CN114919559B - System and method for predicting residual service life of brake system based on digital twin - Google Patents
System and method for predicting residual service life of brake system based on digital twin Download PDFInfo
- Publication number
- CN114919559B CN114919559B CN202210782812.0A CN202210782812A CN114919559B CN 114919559 B CN114919559 B CN 114919559B CN 202210782812 A CN202210782812 A CN 202210782812A CN 114919559 B CN114919559 B CN 114919559B
- Authority
- CN
- China
- Prior art keywords
- data
- module
- braking system
- service life
- digital twin
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60T—VEHICLE BRAKE CONTROL SYSTEMS OR PARTS THEREOF; BRAKE CONTROL SYSTEMS OR PARTS THEREOF, IN GENERAL; ARRANGEMENT OF BRAKING ELEMENTS ON VEHICLES IN GENERAL; PORTABLE DEVICES FOR PREVENTING UNWANTED MOVEMENT OF VEHICLES; VEHICLE MODIFICATIONS TO FACILITATE COOLING OF BRAKES
- B60T17/00—Component parts, details, or accessories of power brake systems not covered by groups B60T8/00, B60T13/00 or B60T15/00, or presenting other characteristic features
- B60T17/18—Safety devices; Monitoring
- B60T17/22—Devices for monitoring or checking brake systems; Signal devices
- B60T17/228—Devices for monitoring or checking brake systems; Signal devices for railway vehicles
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60T—VEHICLE BRAKE CONTROL SYSTEMS OR PARTS THEREOF; BRAKE CONTROL SYSTEMS OR PARTS THEREOF, IN GENERAL; ARRANGEMENT OF BRAKING ELEMENTS ON VEHICLES IN GENERAL; PORTABLE DEVICES FOR PREVENTING UNWANTED MOVEMENT OF VEHICLES; VEHICLE MODIFICATIONS TO FACILITATE COOLING OF BRAKES
- B60T17/00—Component parts, details, or accessories of power brake systems not covered by groups B60T8/00, B60T13/00 or B60T15/00, or presenting other characteristic features
- B60T17/18—Safety devices; Monitoring
- B60T17/22—Devices for monitoring or checking brake systems; Signal devices
- B60T17/221—Procedure or apparatus for checking or keeping in a correct functioning condition of brake systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Transportation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Hardware Design (AREA)
- Mechanical Engineering (AREA)
- Automation & Control Theory (AREA)
- Aviation & Aerospace Engineering (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a system and a method for predicting the residual service life of a braking system based on digital twin. The digital twin module of the braking system can accurately reflect the actual physical characteristics of the braking system, is used for multi-dimensional dynamic modeling of equipment, and interactively updates the model of the system; the residual service life prediction module predicts the residual service life of the braking system; the maintenance decision suggestion module makes maintenance decision suggestions according to the residual service life prediction; the data visualization module performs visualization operation, so that the whole life cycle management of the brake system is convenient to realize. The invention effectively combines the function of digital twin full life cycle management, improves the accuracy of residual service life prediction and improves the reliability of a braking system.
Description
Technical Field
The invention relates to the technical field of train braking, in particular to a system and a method for predicting the residual service life of a braking system based on digital twinning.
Background
The brake system is used as a core component of a plurality of vehicles such as locomotives, high-speed trains and the like, and has safe, quick, high-efficiency and high-standard service in the running process of the trains, however, as the number of actions increases, the system can be gradually aged, performance is degraded until failure can not run due to the influence of various uncertainty factors such as the external environment and the fatigue use of system parts. However, the following problems exist in the current full life cycle monitoring for the aging state of the brake system, equipment diagnosis maintenance information and the like: (1) The running state of the braking system is changed at all times, and an update barrier exists between the historical monitoring information and the real-time detection information, so that serious hysteresis exists in the overall information update of the system; (2) The intelligent degree of operation and maintenance management of the braking system is low, the source of a system operation data set is single, the variety and the quantity of the data sets are small, and the intelligent of operation and maintenance management is difficult to improve.
Digital twinning employs a data-driven approach to update, modify, connect and supplement mathematical models by adding system history and real-time operational data. By constructing the mirror image of the physical entity, the system mechanism and the running data real-time dynamic evaluation system are integrated, so that management personnel can perform intelligent regulation and decision in the operation and maintenance management process of the brake system. The brake system is used as key equipment in the running process of locomotives and high-speed trains, the prediction of the residual service life is always the key point of research, the whole life cycle management digital twin body of the brake system is constructed, the real-time data of the whole running of the equipment can be integrated, the hysteresis of the whole information updating of the system is made up, the accuracy of the prediction of the residual service life is improved, the whole life cycle management of the brake system is realized, and the reliability of the brake system is improved.
Disclosure of Invention
In order to solve the technical problems of the existing brake system in residual service life prediction, the invention provides a digital twin-based brake system residual service life prediction system and a digital twin-based brake system residual service life prediction method, which can effectively improve the accuracy of the brake system residual service life prediction.
