CN115526375A - Space flight equipment predictive maintenance system - Google Patents
Space flight equipment predictive maintenance system Download PDFInfo
- Publication number
- CN115526375A CN115526375A CN202210883327.2A CN202210883327A CN115526375A CN 115526375 A CN115526375 A CN 115526375A CN 202210883327 A CN202210883327 A CN 202210883327A CN 115526375 A CN115526375 A CN 115526375A
- Authority
- CN
- China
- Prior art keywords
- data
- subsystem
- equipment
- space equipment
- health
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
Abstract
The invention provides a space equipment predictive maintenance system, which comprises: the sensing system is used for collecting the operation parameters of the space equipment and transmitting the operation parameters to the processing and analyzing subsystem; the processing and analyzing subsystem is used for processing the data of the operation parameters of the space equipment to obtain characteristic data reflecting the state of the space equipment; the fault prediction and diagnosis subsystem is used for carrying out fault diagnosis and analysis on the characteristic data reflecting the state of the space equipment to obtain the most probable fault type and probability of the space equipment at the current moment; predicting the failure trend of the space equipment in a future preset time period to obtain the failure type and the occurrence probability of the space equipment in the future preset time period; the health and service life prediction subsystem is used for evaluating and predicting the health condition and service life of the space equipment to obtain the health rating and service life curve of the space equipment; and the visualization system is used for displaying the process data and the processing results of the processing and analyzing subsystem, the fault prediction and diagnosis subsystem and the health and service life prediction subsystem.
Description
Technical Field
The invention relates to a novel space equipment predictive maintenance system which is suitable for maintenance guarantee of ground equipment and facility equipment of large space launching sites.
Background
At present, the aerospace launching site in China mainly adopts the combination of scheduled maintenance and after-the-fact maintenance, and meanwhile, part of key ground equipment adopts a predictive maintenance strategy based on a known physical degradation model. The maintenance action has hysteresis by simply selecting a post-maintenance strategy, and serious accidents are easily caused by maintenance after equipment fails, so that great economic loss is brought. The reliability of system equipment is required to be high by space launching, so that a preventive maintenance strategy is combined for task equipment, the actual health state of the equipment is not considered in the periodic maintenance, excessive maintenance is easily caused, and the equipment guarantee task cost is increased.
The existing aerospace equipment fault maintenance system lacks an integrated full-flow predictive maintenance system from perception to decision, and meanwhile, the existing aerospace equipment mostly adopts maintenance strategies of scheduled maintenance and repair and after-maintenance, so that the maintenance cost is high, and the execution success rate of aerospace tasks is often influenced by the after-maintenance.
Disclosure of Invention
The invention solves the technical problems that: the method overcomes the defects of the prior art, provides a predictive maintenance system for space launch typical equipment, accurately predicts the maintenance of the space launch typical equipment, provides a reasonable maintenance strategy, eliminates the hysteresis of the maintenance of the existing space launch typical equipment, and reduces the cost of the maintenance of the space launch typical equipment.
The technical scheme of the invention is as follows: a space equipment predictive maintenance system comprises a perception subsystem, a processing and analyzing subsystem, a fault prediction and diagnosis subsystem, a health and service life prediction subsystem and a visualization subsystem;
the sensing system collects the operation parameters of the space equipment and transmits the operation parameters to the processing and analyzing subsystem;
the system comprises a processing and analyzing subsystem, a health and service life predicting subsystem and a fault diagnosis subsystem, wherein the processing and analyzing subsystem is used for processing the operating parameters of the space equipment to obtain characteristic data reflecting the state of the space equipment, storing the characteristic data reflecting the state of the space equipment into a parameter database and simultaneously sending the characteristic data to the fault prediction and diagnosis subsystem and the health and service life predicting subsystem;
the fault prediction and diagnosis subsystem is used for carrying out fault diagnosis and analysis on the characteristic data reflecting the state of the space equipment to obtain the most probable fault type and probability of the space equipment at the current moment; predicting the failure trend of the space equipment in a future preset time period to obtain the failure type and the occurrence probability of the space equipment in the future preset time period, and simultaneously sending the failure type and the occurrence probability of the space equipment in the current time and the future preset time period to a health and service life prediction subsystem;
the health and service life prediction subsystem is used for evaluating and predicting the health condition and service life of the space equipment according to characteristic data reflecting the state of the space equipment, the current time and the fault type and occurrence probability of the space equipment in a future preset time period to obtain the health rating and service life curve of the space equipment;
and the visualization system is used for displaying the process data and the processing results of the processing analysis subsystem, the fault prediction and diagnosis subsystem and the health and service life prediction subsystem in a visualization mode.
Preferably, the space equipment predictive maintenance system further comprises a safeguard decision management and optimization subsystem;
ensuring decision management and optimization subsystem, obtaining the processing result of processing and analyzing subsystem, i.e. the characteristic data C reflecting the state of space equipment in the past preset time period t And the processing result of the fault prediction and diagnosis subsystem, namely the fault probability F in the future preset time period t And the processing result of the health and life prediction subsystem, namely the health rating L in a future preset time period t Forming a multi-source data matrix, multiplying the multi-source data matrix by an input weight matrix for weighting, and taking the weighted multi-source data matrix as an input vector
Predicting the possible occurrence time m of the aerospace equipment fault by adopting a multi-input multi-output LSTM neural network f Maintenance opportunity m with optimal cost effectiveness opt And the optimal maintenance means a is adopted, and the resource management is planned according to the time of possible occurrence of the faults of the space equipment, so that a requirement form of the quantity of spare parts of the space equipment in a preset time period is obtained.
