CN113919207A - Top-level open type electrical intelligent health monitoring and management system - Google Patents
Top-level open type electrical intelligent health monitoring and management system Download PDFInfo
- Publication number
- CN113919207A CN113919207A CN202111005185.1A CN202111005185A CN113919207A CN 113919207 A CN113919207 A CN 113919207A CN 202111005185 A CN202111005185 A CN 202111005185A CN 113919207 A CN113919207 A CN 113919207A
- Authority
- CN
- China
- Prior art keywords
- health
- data
- equipment
- electrical system
- level
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention designs an upper-level open type electrical intelligent health monitoring and management system, which aims at the health management requirements of fault diagnosis, prediction evaluation and the like of the upper-level electrical system. On the basis of analyzing the working principle of the upper-level electrical system, typical fault modes and fault representations are combed through electrical system fault mechanism analysis. Aiming at the fault diagnosis and prediction evaluation requirements of the upper-level electrical system, algorithm researches and verifications such as rapid fault diagnosis and fault prediction of the electrical system are developed, and technical support is provided for providing functions and realizing the health monitoring of the upper-level electrical system.
Description
Technical Field
The invention relates to an upper-level open type electrical intelligent health monitoring and management system, belonging to the technical research field of intelligent fault diagnosis health management.
Background
The upper stage is generally a relatively independent stage (or stages) added on a base stage carrier rocket, has stronger task adaptability, and can complete the tasks of orbital maneuver, payload separation and the like. The upper level generally has the characteristics of multiple starting, long-time working, autonomous flight and the like, has the capability of multi-satellite launching and orbit deployment, and is one of effective ways for improving the performance and task adaptability of the rocket.
Due to the particularity of tasks, a high-specific impulse propulsion technology, a high-precision autonomous navigation and combined navigation control technology and other advanced technologies need to be deployed at the upper level, so that the complexity and the intelligent degree of an electrical system at the upper level are greatly improved, and a multi-functional complex intelligent system which is provided with a plurality of processors, is subjected to automatic fault diagnosis and is reconstructed is developed by early simple instruction sending, so that a hardware basis is provided for agile control, equipment test and data analysis automation of launch of a carrier rocket. The upper-level electrical system of the novel carrier rocket adopts an integrated design, so that the electrical system scheme has good universality, independence and adaptability.
The upper-level electrical system mainly comprises a control system, a GNC system, a measurement communication system, an on-board power supply, a power supply and distribution system and a ground test system. The functional requirements of the individual systems are as follows:
1) the control system realizes the autonomous control and management of the upper level, monitors, redundancies, fault tolerance and reconfiguration management of the health state of the electrical system, and implements the clock management function of the upper level;
2) the GNC system completes the upper-level attitude stabilization and control, navigation and track deployment tasks;
3) the measuring communication system realizes the functions of remote measurement, remote control and tracking measurement;
4) the power supply and distribution system on the device realizes the functions of power supply and power distribution energy management on the upper-level equipment;
5) the ground test system realizes the flight control and real-time whole-course monitoring of the upper level.
The next generation of upper-stage electrical system of the carrier rocket mainly adopts a 1553B bus architecture, most of the rocket-mounted devices are intelligent devices with independent processing functions, and the overall framework of the electrical system is shown in FIG. 1.
As the space environment is complicated, severe and changeable, the upper-level electrical system is likely to have the situations of device damage, communication interruption and the like. In the test process of the upper-level electrical system, safety faults and non-safety faults can be classified according to the influence degree of faults, and specific fault modes can be shown in table 2. The safety fault belongs to a relatively serious fault, and on one hand, personal safety injury of related personnel can be caused, and a technical safety accident is developed; on the other hand, irreparable damage can be caused to the testing equipment and even the arrow products, and the basic functions of the arrow products are lost.
TABLE 1 common failure modes for upper-level electrical systems
Due to the complexity of the electrical system of the upper level, the location and type of fault occurrence in the system test is unpredictable, and the fault diagnosis has no fixed mode. Therefore, an emergency treatment plan for corresponding problems needs to be made before testing, and because the damage degrees caused by safety faults and non-safety faults are different, the treatment of the two types of faults is different.
The common processing steps for safety failures are: 1) and the power is cut off, a danger source is cut off, and the danger is prevented from being enlarged. 2) Comprehensively recording the field conditions: including the time, place, opportunity, environmental condition of fault, name, type (generation) number, figure number, serial number, batch and other comprehensive information of fault product; summarizing and verifying conditions, making records, and photographing and shooting fault phenomena when necessary. 3) And analyzing and processing problems. And the problems are found and treated in time, and the equipment loss is reduced. The rocket equipment such as the platform, the rocket machine, the servo system and the like and the single machines on the ground have different expression forms when faults occur: voltage, current, frequency, feedback indication, status indication, etc., should find and report problems in time. When severe faults such as disordered swinging of a servo system, oil injection and the like occur, power can be directly cut off for ensuring the safety of equipment and personnel on the rocket. Because such problems occur suddenly and seriously, the power supply of the equipment on the arrow and the ground is required to be cut off immediately. The whole process of system test should be recorded by using multimedia equipment as much as possible, so as to provide reliable basis for analyzing and processing problems.
The processing steps commonly used for the non-safety fault are as follows: 1) protecting the site and freezing; comprehensively recording the fault phenomenon; and (4) calling related personnel, performing preliminary analysis and uniformly describing. 2) According to the situation, fault recurrence is performed. 3) And (4) according to the fault phenomenon, the recurrence result and the analysis result, performing troubleshooting after forming a processing scheme. Non-safety issues account for the majority of system test failures, such issues having sufficient time for analysis and troubleshooting. The states of the equipment on the arrow and the ground are kept as much as possible, a test field is protected well, an interface and an electric connector are not required to be changed easily, the fault phenomenon is not eliminated, and the fault reason is convenient to search and analyze.
