[go: up one dir, main page]

CN114417501B - A predictive modeling approach to health management for airborne deployment - Google Patents

A predictive modeling approach to health management for airborne deployment Download PDF

Info

Publication number
CN114417501B
CN114417501B CN202111663607.4A CN202111663607A CN114417501B CN 114417501 B CN114417501 B CN 114417501B CN 202111663607 A CN202111663607 A CN 202111663607A CN 114417501 B CN114417501 B CN 114417501B
Authority
CN
China
Prior art keywords
model
data
operator
phm
stage
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202111663607.4A
Other languages
Chinese (zh)
Other versions
CN114417501A (en
Inventor
牛伟
王美男
赵建平
韩冰洁
赵洋洋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Aeronautics Computing Technique Research Institute of AVIC
Original Assignee
Xian Aeronautics Computing Technique Research Institute of AVIC
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Aeronautics Computing Technique Research Institute of AVIC filed Critical Xian Aeronautics Computing Technique Research Institute of AVIC
Priority to CN202111663607.4A priority Critical patent/CN114417501B/en
Publication of CN114417501A publication Critical patent/CN114417501A/en
Application granted granted Critical
Publication of CN114417501B publication Critical patent/CN114417501B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/10Geometric CAD
    • G06F30/15Vehicle, aircraft or watercraft design
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00Computer-aided design [CAD]
    • G06F30/20Design optimisation, verification or simulation
    • G06F30/27Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04Ageing analysis or optimisation against ageing
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02ATECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Geometry (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Automation & Control Theory (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The embodiment of the invention discloses an airborne deployment-oriented health management predictive modeling method which comprises the steps of 1, carrying out data processing and model training on airborne data and ground data on a ground platform to obtain a plurality of training models, 2, carrying out evaluation selection and decision making on the training models obtained in the step 1 in a principle prototype to obtain training models meeting task requirements, 3, deploying the training models meeting the task requirements obtained in the step 2 in an airborne, obtaining an airborne full-period PHM modeling strategy through the steps 1-3, and constructing an airborne PHM operator library, and combing according to each stage in the steps 1-3 to obtain the PHM operator library. The technical scheme provided by the embodiment of the invention solves the problem that the complete fault Prediction and Health Management (PHM) predictive maintenance specifications of aviation on-board and ground are not uniform.