Preferably, the multiple-input multiple-output LSTM neural network includes an input layer, an implied layer, and an output layer, wherein an input layer tensor is a three-dimensional matrix composed of a sample number, a time step, and a sample dimension, and an input vector
Comprises the following steps:
wherein, w c As state feature sequence weights, w f For the weight of the fault data, w l A healthy life data weight; samples is the number of samples;
the hidden layer has two layers, which are marked as a first hidden layer and a second hidden layer, each layer consists of 50 neurons, each neuron comprises an input gate, a forgetting gate and an output gate, and the input gate, the forgetting gate and the output gate are provided with different weights;
the activation function of the output layer adopts a sigmoid function, the mean square error minimization is taken as an optimization target, and the result of the hidden layer is mapped and output to obtain maintenance decision data H t Said maintenance decision data H t Comprises the following steps:
wherein the content of the first and second substances,
for the time at which a failure of the aerospace device may occur,
for cost-effective optimum maintenance opportunities, a t Is a maintenance means.
Preferably, for a mechanical structure type aerospace device, the state feature sequence weight w c 0.6, data on healthy Life w l Weight and failure data weight w f 0.2 and 0.2 respectively;
for hydraulic system class devices, health life data weight w l Is 0.6, the state signature sequence weight w c And a fault data weight w f 0.2 and 0.2 respectively;
for control system class devices, the failure data weight w f Is 0.6, the state signature sequence weight w c And health Life data w l Respectively 0.2 and 0.2.
Preferably, the multiple-input multiple-output LSTM neural network is trained by:
s1, acquiring historical data C reflecting state characteristics of aerospace equipment based on equipment number label and time label t ', failure probability historical data F t ', health rating historical data L t ', and its corresponding aerospace device failure possibilityTime of occurrence m f ', maintenance opportunity m with optimal cost effectiveness opt ', optimum maintenance means a';
s2, reflecting the historical data C of the state characteristics of the space equipment t ', failure probability historical data F t ', health rating historical data L tj Respectively carrying out input weighting processing to obtain input vectors
And input the vector
Performing normalization processing to map it to [0,1]An interval;
s3, changing the time period corresponding to the historical data, repeating the step S1 and the step S2 to obtain a data set,
j = 1-J, J being the number of samples;
s4, collecting the data set
The method comprises the following steps of dividing a training set, a verification set and a test set according to a proportion of 8.
Preferably, the health and life prediction subsystem comprises a health condition prediction model and a data-driven RUL model module;
the health condition prediction model module is used for fitting the health condition of the space equipment by adopting a linear regression model, a logistic regression model or a Gaussian process regression model based on historical characteristic data reflecting the state of the space equipment to obtain a time-varying curve of the health condition of the space equipment, and finishing the evaluation and prediction of the health condition of the space equipment;
and the data-driven RUL model module is used for estimating and predicting the residual service life by adopting a Bayesian estimation model based on historical service life data and a curve of the health condition along with time to obtain a residual service life curve, and sending the residual service life curve to the equipment service life prediction model module.
Preferably, the health and life prediction subsystem further comprises a performance failure degradation model module;
and the performance failure degradation model is used for fitting the equipment degradation process by adopting a linear degradation model or a power degradation model based on the historical characteristic data reflecting the state of the space equipment to obtain a degradation curve of the equipment, so that the degradation trend of the equipment is predicted.
Preferably, the operating parameters of the space equipment include equipment performance data, equipment input data, equipment output data, equipment structural data, equipment environment data and equipment alarm data.
Compared with the prior art, the invention has the beneficial effects that:
(1) The system for ensuring, deciding, managing and optimizing the subsystems adopts a multi-data fusion method and calculates the optimal time for ensuring the equipment based on the LSTM neural network, thereby saving the maintenance cost of the equipment and preventing the loss caused by the shutdown of the equipment. The method has certain universality based on the predictive maintenance modeling driven by multi-source data fusion, has better adaptability aiming at the diversity and the complexity of the space equipment, and solves the problems that the current planned maintenance hysteresis of the space equipment influences the launching task, the universality is lacked in the general reliability modeling maintenance, and the current equipment maintenance system lacks in the intelligent guarantee decision and management optimization process, so that the maintenance guarantee of the space equipment is more intelligent and complete;
(2) The invention adopts an integrated platform design aiming at the space equipment, covers the whole process of maintenance of the space equipment, comprises monitoring, processing and analyzing, fault prediction and diagnosis, health life prediction, result visualization and guarantee decision and management optimization, and provides great convenience for maintenance management and decision of the space equipment. The problems that the existing space equipment maintenance system is relatively dispersed in service and the equipment maintenance management technology is not universal are solved;
(3) Aiming at the characteristics of the space equipment, the maintenance decision model is trained based on the equipment state characteristic data, the fault historical data, the health service life historical data and the maintenance decision historical data, the comprehensiveness of the dimensionality of the maintenance decision training data is considered, and the accuracy of the optimal maintenance decision is ensured.
Drawings
FIG. 1 is a block diagram of a spacecraft equipment predictive maintenance system according to an embodiment of the invention;
FIG. 2 is a general block diagram of a spacecraft equipment predictive maintenance system according to an embodiment of the invention;
FIG. 3 is a flowchart of an integrated and platformized aerospace device predictive maintenance technology system according to an embodiment of the present invention;
fig. 4 is a dynamic maintenance decision flow framework for fault and life prediction based on real-time data according to an embodiment of the present invention.
Detailed Description
The invention is further illustrated by the following examples.