With the development of space technology and the demand of space tasks, the upper-level electrical system will develop towards the direction of bearing various space tasks in the future, which puts higher requirements on the reliability of the upper-level electrical system, and meanwhile, the capability of the electrical system to complete related work under the fault condition is more and more important.
Disclosure of Invention
The invention solves the problems that: the defects of the prior art are overcome, the upper-level open type electrical intelligent health monitoring and management system is provided aiming at the working conditions of the upper-level long-time on-orbit multitask requirements, and the system has the capability of performing fault diagnosis, positioning and alarming on the upper-level electrical system.
The technical scheme of the invention is as follows: an upper-level open type electrical intelligent health monitoring and management system comprises a health state data acquisition subsystem, an upper-level health management subsystem and a ground health management subsystem;
the health state data acquisition subsystem is used for carrying out data conversion and time synchronization on health state data of each device in the electrical system and then sending the health state data to the upper-level health management subsystem, wherein the health state data of the devices comprise device operation state data acquired by a sensor and electrical system control calculation data, and the device operation state data comprise temperature, pressure, displacement, strain, voltage current and magnetic field intensity;
the upper level health management subsystem stores the health state data of each device in the electrical system to an upper level database; performing BIT test on single machine equipment in an electrical system; according to the health state data and the BIT test result of each device in the electrical system, adopting an inference machine to carry out fault diagnosis and isolation on each single device, and simultaneously sending the BIT test result of each single device and the obtained health state data of each device in the electrical system to a ground health management subsystem;
the ground health management subsystem is used for storing health state data and BIT test results of all equipment in the electrical system; according to the health state data and the BIT test result of each device in the electrical system, the intelligent diagnosis algorithm is adopted to carry out comprehensive analysis and secondary diagnosis on the health state of the electrical system in the flight process, so that accurate fault detection and positioning are realized, and the overall health state information of the electrical system is given.
Compared with the prior art, the invention has the beneficial effects that:
(1) the invention adopts an open type and layered design method to design an open type health monitoring and management overall architecture of an upper-level intelligent electrical system, forms a layered distributed inference mechanism of a bottom-level sensor and BIT test, regional-level inference diagnosis and platform-level inference diagnosis, and solves the problem of function expansion when the system is compatible with an intelligent advanced electrical technology verification principle prototype and other equipment.
(2) Compared with the traditional health detection method, the method has the advantages that the fault diagnosis and evaluation algorithms based on signal processing, hidden Markov models, multi-signal models, reliability analysis and the like are subjected to fusion analysis, the faults of the electrical system equipment are detected, analyzed and positioned, and the fault detection rate of the electrical system is improved to more than 95%.
(3) And aiming at the fault diagnosis and prediction evaluation requirements of the upper-level electrical system, carrying out algorithm research and verification such as rapid fault diagnosis of the electrical system and the like, and providing technical support for realizing functions of health monitoring of the upper-level electrical system.
Drawings
FIG. 1 is a block diagram of an electrical system of the upper level according to an embodiment of the present invention;
FIG. 2 is a block diagram of an embodiment of a top-level health monitoring and management system architecture;
fig. 3 is a block diagram of wavelet transform according to an embodiment of the present invention.
Detailed Description
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
1. Upper level health monitoring and management system architecture
The upper-level health monitoring and management system can be divided into three levels according to the spatial distribution condition, namely a health state data acquisition subsystem, an upper-level health management subsystem and a ground health management subsystem.
The health state data acquisition subsystem comprises sensors (such as temperature, pressure, displacement, strain, voltage and current, magnetic field intensity and the like) with various types and can acquire running state data of each device in the electrical system in an omnibearing manner; the electric system controls and calculates data and equipment running state data measured by each single sensor to be used as health state data, and the health state data are transmitted to the upper-level health management subsystem through bus transmission after data conversion, time synchronization and other operations;
the upper level health management subsystem stores the health state data of each device in the electrical system to an upper level database; performing BIT test on single machine equipment in an electrical system; according to the running state data and the BIT test result of each device in the electrical system, adopting an inference machine to carry out fault diagnosis and isolation on each single device, and simultaneously sending the BIT test result of each single device and the obtained health state data in the electrical system to a ground health management subsystem;
the ground health management subsystem is used for storing health state data and BIT test results of all equipment in the electrical system; and according to the health state data and the BIT test result of each device in the electrical system, carrying out comprehensive fault detection and accurate positioning on the whole electrical system by adopting an intelligent diagnosis algorithm, and giving the whole health state information of the electrical system.
The intelligent diagnosis and evaluation algorithm for the work of the upper-level electrical system has two research contents, namely, the judgment of fault symptoms, namely, the requirement of high enough resolution sensitivity on weak fault information; secondly, the fault can be effectively detected and accurately positioned. The specific research content is as follows:
The intelligent diagnosis and evaluation algorithm comprises various fault diagnosis algorithms based on signal processing, hidden Markov, multi-signal model fault diagnosis algorithm and the like, aiming at the working principle of an electrical system, more detailed fault information is mined from operation data, the change trend implied by the data is found, the upper-level health evaluation and life prediction are deduced according to the change trend implied by the data, and reliable decision support is provided for the upper-level on-orbit maintenance and design optimization.
The architecture of the upper-level health monitoring and management system is shown in fig. 2. Each subsystem is specifically as follows:
(1) health state data acquisition subsystem
Health status data is the basis of overall health management, and therefore the quality of selected data and the quality of data acquisition techniques directly affect the health prediction and diagnosis part of the overall health management system. The realization of a comprehensive health management system should have the basic functions of acquiring and integrating information from a plurality of system elements and collecting the information into a knowledge base of system health.
The data of the upper-level health monitoring and management system is derived from various sensors and control resolving data of the upper level, and is sent to the upper-level health management subsystem for analysis after signal processing such as data format conversion, A/D conversion, time synchronization and the like. In the task execution process, the sensor continuously acquires the health state data of the upper-level electrical equipment, and the health state data is transmitted to the knowledge database of the upper-level health management subsystem through the data transmission equipment for further analysis.