The technical scheme provided by the invention is as follows: the system comprises a data acquisition module, a data preprocessing module, a braking system digital twin module, a residual service life prediction module, a maintenance decision suggestion module and a data visualization module;
the data acquisition module is used for acquiring monitoring data of the bottom layer sensor from a physical entity of the braking system and transmitting the monitoring data to the data preprocessing module;
the data preprocessing module is used for receiving and processing physical entity monitoring data and storing the processed data;
the digital twin module of the braking system can accurately reflect the actual physical characteristics of the equipment, is used for multi-dimensional dynamic modeling of the equipment, and interactively updates the model of the system;
the residual service life prediction module is used for carrying out feature extraction and feature learning calculation on real-time data of the digital twin model by using a deep learning method to predict the residual service life of the braking system;
the maintenance decision suggestion module is used for fusing the residual service life predicted value with the data of the digital twin module of the braking system and outputting maintenance decision suggestions;
the data visualization module invokes maintenance decision suggestions and system state related information to perform visual display, provides further maintenance decisions and interactive operation, and feeds back the current system state to guide the physical entity to perform maintenance update.
Further, the physical entity parameters of the braking system of the data acquisition module comprise feedback types, command types and control types, and the feedback types comprise: a cylinder equalizing target pressure value Pe, a brake cylinder target pressure value Pb, a total wind pressure value Pta, a train pipe pressure value Ptp, a total wind flow value Fta, a BCU current value Ib, a BCU voltage value Vb; the command class parameters include: emergency braking command Ceb, penalty braking command Cpb, system self-test command Cst, sensor calibration command Csc; the control class parameters include: high-speed solenoid valve group Cve for controlling the pressure of the cylinder and high-speed solenoid valve group Cvb for controlling the pressure of the brake cylinder control the brake valve to be cut off Cvr.
Further, the digital twin module of the braking system comprises a data driving module, a super-realistic simulation environment module and a real-time twin model;
the data driving module is used for analyzing the data of each monitoring module, the state of an input and output passage, the position of a brake handle, balancing and a brake cylinder target pressure value, and carrying out data preprocessing, feature extraction, data labelling and structuring on the data; the super-realistic simulation environment module is used for running a multi-physical-field simulation model of the braking system and analyzing the fault types and fault mechanisms of the monitoring data conversion; the real-time twin model utilizes the data processed by the data driving module to carry out real-time comparison verification with the simulation model of the super-realistic simulation environment module, so that an interactive comparison data stream between the twin model and the physical entity is formed, the running state of the physical entity is continuously approximated, and the real-time correction of the twin model is realized.
Further, the digital twin module of the braking system establishes a three-dimensional structure simulation visualization model corresponding to the braking system according to various structural parameters of a braking control unit, a valve, a cylinder, a sensor, an integrated gas circuit pipeline board and a braking controller in a cab of the braking system, circuit gas circuit attribute parameters and multiple working conditions of the braking system, and combines the electrical action relation of the braking control unit, and receives state data parameters processed in real time to enable the digital twin simulation model to be consistent with the physical entity state.
A method for predicting the residual service life of a brake system based on digital twinning comprises the following steps:
S 1 according to the data acquired by the data acquisition module, the original data is transmitted to the data preprocessing module for processing by utilizing the transmission technology of the Internet of things;
S 2 constructing a corresponding digital twin module of the braking system according to the physical entity; the digital twin module of the braking system comprises a data driving module, a super-realistic simulation environment module and a real-time twin model;
s3, constructing a three-dimensional structure visualization model of the braking system by using CAD, AMESIM or ANSYS software, dynamically rendering the whole appearance of the braking system, receiving the preprocessed data through a data driving module, and interactively mapping the twin real-time model;
S 4 normalizing the values of the system parameters by adopting a min-max method;
S 5 and inputting various parameter data of the digital twin module of the braking system into a residual service life prediction module for prediction.
Further, the method also comprises a step S6, wherein the maintenance decision suggestion module outputs maintenance decision suggestions according to the residual service life value output by the residual service life prediction module and according to the components and the system state output by the digital twin module of the braking system.
Further, step S7 is further included, the data visualization module provides a man-machine interaction interface for displaying related information of the whole life cycle of the braking system, displaying the current braking system diagnosis information and running state information through the user interface, and providing further diagnosis decision operation options and man-machine operation interaction.
The invention has the beneficial effects that:
1. the invention combines the relativity of the real-time data and the historical data, provides the state data of each stage of the whole life cycle management to construct the digital mirror image of the physical entity, and compares the method for predicting the residual service life of the braking system by adopting the historical data or the experimental data;
2. compared with the traditional prediction of the residual service life of the braking system, the method can well utilize the advantages of digital twinning, constructs the digital twinning body perceived by real-time situation through fusing multidimensional monitoring data and a dynamic simulation model, synthesizes various monitoring data of the whole body, compensates the disadvantages of insufficient data samples in the past, and is beneficial to realizing the accurate prediction of the running state, the health state and the residual service life of the braking system.
Drawings
Fig. 1 is a schematic diagram of a system structure according to the present invention.
FIG. 2 is a schematic flow diagram of a digital twin module of the brake system of the present invention.