With the development of artificial intelligence and the technology of the Internet of things, by accumulating the measuring point data, the fault mechanism model and the life cycle related parameters of the equipment, the method helps the maintenance of the aerospace equipment to change from preventive maintenance to predictive maintenance, and is a feasible, efficient and reliable equipment maintenance means.
As shown in fig. 1, fig. 2, and fig. 3, the system for predictive maintenance of aerospace equipment according to the present invention includes a sensing subsystem 1, a processing and analyzing subsystem 3, a failure prediction and diagnosis subsystem 4, a health and life prediction subsystem 5, a visualization subsystem 6, and a safeguard decision management and optimization subsystem 7. The system is connected with equipment monitoring sensors of the space equipment through a sensing subsystem, and is transmitted to a processing and analyzing subsystem, a fault prediction and diagnosis subsystem, a health and service life prediction subsystem, a visualization subsystem and a guarantee decision management and optimization subsystem through a sensing network 2, so that monitoring, analysis, fault diagnosis, health service life prediction, fault result visualization and maintenance decision making of the space equipment are realized, and full-flow intelligent maintenance and guarantee of the space equipment is formed.
Data in the perception subsystem 1 is transmitted to the processing subsystem 3 through the perception network 2, and the processing subsystem 3, the fault prediction and diagnosis system 4, the health and service life prediction system 5, the maintenance visualization system 6 and the guarantee decision management optimization system 7 are connected through network interfaces. The sensing network 2 may be a public information network such as the internet of things.
The predictive maintenance perception system 1 detects relevant measuring point information data in the aerospace equipment which needs to be collected by means of the internet of things collection equipment, and obtains the aerospace equipment operation parameters according to the relevant measuring point data. The sensing system is divided into three layers of a sensing layer, a network layer and an application layer, the sensing layer carries out different measuring point information data acquisition and monitoring according to fault types, physical measurement data such as deformation, pressure, speed, temperature, voltage, current and the like of key measuring point positions are obtained through sensors, physical signals are converted into digital signals, and the digital signals are transmitted to the space equipment predictive maintenance processing subsystem 3 through the sensing network 2 for processing and storage. The measuring point information data comprises but is not limited to operation state information data, fault data, maintenance data and performance data, after the operation parameters of the space equipment are obtained, the processing subsystem 3 unifies and aligns parameter names and dimensions of the operation parameters of the space equipment of different types, clutter and noise signals are filtered through a signal processing and data fusion algorithm, and feature data reflecting the state of the space equipment are obtained through feature extraction and data mining. After the characteristic data reflecting the state of the space equipment is obtained, the data are stored in a database and transmitted to a fault prediction and diagnosis subsystem 4 through a network, fault characteristic analysis and fault trend prediction are carried out based on a signal processing analysis method, a fault tree diagnosis method, an expert scoring system fault diagnosis analysis method, a support vector machine method and a time sequence analysis method, the current fault reason and the fault content of the space equipment are given, the prediction result is given to the fault probability of the space equipment within a certain period of time while the space fault information is monitored, and the calculation results of the processing subsystem 3 and the fault prediction and diagnosis subsystem 4 are converged to a health and service life prediction subsystem 5. In a health and service life prediction system, a performance failure degradation model, a health condition prediction model and a data-driven RUL model are adopted to calculate the current health and service life prediction of the aerospace equipment. Based on the calculation results of the processing subsystem 3, the fault prediction and diagnosis subsystem 4 and the life prediction subsystem 5, the guarantee decision management and optimization subsystem 7 provides corresponding equipment resource management planning contents and automatically provides planning arrangement based on a dynamic maintenance decision optimization model and a reinforcement learning maintenance decision model. The visualized content in the above system can be displayed by a trend chart and a graph through a visualization BI tool of the visualization system 6.
Each subsystem is described in detail below:
1. perception subsystem
The sensing system collects the operation parameters of the space equipment and transmits the operation parameters to the processing and analyzing subsystem; the operating parameters of the space equipment comprise equipment performance data, equipment input data, equipment output data, equipment structural data, equipment environment data and equipment alarm data.
The sensor is connected to the sensing subsystem 1 through ZigBee and CAN bus technology;
device performance data: such as speed, pressure, load, temperature, noise, vibration, etc.;
inputting data by the equipment: such as power, water, air, etc.;
the device outputs data: such as power, tractive effort, pressure, etc.;
device structural data: such as position, material, stiffness, flexibility, fatigue, thermal expansion, etc.;
device environment data: such as water, wind, temperature, altitude, humidity, etc.;
device alarm data: such as over speed, current overload, over voltage or under voltage, etc.;
for the swing rod horizontal rod part, the operation parameters of the aerospace device comprise position, material, rigidity and vibration signals;
for the swing rod hydraulic oil pump equipment, the operation parameters of the aerospace equipment comprise an oil pump operation state, an oil cylinder operation pressure, a swing angle and a swing signal;
for the swing rod control cabinet equipment, the operation parameters of the aerospace equipment comprise operation temperature, voltage values and current values.
For the swing rod control cabinet UPS equipment, the operation parameters of the aerospace equipment comprise an operation temperature value, a voltage current value, a power value, a load value and a capacity value.
The sub-system sets software and hardware measuring points of a perception layer in predictive maintenance. The selection of the measuring point positions is related to the input required by the fault mechanism and the predictive maintenance of the specific aerospace equipment, and an input basis is established for data collection and model modeling.