Description

Airborne deployment-oriented health management predictive modeling method
Technical Field
The invention relates to the field of reliability of an aviation system, relates to ground/airborne health management, and particularly relates to an airborne deployment-oriented health management predictive modeling method.
Background
In the aviation health management and fault diagnosis system, the computing resources and constraint conditions of the airborne and ground platforms are different, different phases of Prediction and Health Management (PHM) processing need to operate in different hardware environments, and particularly, the ground platform has unlimited computing performance, storage space, rich data analysis tool kits and the like, and the airborne environment has limited computing resources, low power consumption and small storage space.
At present, a complete PHM modeling method and a corresponding operator library suitable for aviation onboard deployment do not exist. Therefore, a complete modeling strategy is required to be established aiming at the PHM of the aviation system, so that the standard data acquisition and analysis are realized, the advantages of the ground platform are exerted to perform data processing and model training, the on-board running time and accuracy are ensured during model evaluation and model decision, and finally, the selected optimal model is deployed at an on-board end.
Disclosure of Invention
Aiming at the PHM deployment requirement facing the machine, the invention provides the health management predictive modeling method facing the machine deployment by considering the difference between the ground and the machine environment, and the assessment selection and decision are carried out in a principle model machine by carrying out data processing and model training at the ground stage, so that the PHM modeling strategy facing the machine is finally deployed in the whole period of the machine.
The embodiment of the invention provides a health management predictive modeling method for airborne deployment, which comprises the following steps:
Step 1, carrying out data processing and model training on airborne data and ground data on a ground platform to obtain a plurality of training models;
Step 2, evaluating, selecting and deciding the training model obtained in the step 1 in a principle prototype to obtain a training model meeting task requirements;
Step 3, deploying the training model meeting the task requirement obtained in the step 2 in an onboard machine;
And (3) obtaining the airborne full-period PHM modeling strategy through the steps 1-3.
Optionally, in the airborne deployment-oriented health management predictive modeling method as described above, the step 1 includes a data importing stage, a data processing stage and a model training stage, which run on a ground platform, specifically includes:
Step 1.1, a data importing stage comprises importing various structured data and unstructured data into a ground high-performance platform;
Step 1.2, a data processing stage comprises a data exploration and analysis sub-stage, a data preprocessing sub-stage and a characteristic engineering sub-stage;
Step 1.3, a model training phase comprising obtaining a plurality of algorithm models for a specific PHM task through model training based on a physical or empirical model and model training based on data driving.
Optionally, in the on-board deployment-oriented health management predictive modeling method as described above, in the step 1.2,
The data exploration and analysis sub-stage comprises the steps of intuitively exploring and simply mining the imported data to obtain the number of samples and the number of features, and distributing the features, the feature trend and the correlation of the data;
the data preprocessing sub-stage is used for improving the data quality and comprises data cleaning, data denoising and data standardization;
feature engineering sub-stages include feature mining, feature selection, feature extraction on the data to extract features that are most useful for specific tasks (e.g., fault diagnosis, life prediction).
Optionally, in the airborne deployment-oriented health management predictive modeling method, the PHM task in the step 1.3 comprises state monitoring, fault diagnosis, fault prediction and residual life prediction;
aiming at a specific PHM task type, the problem analysis and deconstructing method is as follows:
In the mode 1, when a physical or empirical model which is easy to solve exists in an object according to the cognition of an analysis object, a corresponding physical model is established for the object aiming at a specific task so as to perform model training and solving;
Mode 2, when the object structure is complex, failure or degradation mechanism is difficult to obtain, adopting a data driving method, selecting an algorithm by using methods such as corresponding machine learning, statistical analysis and the like, and performing model training;
in the model training stage of step 1.3, a plurality of algorithms are adopted to train a plurality of models so as to obtain a plurality of trained algorithm models.
Optionally, in the on-board deployment-oriented health management predictive modeling method as described above, the step 2 includes a model evaluation stage and a model decision stage that operate on a model machine similar to an on-board software and hardware environment, and specifically includes:
Step 2.1, in a model evaluation stage, evaluating a plurality of trained algorithm models in a principle model machine, selecting a plurality of proper model evaluation indexes according to the type of the algorithm models and task requirements, and operating the plurality of algorithm models trained on the ground in the principle model machine to obtain a model evaluation table;
and 2.2, in a model decision stage, making a model decision rule according to the specified task requirement, and carrying out final decision on the model by combining the obtained model evaluation table, so as to select an optimal model for the specific task.
Optionally, in the on-board deployment-oriented health management predictive modeling method as described above, the step 3 includes:
And in the model deployment stage, encapsulating the selected 'optimal model' by adopting a language supported by airborne hardware, and deploying the encapsulated 'optimal model' at an airborne terminal.
Optionally, in the airborne deployment-oriented health management predictive modeling method as described above, the method further includes:
Constructing an onboard PHM operator library, wherein the PHM operator library is obtained by combing according to each stage in the steps 1 to 3;
The onboard PHM operator library is classified according to a data processing flow and comprises a flow module and a supporting module;