FIG. 3 is a flow chart of a remaining life prediction module according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
As shown in FIG. 1, the invention comprises a data acquisition module, a data preprocessing module, a braking system digital twin module, a residual service life prediction module, a maintenance decision suggestion module and a data visualization module. The functions of the respective modules are as follows: the data acquisition module is used for acquiring bottom sensor monitoring data from a physical entity of the braking system and transmitting the bottom sensor monitoring data to the data preprocessing module; the data preprocessing module is used for receiving and processing physical entity monitoring data and storing the processed data; the digital twin module of the braking system can accurately reflect the actual physical characteristics of the equipment, is used for multi-dimensional dynamic modeling of the equipment, and interactively updates the model of the system; the residual service life prediction module is used for carrying out feature extraction and feature learning calculation on real-time data of the digital twin model by using a deep learning method to predict the residual service life of the braking system; the maintenance decision suggestion module is used for fusing the residual service life predicted value with the data of the digital twin module of the braking system and outputting maintenance decision suggestions; the data visualization module invokes maintenance decision suggestions and system state related information to perform visual display, provides further maintenance decisions and interactive operation, and feeds back the current system state to guide the physical entity to perform maintenance update.
The above modules complement each other, data driving is used as a main component, simulation is used as an auxiliary component, a digital twin body of the braking system is constructed, the running state information of the whole system is mapped into a digital mirror image, closed loop feedback processing of the whole system is formed, full life cycle management of the braking system is realized, and the residual service life of the braking system is accurately predicted.
Further, the related parameters of the physical entity of the braking system of the data acquisition module comprise feedback types, command types and control types. The feedback class parameters include: a cylinder equalizing target pressure value Pe, a brake cylinder target pressure value Pb, a total wind pressure value Pta, a train pipe pressure value Ptp, a total wind flow value Fta, a BCU current value Ib, a BCU voltage value Vb; the command class parameters include: emergency braking command Ceb, penalty braking command Cpb, system self-test command Cst, sensor calibration command Csc; the control class parameters include: high-speed solenoid valve group Cve for controlling the pressure of the cylinder and high-speed solenoid valve group Cvb for controlling the pressure of the brake cylinder control the brake valve to be cut off Cvr.
Further, the digital twin module of the braking system comprises a data driving module, a super-realistic simulation environment module and a real-time twin model.
The data driving module is used for analyzing the data of each monitoring module, the state of an input/output passage, the position of a brake handle, equalization and a brake cylinder target pressure value, performing data preprocessing, feature extraction, data tagging and structuring on the data, and analyzing the data by adopting one or more algorithms which are arbitrarily combined and comprise an artificial intelligent algorithm such as a BP neural network algorithm, an LSTM deep learning algorithm and the like; the super-realistic simulation environment module is used for running a multi-physical-field simulation model of the braking system, adopts one or a plurality of model combinations in modeling software such as MATLAB, CAD, AMESIM, ANSYS and the like to construct a model, and analyzes the fault types and fault mechanisms of the monitoring data conversion; the real-time twin model utilizes the data processed by the data driving module to carry out real-time comparison verification with the simulation model of the super-realistic simulation environment module, so that an interactive comparison data stream between the twin model and the physical entity is formed, the running state of the physical entity is continuously approximated, and the real-time correction of the twin model is realized.
The simulation system comprises a simulation module, a conversion valve, a brake screen and a controller, wherein the simulation module is used for simulating a brake screen, the simulation module is used for simulating the brake screen, the conversion valve is used for realizing the functions of manually cutting off and linking a brake screen and a brake screen, and the simulation module is controlled by the controller of MATLAB; the simulation platform realizes the function of an equalization module through a two-position three-way electromagnetic valve, and then simulates the flow of the electromagnetic valve according to the Cv value of the actual conversion valve; the relay valve and the pressure reducing valve reproduce the functions of the relay valve and the pressure reducing valve in the simulation platform through the functions of the components of the schematic diagram.
Constructing a digital twin model of various faults through the fault injection function of the simulation model, wherein the fault types comprise: solenoid valve failure, controlled valve failure, cylinder line failure, sensor failure. The solenoid valve failure includes: the brake electromagnetic valve is accidentally actuated/disconnected, the protection electromagnetic valve is accidentally actuated/disconnected, the inflation electromagnetic valve is accidentally actuated/disconnected, the exhaust electromagnetic valve is accidentally actuated/disconnected, the neutral valve is accidentally actuated/applied, and the like; the controlled valve failure includes: ER/BP relay valve clamping piece, ER/BP relay valve air charging speed slow, ER/BP relay valve air leakage, ER/BP relay valve insensitivity, BC distributing valve clamping piece, BC distributing valve air charging speed slow, BC distributing valve air leakage, BC distributing valve insensitivity and the like; the cylinder line fault includes: equalizing reservoir leakage, train pipe leakage, brake cylinder average pipe leakage, etc.; the sensor failure includes: brake cylinder 1 sensor failure, brake cylinder 2 sensor failure, total wind sensor failure, train pipe sensor failure, equalizing reservoir sensor failure, brake cylinder pre-control sensor failure, service pipe sensor failure, flow sensor failure, etc. And simulating the influence of different fault degrees on the gas circuit by adjusting corresponding parameters, acquiring a large number of simulation data sets from simulation, performing sample training with the real-time data sets, supplementing a digital twin sample set, and interactively synchronizing to a digital twin model of the braking system.