2. Processing analysis subsystem
The system processes the operation parameters of the space equipment so as to meet the processing requirements of state monitoring, subsequent diagnosis and evaluation. Therefore, the processing and analyzing subsystem carries out data processing on the operation parameters of the space equipment to obtain characteristic data reflecting the state of the space equipment, stores the characteristic data reflecting the state of the space equipment into the parameter database, and simultaneously sends the characteristic data to the fault prediction and diagnosis subsystem and the health and service life prediction subsystem.
The data processing comprises one or more of feature extraction, signal processing, data mining and data fusion.
The signal processing method comprises the following steps: missing value filling processing, interpolation value supplementing, abnormal value processing, noise reduction, data smoothing, data extrapolation and the like;
the feature extraction method comprises the following steps: time domain and frequency domain feature extraction, statistical feature extraction, fast Fourier transform, PCA and the like;
the data mining comprises the following steps: a decision tree algorithm, a genetic algorithm, a nearest neighbor algorithm, a bayesian network, a maximum expectation algorithm, etc.;
the data fusion comprises the following steps: competitive fusion, complementary fusion, synergistic fusion and the like.
The characteristic data reflecting the state of the aerospace device may be an average value and a variance of device performance data, a maximum value of device output data, a frequency spectrum of device environment data, and the like.
For the horizontal rod part of the swing rod, the characteristic data corresponding to the state of the space equipment is vibration signal data, threshold value filtering noise information is set, and Fourier transform is adopted to extract and monitor the non-stationary signal, so that the characteristic data reflecting the state of the space equipment can be obtained.
For the swing rod hydraulic oil pump equipment, the corresponding characteristic data reflecting the state of the aerospace equipment are historical data of oil pump starting time, swing rod swing angle and a swing rod in-place mark, and the characteristic data reflecting the state of the aerospace equipment are extracted by adopting a mean standard deviation calculation method.
For the swing rod control cabinet equipment, the corresponding characteristic data reflecting the state of the space equipment are high-temperature alarm mark data, current overload mark data and voltage overpressure or low-voltage mark data, and the characteristic data reflecting the state of the space equipment are extracted in a collaborative fusion mode.
For the UPS equipment of the swing rod control cabinet, the corresponding characteristic data reflecting the state of the space equipment are voltage value data, current value data and power value data, and the characteristic data reflecting the state of the space equipment are extracted by adopting a Bayesian method.
3. Fault predicting and diagnosing subsystem
The fault prediction and diagnosis subsystem is used for carrying out fault diagnosis and analysis on the characteristic data reflecting the state of the space equipment to obtain the most probable fault type and probability of the space equipment at the current moment; predicting the failure trend of the space equipment in a future preset time period to obtain the failure type and the occurrence probability of the space equipment in the future preset time period, and simultaneously sending the failure type and the occurrence probability of the space equipment in the current time and the future preset time period to a health and service life prediction subsystem;
the fault prediction and diagnosis subsystem comprises a fault diagnosis subsystem and a fault prediction subsystem.
The fault diagnosis subsystem adopts signal processing analysis, fault tree diagnosis or expert scoring system fault diagnosis analysis.
The fault prediction subsystem predicts the fault trend of the space equipment in the future time period by adopting a support vector machine and time sequence analysis.
The system utilizes the working state data of the equipment in the states of health, abnormity, fault and the like accumulated in the long-term running process of the equipment, utilizes a large amount of sample data in the health state, adopts a machine learning method to extract the fault characteristics of the equipment, constructs an equipment fault classification and identification model, and is applied to subsequent identification and diagnosis of similar faults.
4. Health and life prediction subsystem
The health and service life prediction subsystem evaluates and predicts the health condition and service life of the space equipment according to the characteristic data reflecting the state of the space equipment, the current time and the fault type and occurrence probability of the space equipment in a future preset time period to obtain the health rating and service life curve of the space equipment.
And the health and service life prediction subsystem is used for evaluating and predicting the health state and service life of the equipment based on the performance failure degradation modeling, the health condition prediction model and the data-driven RUL model.
The health condition prediction model module is used for fitting the health condition of the space equipment by adopting a linear regression model, a logistic regression model or a Gaussian process regression model based on historical characteristic data reflecting the state of the space equipment to obtain a time-varying curve of the health condition of the space equipment, so that the evaluation and prediction of the health condition of the space equipment are completed;
and the data-driven RUL model module is used for estimating and predicting the residual service life (RUL) by adopting a Bayesian estimation model based on historical service life data and a curve of the health condition along with time to obtain a residual service life curve, and sending the residual service life curve to the equipment service life prediction model module.
And the performance failure degradation model is used for fitting the equipment degradation process by adopting a linear degradation model or a power degradation model based on the historical characteristic data reflecting the state of the space equipment to obtain a degradation curve of the equipment, so that the degradation trend of the equipment is predicted.
The linear degradation model is applicable to equipment with loss in the linear degradation characteristic;
the power degradation model is applicable to equipment with the loss presenting the characteristic of nonlinear exponential degradation;
5. guarantee decision management optimization subsystem
The predictive maintenance and guarantee decision and management optimization technology is a maintenance decision mode based on the working state of equipment, and aims to reduce maintenance cost and ensure continuous and efficient operation of a production system. The system determines the maintenance time with optimal cost benefit before the fault occurs by continuously monitoring the running state of the equipment, and adopts proper maintenance activities to prevent the running state of the equipment from being degraded, thereby reducing the unplanned shutdown time of the equipment to the maximum extent and reducing the maintenance cost.