The onboard PHM operator library is classified according to operator functions and comprises a data processing general operator, a PHM task special operator and an integrated modularized airplane component level/system level/full-airplane level PHM operator.
Optionally, in the on-board deployment oriented health management predictive modeling method described above,
The data processing universal operator comprises a data importing operator unit, a data basic operation operator unit, a data preprocessing operator unit, a data exploration analysis operator unit, a characteristic engineering operator unit, a machine learning operator unit and a super parameter optimization operator unit;
The PHM task special operator comprises an expert system operator unit, an anomaly monitoring operator unit, a fault diagnosis operator unit, a life prediction operator unit and other PHM task special operators, wherein the expert system operator unit comprises expert knowledge, an experience-based model, a model based on a physical failure mechanism and the like.
The integrated modularized PHM operator is a high-integration state monitoring, fault diagnosis and service life predicting operator aiming at specific parts, specific systems or all-aircraft of the aircraft, wherein the aircraft part-level/system-level/all-aircraft-level PHM operator is a modularized operator for selecting specific data processing general operators and PHM special operators from corresponding functional operators according to a full-flow airborne PHM algorithm processing strategy, and the built parts or systems.
Optionally, in the on-board deployment oriented health management predictive modeling method described above,
The flow module comprises a data importing operator unit, a data exploration and analysis operator unit, a data preprocessing operator unit, a characteristic engineering operator unit, a model training operator unit, a model evaluating operator unit, a model decision operator unit and a model deployment operator unit;
the supportability module comprises a data basic operation operator unit, a machine learning operator unit, an expert system operator unit and a super parameter optimization operator unit;
The machine learning operator unit, the expert system operator unit and the super parameter optimization operator unit are used as supporting operators for supporting the model training operator unit.
The health management predictive modeling method for the airborne deployment has the advantages that on one hand, analysis by utilizing massive heterogeneous data, problem modeling and model decision making are achieved through a high-performance computer of a ground platform and a principle model machine which is the same as an airborne environment, and finally, an optimal model for specific tasks is selected to conduct model deployment, so that a full-period PHM modeling strategy for the airborne deployment is achieved, on the other hand, the full-period PHM modeling strategy for the airborne deployment is imaged as a specific operator, a PHM operator library is provided, and an airborne PHM modeling process is completely supported.
Drawings
The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate and do not limit the application.
Fig. 1 is a schematic diagram of a health management predictive modeling method for airborne deployment according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a model evaluation stage and a model decision stage according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an onboard PHM operator library in an embodiment of the invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present application more apparent, embodiments of the present application will be described in detail hereinafter with reference to the accompanying drawings. It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be arbitrarily combined with each other.
The following specific embodiments may be combined with each other, and some embodiments may not be repeated for the same or similar concepts or processes.
The embodiment of the invention aims to establish a PHM modeling method oriented to airborne deployment and build a PHM operator library, which can completely support the aviation ground/on-board PHM data analysis and modeling process, quickly and conveniently realize the fault diagnosis and service life analysis of aircraft component level/system level/all-plane level members and evaluate the health state of an aircraft.
Fig. 1 is a schematic diagram of a health management predictive modeling method for airborne deployment according to an embodiment of the present invention. The health management predictive modeling method for airborne deployment provided by the embodiment of the invention is realized by adopting the following technical scheme, and specifically comprises two parts of contents, namely, an airborne PHM algorithm processing strategy and a PHM operator library are formed.
The first part is to form an onboard PHM algorithm processing strategy:
An embodiment of this section includes:
Step 1, carrying out data processing and model training on airborne data and ground data on a ground platform to obtain a plurality of training models;
And 2, performing evaluation selection and decision making on the training model obtained in the step 1 in the principle prototype to obtain the training model meeting the task requirements.
And 3, deploying the training model meeting the task requirements obtained in the step 2 in an onboard machine.
The first part specifically comprises a data importing stage, a data processing stage and a model training stage which are operated on a ground platform, a model evaluating stage and a model deciding stage which are operated in a principle model machine, and a model deployment stage which is operated on an airborne machine, wherein the operations of the stages are shown in figure 1.
In the step 1, through massive heterogeneous data such as airborne operation history data, test data generated by an experiment bench, simulation data generated by a simulation model platform, maintenance and repair data and the like, deep mining analysis of the data is realized by utilizing a data analysis tool kit rich in the ground platform and a high-performance computer of the ground platform, model training for specific PHM tasks (such as fault diagnosis, life prediction and the like) is realized, a trained model is provided for the specific PHM tasks on the machine, and a foundation is provided for realizing the PHM tasks on the machine. The following describes the specific embodiments of each stage in step 1.
Step 1.1, data import stage.