Further, the digital twin module of the braking system establishes a three-dimensional structure simulation visualization model corresponding to the braking system according to various structural parameters of a braking control unit, a valve, a cylinder, a sensor, an integrated gas circuit pipeline board and a braking controller in a cab of the braking system, circuit gas circuit attribute parameters and multiple working conditions of the braking system, combines the electrical action relation of the braking control unit, and receives state data parameters processed in real time to enable the digital twin simulation model to be consistent with the physical entity state.
The model of the digital twin module of the braking system adopts a CAD drawing method, a ANSYS Twin Builder construction method and an approximate numerical analysis method to digitally model each component in the braking system, and builds a complete digital twin model in a simulation platform.
The model running state data of the digital twin module of the braking system comprises: structured operational data, health prediction data. The structured operating data includes pressure, flow, current, voltage, and brake system transfer functions; the health prediction data includes a number of remaining useful lives and a system health status.
The model structure parameters of the digital twin module of the braking system comprise: total reservoir volume, equalization reservoir volume, operating position pressure, primary position pressure, full position pressure, emergency position pressure, equalization reservoir rise time from 0kPa to 600kPa, etc.
The model multi-working condition parameters of the digital twin module of the braking system comprise: the number of braking, the braking pressure, the number of solenoid valve use, the number of pipeline ventilation, etc.
Further, the maintenance decision suggestion module automatically searches the expert experience database for corresponding maintenance decision suggestions according to the residual service life value output by the residual service life prediction module and according to the components and the system state output by the digital twin module of the braking system.
Further, the data visualization module provides a man-machine interaction interface for displaying related information of the whole life cycle of the braking system, displaying the current braking system diagnosis information and running state information through a user interface, and providing further diagnosis decision operation options and man-machine operation interaction.
A method for predicting the residual service life of a brake system based on digital twinning comprises the following steps:
S 1 and according to the data acquired by the data acquisition module, transmitting the original data to the data preprocessing module for processing by utilizing the transmission technology of the Internet of things. The processing comprises the following steps: noise processing, missing value processing, outlier processing, and structuring processing. The noise processing uses smooth filtering or median filtering to filter useless noise signals received by the sensor in the acquisition process; the missing value processing uses a missing judging method to judge the NaN value or null value in the transmission data, and fills in a default value to process the missing value; the outlier processing uses 6 sigma criterion to judge outlier data, cleans abnormal data, and fills abnormal data positions by using median and average; the structuring process loads data into the structure body according to a digital twin definition structure protocol and a protocol rule;
S 2 constructing a corresponding digital twin module of the braking system according to the physical entity; the digital twin module of the braking system comprises a data driving module, a super-realistic simulation environment module and a real-time twin model;
the S is 2 Concrete embodimentsThe steps are as shown in fig. 2, including the steps of:
S 21 and establishing a model of a brake cylinder pressure control subsystem, wherein the pressure characteristic formula of the brake cylinder is as follows:
wherein the method comprises the steps of
Is the pressure of the brake cylinder, K is the general gas constant, T is the temperature of the gas, V is the total volume of the brake cylinder, f m Is the mass flow of compressed gas through the valve, and the formula is as follows:
wherein the method comprises the steps of
Is the pressure of the total reservoir, G 1 ∈[0,1]Is the valve flow coefficient 1, G 2 ∈[0,1]Is the valve flow coefficient 2 and s is the valve cross-sectional area.