In the using process of the health monitoring and management system, the health state data acquisition subsystem needs to be timely improved and updated according to the using condition, and the improvement content comprises the aspects of newly added sensor types, optimization of sensor position distribution, improvement of data acquisition precision and the like.
2) Top level load health management subsystem
The upper-level load health management system comprises two main functions of BIT, inference engine and the like,
the working mode of the upper level loading health management system BIT module comprises the working BIT and the working before BIT: the BIT is mainly used for acquiring various performance index parameters of each device of the electrical system during working, storing state data in real time and feeding back the state data to the inference machine in time; the BIT before work is mainly used for checking whether the technical state of each device of the electronic system before work is normal or not, and corresponds to the testing, launching and controlling process before the launching of the upper level. The upper-level BIT is in a centralized structure, is arranged on each piece of upper-level electrical system equipment such as a switching amplifier, an optical fiber inertial measurement unit and a star sensor and is used for monitoring bottom-layer equipment, the upper-level BIT module of the health management system runs in an upper-level core processor, the upper-level core processor controls the diagnostic program of each subsystem, and sensor signals are transmitted to the core processor.
The inference machine is composed of a group of computer functional program modules, receives performance indexes of each device, extracts fault signs of sensor signals output by BIT detection, performs fault inference through knowledge matching, and finally enters a knowledge base for storage and is transmitted to the ground.
When the inference engine of the upper-level load health management subsystem operates, the fault analysis needs to be carried out on the continuously optimized fault diagnosis algorithm. In the verification process, the functional requirement analysis of the fault diagnosis and prediction capability is carried out according to the influence relation between the fault diagnosis and prediction capability and the specific use guarantee task, an acceptable threshold value is defined for the performance measurement index on the basis of considering the requirements of timeliness and accuracy, and when the diagnosis and prediction performance measurement index does not meet the threshold value requirement, the verification result is returned to an algorithm developer for targeted adjustment.
In a diagnosis and prediction performance metric index system, different performance metric indexes describe the capability of the algorithm from different aspects, for example, indexes such as successful detection rate, false alarm rate, accuracy and the like represent the correct diagnosis capability of the system algorithm; the average fault detection time represents the running timeliness of a system algorithm, and has higher reference value in tasks with real-time requirements; the working condition sensitivity and the noise sensitivity represent the adaptability of the system algorithm to different use environments, and have higher reference value in complex tasks. Therefore, when the index does not meet the threshold requirement and the system algorithm needs to be adjusted, a design adjustment strategy according to the specific index and the specific task pertinence is needed.
3) Ground health management subsystem
The ground health management system module is required to have the capability of comprehensively analyzing health state data and secondarily diagnosing the health state data in the flight process, a targeted maintenance strategy is made according to the fault and health state information of each flight task, the informationized maintenance and guarantee work is supported, the evaluation on the health state of the whole aircraft is completed on the basis of the platform data analysis result and the maintenance evaluation result, the suggestion whether the next flight task can be entered is given, and the decision of a commander is assisted.
The process management system in the ground health management subsystem can optimize the fault diagnosis method of the whole system by continuously updating or adding new modules or fault information, and improve the fault diagnosis capability; the ground health management subsystem also has the capability of exchanging and integrating information with other systems (such as the existing equipment comprehensive maintenance information), and can continuously acquire fault information data from the outside. The ground health management subsystem mainly comprises five major functions, namely, machine-ground cooperative health management, daily detection and maintenance, pre-emission test decision, health management system maintenance and process management. The functions cover the whole processes of pre-shooting test, airplane-ground cooperation in the flying process and maintenance and guarantee after return.
Mechanical and ground cooperative health management
In the time period from the transmission to the return to the ground, the upper level load health management subsystem finishes data acquisition, data processing, state monitoring and health evaluation, state information is downloaded to the ground, and the ground health management system predicts the health state.
The upper-level on-board health management subsystem memory imports data obtained by resolving of the inference engine into the ground health management system for comprehensive analysis, and then the upper-level on-board health management subsystem diagnosis algorithm is iteratively optimized according to the health detection result of the comprehensive information analysis.
② daily inspection and maintenance
After returning to the ground, a large amount of inspection and maintenance work is required before the next task. One of the important functions of the ground health management subsystem is the daily detection and maintenance function, and secondary diagnosis and analysis of flight process data and faults need to be realized, so that the proposal of the current maintenance and test is provided based on the function, and the maintenance and test efficiency is improved. And analyzing and managing the overhaul and test data to form a complete health management database.
(iii) Pre-launch test decision
The pre-emission test decision system is a system which carries out the fault diagnosis of the whole system in the upper-level test emission section and provides decision information according to the diagnosis result. The method carries out measuring point detection and fault diagnosis before the upper-level transmission, and gives transmitted decision information, thereby being an effective means for reducing transmission accidents and improving transmission safety.
The pre-emission whole system test and health evaluation function is the core function of a ground subsystem, so that the comprehensive analysis and quantitative evaluation of the whole machine health state are realized, the suggestion of whether the next flight can be carried out is given, and the decision reference is provided for the emission commander.
Maintenance of health management system
The maintenance of the health management system is the basis for ensuring the health management system to normally exert the functions thereof, and the health management system needs to be optimized and improved due to the increase of the flight times, the continuous increase of the maintenance times and the continuous accumulation of various data volumes. Meanwhile, with the deepening of problems and technical understanding, the model base, the knowledge base and the decision base are inevitably subjected to continuous maintenance and upgrading, so that each system engineer can conveniently maintain and upgrade the system, and the capability of the health management system is continuously enhanced.
Process management module
The upper level health monitoring and management system needs to frequently interact with other systems in the process of function realization, the system is complex in function structure and large in data volume, and a full task profile is covered, so that a process management module is needed to monitor the system flow, tasks are managed in a unified manner, the tasks are scheduled according to task requirements, system resources are managed in a unified manner, and the upper level health monitoring and management system is efficiently operated.
2. Rapid fault diagnosis and health assessment algorithm for electrical system
The upper-level electrical system comprises control equipment, navigation guidance equipment, measurement communication equipment, a power supply and distribution equipment;
for the control equipment and the navigation guidance equipment, fault diagnosis is carried out by adopting a fault diagnosis algorithm based on a hidden Markov model;
for the measurement communication equipment, fault diagnosis is carried out by adopting a fault diagnosis algorithm based on signal processing, wherein the fault diagnosis algorithm based on signal processing comprises a time-frequency analysis method;
and for the power supply and distribution equipment, fault diagnosis and health evaluation algorithms based on a Gaussian mixture model or fault diagnosis and health evaluation algorithms based on a deep sequential network are adopted for fault diagnosis.
2.1 Markov model-based fault diagnosis algorithm
(1) Data preprocessing
Carrying out data standardization processing on the health state data acquired by the control equipment and the navigation guidance equipment, namely processing all the data into a uniform data format;
because different parameters have different dimensions and dimension units, such a situation may affect the result of data analysis, and in order to eliminate the dimension influence between the parameters, data normalization is required, and in the algorithm, data normalization processing is performed on the original data. The data standardization processing mode is as follows: all data are processed into a unified data format, when all original data are processed, a header 2 byte (1C 1D), a data identification bit 4 byte (a 1 st byte represents a signaling subsystem, a 2 nd byte represents a receiving subsystem, a 3 rd byte and a 4 th byte all represent frame types), a data length 2 byte (determined by the length of the next column of data content), a data content X byte (determined by the transmission data content and represented by 16-bit binary data), a check code 2 byte (Y ═ X16+ X12+ X5+1, an initial state value is 1) and a frame tail 2 byte (C3C 3) are processed.
(2) Data smoothing processing
Smoothing the health state data after the standardization processing by adopting an exponential moving average smoothing method;
after the raw data are standardized, although the change trend can be seen, the data still have great influence on the construction of the model and the guarantee of the precision. The health assessment work is greatly influenced by sudden change of data points and change of data trend in the original data, so that the parameter smoothing processing is performed by adopting an exponential moving average smoothing method.
(3) Feature extraction
Because the original parameter data has numerous parameters, coupling phenomena exist among the parameters inevitably, and the characterization capabilities of different parameters to the health state are different, if the original parameter data is directly used for model construction, not only can the complexity of the model be overlarge, but also the situation that the model cannot be fitted occurs, so that a Principal Component Analysis (PCA) method is selected for feature extraction of the health state data after smoothing processing, and a feature parameter set after PCA dimension reduction is obtained.
(4) Fault pre-processing
The method for judging the threshold value of the feature parameter set after the PCA dimension reduction is used for initially judging the control equipment and the navigation guidance equipment and estimating the health states of the control equipment and the navigation guidance equipment, wherein the estimated health states comprise three types: normal working state, control equipment failure and control equipment failure;
(5) health state evaluation
And (4) the health state estimation results of the control equipment and the navigation guidance equipment send the feature parameter set subjected to PCA dimensionality reduction into hidden Markov models in different health states to obtain possible health states in the current state, and complete health state evaluation.
The hidden Markov model libraries under different health states comprise a hidden Markov model library under a normal working state, a hidden Markov model library after the control equipment fails and a hidden Markov model library after the navigation guidance equipment fails.
The invention utilizes the characteristic parameter set after PCA dimensionality reduction to construct a hidden Markov model, and carries out parameter training to construct hidden Markov model libraries in different health states.
In the invention, due to the limitation of the original data volume, a test data set is constructed in a mode of adding white noise to the original data so as to complete the verification of the algorithm.
2.2 Fault diagnosis Algorithm based on Signal processing
The non-stationary signal generally has the characteristics that the amplitude change is particularly obvious, and a frequency map obtained after the non-stationary signal is subjected to frequency spectrum conversion does not have obvious periodic characteristic frequency. It is known that a large number of non-stationary components such as offset, trend, sudden change and the like exist in a vibration signal generated in a real working condition, and important characteristics of the signal are often represented by the non-stationary components. An analysis method capable of having both time resolution and frequency domain resolution becomes particularly important. Therefore, the fault diagnosis algorithm based on signal processing of the present invention is referred to as a time-frequency analysis method.
The wavelet transform is a time-frequency analysis method with changeable time window and frequency window, and has higher frequency resolution and lower time resolution in the low frequency part and higher time resolution and lower frequency resolution in the high frequency part, so that the wavelet transform has strong capability of representing local characteristics of signals in both time domain and frequency domain. The typical characteristics of wavelet transform can be analyzed from coarse resolution to fine resolution on signals by using different scales through the wavelet transform, and quality factors are constant and invariable under different resolutions, so the inherent wavelet transform is known as the beauty of a mathematical microscope; the other characteristic is the ability to highlight the local characteristics of the signal in the time domain and the frequency domain, so that the method can be used as a good tool for detecting signal transients and edges.
The core idea of wavelet transform is to decompose the most original signal into an approximation signal and a detail signal first, and then to decompose the most original signal into an approximation signal and a detail signal continuously. And so on, the approximation signal and the detail signal can be decomposed into n layers finally. In order to increase the temporal resolution of the decomposed low-frequency signal and high-frequency signal to that of the original signal, the decomposed signals are reconstructed, and the inverse process of the decomposition is actually the reconstruction process thereof. The low frequency part of the wavelet decomposed signal is represented in the approximate signal, and the high frequency part of the signal is represented in the detail signal. In the vibration signal, the characteristics of the signal itself are represented in the low frequency portion, and the nuances of the signal are represented in the high frequency portion. As shown in fig. 3.
The method based on signal processing avoids the difficulty of extracting a mathematical model of an object, and the wavelet transform-based electrical equipment fault diagnosis can provide local description of signals in time and frequency domains by utilizing the wavelet transform so as to analyze non-stationary signals. The wavelet analysis can also divide the frequency spectrum of the interference signal after the noise is superimposed into a useful frequency part and an interference part, thereby achieving the purpose of effectively suppressing the noise. The method can effectively extract fault characteristics and realize effective and reliable online fault diagnosis.
Since the local variations involved in the continuous wavelet transform are variable, a higher time resolution is exhibited in the high frequency part, whereas a higher frequency resolution is exhibited in the low frequency part, i.e. the wavelet transform has the property of "zooming", all of which is particularly suitable for using the wavelet transform when dealing with abrupt signals.
Meanwhile, due to good video localization characteristics, wavelet analysis can accurately present the characteristics of dynamic signals, and the method has obvious advantages in dynamic signal analysis and is suitable for online fault diagnosis of communication measuring equipment.
2.3 Fault diagnosis and health assessment algorithm based on Gaussian mixture model
And establishing a Gaussian mixture model by taking the performance parameters of the electrical system as input based on the fault diagnosis and the health assessment of the upper-level electrical system to obtain the health degree of the power supply and distribution equipment. The overall health assessment scheme is shown in the above figure, and the specific implementation flow of the algorithm is described below.
(1) Obtaining power performance parameter data
For the power supply and distribution equipment of the upper level, the degradation sensitive parameters mainly comprise signals such as the maximum discharge capacity of a power supply, the cut-off voltage of a power supply battery pack and the like. According to a built-in data interface of the software, required parameter data is read from a historical parameter database or a simulation model, and is analyzed according to a protocol and converted into the standard data format.
(2) Data preprocessing
Because the upper-level power supply is interfered by environment influence, running condition fluctuation, measurement error and the like during running, actual parameter data usually contains a large number of interference factors such as outliers, noise and the like, and the training convergence and the evaluation accuracy of the health evaluation model are influenced. Therefore, some pre-processing operation on the actual parameter data is required before using the data. And eliminating outlier points in the original data by adopting statistical criteria indexes such as a Laite criterion, a Neel criterion, a Grabbs criterion and the like. If the statistical index of the data point exceeds the threshold range required by the above criteria, the data point is considered to belong to the outlier, and the parameter data is removed or subjected to smooth noise reduction by adopting a five-point smoothing method, a linear corner-smearing method, a positive axis parabolic weighted average method, an oblique axis parabolic weighted average method, a local weighted regression scatter point smoothing method (LOWESS) and other methods, so as to remove local noise or interference contained in the data. In addition, in order to adapt to the performance characteristics of the Gaussian mixture model, normalization processing needs to be carried out on the parameter data.
(3) Calculating the overlapping degree between the first GMM model and a second GMM model established by taking the degradation sensitive parameters in the normal state as data samples by taking the degradation sensitive parameters after the current normalization processing as the data samples;
the two GMM model overlap function is as follows:
wherein, g1(x) Is a first GMM modelA density distribution function of;
g2(x) Is the density distribution function of the second GMM model, and x is the operating parameter of the device.
(4) Normalizing the distance to the health degree
After the calculation result of the degree of overlap is obtained, the relative magnitude of the distance measurement result value reflects the health state of the upper stage power supply, but the absolute magnitude thereof cannot represent the health degree thereof. Therefore, a normalization method is needed to map the overlapping degree to the interval of 0-1 and convert the overlapping degree into a health degree CV value to carry out quantitative characterization on the power supply health degree.
And normalizing the overlapping degree by adopting an arctan function normalization formula as follows to obtain a CV value.
Wherein overlap is the overlapping degree of the two GMM models, and a is the maximum value of the error parameter.
(5) And evaluating the health state of the power supply and distribution equipment according to the health degree, wherein the health degree value is closer to 1, which indicates that the power supply and distribution equipment is healthier, and the health value is closer to 0, which indicates that the power supply and distribution equipment is unhealthy.
The Gaussian mixture model is constructed as follows:
the performance parameters of the upper-level power supply are multidimensional characteristics and have complex statistical distribution, and effective fitting is difficult to realize by using single statistical distribution, while the method of the Gaussian mixture model GMM is linear combination of a plurality of Gaussian distribution functions, can realize effective fitting of any type of distribution, and is usually used for solving the problem that data under the same set contains a plurality of different distributions. Therefore, for the health assessment of the power distribution equipment, a deviation measurement method based on a Gaussian mixture model is adopted. After the preprocessed typical component performance parameters are obtained, a parameter vector formed by the performance parameters at each moment is used as a feature, parameters in the GMM model are estimated by using an EM algorithm introduced in the algorithm principle, and a parameter space formed by the estimated model parameters is used as a high-dimensional space of the key components in the health state.
The data to be evaluated are actually measured parameter data, including upper-level position attitude data, time sequence data and power supply output voltage and current signals. After actual parameter data are obtained according to a data acquisition mode, data preprocessing is carried out by using the same preprocessing method, then a characteristic vector formed by performance parameters at each moment is used as input, parameters of a Gaussian mixture model are estimated, and the parameters are used as a high-dimensional space where a data state to be evaluated is located.
2.5 multiple signal model fault diagnosis algorithm
Before modeling a multi-signal model, specific information of a research object needs to be combined, and information such as a test/test point, a tested object composition unit, a fault class, fault-test correlation and the like is defined, wherein the specific information is as follows:
testing and testing points: testing refers to the process of operations taken to determine the performance, characteristics, or whether a system or device is functioning properly and effectively. The Unit Under Test (UUT) is in normal state or fault state, which is determined by whether the response of the excitation and control used in the Test process is expected, if the expected value is reached, the UUT is considered to be in normal working state, otherwise, the UUT is considered to be in fault state. Before testing, test points need to be determined. The definition of a site is any physical location where the required state information can be obtained. It is emphasized that a test may be performed using one or more test points, and that a test point may likewise be used by one or more tests. In the multi-signal modeling process of the upper-level electrical system, one test point is only used by one test.
The tested object comprises a unit and a fault class: the component units are units which need to be replaced when the components are repaired after the fault occurs. In fact, the failure diagnosis mainly focuses on the failure of the component units, so the component units can be represented by the failure units, and if the performance characteristics of the component units are the same or similar, the component units are called failure classes.
Fault-test correlation: faults and tests in a system are often of relevance. If there is a fault siOccurrence, derivation testtjDoes not pass through and is composed ofjBy deducing siIf not, it is called tjAnd siAre interrelated, or tjIs a symmetry test. If only can be composed of siOccurrence of push-out tjDo not pass through, but cannot pass throughjBy pushing out siIf not, it is called tjAn asymmetric test.
Correlation matrix: the dependency matrix (also called D matrix) is a matrix reflecting the dependency relationship between the fault and the test, and is generally denoted as:
in the formula, matrix element dijIs a binary variable, if test tjCapable of detecting a component siIn case of failure, then dij1 is ═ 1; otherwise, let dij＝0；
Any row vector [ D ] in D matrixi1,di2,…,din]TMeans a component siThe result of the detection of each test in the event of a fault, the result of this row vector being taken as component siA sign of failure; any column vector [ D ] in D matrix1j,d2j,…,dmj]TShows the test tjAll the fault states that can be detected reflect the test t to a certain extentjFault detection capability of;
s2, constructing a multi-signal model according to the type, the quantity and the working characteristics of the electrical system equipment, wherein the multi-signal model consists of the following elements:
C＝{c1,c2,…,cm}: m unit parts that may fail;
T＝{t1,t2,…,tn}: a finite set of n available tests; PT is the subset of T without alarm test, FT is the subset of T with alarm; the mathematical meaning of PT and FT is PT ═ ti|tiNo alarm, ti∈T}，FT＝{tj|tjAlarm, tj∈T}，T＝PT∪FT(i＝1,2,…,m；j＝1,2,…,n)；
D＝[dij]: correlation matrix between electronic system unit components and tests, dijIf an electronic system unit component s is denoted by 1iIf a fault occurs, test tjConfirming the alarm; and dij0 means that the electronic system unit component siFailed and can not be tested tjDetecting;
C(ti): can be determined by the test tiA set of detected finite elements; each cell or component is associated with four different states: normal Good, fault Bad, Suspected and Unknown uknown;
initially, the state of all cells or components is unknown. If the test detects that the unit or the component is normal, the state of the corresponding unit or component is updated to be normal; otherwise, the status of these units or components will be set to suspect.
S3, after the multi-signal model is constructed, the diagnosis reasoning based on the multi-signal model can be realized by the following algorithm:
for ti∈PT，G←Uiε∈PTC(ti)；
For tjE.g. FT, set S equals C (t)j) Limit value of G, by sjDenotes, due to the set sjTends towards zero, so the sets Su and s arejThe union of } finally tends to itself, as shown by the following equation:
S＝{sj}←C(tj)-G,Su←Su∪{sj}；
if | { sj} | ═ 1, then B ← B £ u(s)j),Su←Su-B
Here, the term "in ← denotes an update setting state, and S ═ SjIs from the set C (t)i) After all normal elements are deleted. G represents a normal component set, Su represents a suspect component set, and B represents a faulty component set. { S | { SjIs { s } |jThe potential of the electrode.
Intelligent monitoring management of electrical system based on information fusion
According to the characteristics of transmissible faults, various monitoring parameter types and the like of the upper-level electrical equipment, an ARIMA information fusion calculation intelligent monitoring management algorithm for an electrical system is researched by combining various fault diagnosis algorithms such as signal processing, hidden Markov, multi-information models and the like.
The invention also researches an electric system fault prediction method based on the weighted fusion ARIMA. According to the actual data characteristics of the electrical system at the upper level, based on the sensitive parameters of the electrical system, by means of the robust prediction performance of a weighted fusion algorithm and the uncertainty management capability of a statistical time sequence modeling algorithm, fusion prediction of electrical system equipment is carried out by utilizing a multi-ARIMA model, the degradation difference of the electrical equipment is restrained, fault interval prediction with probability significance is provided, and further extrapolation prediction of power supply fault occurrence time with good interpretability is realized.
After the multiple ARIMA models acquire the fault information of the power supply system, information fusion analysis needs to be performed on the acquired data by using an information fusion technology, and an information data set acquired by the multiple ARIMA models is set as SiThe credibility data and the plausibility data obtained in the information acquisition process are respectively Ai、BiThe corresponding reliability function and the similarity function are m (A)i) And m (B)i). The average support degree of the electrical system state variable attribute for the information data of time, upper-level position attitude, navigation, and the like acquired by the ith ARIMA model is shown as follows.
Wherein J (A)l) And the credibility analysis probability of the representation model to the real data is shown.
The uncertainty entropy introduced by the information data acquired in the ith ARIMA model is thus defined as follows
The larger the obtained uncertain entropy value is, the larger the unknown degree of the data information corresponding to the uncertain entropy is proved to be, and the credibility of the information data acquired by the ARIMA models can be correspondingly expressed as follows
The value α represents timely reliability data for the information data, and may be obtained by time-frequency resolution through the aforementioned wavelet transform.
When information data analysis and fusion are carried out, the weighting coefficients are defined as follows:
in the formula MlThe first classification result includes information data (corresponding to each single machine system of the electrical system) acquired by the ith ARIMA model.
And performing weighted average on all the classified synthetic information to obtain final multi-ARIMA model fusion information which can reflect the whole health information index of the electrical system.
In the formula
Is the synthetic information of the ith ARIMA model under the first classification.
mFAnd detecting the acquired column vector state data information by the corresponding power system.
And then, adopting an FMECA analysis method, aiming at all possible faults of the product, determining the influence of each fault mode on the work of the product according to the analysis of the fault modes, finding out single-point faults, and determining the hazard of the single-point faults according to the severity and the occurrence probability of the fault modes. The FMECA analysis result can assist in selecting a high-reliability design method in the product design process and can also provide a basis for product test maintenance work planning.
In the FMECA analysis of the electrical equipment of the upper class, the first step is to collect basic information of the electrical equipment, including: the composition structure diagram of the electrical system; the flow of each functional profile and process step; the functions of the system components and the functional connection relationship with other components; environmental parameters that may have an impact on the operating conditions; specific failure modes of the upper level electrical equipment and their consequences.
In the second step, the system components are listed, confirming the following problems: the manner in which the various components fail significantly, the failure mechanism that caused such failure, the impact that the failure may have (with no significant impact or harm to system functionality), the manner in which the failure is detected.
Thirdly, classifying the failure modes, wherein the classification standard can adopt failure severity and generally comprises four categories: type I catastrophic failure: causing death or permanent total disability or irreversible severe environmental damage to personnel; type II severe accidents: can cause permanent partial disability of personnel and cause reversible environmental damage; critical accident class III: can cause injury to personnel or occupational diseases or environmental damage without violating laws or regulations; mild accident of class IV: can cause injury to personnel or occupational diseases, but can still work, or cause minor environmental damage without violating laws or regulations.
The failure mode, the failure mechanism and its manifest of impact on each component or system or process step (possibly including information on the likelihood of failure) can provide information on the cause of the failure and its impact on the overall system.
After the FMECA is built, a training set is built by adopting the existing normal data and fault data of the electrical equipment. When the collected fault data is increased and the fault mode is increased, the current fault diagnosis model can be retrained and updated through the fault diagnosis model, so that the latest fault diagnosis is realized or the diagnosis precision of the existing fault mode is improved. When performing a fault diagnosis, the corresponding signal should be selected. When the current electrical equipment is determined to be faulty or the state of the current electrical equipment cannot be analyzed clearly through preliminary analysis, the fault mode of the current state can be determined by adopting the convolutional neural network to carry out the fault diagnosis method.
The above-described technical solutions for detailed description belong to the common general knowledge of the skilled person.
Claims (7)
1. An upper-level open type electrical intelligent health monitoring and management system is characterized by comprising a health state data acquisition subsystem, an upper-level health management subsystem and a ground health management subsystem;
the health state data acquisition subsystem is used for carrying out data conversion and time synchronization on health state data of each device in the electrical system and then sending the health state data to the upper-level health management subsystem, wherein the health state data of the devices comprise device operation state data acquired by a sensor and electrical system control calculation data, and the device operation state data comprise temperature, pressure, displacement, strain, voltage current and magnetic field intensity;
the upper level health management subsystem stores the health state data of each device in the electrical system to an upper level database; performing BIT test on single machine equipment in an electrical system; according to the health state data and the BIT test result of each device in the electrical system, adopting an inference machine to carry out fault diagnosis and isolation on each single device, and simultaneously sending the BIT test result of each single device and the obtained health state data of each device in the electrical system to a ground health management subsystem;
the ground health management subsystem is used for storing health state data and BIT test results of all equipment in the electrical system; according to the health state data and the BIT test result of each device in the electrical system, the intelligent diagnosis algorithm is adopted to carry out comprehensive analysis and secondary diagnosis on the health state of the electrical system in the flight process, so that accurate fault detection and positioning are realized, and the overall health state information of the electrical system is given.
2. The upper-level open type electrical intelligent health monitoring and management system of claim 1, wherein the upper-level electrical system comprises a control device, a navigation guidance device, a measurement communication device and a power supply and distribution device;
for the control equipment and the navigation guidance equipment, the ground health management subsystem carries out fault diagnosis by adopting a fault diagnosis algorithm based on a hidden Markov model;
for the measurement communication equipment, the ground health management subsystem carries out fault diagnosis by adopting a fault diagnosis algorithm based on signal processing;
and for the power supply and distribution equipment, fault diagnosis is carried out by adopting a fault diagnosis algorithm based on a Gaussian mixture model.
3. The above-level open electrical intelligent health monitoring and management system of claim 2, wherein the hidden markov model based fault diagnosis algorithm comprises the steps of:
s1.1, carrying out data standardization processing on health state data acquired by control equipment and navigation guidance equipment, namely processing all data into a uniform data format;
s1.2, smoothing the health state data after the standardization treatment by adopting an exponential moving average smoothing method;
s1.3, selecting a Principal Component Analysis (PCA) method for feature extraction on the health state data after smoothing processing to obtain a feature parameter set after the dimensionality of the PCA is reduced;
s1.4, a method for judging a threshold value of the feature parameter set after the PCA dimension reduction is carried out, initial judgment is carried out on the control equipment and the navigation guidance equipment, and the health states of the control equipment and the navigation guidance equipment are estimated, wherein the estimated health states comprise three types: normal working state, control equipment failure and control equipment failure;
s1.5, the health state estimation results of the control equipment and the navigation guidance equipment send the feature parameter set subjected to PCA dimensionality reduction into hidden Markov models in different health states to obtain possible health states in the current state, and health state assessment is completed.
The hidden Markov model libraries under different health states comprise a hidden Markov model library under a normal working state, a hidden Markov model library after the control equipment fails and a hidden Markov model library after the navigation guidance equipment fails.
4. The above-level open electrical intelligent health monitoring and management system of claim 2, wherein said fault diagnosis algorithm based on signal processing is time-frequency analysis.
5. The above-mentioned system of claim 2, wherein the fault diagnosis and health assessment algorithm based on Gaussian mixture model comprises the following steps:
s2.1, obtaining degradation sensitive parameters of power supply and distribution equipment, wherein the degradation sensitive parameters comprise the maximum discharge capacity and cut-off voltage of a power supply;
s2.2, performing smooth noise reduction and normalization processing on degradation sensitive parameters of the power supply and distribution equipment;
s2.3, taking the degradation sensitive parameters after the current normalization processing as data samples, establishing a first GMM model, and calculating the overlapping degree between the first GMM model and a second GMM model established by taking the degradation sensitive parameters in a normal state as the data samples;
s2.4, normalizing the overlapping degree of the first GMM model and the second GMM model, and mapping the overlapping degree to a 0-1 interval to obtain the health degree of the power supply and distribution equipment;
and S2.5, evaluating the health state of the power supply and distribution equipment according to the health degree, wherein the health degree value is closer to 1, which indicates that the power supply and distribution equipment is healthier, and the health value is closer to 0, which indicates that the power supply and distribution equipment is unhealthy.
6. The above-level open electrical intelligent health monitoring and management system of claim 5, wherein the smoothing and noise reduction step in step S2 comprises: and eliminating outlier points in the original data by adopting a Laplace criterion, a Neel criterion or a Grebs criterion statistical criterion index, or smoothing and denoising the degradation sensitive parameters by adopting a five-point smoothing method, a linear corner smearing method, a positive axis parabolic weighted average method, an oblique axis parabolic weighted average method and a local weighted regression scattered point smoothing method, so as to remove noise or interference contained in the data.
7. The above-level open electrical intelligent health monitoring and management system of claim 1, wherein said intelligent diagnostic algorithm further comprises a multiple signal model fault diagnosis algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111005185.1A CN113919207A (en) | 2021-08-30 | 2021-08-30 | Top-level open type electrical intelligent health monitoring and management system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111005185.1A CN113919207A (en) | 2021-08-30 | 2021-08-30 | Top-level open type electrical intelligent health monitoring and management system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113919207A true CN113919207A (en) | 2022-01-11 |
Family
ID=79233746
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111005185.1A Pending CN113919207A (en) | 2021-08-30 | 2021-08-30 | Top-level open type electrical intelligent health monitoring and management system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113919207A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114996843A (en) * | 2022-05-25 | 2022-09-02 | 上海航天控制技术研究所 | Design method of electric aircraft power system fault diagnosis and health management system |
| CN115391083A (en) * | 2022-10-27 | 2022-11-25 | 中国航空工业集团公司金城南京机电液压工程研究中心 | Airborne electromechanical equipment health management method and system |
| CN115615466A (en) * | 2022-12-20 | 2023-01-17 | 中国人民解放军火箭军工程大学 | Complex engineering system health state determination method and system |
-
2021
- 2021-08-30 CN CN202111005185.1A patent/CN113919207A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114996843A (en) * | 2022-05-25 | 2022-09-02 | 上海航天控制技术研究所 | Design method of electric aircraft power system fault diagnosis and health management system |
| CN115391083A (en) * | 2022-10-27 | 2022-11-25 | 中国航空工业集团公司金城南京机电液压工程研究中心 | Airborne electromechanical equipment health management method and system |
| CN115391083B (en) * | 2022-10-27 | 2023-02-03 | 中国航空工业集团公司金城南京机电液压工程研究中心 | Health management method and system for airborne electromechanical equipment |
| CN115615466A (en) * | 2022-12-20 | 2023-01-17 | 中国人民解放军火箭军工程大学 | Complex engineering system health state determination method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tariq et al. | Detecting anomalies in space using multivariate convolutional LSTM with mixtures of probabilistic PCA | |
| EP1374167B1 (en) | Real-time spatio-temporal coherence estimation for autonomous mode identification and invariance tracking | |
| CN113919207A (en) | Top-level open type electrical intelligent health monitoring and management system | |
| US8478479B2 (en) | Predicting time to maintenance by fusion between modeling and simulation for electronic equipment on board an aircraft | |
| Uckun et al. | Standardizing research methods for prognostics | |
| CN114282579A (en) | Aviation bearing fault diagnosis method based on variational modal decomposition and residual error network | |
| CN109581871B (en) | Industrial control system intrusion detection method of immune countermeasure sample | |
| US20230013544A1 (en) | Method, Apparatus and System for Detecting Abnormal Operating States of a Device | |
| Adhikari et al. | Machine learning based data driven diagnostics & prognostics framework for aircraft predictive maintenance | |
| CN113359664B (en) | Fault diagnosis and maintenance system, method, equipment and storage medium | |
| CN110795309A (en) | Hierarchical software health management method and system for microsatellite | |
| CN114139274B (en) | Health management system | |
| CN114267178A (en) | Intelligent operation maintenance method and device for station | |
| CN113343364A (en) | Aviation big data analysis method and tool | |
| Mosallam | Remaining useful life estimation of critical components based on Bayesian Approaches. | |
| EP1752898A2 (en) | Exception analysis for multimissions | |
| KR102694136B1 (en) | Oil gas plant equipment failure prediction and diagnosis system based on artificial intelligence | |
| EP1724717A2 (en) | Real-time spatio-temporal coherence estimation for autonomous mode identification and invariance tracking | |
| Calderano et al. | An enhanced aircraft engine gas path diagnostic method based on upper and lower singleton type-2 fuzzy logic system | |
| CN117367807A (en) | Method, system, equipment and medium for diagnosing faults of aero-engine | |
| US20100161307A1 (en) | Software health management testbed | |
| CN114104329A (en) | Automated prediction of fixes based on sensor data | |
| Burnaev | Rare failure prediction via event matching for aerospace applications | |
| Sankararaman et al. | Uncertainty in prognostics: Computational methods and practical challenges | |
| Rabenoro et al. | A methodology for the diagnostic of aircraft engine based on indicators aggregation |
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 |