As shown in fig. 4, the safeguard decision management and optimization subsystem obtains the processing result of the processing and analysis subsystem, that is, the feature data C reflecting the state of the space equipment in the past preset time period t The processing result of the failure prediction and diagnosis subsystem, namely the failure probability F in the future preset time period t And the processing result of the health and life prediction subsystem, namely the health rating L in a preset time period in the future t Forming a multi-source data matrix, multiplying the multi-source data matrix by an input weight matrix, and taking the weighted multi-source data matrix as an input vector
Predicting the possible occurrence time m of the aerospace equipment fault by adopting a multi-input multi-output LSTM neural network f Maintenance opportunity m with optimal cost effectiveness opt And the optimal maintenance means a is adopted, and the resource management is planned according to the time of possible occurrence of the faults of the space equipment, so that a requirement form of the quantity of spare parts of the space equipment in a preset time period is obtained.
The multi-input multi-output LSTM neural network comprises an input layer, a hidden layer and an output layer, wherein the input layer tensor is a three-dimensional matrix formed by the number of samples, the time step and the dimension of the samples.
In a specific embodiment of the present invention, a layer tensor input = (samples, times, data _ dim) is input, and a three-dimensional matrix of samples × times × data _ dim is formed, where a sample value range is at least greater than 100; the value range of the timetags is 1-3, and the value range of the data \dimis 3.
Input vector
Comprises the following steps:
for the mechanical structure type aerospace equipment (such as a transmission part of a launching tower swing rod and the like), training input of the equipment state characteristic data is mainly based on time sequence, and the state characteristic sequence weight w c 0.6, health life history data and fault history data as auxiliary input, health life data w l Weight and failure data weight w f 0.2 and 0.2 respectively;
for hydraulic system type equipment (for example, a hydraulic pump and an overflow valve), training input of the equipment is health life history data which is mainly based on time sequence, and health life data w l The weight is 0.6, the equipment state characteristic data and the fault history data are used as auxiliary input, and the weight w of the state characteristic sequence is c And a fault data weight w f 0.2 and 0.2 respectively;
for a control system type device (for example, a control cabinet), the training input of the control system type device is mainly fault history data based on a time sequence, the weight of the fault data is 0.6, device state characteristic data and health life history data serve as auxiliary input, and the weight w of the state characteristic sequence c And health life data w l 0.2 and 0.2 respectively.
The hidden layer has two layers, which are marked as a first hidden layer and a second hidden layer, each layer consists of 50 neurons, each neuron comprises an input gate, a forgetting gate and an output gate, and the input gate, the forgetting gate and the output gate are provided with different weights;
the activation function of the output layer adopts a sigmoid function, the mean square error minimization is taken as an optimization target, and the result of the hidden layer is mapped and output to obtain maintenance decision data H t Said maintenance decision data H t Comprises the following steps:
wherein the content of the first and second substances,
for the time at which a failure of the aerospace device may occur,
for cost-effective maintenance opportunities, a t Is a maintenance means.
The multi-input multi-output LSTM neural network is obtained by training through the following method:
s1, acquiring historical data C reflecting aerospace equipment state characteristics based on equipment number labels and time labels t ', failure probability historical data F t ', health rating historical data L t ', and the time m at which its corresponding aerospace device failure may occur f ', maintenance opportunity m with optimal cost effectiveness opt ', optimum maintenance means a';
s2, reflecting the historical data C of the state characteristics of the space equipment t ', failure probability historical data F t ', health rating historical data L tj Respectively carrying out input weighting processing to obtain input vectors
And input the vector
Go on to unityMapping the image to [0,1]]An interval;
s3, changing the time period corresponding to the historical data, repeating the step S1 and the step S2 to obtain a data set,
j = 1-J, J being the number of samples;
s4, collecting the data set
The method comprises the following steps of dividing a training set, a verification set and a test set according to a proportion of 8. The health and service life prediction subsystem comprises a health condition prediction model, a data-driven RUL model module and a performance failure degradation model module;
the health condition prediction model module is used for fitting the health condition of the space equipment by adopting a linear regression model, a logistic regression model or a Gaussian process regression model based on historical characteristic data reflecting the state of the space equipment to obtain a time-varying curve of the health condition of the space equipment, so that the evaluation and prediction of the health condition of the space equipment are completed;
and the data-driven RUL model module is used for estimating and predicting the residual service life (RUL) by adopting a Bayesian estimation model based on historical service life data and a curve of the health condition along with time to obtain a residual service life curve, and sending the residual service life curve to the equipment service life prediction model module.
And the performance failure degradation model is used for fitting the equipment degradation process by adopting a linear degradation model or a power degradation model based on the historical characteristic data reflecting the state of the space equipment to obtain a degradation curve of the equipment, so that the degradation trend of the equipment is predicted.
The linear degradation model is applicable to equipment with loss in the linear degradation characteristic;
the power degradation model is applicable to equipment with the loss presenting the characteristic of nonlinear exponential degradation;
5. visualization system
And the visualization system displays the process data and the processing results of the processing and analyzing subsystem, the fault prediction and diagnosis subsystem and the health and service life prediction subsystem in a visualization mode.
Aerospace device data visualization is an important tool for predictive maintenance. Can help non-professional maintenance engineer in time to understand the equipment state. With the aid of visualization elements such as charts, graphs, maps and the like, the data visualization tool can conveniently view and understand trends, abnormal values and patterns in the equipment data, and is very important for analyzing massive information and making data-driven decisions. The system analyzes the pertinence of multiple faults, long-stop faults and high-cost faults by visually comparing and analyzing the change trend of the overall availability of the equipment, the descending trend of maintenance cost, the descending trend of maintenance inventory, the descending trend of after-the-fact maintenance times, point inspection and periodic maintenance execution condition analysis.