By combining the two formulas, the differential pressure formula of the brake cylinder is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,
S 22 the data accessed through the data driving module is linked with the mathematical model, and the parameter G which is difficult to directly measure is fitted on line by using the preprocessed data 1 And G 2 . The invention applies least square method to estimate parameters on line, and the formula is as follows:
g(n)＝λ -1 H(n-1)ξ(n)[1+λ -1 ξ T (n)H(n-1)ξ] -1
H(n)＝λ -1 H(n-1)-λ -1 g(n)ξ T (n)H(n-1)
where λ is the fitting constant, g (n) is the gain vector, H (n) is the correlation matrix,
the method is characterized in that the method is a vector of estimated parameters, G (n) is a parameter to be measured, and a digital twin data set is supplemented and mapped to a digital twin model by inputting preprocessing data and continuously fitting parameters which are difficult to measure in a characteristic formula;
S 23 in a simulation environment, building a simulation module of each component of the braking system; the simulation module is constructed by a simulation module, the simulation module comprises a simulation module, a relay valve, a pressure reducing valve and the like, the simulation module uses an electromagnetic valve provided by software in a simulation platform to simulate, the conversion valve realizes the functions of manually cutting off and linking a balancing air cylinder and a brake screen balance pipe in a brake system, the control function is controlled by a simulation controller of MATLAB in the simulation, the function of the simulation module is realized by a two-position three-way electromagnetic valve, and then the flow of the electromagnetic valve is simulated according to the Cv value of an actual conversion valve; the relay valve and the pressure reducing valve form a complete function in the simulation platform through the components of the schematic diagram, and then the simulation function of the valve is completed through simulating the pressure flow;
S 24 the influence of different fault degrees on the gas circuit, the components and the system is simulated by adjusting corresponding parameters, a large number of simulation data sets are obtained, sample training is carried out on the simulation data sets and the real-time data sets, and the digital twin sample sets are continuously supplemented; the fault types in the simulation include: solenoid valve failure, controlled valve failure, cylinder line failure, sensor failure. The solenoid valve failure includes: unexpected actuation/disconnection of brake solenoid valve, relief of unexpected actuation/disconnection of solenoid valve, protection of unexpected actuation/disconnection of solenoid valve, and inflationUnexpected actuation/de-actuation of the solenoid valve, unexpected actuation/de-actuation of the exhaust solenoid valve, unexpected relief/application of the neutral valve, etc.; the controlled valve failure includes: ER/BP relay valve clamping piece, ER/BP relay valve air charging speed slow, ER/BP relay valve air leakage, ER/BP relay valve insensitivity, BC distributing valve clamping piece, BC distributing valve air charging speed slow, BC distributing valve air leakage, BC distributing valve insensitivity and the like; the cylinder line fault includes: equalizing reservoir leakage, train pipe leakage, brake cylinder average pipe leakage, etc.; the sensor failure includes: brake cylinder 1 sensor failure, brake cylinder 2 sensor failure, total wind sensor failure, train pipe sensor failure, equalizing reservoir sensor failure, brake cylinder pre-control sensor failure, service pipe sensor failure, flow sensor failure, etc.
S 25 Searching a corresponding real-time twin module according to the braking fault position, the fault type and the fault type simulation model information of the super-realistic simulation environment module, supplementing a digital twin data sample, and interactively synchronizing a digital twin model of the braking system;
S 26 according to various physical structure parameters, circuit and gas path attribute parameters and multi-working-condition parameters of a brake system of a brake control unit, a valve, a cylinder, a sensor, an integrated gas path pipeline board and a brake controller in a cab of the brake system, a three-dimensional structure simulation visual model corresponding to the brake system is established in combination with the electrical action relation of the brake control unit, and real-time state data is received to enable the digital twin simulation model to be consistent with a physical entity; wherein the physical structure parameters include: total reservoir volume, equalization reservoir volume, operating position pressure, primary position pressure, full position pressure, emergency position pressure, equalization reservoir rise time from 0kPa to 600kPa, etc.; the multiple working condition parameters include: the braking times, the braking pressure, the using times of the electromagnetic valve, the pipeline ventilation times and the like;
S 3 and constructing a three-dimensional structure visualization model of the braking system by using CAD, AMESIM, ANSYS software, dynamically rendering the whole appearance of the braking system, receiving the preprocessed data through a data driving module, and interactively mapping the twin real-time model. The implementation method is as follows: first of all the physical entity of the braking systemAfter receiving a data request of a data acquisition module, sending all original parameters of a physical entity to a data preprocessing module for data preprocessing, and storing preprocessed data into a database; and secondly, transmitting the preprocessed data to a digital twin module to realize interactive mapping of physical entities and digital twin, and interactively transmitting the digital twin three-dimensional model and the data to form a digital twin real-time mapping relation.
S 4 Because the value ranges of the system parameters are very different, normalization processing is needed to be carried out on the parameters, the invention adopts a min-max method to carry out normalization processing and outputs the parameters x in digital twin operation i ＝{x 1 ,x 2 ,...,x n Normalized as shown in the formula:
wherein the method comprises the steps of
Is normalized data ranging from [0,1 ]]；min(x i ) And max (x) i ) Is x i N is the length of the signal parameter;
S 5 inputting various parameter data of the digital twin module of the braking system into a residual service life prediction module for prediction;
the S is 5 As shown in fig. 3, comprising the steps of:
S 51 and establishing a mapping relation between the residual service life and the parameter data of the braking system, wherein the formula is as follows:
RUL＝f(x t ),
wherein x is t Is real-time data of a digital twin model of the braking system during operation, f is a mapping function, RUL is a real-time residual service life predicted by f, n is the length of a signal parameter,
S 52 and extracting the characteristics of the parameter signals by using a deep convolutional neural network. Firstly, transforming an original current signal of a digital twin module of a braking system by adopting ensemble empirical mode decomposition to obtain frequency domain energy, obtaining 8 different IMF components by adopting ensemble empirical mode decomposition, solving the energy entropy of the 8 IMF components, segmenting the energy entropy and other parameter signal data, wherein the segmented braking system parameter sequence is x i ＝{x 1 ,x 2 ,...,x n Where n is the length of the signal parameter. The convolution operation for each sequence is defined as follows:
fea i ＝g(cx i +b)
where b and g represent the bias term and the nonlinear activation function, respectively. Convolutional layer output fea i Is derived from x by a convolution kernel c i Is obtained by sliding the first point to the last point, is the convolution kernel in the corresponding sequence x i The learned features.
The pooling layer is applied to the output features generated by the convolution layer, the most important local information in each feature map is extracted in a pooling mode, feature dimensions are reduced, the maximum pooling function is used in the invention, and the pooling operation of each sequence is expressed as follows:
where r is the length of the pooling window, fo i Representing the output of the ith sequence feature after the pooling operation, and s represents the dimension;
S 53 in order to make the model acquire useful information from the multi-element parameter sequence as much as possible, the invention adopts a sliding window with 50 size to do fo on the data i Segmentation pretreatment, wherein the sliding step length is 1;
S 54 the invention adopts a two-way long-short-period memory network to realize the prediction of the residual service life, and the two-way long-short-period memory network is adopted for the prediction of the residual service lifeThe memory network is an improvement on a long-term memory network, can capture more valuable information, and has the following specific calculation formula:
f t ＝v(W f x t +K f y t-1 +b f )
i t ＝σ(W i x t +K i y t-1 +b i )
o t ＝σ(W o x t +K o y t-1 +b o )
wherein W is f ，W i ，W o And W is e Representing weights between three gates and cells; k (K) f 、K i 、K o And K c Weights representing loop connections; b f 、b i 、b o And b c Representing a bias vector; y is the state of the hidden layer; c (C) t And C t-1 Representing the output states of the current and last cell, respectively;
representing a cell input state; />
Multiplication of representative elements; tan h is the hyperbolic tangent function and σ is the sigmoid activation function.
The core structure of the two-way long-short-term memory network is to combine two opposite long-short-term memory network layers to process the time sequence, output is calculated by connecting the forward and backward long-term memory network outputs, and the front-rear long-term dependency relationship of the current state of the time sequence is captured. The hidden state calculation method of the forward and backward layers is as follows:
S 55 To prevent overfitting, dropout is used to increase the generalization ability of the model; and update the model parameters using Adam optimization algorithm. In the invention, dropout is 0.5, learning rate is 0.001, batch-size is 512, and maximum training period is set to 200. And (3) the processed characteristic data passes through the two-way long-short-term memory network to obtain the output RUL with the residual service life.
S 56 The invention adopts Score index to verify the effectiveness of the method, and is defined as follows:
where n is the number of test samples, RUL' i Is a predicted RUL i Is the true RUL for the i-th sample. For Score index, the lower the value, the higher the prediction accuracy.
S 57 Comparing the model Score values, and repeating S if the model Score values do not meet the preset Score values 54 ；
S 58 And structuring the residual service life prediction data meeting the system requirements, and updating the residual service life prediction data into a digital twin module of the braking system to ensure that a physical entity and digital twin are kept consistent in real time.
S 6 The maintenance decision suggestion module gives maintenance decision suggestions according to the residual service life value output by the residual service life prediction module, the component and system state output by the digital twin module;
S 7 the data visualization module provides a man-machine interaction interface for displaying related information of the whole life cycle of the braking system, displaying maintenance decision information and running state information through a user interface, and providing further maintenance decision operation options and man-machine operation interaction.
Claims (4)
1. A method for predicting the residual service life of a brake system based on digital twinning is characterized by comprising the following steps: the prediction method is completed by depending on a prediction system; the prediction system comprises a data acquisition module, a data preprocessing module, a braking system digital twin module, a residual service life prediction module, a maintenance decision suggestion module and a data visualization module;
the data acquisition module is used for acquiring monitoring data of the bottom layer sensor from a physical entity of the braking system and transmitting the monitoring data to the data preprocessing module;
the data preprocessing module is used for receiving and processing physical entity monitoring data and storing the processed data;
the digital twin module of the braking system can accurately reflect the actual physical characteristics of the equipment, is used for multi-dimensional dynamic modeling of the equipment, and interactively updates the model of the system;
the residual service life prediction module is used for carrying out feature extraction and feature learning calculation on real-time data of the digital twin model by using a deep learning method to predict the residual service life of the braking system;
the maintenance decision suggestion module is used for fusing the residual service life predicted value with the data of the digital twin module of the braking system and outputting maintenance decision suggestions;
the data visualization module invokes maintenance decision suggestions and system state related information to perform visual display, provides further maintenance decisions and interactive operation, and feeds back the current system state to guide a physical entity to perform maintenance update;
the prediction method comprises the following steps:
S 1 according to the data acquired by the data acquisition module, the original data is transmitted to the data preprocessing module for processing by utilizing the transmission technology of the Internet of things;
S 2 constructing a corresponding digital twin module of the braking system according to the physical entity; the digital twin module of the braking system comprises a data driving module, a super-realistic simulation environment module and a real-time twin model; the method comprises the following specific steps:
S 21 and establishing a model of a brake cylinder pressure control subsystem, wherein the pressure characteristic formula of the brake cylinder is as follows:
wherein the method comprises the steps of
Is the pressure of the brake cylinder, K is the general gas constant, T is the temperature of the gas, V is the total volume of the brake cylinder, f m Is the mass flow of compressed gas through the valve,
S 22 the data accessed through the data driving module is connected with a model of a brake cylinder pressure control subsystem, and the parameter G which is difficult to directly measure is fitted on line by using the preprocessed data 1 And G 2 The least square method is adopted to estimate parameters on line, and the formula is as follows:
g(n)＝λ -1 H(n-1)ξ(n)[1+λ -1 ξ T (n)H(n-1)ξ] -1
H(n)＝λ -1 H(n-1)-λ -1 g(n)ξ T (n)H(n-1)
where λ is the fitting constant, g (n) is the gain vector, H (n) is the correlation matrix,
the method is characterized in that the method is a vector of estimated parameters, G (n) is a parameter to be measured, and a digital twin data set is supplemented and mapped to a digital twin model by inputting preprocessing data and continuously fitting parameters which are difficult to measure in a characteristic formula;
S 23 in a simulation environment, building a simulation module of each component of the braking system;
S 24 the influence of different fault degrees on the gas circuit, the components and the system is simulated by adjusting corresponding parameters, a simulation data set is obtained, sample training is carried out on the simulation data set and the real-time data set, and a digital twin sample set is continuously supplemented;
S 25 searching a corresponding real-time twin module according to the braking fault position, the fault type and the fault type simulation model information of the super-realistic simulation environment module, supplementing a digital twin data sample, and interactively synchronizing a digital twin model of the braking system;
S 26 according to various physical structure parameters, circuit and gas path attribute parameters and multi-working-condition parameters of a brake system of a brake control unit, a valve, a cylinder, a sensor, an integrated gas path pipeline board and a brake controller in a cab of the brake system, a three-dimensional structure simulation visual model corresponding to the brake system is established in combination with the electrical action relation of the brake control unit, and real-time state data is received to enable the digital twin simulation model to be consistent with a physical entity;
S 3 constructing a three-dimensional structure simulation visual model of the braking system by using CAD, AMESIM or ANSYS software, dynamically rendering the whole appearance of the braking system, receiving the preprocessed data by a data driving module, and interactively mapping the twin real-time model;
S 4 normalizing the values of the system parameters by adopting a min-max method;
S 5 will brakeAnd inputting various parameter data of the system digital twin module into a residual service life prediction module for prediction.
2. A method for predicting the remaining service life of a digitally twinned brake system as claimed in claim 1, wherein: the S is 5 The method comprises the following steps:
S 51 establishing a mapping relation between the residual service life and the parameter data of the braking system;
S 52 extracting characteristics of the parameter signals by using a deep convolutional neural network; transforming an original current signal of a digital twin module of a braking system by adopting ensemble empirical mode decomposition to obtain frequency domain energy, obtaining 8 different IMF components by adopting ensemble empirical mode decomposition, solving the energy entropy of the 8 IMF components, segmenting the energy entropy and other parameter signal data, wherein the segmented braking system parameter sequence is x i ＝{x 1 ,x 2 ,...,x n -where n is the length of the signal parameter; the convolution operation for each sequence is defined as follows:
fea i ＝g(cx i +b)
wherein b and g represent bias terms and nonlinear activation functions, respectively, the convolutional layer output fea i Is derived from x by a convolution kernel c i Is obtained by sliding the first point to the last point, is the convolution kernel in the corresponding sequence x i The characteristics of the features that were learned in the above,
the pooling layer is applied to the output features generated by the convolution layer, the most important local information in each feature map is extracted in a pooling mode, feature dimensions are reduced, the maximum pooling function is used, and the pooling operation of each sequence is expressed as follows:
where r is the length of the pooling window, fo i Representing the output of the ith sequence feature after the pooling operation, and s represents the dimension;
S 53 the data is subjected to segmentation pretreatment by adopting a sliding window;
S 54 the residual service life prediction is realized by adopting a two-way long-short-period memory network; the specific calculation formula is as follows:
f t ＝σ(W f x t +K f y t-1 +b f )
i t ＝σ(W i x t +K i y t-1 +b i )
o t ＝σ(W o x t +K o y t-1 +b o )
wherein W is f ，W i ，W o And W is c Representing weights between three gates and cells; k (K) f 、K i 、K o And K c Weights representing loop connections; b f 、b i 、b o And b c Representing a bias vector; y is the state of the hidden layer; c (C) t And C t-1 Representing the output states of the current and last cell, respectively;
representing a cell input state; />
Multiplication of representative elements; tan h is a hyperbolic tangent function, σ is a sigmoid activation function;
the core structure of the two-way long-short-term memory network is to combine two opposite long-short-term memory network layers to process a time sequence, output is calculated by connecting a forward long-term memory network output and a backward long-term memory network output, the front-rear long-term dependency relationship of the current state of the time sequence is captured, and the hidden state calculation method of the forward layer and the backward layer is as follows:
S 55 the generalization capability of the model is improved by using dropout; updating model parameters by using an Adam optimization algorithm;
S 56 a Score index is adopted to verify the effectiveness of the method;
S 57 comparing the model Score values, and repeating S if the model Score values do not meet the preset Score values 54 ；
S 58 And structuring the residual service life prediction data meeting the system requirements, and updating the residual service life prediction data into a digital twin module of the braking system to ensure that a physical entity and digital twin are kept consistent in real time.
3. The method for predicting remaining service life of a digitally twinned brake system according to claim 1, wherein: and step S6, the maintenance decision suggestion module outputs maintenance decision suggestions according to the residual service life value output by the residual service life prediction module and according to the components and the system state output by the digital twin module of the braking system.
4. A method for predicting the remaining service life of a digitally twinned brake system as claimed in claim 1, wherein: and step S7, the data visualization module provides a man-machine interaction interface for displaying related information of the whole life cycle of the braking system, displaying the current braking system diagnosis information and running state information through a user interface, and providing further diagnosis decision operation options and man-machine operation interaction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210782812.0A CN114919559B (en) | 2022-07-05 | 2022-07-05 | System and method for predicting residual service life of brake system based on digital twin |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210782812.0A CN114919559B (en) | 2022-07-05 | 2022-07-05 | System and method for predicting residual service life of brake system based on digital twin |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114919559A CN114919559A (en) | 2022-08-19 |
| CN114919559B true CN114919559B (en) | 2023-05-12 |
Family
ID=82816057
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210782812.0A Active CN114919559B (en) | 2022-07-05 | 2022-07-05 | System and method for predicting residual service life of brake system based on digital twin |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114919559B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116050190B (en) * | 2023-03-31 | 2023-06-23 | 中国海洋大学 | Product performance and degradation state analysis method based on digital twinning |
| CN116127821A (en) * | 2023-04-14 | 2023-05-16 | 浪潮软件科技有限公司 | Three-dimensional visual presentation method and platform for operation and maintenance data |
| CN117952323A (en) * | 2024-03-26 | 2024-04-30 | 大连豪森软件有限公司 | Product creation system, method, equipment and medium based on digital twin |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111399474B (en) * | 2020-02-29 | 2021-05-14 | 中南大学 | Health index-based life prediction method and device for balance control module |
| CN112528533B (en) * | 2020-11-19 | 2022-02-25 | 中国矿业大学 | Method for intelligently evaluating reliability and predicting service life of brake of kilometer deep well elevator |
| CN112904220B (en) * | 2020-12-30 | 2021-12-31 | 厦门大学 | UPS (uninterrupted Power supply) health prediction method and system based on digital twinning and machine learning, electronic equipment and storable medium |
| CN113119937B (en) * | 2021-03-31 | 2022-04-08 | 南京航空航天大学 | Intelligent brake-by-wire system based on digital twins and prediction control method thereof |
| CN113567858A (en) * | 2021-07-30 | 2021-10-29 | 北京航空航天大学 | Control moment gyro residual life prediction system |
-
2022
- 2022-07-05 CN CN202210782812.0A patent/CN114919559B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114919559A (en) | 2022-08-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114919559B (en) | System and method for predicting residual service life of brake system based on digital twin | |
| Luo et al. | Convolutional neural network based on attention mechanism and Bi-LSTM for bearing remaining life prediction | |
| Byington et al. | Data-driven neural network methodology to remaining life predictions for aircraft actuator components | |
| CN112131760A (en) | CBAM model-based prediction method for residual life of aircraft engine | |
| CN111539515B (en) | Complex equipment maintenance decision method based on fault prediction | |
| CN111596604A (en) | Intelligent fault diagnosis and self-healing control system and method for engineering equipment based on digital twinning | |
| Chen et al. | Application of generalized frequency response functions and improved convolutional neural network to fault diagnosis of heavy-duty industrial robot | |
| CN108053052B (en) | A kind of oil truck oil and gas leakage speed intelligent monitor system | |
| CN109242147A (en) | Signal fused fan condition prediction technique based on Bp neural network | |
| CN105825271A (en) | Satellite fault diagnosis and prediction method based on evidential reasoning (ER) | |
| CN113469470B (en) | Energy consumption data and carbon emission correlation analysis method based on electric brain center | |
| CN115438726A (en) | Device life and fault type prediction method and system based on digital twin technology | |
| CN116010900A (en) | Multi-scale feature fusion gearbox fault diagnosis method based on self-attention mechanism | |
| CN114266201B (en) | Self-attention elevator trapping prediction method based on deep learning | |
| CN115205782A (en) | Rice feeding robot cluster monitoring and management method based on digital twin technology | |
| CN116822652A (en) | Subway fault prediction method, subway fault prediction device, electronic equipment, subway fault prediction system and storage medium | |
| CN117252051A (en) | Cable tunnel monitoring and early warning method and system based on digital twinning | |
| CN115409369A (en) | Comprehensive energy system reliability evaluation method based on mechanism and data hybrid driving | |
| CN117371207A (en) | Extra-high voltage converter valve state evaluation method, medium and system | |
| Li et al. | A LSTM-based method for comprehension and evaluation of network security situation | |
| CN114720129B (en) | Rolling bearing residual life prediction method and system based on bidirectional GRU | |
| CN115330082A (en) | PM2.5 concentration prediction method of LSTM network based on attention mechanism | |
| CN109359733A (en) | A kind of dynamical system operating status modeling method based on variation self-encoding encoder | |
| CN112598186B (en) | Improved LSTM-MLP-based small generator fault prediction method | |
| Virk et al. | Fault prediction using artificial neural network and fuzzy logic |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |