy (Accuracy), precision (Precision), timeliness of prediction to failure time (TTF), confidence of prediction (Confidence), similarity (Similarity), sensitivity (Sensitivity), and the like, as shown in fig. 3;

furthermore, the system of the invention provides a large number of universal PHM algorithm models, and in practical application, the diagnosis, prediction and evaluation models need to be verified and iteratively corrected through training. The function can complete training of the PHM algorithm model according to the verification result (precision, error and the like) expected by the user, and continuous iterative optimization is realized.

The general algorithm library in the system comprises a PHM general algorithm, the algorithm completes the designated function through a machine learning method, and the algorithm is divided into a training operator and a prediction operator. Training an operator: training an algorithm through training data to obtain a training result after fitting/approximation, wherein the training result is usually stored in a parameter value mode; a predictor: the operator of the training result is used to input the training result, initial parameter configuration is provided for the algorithm according to the parameters in the result, and the service requirement in the scene is completed.

The specific use process is as shown in fig. 4, in the PHM model training task, a training operator is added to complete training through a newly-built training task, and the training operator is input to a prediction operator to obtain a test result, a verification operator is responsible for verifying the test result (including fault accuracy, life prediction accuracy, health assessment accuracy and the like), and a user can perform algorithm optimization through the verification result.

The general algorithm library manages operators packaged according to operator standards for various calculation algorithms in PHM modeling analysis and data analysis, the operators are used as minimum combinable units for PHM modeling, a PHM service model is a data flow model consisting of a plurality of operators, and necessary functions of equipment state monitoring, fault diagnosis, service life prediction and the like are realized by loading data.

The PHM business model algorithm development module can be further included, open source algorithms such as Python functions can be adopted as the primitive algorithms, and users can conveniently improve the algorithms. Meanwhile, the case base and the secondary development sub-environment support the addition of new cases and the packaging of new algorithms so as to expand the engineering application range. Two technologies of Python script encapsulation and FMI model encapsulation are selected to realize ideas, and taking Python encapsulation as an example, a user can package Python codes through the function to package the Python codes into primitives which complete specific functions. And when the system runs, calling python by software, sending the input variable of the primitive into a python environment, running the script, and returning the obtained output variable to the software to serve as the output of the primitive.

The general algorithm is designed around PHM service, and comprises algorithms of data processing, state monitoring, fault diagnosis, fault prediction and health assessment, the algorithms are converted into operators in the system, and the operators can carry out input and output operations.

And the PHM data management module is mainly used for managing the equipment data under the engineering. The data mainly comprises basic data of a single equipment object, operation data, fault data and task operation result data of the single equipment object, and simulation data management of a project.

PHM data is the core asset of research equipment PHM, equipment PHM data is the general name of equipment various types of sensor data, all models and PHM work development need to reflect the equipment state by the sensor data, and in the module, various types of sensor data of a research object are managed in an engineering range.

PHM data is data in a specific format of the system, and supports the import of txt, csv and other format data, and the imported data needs to be filled with necessary Phm information. And functions of adding, editing, deleting and viewing engineering data are provided.

Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, but various changes may be apparent to those skilled in the art, and it is intended that all inventive concepts utilizing the inventive concepts set forth herein be protected without departing from the spirit and scope of the present invention as defined and limited by the appended claims.

Claims (4)

1. A generalized equipment failure prediction and health management modeling system, comprising:

the PHM project management module is used for completing the establishment of PHM projects, supporting the structural design of equipment with multi-level system level, regional level and component level, configuring and managing related parameters of the equipment, forming a sensor parameter matrix of the equipment and associating the sensor parameter matrix with PHM data;

the PHM data management module is responsible for importing and managing various types of sensor data of equipment and supporting a user-defined data format, and equipment state data provide input data for visual PHM modeling and visual PHM model training and are necessary data for fault prediction and monitoring management of the equipment;

the visual PHM modeling module is based on a visual modeling technology of a graph theory, utilizes different types of primitives to express PHM model elements, realizes open modeling of models such as data processing, fault diagnosis, prediction, health assessment and the like, provides modeling support for different objects and various types of PHM functional models, provides basic algorithm primitives for visual PHM modeling, provides basic index verification algorithm primitives for PHM model verification indexes, and uses and operates the results trained by the visual PHM model training module in the PHM modeling module;

the visual PHM model training module is used for finishing multiple training and optimization of the model aiming at machine learning algorithms, providing basic parameter adjusting parameters in various algorithm models, supporting the classification of training data and test data of data in the PHM data management module, providing output data for the training model by the training data, helping the algorithm to be finished and optimized, evaluating the training result in a test operator, providing an evaluation index by a PHM model verification index, citing, loading and operating the algorithm result trained by the visual PHM model training module, and finishing the issuing work of the algorithm.

2. The generalized equipment failure prediction and health management modeling system of claim 1, further comprising:

and the PHM model verification index module is used for providing an evaluation and verification operator for the PHM in the PHM model verification index after the model development is finished, and the evaluation result can be quickly obtained by connecting the evaluation operator with the PHM model so as to help a user finish quick debugging on PHM model training and optimization.

3. The generalized equipment failure prediction and health management modeling system of claim 1, further comprising:

and the generalized algorithm library provides a generalized primitive algorithm required by modeling for the visualized PHM modeling module and the visualized PHM model training module, and provides functions of adding and deleting, inquiring, modifying, upgrading and expanding the algorithm.

4. A generalized equipment failure prediction and health management modeling system, comprising the steps of:

step 1: the PHM engineering management module creates a PHM engineering, creates specific equipment and a multi-level 'system level-regional level-component level' structure of the equipment in the engineering, describes health key information in the structure, binds the equipment and the PHM model after establishing the PHM model, and carries out PHM work aiming at the specific equipment;

step 2: the PHM data management module is used for importing sensor data of single equipment, supporting the time sequence data import of multiple sensors and a single sensor, defining the acquisition characteristics and the data characteristics of the sensors after one file is a data set;

and step 3: the visual PHM modeling module is used for creating a PHM model, dragging operators in the generalized algorithm library into the PHM model, and combining the operators into a complete DAG data stream to finish algorithm editing; after the editing is finished, the PHM model is released, and the released PHM model is bound with equipment; after binding, loading equipment sensor data and operating a PHM (physical fitness management) model to obtain state monitoring, fault diagnosis, fault prediction and health assessment tasks aiming at the equipment;

and 4, step 4: for the model needing to be trained, the visual PHM model training module drags a machine learning algorithm and a PHM model verification index, the verification index gives a training effect, multiple times of training and optimization of the model are completed, and finally a model training result is obtained, and the result can be related and quoted in the visual PHM modeling module;

and 5: when a user needs to expand the algorithm, a new algorithm is created in the general algorithm library, and the algorithm is converted into an operator through release and is used as a generalized operator in the visual PHM modeling module.

Images