The dynamic action process of the system is as follows:
(1) And the sensing subsystem reports the state of the relevant measuring points through the equipment sensor through a sensing network and reports the state to the processing and analyzing system.
(2) And the processing and analyzing system performs signal processing, data fusion and data mining on the data, extracts characteristic data reflecting the state of the space equipment and stores the characteristic data reflecting the state of the space equipment in a database.
(3) And the fault prediction and diagnosis system acquires corresponding characteristic parameter data, performs fault characteristic analysis on the signals and performs fault trend prediction.
(4) And the health and service life prediction system comprehensively processes the characteristic data provided by the analysis system and the fault characteristic analysis and trend prediction data provided by the fault prediction diagnosis system to predict the service life of the equipment component.
(5) And according to the real-time fault trend prediction and service life prediction results, the guarantee decision and management optimization system provides a corresponding dynamic maintenance decision conclusion and a reasonable resource management plan.
The invention comprises the following key technologies:
(1) Integrated and platform-based aerospace equipment predictive maintenance technical system
The technology realizes the whole-process guarantee of the maintenance of the space equipment, and comprises monitoring, analysis and processing, feature extraction, fault feature analysis, fault trend prediction, dynamic maintenance decision and resource management planning. The method covers the corresponding historical maintenance guarantee, real-time state and predictive maintenance of the space equipment. The maintenance of the aerospace ground equipment plays a crucial role in reducing the failure frequency, improving the operation efficiency, guaranteeing the emission quality and the like. In the traditional maintenance method, fault characteristic analysis, fault prediction, health state and maintenance management are different subsystems. In the system design, fault analysis is integrated and the current equipment health state is combined, dynamic fault life prediction is given, and dynamic maintenance decision is given based on the prediction result. The utilization rate of data among all systems is improved, the scattered system analysis results are integrated, the manual analysis cost of equipment maintenance is reduced, and the decision-making power is improved.
The technology aims to reduce maintenance cost and ensure continuous and efficient operation of a production system. Specifically, by continuously monitoring the operating state of the equipment, prejudging the time when a fault may occur, determining a cost-effective maintenance opportunity before the fault occurs, and adopting appropriate maintenance activities, the degradation of the operating state of the equipment is prevented, the unplanned downtime of the equipment is minimized, and the maintenance cost is reduced. Maintenance cost can be effectively saved and task success rate can be guaranteed based on the platform.
(2) Dynamic maintenance decision-making technology for predicting fault and service life based on real-time data
In the operation process of the space equipment, the service life prediction and maintenance decision conclusion obtained from the real-time data at each moment can be used for calculating the service life prediction and maintenance decision at the next moment. The technology is divided mainly according to different time intervals, and whether the current equipment needs to be maintained and repaired and corresponding maintenance time and strategy can be dynamically given based on the state of the current time interval. After maintenance is performed each time, the maintenance strategy of the equipment can reach a new probability state, and the traditional prediction algorithm has certain uncertainty and is difficult to be used universally due to manual participation in feature extraction. The decision problem based on the time sequence is suitable for modeling realization of an LSTM algorithm, artificial signal feature extraction is not relied on during modeling, and the problems of diversity, complexity and high modeling cost of space equipment can be solved.
The algorithm mainly adopts an LSTM neural network, the sensing data of each device in each time window is used as the input of a classifier, the failure probability and the service life prediction result are output by the classifier, and the corresponding maintenance decision is obtained. A dynamic maintenance decision framework for fault and life prediction based on real-time data is shown in the following figure.
Historical data of the equipment sensor, historical data of a fault state and historical data of a maintenance decision are input into a training classifier, firstly corresponding data are subjected to normalization processing and mapped to a [0,1] interval, then three kinds of data are subjected to alignment processing based on an equipment number label and a time label, and the three kinds of data are input into an LSTM classifier for training. The data set is divided into a training set, a verification set and a test set according to a proportion of 8. The LSTM classifier comprises an input layer, a hidden layer and an output layer, wherein the hidden layer is composed of two layers of 50-unit neurons, mean square error minimization is used as an optimization target, and a sigmoid activation function is used by the output layer to output failure probability life prediction and corresponding maintenance decisions.
The two key technologies provide technical guarantee for innovation points of the intelligent aerospace equipment predictive maintenance system, intelligent aerospace equipment predictive maintenance is systematically realized, and development and innovation of aerospace equipment maintenance are realized.
The invention has the advantages that:
1) The integrated platform design covers the whole process of maintenance of the space equipment, and comprises monitoring, processing analysis, failure prediction diagnosis, health life prediction, result visualization, guarantee decision making and management optimization, so that great convenience is provided for maintenance and management of the space equipment.
2) The predictive maintenance modeling based on the data driving has certain universality and has better adaptability to the diversity and complexity of the space equipment.
3) The system design module is clearly divided, so that the device is convenient to increase, access and modify, and the maintainability and the expandability are high.
4) The visualization module provides visual result analysis and display functions, provides intuitive result display for professional or non-professional personnel, and facilitates decision making of users.
5) Based on advanced internet of things, data processing and artificial intelligence technologies, the accuracy of the space equipment predictive maintenance is effectively improved, the fault recognition capability is greatly improved, and meanwhile, a more reasonable decision suggestion is provided.
In conclusion, the invention provides a set of one-stop predictive maintenance system for the space equipment, and solves the problems that the current planned maintenance hysteresis of the space equipment affects the launching task, the general reliability modeling maintenance lacks universality, and the current equipment maintenance system lacks an intelligent guarantee decision and management optimization process, so that the maintenance guarantee of the space equipment is more intelligent and complete.
Although the present invention has been described with reference to the preferred embodiments, it is not intended to limit the present invention, and those skilled in the art can make variations and modifications of the present invention without departing from the spirit and scope of the present invention by using the methods and technical contents disclosed above.
Claims (8)
1. A space equipment predictive maintenance system is characterized by comprising a perception subsystem, a processing and analyzing subsystem, a fault prediction and diagnosis subsystem, a health and service life prediction subsystem and a visualization subsystem;
the sensing system is used for collecting the operation parameters of the space equipment and transmitting the operation parameters to the processing and analyzing subsystem;
the system comprises a processing and analyzing subsystem, a health and service life predicting subsystem and a fault diagnosis subsystem, wherein the processing and analyzing subsystem is used for processing the operating parameters of the space equipment to obtain characteristic data reflecting the state of the space equipment, storing the characteristic data reflecting the state of the space equipment into a parameter database and simultaneously sending the characteristic data to the fault prediction and diagnosis subsystem and the health and service life predicting subsystem;
the fault prediction and diagnosis subsystem is used for carrying out fault diagnosis and analysis on the characteristic data reflecting the state of the space equipment to obtain the most probable fault type and probability of the space equipment at the current moment; predicting the failure trend of the space equipment in a future preset time period to obtain the failure type and the occurrence probability of the space equipment in the future preset time period, and simultaneously sending the failure type and the occurrence probability of the space equipment at the current moment and in the future preset time period to a health and service life prediction subsystem;
the health and service life prediction subsystem is used for evaluating and predicting the health condition and service life of the space equipment according to characteristic data reflecting the state of the space equipment, the current moment and the fault type and occurrence probability of the space equipment in a future preset time period to obtain the health rating and service life curve of the space equipment;
and the visualization system is used for displaying the process data and the processing results of the processing analysis subsystem, the fault prediction and diagnosis subsystem and the health and service life prediction subsystem in a visualization mode.
2. The system of claim 1, further comprising a security decision management optimization subsystem;
ensuring decision management and optimization subsystem, obtaining the processing result of processing and analyzing subsystem, i.e. the characteristic data C reflecting the state of space equipment in the past preset time period t The processing result of the failure prediction and diagnosis subsystem, namely the failure probability F in the future preset time period t And the processing result of the health and life prediction subsystem, namely the health rating L in a preset time period in the future t Forming a multi-source data matrix, multiplying the multi-source data matrix by an input weight matrix for weighting, and taking the weighted multi-source data matrix as an input vector
Predicting the possible occurrence time m of the aerospace equipment fault by adopting a multi-input multi-output LSTM neural network f Maintenance opportunity m with optimal cost effectiveness opt And the optimal maintenance means a is adopted, and the resource management is planned according to the time of possible occurrence of the faults of the space equipment, so that a requirement form of the quantity of spare parts of the space equipment in a preset time period is obtained.
3. The system of claim 2, wherein the LSTM neural network comprises an input layer, a hidden layer, and an output layer, wherein the input layer tensor is a three-dimensional matrix comprising a number of samples, a time step, and a dimension of the samples, and the input vector tensor is a vector of the three-dimensional matrix
Comprises the following steps:
wherein, w c Is a state signature sequence weight, w f For the weight of the fault data, w l A healthy life data weight; samples is the number of samples;
the hidden layer has two layers, which are marked as a first hidden layer and a second hidden layer, each layer consists of 50 neurons, each neuron comprises an input gate, a forgetting gate and an output gate, and the input gate, the forgetting gate and the output gate are provided with different weights;
the activation function of the output layer adopts a sigmoid function, the mean square error minimization is taken as an optimization target, and the result of the hidden layer is mapped and output to obtain maintenance decision data H t Said maintenance decision data H t Comprises the following steps:
wherein, the first and the second end of the pipe are connected with each other,
for the time that a failure of the aerospace device may occur,
for cost-effective optimum maintenance opportunities, a t Is a maintenance means.
4. A predictive maintenance system for aerospace devices according to claim 3 wherein:
for the mechanical structure type space equipment, the state characteristic sequence weight w c Is 0.6, healthy Life data w l Weight and failure data weight w f 0.2 and 0.2 respectively;
for hydraulic system class devices, healthy life data weight w l Is 0.6, the state signature sequence weight w c And a fault data weight w f 0.2 and 0.2 respectively;
for control system class devices, the failure data weight w f Is 0.6, the state signature sequence weight w c And health Life data w l 0.2 and 0.2 respectively.
5. The predictive maintenance system for aerospace devices of claim 4 wherein said multiple-input multiple-output (LSTM) neural network is trained by:
s1, acquiring historical data C reflecting aerospace equipment state characteristics based on equipment number labels and time labels t ', failure probability historical data F t ', health rating historical data L t ', and the time m at which its corresponding aerospace device failure may occur f ' maintenance opportunity m with optimal cost effectiveness opt ', optimum maintenance means a';
s2, reflecting the historical data C of the state characteristics of the space equipment t ', failure probability historical data F t ', health rating historical data L tj Respectively carrying out input weighting processing to obtain input vectors
And inputting the vector
Performing normalization processing to map it to [0,1]An interval;
s3, changing the time period corresponding to the historical data, repeating the step S1 and the step S2 to obtain a data set,
j is the number of samples;
s4, collecting the data set
The method comprises the following steps of dividing a training set, a verification set and a test set according to a proportion of 8.
6. The system of claim 1, wherein the health and life prediction subsystem comprises a health prediction model, a data-driven RUL model module;
the health condition prediction model module is used for fitting the health condition of the space equipment by adopting a linear regression model, a logistic regression model or a Gaussian process regression model based on historical characteristic data reflecting the state of the space equipment to obtain a time-varying curve of the health condition of the space equipment, so that the evaluation and prediction of the health condition of the space equipment are completed;
and the data-driven RUL model module is used for estimating and predicting the residual service life by adopting a Bayesian estimation model based on historical service life data and a curve of the health condition along with time to obtain a residual service life curve, and sending the residual service life curve to the equipment service life prediction model module.
7. The system of claim 1, wherein the health and life prediction subsystem further comprises a performance failure degradation model module;
and the performance failure degradation model is used for fitting the degradation process of the equipment by adopting a linear degradation model or a power degradation model based on historical characteristic data reflecting the state of the space equipment to obtain a degradation curve of the equipment, so that the degradation trend of the equipment is predicted.
8. A predictive maintenance system for aerospace devices according to claim 1 wherein said aerospace device operating parameters include device performance data, device input data, device output data, device structural data, device environmental data, device alarm data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210883327.2A CN115526375A (en) | 2022-07-26 | 2022-07-26 | Space flight equipment predictive maintenance system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210883327.2A CN115526375A (en) | 2022-07-26 | 2022-07-26 | Space flight equipment predictive maintenance system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115526375A true CN115526375A (en) | 2022-12-27 |
Family
ID=84695604
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210883327.2A Pending CN115526375A (en) | 2022-07-26 | 2022-07-26 | Space flight equipment predictive maintenance system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115526375A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116643517A (en) * | 2023-05-17 | 2023-08-25 | 青岛哈尔滨工程大学创新发展中心 | Accident prevention equipment for underwater manned submersible vehicle |
| CN116827802A (en) * | 2023-08-29 | 2023-09-29 | 青岛海瑞达网络科技有限公司 | Industrial Internet of things maintenance method and monitoring system based on data monitoring |
| CN116976862A (en) * | 2023-09-20 | 2023-10-31 | 山东国研自动化有限公司 | Factory equipment informatization management system and method |
-
2022
- 2022-07-26 CN CN202210883327.2A patent/CN115526375A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116643517A (en) * | 2023-05-17 | 2023-08-25 | 青岛哈尔滨工程大学创新发展中心 | Accident prevention equipment for underwater manned submersible vehicle |
| CN116643517B (en) * | 2023-05-17 | 2024-02-23 | 青岛哈尔滨工程大学创新发展中心 | Accident prevention equipment for underwater manned submersible vehicle |
| CN116827802A (en) * | 2023-08-29 | 2023-09-29 | 青岛海瑞达网络科技有限公司 | Industrial Internet of things maintenance method and monitoring system based on data monitoring |
| CN116827802B (en) * | 2023-08-29 | 2023-11-24 | 青岛海瑞达网络科技有限公司 | Industrial Internet of things maintenance method and monitoring system based on data monitoring |
| CN116976862A (en) * | 2023-09-20 | 2023-10-31 | 山东国研自动化有限公司 | Factory equipment informatization management system and method |
| CN116976862B (en) * | 2023-09-20 | 2024-01-02 | 山东国研自动化有限公司 | Factory equipment informatization management system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109886430B (en) | Equipment health state assessment and prediction method based on industrial big data | |
| CN115526375A (en) | Space flight equipment predictive maintenance system | |
| US6871160B2 (en) | Intelligent condition-based engine/equipment management system | |
| CN110320892A (en) | The sewage disposal device fault diagnosis system and method returned based on Lasso | |
| Garcia et al. | SIMAP: Intelligent System for Predictive Maintenance: Application to the health condition monitoring of a windturbine gearbox | |
| CN113156917B (en) | Power grid equipment fault diagnosis method and system based on artificial intelligence | |
| CN102520697B (en) | Onsite information preprocessing method of remote cooperative diagnosis | |
| CN112052979B (en) | Equipment spare part demand prediction system based on fault prediction and health management | |
| US20040049715A1 (en) | Computer networked intelligent condition-based engine/equipment management system | |
| CN111401583A (en) | Escalator full life cycle health management system based on predictive maintenance | |
| CN108873859B (en) | Bridge type grab ship unloader fault prediction model method based on improved association rule | |
| CN113339204B (en) | Wind driven generator fault identification method based on hybrid neural network | |
| CN109492790A (en) | Wind turbines health control method based on neural network and data mining | |
| CN116228186A (en) | Ship cabin intelligent operation and maintenance system based on human engineering | |
| CN113780375B (en) | Digital twinning-based virtual-real interaction wind power plant wind power monitoring system and application | |
| CN110757510A (en) | Method and system for predicting remaining life of robot | |
| CN114519923A (en) | Intelligent diagnosis and early warning method and system for power plant | |
| CN114002597A (en) | Motor fault diagnosis method and system based on GRU network stator current analysis | |
| Zhang et al. | A novel fault diagnosis method for wind turbine based on adaptive multivariate time-series convolutional network using SCADA data | |
| Wang | Key techniques in intelligent predictive maintenance (IPdM)–a framework of intelligent faults diagnosis and prognosis system (IFDaPS) | |
| Zhi | Remote performance evaluation, life prediction and fault diagnosis of RV reducer for industrial robot | |
| CN115600695B (en) | Fault diagnosis method for metering equipment | |
| CN116483042A (en) | Digital lean diagnosis method for lean production control platform | |
| Chen et al. | BIM-and IoT-Based Data-Driven Decision Support System for Predictive Maintenance of Building Facilities | |
| Gray et al. | A practical approach to the use of scada data for optimized wind turbine condition based maintenance |
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 |