The data importing stage operates on a ground high-performance platform and can comprise importing of various structured data (such as txt, csv and other data files) and unstructured data (such as forms, images and other files).
Step 1.2, data processing stage.
The data processing stage operates on a ground high-performance platform and comprises a data exploration and analysis sub-stage, a data preprocessing sub-stage and a characteristic engineering sub-stage.
The data exploration and analysis sub-stage can be realized through a plurality of mature algorithms, for example, the data correlation analysis can be realized through calculating the pearson correlation coefficient, the Kendell correlation coefficient and the like between every two features of the data, and simultaneously, the linear correlation between every two features of the data can be intuitively displayed through drawing a scattered point pair diagram, a correlation hot zone diagram and the like, and the trend analysis can be intuitively displayed through drawing a line diagram of each feature. The initial detection of the data distribution condition can be realized by calculating various statistical characteristics (mean, variance, mode, kurtosis, skewness, root mean square and the like) of the data, and the data distribution condition can be visually displayed by drawing a data box diagram, a frequency histogram and the like.
The data preprocessing sub-stage can realize the improvement of data quality, including data cleaning, data denoising, data standardization and the like. The data cleaning is mainly performed from three aspects of consistency check, invalid value and missing value processing and repeated value processing. And adopting three processing methods according to different conditions for invalid values and missing values, wherein when the missing values are few and the importance degree of the attributes is low, if the attributes are numerical data, the average value, the median and the like are adopted for simple filling according to the distribution condition of the data, if the missing rate is high and the importance degree of the attributes is low, the attributes can be directly deleted, and if the missing rate is high and the importance degree of the attributes is high, an interpolation method and a modeling method are adopted. And for judging the repetition value, firstly ordering the records in the data set according to a certain rule, and then comparing adjacent records to see whether the records are similar or not so as to detect whether the records are repeated or not. Common methods for data normalization are min-max normalization (normalize data of different features to within the [0,1] interval) and z-score normalization (normalize data of different features to data that obey a standard normal distribution).
The feature engineering sub-stage may employ a number of algorithms, specifically divided into empirical and physical model based methods and data driven based methods. The method based on data driving selects or constructs the most significant features for the target only from the data perspective by analyzing the trend of different feature data, the relevance with the target label and the like, and the specific feature selection method comprises a variance-based method, a relief, a fisher-score-based method, a sparse learning-based method and the like, wherein the feature extraction method comprises principal component analysis PCA, linear discriminant analysis LDA, independent component analysis ICA, local linear embedding LLE, equidistant mapping Isomap and the like.
And step 1.3, model training stage.
The model training stage operates on a ground high-performance platform and comprises the steps of obtaining a plurality of algorithm models aiming at specific PHM tasks through model training based on a physical or experience model and model training based on data driving.
PHM tasks in this step include, for example, status monitoring, fault diagnosis (identifying whether a fault exists, identifying the type of fault, identifying the location of the fault), fault prediction (predicting when a fault will occur), residual life prediction (predicting component degradation trend, predicting when a component will fail), etc.
The problem analysis and deconstruction method for specific PHM task types can be that firstly, according to cognition of an analysis object, if a physical or experience model which is easy to solve exists in the object, a corresponding physical model is established for the object for specific tasks so as to perform model training and solving, if the object structure is complex, failure or degradation mechanism is difficult to obtain, a data driving method is adopted, and algorithms are selected and model training is performed by using methods such as corresponding machine learning, statistical analysis and the like.
It should be noted that in the model training stage of step 1.3, the embodiment of the present invention may adopt a plurality of algorithms to perform training of a plurality of models, so as to obtain a plurality of trained models.
For each stage in the step 2, namely a model evaluation stage and a model decision stage, the model evaluation stage and the model decision stage operate in a principle model machine similar to an onboard software and hardware environment. After the model training stage of step 1 is completed, several trained models are obtained, and decisions need to be made on the several models to obtain an "optimal model" which is finally deployed on board. Because the concept of the optimal model is closely related to practical application and aims at airborne deployment, the model is evaluated by specifying the hardware environment of the model operation, and the same principle model machine simulation on-board operation environment as the airborne software/hardware environment is constructed. The following describes the specific embodiments of each stage in the above step 2.
And 2.1, model evaluation stage.
The model evaluation runs in a principle model machine, and for different task types, various model evaluation indexes reflecting model prediction accuracy can be adopted for model evaluation. The specific implementation mode is that for a regression problem model, common model evaluation indexes comprise Root Mean Square Error (RMSE), average absolute error (MAE), fitting goodness (R2) and the like, and for a classification problem model, common model evaluation indexes comprise accuracy acc, precision, recall rate, F1 fraction, ROC curve and the like. Besides the index reflecting the model prediction accuracy, there is also an index reflecting the model running efficiency, the model prediction time, which is necessary for a strong real-time task. The evaluation of a plurality of trained models is performed in a principle model machine, and a model evaluation table is obtained, as shown in fig. 2, which is a schematic diagram of a model evaluation stage and a model decision stage in the embodiment of the present invention.
And 2.1, model decision stage.
The model decision stage operates on a principle prototype. Firstly, making a model decision rule according to specific task requirements, and finally deciding the model by combining the obtained model evaluation table, and selecting an optimal model for the specific task, namely a training model meeting task requirements.
The process of the model evaluation phase and the model decision phase is shown in fig. 2.
Step 3 of the embodiment of the invention is specifically a model deployment stage.
And in the model deployment stage, packaging the optimal model selected in the step 2 by adopting a language supported by airborne hardware, and deploying the packaged optimal model at an airborne terminal.
And (3) obtaining the full-period PHM modeling strategy carried in the first part through the steps 1-3.
The second part is to construct an onboard PHM operator library;
according to each stage and each sub-stage of the onboard PHM algorithm processing strategy in the first part, the corresponding content is carded to obtain a PHM operator library, as shown in FIG. 3, which is a schematic diagram of the onboard PHM operator library in the embodiment of the invention.
The airborne PHM operator library in the embodiment of the invention is classified according to the data processing flow and can comprise a flow module and a supportability module.
The airborne PHM operator library in the embodiment of the invention is classified according to operator functions and can comprise a data processing general operator, a PHM task special operator and an integrated modularized airplane component level/system level/full-airplane level PHM operator.
As shown in FIG. 3, the onboard PHM operator library comprises four supportability modules, namely a data basic operation operator unit, a machine learning operator unit, an expert system operator unit and a super parameter optimization operator unit.
The data basic operation operator unit is used as a basic operator for supporting the whole flow. Including logical operations, per-element operations, select/replace rows/columns, etc.
The machine learning operator unit, the expert system operator unit and the super parameter optimization operator unit are used as supporting operators of the supporting model training operator unit. The machine learning operator unit provides various classification, regression and artificial neural network operators which can be flexibly selected for subsequent fault diagnosis and life prediction. The expert system operator unit contains an empirical solution for a particular specific component. The super-parameter optimization is a parameter optimization method provided for super-parameters in model training, wherein common model super-parameters comprise the number of network layers and nodes in a neural network, the maximum tree depth in a decision tree model, regularization coefficients and the like, and the super-parameter optimization method comprises particle swarm optimization, genetic algorithm, simulated annealing algorithm and the like.
According to the strategy of the onboard PHM algorithm processing method, a specific data processing general operator and a PHM special operator are selected from an operator library according to the sequence of flow modules in the operator library, and the whole flow state monitoring, fault diagnosis and service life prediction process aiming at a specific component or a specific system is built to serve as an integrated modularized operator of the component or the system.
The technical scheme provided by the embodiment of the invention aims at aviation predictive maintenance, solves the problem that the complete fault prediction and health management (PHM for short) predictive maintenance specifications of aviation on-board and ground are not uniform, and comprises a ground platform, a principle model machine and a full-period PHM algorithm strategy for on-board deployment. The PHM operator library supporting data analysis and modeling is formed and comprises a data importing stage, a data processing stage, a model training stage, a model evaluating stage, a model decision stage and a model deployment stage, and is realized as a specific PHM operator library. The PHM operator library is complete in PHM modeling and data analysis processes on the supporting machine, and meanwhile flexibility and integrated modularization are considered. Through theoretical analysis and experiments, the operator library framework developed under the patent of the invention can meet the aviation PHM requirement, and PHM processing of aircraft component level/system level/full aircraft level members facing airborne deployment can be realized based on the operator library framework.
The health management predictive modeling method for the airborne deployment provided by the embodiment of the invention realizes analysis, problem modeling and model decision making by utilizing massive heterogeneous data through a high-performance computer of a ground platform and a principle model machine which is the same as an airborne environment, and finally selects an optimal model for specific tasks to carry out model deployment so as to realize a full-period PHM modeling strategy for the airborne deployment, and on the other hand, the PHM operator library is provided by embodying the full-period PHM modeling strategy for the airborne deployment into specific operators so as to completely support the PHM modeling process for the airborne.
The following describes schematically, by means of a specific embodiment, the implementation of the airborne deployment-oriented health management predictive modeling method provided by the embodiment of the present invention.
The health status of the aircraft components can be quickly and conveniently assessed by using the integrated modularized aircraft component level operators for specific components of the aircraft, such as engines, lubricating oil modules, rotating components and the like.
Aiming at data and faults which are not frequently generated on the machine, according to the onboard PHM algorithm processing strategy provided by the embodiment of the invention, the abundant data processing universal operators and PHM task special operators in the operator library are supported to be used by themselves, the data are explored and analyzed according to the sequence of the flow modules in the operator library, and the whole diagnosis or prediction flow aiming at the data is flexibly built. The method comprises the following specific steps:
First, data import, data processing and model training are performed in a ground high performance platform. Take the python language as an example for windows platforms. Both data analysis and model training performed in the ground platform invoke operators in the PHM operator library. And (3) exporting the numerical data collected by the test bed into file forms such as txt, excel, csv, importing the collected file data into a python environment, and converting the acquired file data into dataframe-format data by using a data importing operator.
And data cleaning is carried out on the original data. And using a data cleaning operator of the data preprocessing part in the PHM operator library to check whether missing data and repeated data exist or not and check whether the data are inconsistent or not, and processing the data problems according to a cleaning algorithm to improve the data quality. And (3) using operators for data exploration and analysis in the PHM operator library to perform preliminary exploration on the data with improved quality, obtaining data characteristics such as data scale, distribution characteristics, correlation and the like, and providing ideas and guidance for the subsequent steps. The operators of the data exploration and analysis part comprise a correlation thermodynamic diagram, a line diagram, a box diagram, data statistical characteristics and the like.
After the data exploration analysis step, the analysis modeling thought after the preliminary definition is performed. First, a task is used as a guide to determine which model to build. According to the actual situation of the task, the method can be divided into regression problems, classification problems, clustering problems, prediction problems and the like, and each problem needs to establish a different model. Regression and classification belong to the category of supervised learning, which means that the feature to be solved (sample label) in the training set data is known, wherein the feature to be solved of the regression problem is a continuous value, and the feature to be solved of the classification problem is a discrete value. The clustering belongs to the category of unsupervised learning, namely, the feature (sample label) to be solved in the training set data is unknown, and a clustering model is commonly used for abnormality detection and other problems. The prediction problem refers to the characteristic value trend of a given period of time, the value of the characteristic at the future time is predicted, or the failure time of the equipment is predicted according to the value of some characteristics in a period of time.
And carrying out pretreatment and characteristic engineering operation on the data in combination with a conclusion obtained by data exploration analysis. If the model condition is unknown in the actual situation, the data merging operators in the operator library are used in the preprocessing step, all the input working condition data files are merged, and the working condition is not input as the model. If the selected algorithm model is sensitive to the range differences of different input features, the data features are normalized into data conforming to standard normal distribution by using a data normalization operator in a preprocessing operator library. In the feature engineering step, input features of the model need to be selected or constructed from the raw data. If the variance of a feature is found to be small in the exploratory analysis step, the feature is deleted if there is little trend of change.
In the model training stage, proper algorithm is selected to train the model, and the operator library of the part comprises algorithms such as linear regression, decision trees, support vector machines, gradient lifting trees, random forests, artificial neural networks and the like, and various regression, classification and prediction models are supported to be constructed. Selecting several proper algorithms to construct models according to ideas provided in data exploration and analysis, dividing all samples into training sets and test sets according to proper proportion, and sending the training set samples into the constructed several models to perform model training to obtain several specific trained models respectively.
The data analysis and model training phase at the ground platform using the high performance server ends up here. Because the airborne computing resources are limited, the prediction part of the model is only considered to be carried out on the aircraft, namely, the model trained on the ground is subjected to light-weight encapsulation by using a C language and is deployed in an embedded platform on the aircraft, so that real-time diagnosis and prediction on the aircraft are carried out. In order to simulate an onboard embedded environment, a proper embedded development board is selected or a proper development board is built by the device as a principle model machine, firstly, several trained models are packaged in a light-weight mode by using a C language, the obtained several trained models are respectively burnt into the principle model machine, test set data are input, the prediction of the models is carried out in the board, and the step is used as the environment simulation of the onboard prediction so as to realize the verification link of an algorithm.
And evaluating the results obtained by predicting the several models in the plate by using various evaluation indexes. The accuracy evaluation indexes of the common regression model comprise RMSE, MAE and R2, and the accuracy evaluation indexes of the common classification model comprise precision, recall ratio, accuracy, F1 value and the like. Besides the model accuracy evaluation indexes, indexes such as model training time, model testing time and the like are needed for evaluating the prediction instantaneity of the model. The test set data are respectively input into training models in principle prototypes, prediction is carried out in the board, and the numerical values of several model evaluation indexes are calculated to obtain a model evaluation table shown in fig. 2. The evaluation table is used as a reference for selecting an optimal model and on-board deployment, and the optimal model which best meets the actual requirements is selected according to the actual condition requirements, so that the model can be deployed on the machine.
Although the embodiments of the present invention are described above, the embodiments are only used for facilitating understanding of the present invention, and are not intended to limit the present invention. Any person skilled in the art can make any modification and variation in form and detail without departing from the spirit and scope of the present disclosure, but the scope of the present disclosure is to be determined by the appended claims.

Claims (8)

1. An airborne deployment-oriented health management predictive modeling method is characterized by comprising the following steps:
Step 1, carrying out data processing and model training on airborne data and ground data on a ground platform to obtain a plurality of training models;
Step 2, evaluating, selecting and deciding the training model obtained in the step 1 in a principle prototype to obtain a training model meeting task requirements;
Step 3, deploying the training model meeting the task requirement obtained in the step 2 in an onboard machine;
obtaining an onboard full-period PHM modeling strategy through the steps 1-3;
The step 2 comprises a model evaluation stage and a model decision stage which are operated in a principle model machine similar to an onboard software and hardware environment, and specifically comprises the following steps:
Step 2.1, in a model evaluation stage, evaluating a plurality of trained algorithm models in a principle model machine, selecting a plurality of proper model evaluation indexes according to the type of the algorithm models and task requirements, and operating the plurality of algorithm models trained on the ground in the principle model machine to obtain a model evaluation table;
And 2.2, in a model decision stage, making a model decision rule according to the designated task requirement, and carrying out final decision on the model by combining the obtained model evaluation table to select an optimal model for a specific task.
2. The airborne deployment-oriented health management predictive modeling method of claim 1, wherein the step 1 comprises a data importing stage, a data processing stage and a model training stage which run on a ground platform, and specifically comprises the following steps:
Step 1.1, a data importing stage comprises importing various structured data and unstructured data into a ground high-performance platform;
Step 1.2, a data processing stage comprises a data exploration and analysis sub-stage, a data preprocessing sub-stage and a characteristic engineering sub-stage;
Step 1.3, a model training phase comprising obtaining a plurality of algorithm models for a specific PHM task through model training based on a physical or empirical model and model training based on data driving.
3. The on-board deployment-oriented health management predictive modeling method of claim 2, wherein, in step 1.2,
The data exploration and analysis sub-stage comprises the steps of intuitively exploring and simply mining the imported data to obtain the number of samples and the number of features, and distributing the features, the feature trend and the correlation of the data;
the data preprocessing sub-stage is used for improving the data quality and comprises data cleaning, data denoising and data standardization;
feature engineering sub-stages include feature mining, feature selection, feature extraction on the data to extract features that are most useful for a particular task.
4. The on-board deployment-oriented health management predictive modeling method of claim 3, wherein the PHM tasks in step 1.3 include state monitoring, fault diagnosis, fault prediction, residual life prediction;
aiming at a specific PHM task type, the problem analysis and deconstructing method is as follows:
In the mode 1, when a physical or empirical model which is easy to solve exists in an object according to the cognition of an analysis object, a corresponding physical model is established for the object aiming at a specific task so as to perform model training and solving;
mode 2, when the object structure is complex, failure or degradation mechanism is difficult to obtain, adopting a data driving method, selecting an algorithm by using a corresponding machine learning and statistical analysis method and performing model training;
in the model training stage of step 1.3, a plurality of algorithms are adopted to train a plurality of models so as to obtain a plurality of trained algorithm models.
5. The on-board deployment-oriented health management predictive modeling method of claim 4, wherein the step 3 comprises:
And in the model deployment stage, encapsulating the selected optimal model by adopting a language supported by airborne hardware, and deploying the encapsulated optimal model at an airborne end.
6. The on-board deployment-oriented health management predictive modeling method of any of claims 1-5, further comprising:
Constructing an onboard PHM operator library, wherein the PHM operator library is obtained by combing according to each stage in the steps 1 to 3;
The onboard PHM operator library is classified according to a data processing flow and comprises a flow module and a supporting module;
The onboard PHM operator library is classified according to operator functions and comprises a data processing general operator, a PHM task special operator and an integrated modularized airplane component level/system level/full-airplane level PHM operator.
7. The on-board deployment-oriented health management predictive modeling method of claim 6, wherein,
The data processing universal operator comprises a data importing operator unit, a data basic operation operator unit, a data preprocessing operator unit, a data exploration analysis operator unit, a characteristic engineering operator unit, a machine learning operator unit and a super parameter optimization operator unit;
the PHM task special operator comprises an expert system operator unit, an anomaly monitoring operator unit, a fault diagnosis operator unit and a life prediction operator unit, wherein the expert system operator unit comprises expert knowledge, an experience-based model and a model based on a physical failure mechanism;
The integrated modularized PHM operator is a high-integration state monitoring, fault diagnosis and service life predicting operator aiming at specific parts, specific systems or all-aircraft of the aircraft, wherein the aircraft part-level/system-level/all-aircraft-level PHM operator is a modularized operator for selecting specific data processing general operators and PHM special operators from corresponding functional operators according to a full-flow airborne PHM algorithm processing strategy, and the built parts or systems.
8. The on-board deployment-oriented health management predictive modeling method of claim 6, wherein,
The flow module comprises a data importing operator unit, a data exploration and analysis operator unit, a data preprocessing operator unit, a characteristic engineering operator unit, a model training operator unit, a model evaluating operator unit, a model decision operator unit and a model deployment operator unit;
the supportability module comprises a data basic operation operator unit, a machine learning operator unit, an expert system operator unit and a super parameter optimization operator unit;
The machine learning operator unit, the expert system operator unit and the super parameter optimization operator unit are used as supporting operators for supporting the model training operator unit.
CN202111663607.4A 2021-12-30 2021-12-30 A predictive modeling approach to health management for airborne deployment Active CN114417501B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111663607.4A CN114417501B (en) 2021-12-30 2021-12-30 A predictive modeling approach to health management for airborne deployment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111663607.4A CN114417501B (en) 2021-12-30 2021-12-30 A predictive modeling approach to health management for airborne deployment

Publications (2)

Publication Number Publication Date
CN114417501A CN114417501A (en) 2022-04-29
CN114417501B true CN114417501B (en) 2025-06-17

Family

ID=81271043

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111663607.4A Active CN114417501B (en) 2021-12-30 2021-12-30 A predictive modeling approach to health management for airborne deployment

Country Status (1)

Country Link
CN (1) CN114417501B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115599002B (en) * 2022-12-16 2023-04-14 中国航空工业集团公司西安飞机设计研究所 A method and device for monitoring the status and life of an airborne PHM system
CN117454232A (en) * 2023-12-22 2024-01-26 山东未来集团有限公司 Production network construction fault diagnosis, prediction and health management system and method
CN119129230A (en) * 2024-08-27 2024-12-13 北京航空航天大学 PHM system architecture design method, electronic device and storage medium

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106598033A (en) * 2016-12-31 2017-04-26 清华大学深圳研究生院 PHM (Prognostic and Health Management) system for unmanned aerial vehicle based on middleware
CN111896244A (en) * 2020-07-29 2020-11-06 北京天地龙跃科技有限公司 Fully mechanized mining equipment PHM system

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6871160B2 (en) * 2001-09-08 2005-03-22 Scientific Monitoring Inc. Intelligent condition-based engine/equipment management system
CN112464569B (en) * 2020-12-10 2024-11-15 南京明略科技有限公司 A machine learning method and system

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106598033A (en) * 2016-12-31 2017-04-26 清华大学深圳研究生院 PHM (Prognostic and Health Management) system for unmanned aerial vehicle based on middleware
CN111896244A (en) * 2020-07-29 2020-11-06 北京天地龙跃科技有限公司 Fully mechanized mining equipment PHM system

Also Published As

Publication number Publication date
CN114417501A (en) 2022-04-29

Similar Documents

Publication Publication Date Title
CN114417501B (en) A predictive modeling approach to health management for airborne deployment
Filz et al. Data-driven failure mode and effect analysis (FMEA) to enhance maintenance planning
CN113469060A (en) Multi-sensor fusion convolution neural network aeroengine bearing fault diagnosis method
CN114282579A (en) A Fault Diagnosis Method for Aerospace Bearings Based on Variational Mode Decomposition and Residual Networks
CN115422814B (en) A closed-loop optimization design method for complex electromechanical products driven by digital twins
Bisanti et al. Digital twins for aircraft maintenance and operation: A systematic literature review and an IoT-enabled modular architecture
US11954923B2 (en) Method for rating a state of a three-dimensional test object, and corresponding rating system
CN117669176A (en) Power distribution network reliability assessment method and device based on network simplification method
CN113989550A (en) Prediction method of electric vehicle charging pile operation state based on CNN and LSTM hybrid network
CN111160393B (en) Modularized modeling method of carrier rocket health evaluation model based on data driving
Nguyen-Da et al. Real-time AIoT anomaly detection for industrial diesel generator based an efficient deep learning CNN-LSTM in industry 4.0
CN116910574A (en) Mechanical equipment diagnosis method and device based on meta-learning and time convolution network
Azyus et al. Prediction of remaining useful life using the CNN-GRU network: A study on maintenance management
CN117196575A (en) Ground equipment fault prediction and health management system universal architecture and use method thereof
Li et al. Civil aircraft health management research based on big data and deep learning technologies
CN117664575A (en) Remaining service life prediction method of rolling bearings based on dual-graph attention and GRU
Jose et al. Solving time alignment issue of multimodal data for accurate prognostics with CNN-Transformer-LSTM network
CN120217029A (en) Damage trend prediction method based on hierarchical graph structure and Markov residual correction
Castro et al. Artificial Intelligence Models: A literature review addressing Industry 4.0 approach
Bond et al. A hybrid learning approach to prognostics and health management applied to military ground vehicles using time-series and maintenance event data
CN119276728A (en) A 3D twin model-based IoT device monitoring method, system and terminal
CN117892192A (en) Data-driven-based flow prediction method for onboard hydraulic system
CN115048975B (en) A training method, system and detection method for anomaly recognition model
Duan et al. Uncertain Data Driven Predictive Maintenance: A Cost-oriented Implementation Method on Aircraft System
Amri et al. Industrial Systems Fault's Diagnosis Approaches Adapted to the Context of the Smart Industry

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant