CN110361180A - Intelligent train pantograph military service performance dynamic monitoring and appraisal procedure and its system - Google Patents
Intelligent train pantograph military service performance dynamic monitoring and appraisal procedure and its system Download PDFInfo
- Publication number
- CN110361180A CN110361180A CN201910676732.5A CN201910676732A CN110361180A CN 110361180 A CN110361180 A CN 110361180A CN 201910676732 A CN201910676732 A CN 201910676732A CN 110361180 A CN110361180 A CN 110361180A
- Authority
- CN
- China
- Prior art keywords
- pantograph
- service life
- train
- prediction
- model
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000012544 monitoring process Methods 0.000 title claims abstract description 42
- 230000036541 health Effects 0.000 claims abstract description 41
- 238000012423 maintenance Methods 0.000 claims description 40
- 238000012549 training Methods 0.000 claims description 31
- 230000032683 aging Effects 0.000 claims description 27
- 238000012360 testing method Methods 0.000 claims description 25
- 230000000630 rising effect Effects 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 20
- 238000012800 visualization Methods 0.000 claims description 13
- 230000006378 damage Effects 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000001514 detection method Methods 0.000 claims description 9
- 238000002474 experimental method Methods 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 6
- 230000003862 health status Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 5
- 230000008439 repair process Effects 0.000 claims description 5
- 238000013526 transfer learning Methods 0.000 claims description 5
- 230000000007 visual effect Effects 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 241000208340 Araliaceae Species 0.000 claims description 3
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 claims description 3
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims description 3
- 235000003140 Panax quinquefolius Nutrition 0.000 claims description 3
- 208000027418 Wounds and injury Diseases 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 235000008434 ginseng Nutrition 0.000 claims description 3
- 208000014674 injury Diseases 0.000 claims description 3
- 230000007787 long-term memory Effects 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 238000013139 quantization Methods 0.000 claims 1
- 239000000284 extract Substances 0.000 abstract description 4
- 230000005611 electricity Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 230000003203 everyday effect Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000013210 evaluation model Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000009798 acute exacerbation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000013016 damping Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003412 degenerative effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
-
- 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
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Educational Administration (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of intelligent train pantograph military service performance dynamic monitoring and appraisal procedure, the working parameters including obtaining train pantograph；Extract the health indicator timing of all parts of train pantograph；Prediction obtains the remaining service life of each component of pantograph and pantograph entirety and corrects；Obtain the military service performance dynamic monitoring assessment result of final train pantograph.The invention also discloses the systems for realizing intelligent train the pantograph dynamic monitoring of military service performance and appraisal procedure.The present invention can be realized the monitoring to pantograph service state and external influence factor, completing extraneous factor influences modeling to the uncertain of remaining life, it can be realized the predicting residual useful life to pantograph, pantograph military service performance is effectively improved, and high reliablity, applicability are good and assessment is comprehensively accurate.
Description
Technical field
Present invention relates particularly to a kind of intelligent train pantograph military service performance dynamic monitoring and appraisal procedure and its systems.
Background technique
With the development and the improvement of people's living standards of economic technology, it is mostly important that traffic has become modern society
One of component part.And with the development of technology, the railway systems in China have also obtained great development.
Power train is the prevailing traffic vehicle of China railways industry.And train pantograph is the important composition portion of train
Point, it is responsible for the task for train energy supply.The holding of good pantograph contact line relation is for ensureing that train operating safety is most important.Such as
Fruit pantograph breaks down, then it is unstable to will lead to train energy supply, even result in it is unexpected stop taking turns, and then it is reliable to influence train operation
Property.The military service performance dynamic monitoring of pantograph and appraisal procedure can monitor the military service performance of pantograph, and predict pantograph event
Barrier, provides maintenance decision foundation for pantograph administrative staff.
Existing pantograph monitoring system focuses on the military service performance monitoring in pantograph pan part, such as CN more
109,143,001 108680104 A etc. of A, CN of 108169616 A, CN, these systems, which have ignored, is integrally on active service shape to pantograph
The monitoring and assessment of state.Since these systems have ignored coupling of each component part of pantograph for pantograph military service performance
It influences, therefore accurately can not decline progress modeling and forecasting to by other component failure bring pantograph military service performance.In addition, this
A little systems have ignored influence of the extraneous factors such as weather, passenger capacity for pantograph military service performance.The naked leakage of pantograph is on train top
End, and the electric current of its conduction, voltage are influenced by train load, therefore the factors such as weather, passenger capacity in future time
Uncertainty can be brought to the military service performance of pantograph.Since these systems have ignored for these probabilistic modelings, because
This cannot achieve the Accurate Prediction to pantograph military service performance.
Summary of the invention
One of the objects of the present invention is to provide a kind of high reliablity, applicability is good and assesses comprehensive and accurate intelligent train
Pantograph military service performance dynamic monitoring and appraisal procedure.
The second object of the present invention be to provide it is a kind of realize the intelligent train pantograph military service performance dynamic monitoring with
The system of appraisal procedure.
This intelligent train pantograph military service performance dynamic monitoring provided by the invention and appraisal procedure, including walk as follows
It is rapid:
S1. the working parameters of train pantograph are obtained；
S2. according to the working parameters of the step S1 train pantograph obtained, all parts of train pantograph are extracted
Health indicator timing；
S3. the healthy timing indicator obtained according to step S2, it is whole that prediction obtains each component of pantograph and pantograph
Remaining service life；
S4. the remaining service life of the step S3 each component of pantograph predicted and pantograph entirety is repaired
Just；
S5. the correction result obtained according to step S4, the military service performance dynamic monitoring for obtaining final train pantograph are commented
Estimate result.
A kind of described intelligent train pantograph military service performance dynamic monitoring and appraisal procedure, further include following steps:
S6. the military service performance dynamic monitoring assessment result of the step S5 final train pantograph obtained is visualized
It shows, to realize that the dynamic and visual of the military service performance of train pantograph is shown.
The working parameters of train pantograph described in step S1 specially obtain working performance ginseng using following steps
Number:
A. train pantograph system is divided into sledge system, rising bow system, Four-connecting-rod hinge system and damper system；
B. it is directed to pantograph pan system, the wearing valve, disalignment amount and pantograph for obtaining pantograph pan are inclined
Gradient is as military service performance parameter；
For rising bow system, the pressure value of pantograph cylinder pressure is obtained as military service performance parameter；
For Four-connecting-rod hinge system, obtains the vibration signal of the transverse direction of each bearing and longitudinal direction when rising bow and be used as clothes
Use as a servant performance parameter；
For damper system, the vibration signal of damper is as military service performance parameter when obtaining rising bow；
C. according to the real-time positioning information of train, the meteorological data in train location is obtained as ambient condition parameter；
D. according to the real-time passenger capacity information of train, effective passenger loading data of train is obtained as load condition parameter.
The health indicator timing that all parts of train pantograph are extracted described in step S2, specially uses following steps
Extract the health indicator timing of all parts of train pantograph:
A. all parts of pantograph are obtained from T by accelerated aging tests0Moment is to TNThe military service performance parameter at moment；
Shown in T0Moment is original state moment, TNMoment is retired state moment, T0Moment is to TNWhen operating status between the moment
Quarter is defined as Ti；The all parts of the pantograph include sledge system, rising bow system, Four-connecting-rod hinge system and damper
System；
B. by Hallbreg-Peck model be calculated between accelerated aging tests time and actual run time etc.
Effect accelerates scale parameter AF；
C. after the completion of accelerated aging tests, the injury experiment data training variation using all parts of pantograph is automatic
Encoder VAE obtains high-order feature, and calculates the quadrant angle of each high-order feature；
D. accurate adjustment is carried out using sequence of differences of the multilayer perceptron MLP model to the obtained quadrant angle of step c, obtained by electricity
The all parts of bow are in TiThe health indicator H at momenti, to extract the health indicator timing of all parts of train pantograph.
Hallbreg-Peck model described in step b, specially using following formula as Hallbreg-Peck model:
AF=(RHa/RH0)2×exp{(Ef/K)×[(1/T0)-(1/Ta)]}
RH in formulaaFor the relative humidity of accelerated aging tests；RH0For the relative humidity in actual motion；TaFor accelerated ageing
Temperature in experiment；T0For the temperature in actual motion；K is Boltzmann constant；EfIt can and be obtained by experiment for fault activation
?；Exp () is exponential function.
Variation autocoder VAE described in step c, specially outputting and inputting for variation autocoder VAE are
The status data of all parts of pantograph, the expression formula of prior distribution are as follows:Beta is in formula
Beta distribution, Uni are to be uniformly distributed；The high-order feature approximation that VAE is generated after training obeys above-mentioned prior distribution；In running
Component working parameters be input in trained VAE model and obtain high-order feature, which is sat to circle
It is (r, θ) that coordinate is obtained in mark system；Then the radial width of override feature, so that the coordinate of feature is reduced to θ.
Prediction described in step S3 obtains the remaining service life of each component of pantograph and pantograph entirety, specially adopts
Remaining service life is predicted with following steps:
(1) decomposition prediction model of the health indicator of all parts of pantograph is established；
(2) it establishes from component residue service life to the mapping model of pantograph residue service life；
(3) environment and load condition parametric prediction model are established；
(4) all parts and environment/load parameter during whole remaining service life for establishing pantograph influence
Quantitative model, so that prediction obtains the remaining service life of each component of pantograph and pantograph entirety.
The decomposition prediction model that the health indicator of all parts of pantograph is established described in step (1), specially uses
Following steps establish decomposition prediction model:
(1)-A. obtains the health indicator timing of all parts of pantograph；The all parts of the pantograph include sliding
Plate system, rising bow system, Four-connecting-rod hinge system and damper system；
(1) when-B. is using health indicator of the Algorithms of Discrete Wavelet Transform MODWT by all parts of pantograph is greatly overlapped
Sequence is decomposed into each wavelet coefficient layer；The s layers of wavelet coefficient wherein decomposed are
(1) it is IN that-C., which takes input number, and output number is OUT, constructs the input square of s layers of wavelet coefficient prediction model
Battle array IsWith output matrix Os；
(1)-D. is with IsFor input, OsFor output, the full convolutional neural networks prediction of the non-extraction of s layers of wavelet coefficient of training
Model；
(1)-E. carries out transfer learning to the prediction model that step (1)-D is obtained in real time；
(1)-F. carries out remaining life using the model after step (1)-E study and rolls iteration prediction.
The real-time prediction model obtained to step (1)-D described in step (1)-E carries out transfer learning, specially in train
Actual moving process summarizes, and records from T0The military service performance parameter of all parts for the pantograph that moment starts, wherein T0For component
At the time of after maintenance；T in component operational processiMoment obtains the health index parameters H at the momenti, and using greatly weight
Folded Algorithms of Discrete Wavelet Transform is decomposed into S layers of wavelet coefficientAnd in i≤X1When use master mould
It carries out remaining life and rolls iteration prediction, and in i > X1When, in s layers of wavelet coefficient withTo input, withThe parameter of s layers of prediction model is refreshed for output, and
Remaining life, which is carried out, using the model after refreshing rolls iteration prediction.
Remaining life described in step (1)-F rolls iteration prediction, will specially for each subsequenceIt is input in s layers of prediction model, obtains wavelet coefficient predicted valueThen will
Wavelet coefficient predicted value is incorporated into s layers of wavelet coefficient sequence, and deletes first wavelet coefficient in timing, is obtainedThen the time is reconstructed into using change commanders each layer of wavelet coefficient of the big overlapping discrete wavelet transformer of inverse pole
SequenceAnd it takes thereinFor one-step prediction value；Further, willS layers of prediction model are input to, wavelet coefficient predicted value is obtainedThen
By wavelet coefficient predicted valueIt is incorporated into s layers of wavelet coefficient sequence, and deletes first wavelet coefficient in timing, obtain
?Then using inverse pole it is big overlapping discrete wavelet transformer change commanders each layer wavelet coefficient reconstruct
For time seriesAnd it takes thereinFor two-staged prediction value；It repeats the above steps,
Until obtaining advanced multi-step prediction value；And the stop condition to repeat the above steps isWherein h-2 is TiThe remaining service life L at momenti。
Foundation described in step (2) is from component residue service life to the mapping model of pantograph residue service life, tool
Body is all parts that pantograph is obtained using the accelerated aging tests of pantograph entirety and the remaining service life of entirety；So
Equivalent remaining clothes corresponding to the obtained remaining service life of accelerated aging tests are calculated using Hallbreg-Peck model afterwards
Use as a servant the service life.
Environment and load condition parametric prediction model are established described in step (3), specially by y, the ginseng of the m month
Number is defined as Dy,m；Parameter is divided into 12 months by annual prediction model individually to be modeled；The training dataset of model is history
M month parameterWherein NyFor historical years quantity；History parameters are reconstructed into input matrixAnd output matrixAnd
With ImTo input,For output, the shot and long term Memory Neural Networks LSTM prediction model of training m month parameter；Then it uses
The model that training is completed carries out rolling iteration prediction, to obtain the prediction result of several steps；For ambient condition parameter and bear
State parameter is carried, prediction model is established using above-mentioned process respectively.
The all parts of pantograph and environment during whole remaining service life/negative are established described in step (4)
Carrying parameter influences quantitative model, specially T at the time of train operationi, remaining service life is Li, calculate from TiTo Ti+LiIt
Between all months ambient condition parameter prediction value and load condition parameter prediction value；Then it is calculated by method of weighting remaining
The influence quantizating index of state parameter during service life.
The remaining military service longevity whole to the step S3 each component of pantograph predicted and pantograph described in step S4
Life is modified, and specially uses the historical data of pantograph actual moving process, will predict remaining service life, ambient condition
Parameter and load condition parameter are used as the real surplus service life of input, train pantograph to be exported, for pantograph
All parts and pantograph are whole, and tree-shaped Gaussian process TGP model is respectively trained；The pantograph that step (1) is obtained it is each
The prediction residue service life for the pantograph entirety that the prediction residue service life of a component, step (2) obtain, step (3) obtain
The prediction arrived environment and load condition parameter input training after the completion of tree-shaped Gaussian process TGP model, thus obtain by
The amendment residue service life prediction result of all parts and the pantograph entirety of pantograph.
To the military service performance dynamic monitoring assessment result of the step S5 final train pantograph obtained described in step S6
It is visualized, is specially visualized using following steps:
1) using time, the health status of pantograph, environment and load condition parameter and remaining service life as axis,
It is to visualize content with all parts of pantograph and the remaining service life of pantograph entirety, realizes that train is arbitrarily being transported
The remaining service life at row moment, all parts and the pantograph entirety of pantograph visualizes；
2) maintenance decision of all parts of the pantograph of train is visualized；
3) the whole maintenance decision of the pantograph of train is visualized.
The maintenance decision of all parts of the pantograph of train is visualized described in step 2), is specially adopted
The maintenance decision of all parts of pantograph is carried out with step:
2) the holistic health correlation and accidental damage cost of -1. calculating each components of pantograph；
2) -2. pareto analysis method is used, finds out Pareto optimality face；The corresponding Pareto optimality face is whole
Body health correlation is most strong and damages the highest component of cost；Then the component is divided into the first important component；
2) after -3. the first important components of removing, step 2) -2 is repeated, until all components are divided and finish；
2) -4. for the component after division importance, using following Rulemaking maintenance decision:
For the first important component, the decision residue service life of component takes 0.5% quantile of service life probability；And every
It is a repair journey at the end of, if component predicts that remaining service life is less than next length for repairing journey, carried out at the end of repairing journey
Maintenance；
For the second important component, the decision residue service life of component takes 2.5% quantile of service life probability, and surplus
Remaining service life is replaced the previous moon；
For the remainder in addition to the first important component and the second important component, the decision residue service life of component
5% quantile of service life probability is taken, and is replaced in the first half of the month of remaining service life.
The whole maintenance decision of the pantograph of train is visualized described in step 3), specially to train by
The remaining service life probability distribution of pantograph and maintenance time are visualized；And the amendment residue of pantograph is taken to be on active service the longevity
0.5% quantile of probability distribution is ordered as decision residue service life, in decision residue service life the first two months to by electricity
Bow carries out depot repair.
A kind of intelligent train pantograph military service performance dynamic monitoring and assessment side are realized the present invention also provides a kind of
The system of method, including pantograph state parameter detection module, pantograph residue service life prediction module and maintenance visualization mould
Block；Pantograph state parameter detection module, pantograph residue service life prediction module and maintenance visualization model are sequentially connected in series；
Pantograph state parameter detection module is used to detect the state parameter of pantograph and uploads pantograph residue service life prediction mould
Block；Pantograph residue service life prediction module is used to integrally carry out life prediction to all parts and pantograph of pantograph,
And prediction result is uploaded into maintenance visualization model；Maintenance visualization model is for visualizing the bimetry of pantograph
It shows, generate maintenance measures data and is visualized.
This kind of intelligent train pantograph military service performance dynamic monitoring provided by the invention and appraisal procedure and its system,
It is measured in real time and analyzes by the running parameter to pantograph, can be realized on pantograph service state and extraneous influence
The monitoring of factor is established and is integrally on active service Performance Evaluation Model to pantograph, complete extraneous factor to remaining life not really
Qualitative effect modeling, and comprehensive visual information is provided；The method of the present invention can be realized the predicting residual useful life to pantograph,
Pantograph military service performance is effectively improved, and high reliablity, applicability are good and assessment is comprehensively accurate.
Detailed description of the invention
Fig. 1 is the method flow schematic diagram of the method for the present invention.
Fig. 2 is the functional block diagram of present system.
Specific embodiment
As shown in Figure 1 be the method flow schematic diagram of the method for the present invention: this kind of intelligent train provided by the invention by
Pantograph military service performance dynamic monitoring and appraisal procedure, include the following steps:
S1. the working parameters of train pantograph are obtained；Specially working parameters are obtained using following steps:
A. train pantograph system is divided into sledge system, rising bow system, Four-connecting-rod hinge system and damper system；
B. it is directed to pantograph pan system, the wearing valve, disalignment amount and pantograph for obtaining pantograph pan are inclined
Gradient is as military service performance parameter；High-speed camera is set at pantograph pan system, is calculated in rising bow by image procossing
Method detects abrasion of pantograph pan value, disalignment amount and pantograph tilt quantity as military service performance parameter；
For rising bow system, the pressure value of pantograph cylinder pressure is obtained as military service performance parameter；In pantograph rising bow
At system be arranged baroceptor, calculate time span be 1 minute pantograph cylinder pressure sequence average value and square
Average is as military service performance parameter；
For Four-connecting-rod hinge system, obtains the vibration signal of the transverse direction of each bearing and longitudinal direction when rising bow and be used as clothes
Use as a servant performance parameter；Vibrating sensor is set at pantograph Four-connecting-rod hinge system, detects the transverse and longitudinal side of each bearing when rising bow
To vibration signal, the kurtosis for calculating the vibration acceleration sequence that time span is 1 minute and average frequency are as military service performance
Parameter；
For damper system, the vibration signal of damper is as military service performance parameter when obtaining rising bow；It is hindered in pantograph
Vibrating sensor is set at Buddhist nun's device system, and the shock conditions of damper when detecting rising bow calculate the vibration that time span is 1 minute
The kurtosis and average frequency for accelerating degree series are as military service performance parameter；
In order to eliminate the influence of the speed of service, operating status for pantograph military service performance parameter, pantograph system it is each
A parameter is effective in train rated power operation；And in order to eliminate influence of the track condition to pantograph military service performance parameter,
All actual parameters of every day are averaged, train is obtained and is effectively on active service performance parameter；
C. according to the real-time positioning information of train, the meteorological data in train location is obtained as ambient condition parameter；
Train every 1 hour determines position using GPS positioning, and obtains wind speed, rainfall from the local weather bureau in location
Amount, snowfall are as ambient condition parameter；Ambient condition parameter is effective in train operation, stops library Shi Wuxiao in train；Environment
State parameter has significant annual cycles, while having randomness in a short time；In short-term environmental parameter acute exacerbation
When, railway department can take measures the damage for reducing environmental degradation to train；Therefore, compared to short-term randomness, the above ring
The long term periodicities of border parameter are even more important for the damage of pantograph military service performance；In order to weaken the short-term of ambient condition parameter
Ambient condition parameter in 1 effective time in the middle of the month is averaging and obtains effective environment state parameter by randomness；
D. according to the real-time passenger capacity information of train, effective passenger loading data of train is obtained as load condition parameter；
Assuming that train one day passes through N number of website, from i-th of website to i+1 website, passenger capacity Ai, mileage Li,
Then the passenger capacity of train every day isEffective passenger capacity in 1 month is taken into the average effective passenger capacity of acquisition；
S2. according to the working parameters of the step S1 train pantograph obtained, all parts of train pantograph are extracted
Health indicator timing；The health indicator timing of all parts of train pantograph is specially extracted using following steps:
A. all parts of pantograph are obtained from T by accelerated aging tests0Moment is to TNThe military service performance parameter at moment；
Shown in T0Moment is original state moment, TNMoment is retired state moment, T0Moment is to TNWhen operating status between the moment
Quarter is defined as Ti；The all parts of the pantograph include sledge system, rising bow system, Four-connecting-rod hinge system and damper
System；
In accelerated aging tests, train pantograph is with rated power operation, between the sampling time of accelerated aging tests
Every Δ T=Ti-Ti-1It is equivalent to the Δ T of actual run timerun×1.5；Wherein Δ TrunIt is transported for train every day with rated power
Capable mean up time, this parameter are related to train operation route and type of train；Mean up time is multiplied by 1.5 can
The case where to cover in 1 day with non-rated power operation；By this method, so that being spaced in accelerated aging tests with Δ T
The middle obtained data that sample are equal to the data obtained in actual moving process with 1 day for interval sampling；
B. by Hallbreg-Peck model be calculated between accelerated aging tests time and actual run time etc.
Effect accelerates scale parameter AF；The Hallbreg-Peck model, specially using following formula as Hallbreg-Peck
Model:
AF=(RHa/RH0)2×exp{(Ef/K)×[(1/T0)-(1/Ta)]}
RH in formulaaFor the relative humidity of accelerated aging tests；RH0For the relative humidity in actual motion；TaFor accelerated ageing
Temperature in experiment；T0For the temperature in actual motion；K is Boltzmann constant；EfIt can and be obtained by experiment for fault activation
?；Exp () is exponential function；
C. after the completion of accelerated aging tests, the injury experiment data training variation using all parts of pantograph is automatic
Encoder VAE obtains high-order feature, and calculates the quadrant angle of each high-order feature；Specially variation autocoder VAE's is defeated
Enter and export be pantograph all parts status data, the expression formula of prior distribution are as follows:Beta is Beta distribution in formula, and Uni is to be uniformly distributed；The high-order feature that VAE is generated after training
Approximation obeys above-mentioned prior distribution；It is (r, θ) that the high-order Feature Mapping, which is obtained coordinate into circle coordinates system,；Then ignore spy
The radial width of sign, so that the coordinate of feature is reduced to θ；
For the sampling point distributions of the prior distribution in annulus, the high-order feature approximation that VAE is generated after training obeys the priori
Distribution.By high-order Feature Mapping into circle coordinates system coordinate [r, θ], since the radial width of feature is relatively narrow, it can ignore, therefore
The coordinate of feature can simplify as θ；By initial time θ0Status data be input in the VAE of training completion that obtain high-order special
Sign, the quadrant angle for calculating feature is θ0；For TiThe status data of component is inputted the VAE trained by the moment, and it is special to obtain high-order
Sign, the quadrant angle for calculating feature is θi；Quadrant angle difference θi-θ0The difference that component operating status and original state can be embodied, because
This can characterize the health condition of component；
D. accurate adjustment is carried out using sequence of differences of the multilayer perceptron MLP model to the obtained quadrant angle of step c, obtained by electricity
The all parts of bow are in TiThe health indicator H at momenti, to extract the health indicator timing of all parts of train pantograph；
In training, the input of MLP model is quadrant angle difference [θ0-θ0,...,θi-θ0,...,θN-θ0], it exports and is
[1,...,(N-i)/N,...,0].By θi-θ0It is input to the MLP model of training completion, can be obtained TiThe health indicator at moment
Hi；Quadrant angle sequence of differences is mapped to a polynomial sequence by MLP by the accurate adjustment process；The hidden layer number of MLP can be with
It is taken as 10；One order polynomial can describe the monotone variation process of the remaining life of pantograph system wear process；The processing energy
It is enough that the damage priori knowledge of pantograph system is brought into health indicator developing algorithm, health indicator is improved for train part
The descriptive power of degenerative process；
S3. the healthy timing indicator obtained according to step S2, it is whole that prediction obtains each component of pantograph and pantograph
Remaining service life；Specially remaining service life is predicted using following steps:
(1) decomposition prediction model of the health indicator of all parts of pantograph is established；Specially using as follows
(1)-A. obtains the health indicator timing of all parts of pantograph；The all parts of the pantograph include sliding
Plate system, rising bow system, Four-connecting-rod hinge system and damper system；
Off-line model is trained first, military service performance parameter timing is input in VAE the and MLP model of training completion,
Obtain the health indicator timing of component
(1) when-B. is using health indicator of the Algorithms of Discrete Wavelet Transform MODWT by all parts of pantograph is greatly overlapped
Sequence is decomposed into each wavelet coefficient layer；The s layers of wavelet coefficient wherein decomposed are
In the specific implementation, wavelet mother function is taken as " morlet ", and Decomposition order is taken as 3 layers, and available 4 sons decompose
Layer；
(1) it is IN that-C., which takes input number, and output number is OUT, constructs the input square of s layers of wavelet coefficient prediction model
Battle array IsWith output matrix Os；
In the specific implementation, taking input number is 5, and output number is 1, constructs the defeated of s layers of wavelet coefficient prediction model
Enter matrix IsAnd output matrix OsIt is shown below:
(1)-D. is with IsFor input, OsFor output, the full convolutional neural networks prediction of the non-extraction of s layers of wavelet coefficient of training
Model；
In the specific implementation, model uses cause and effect convolution, and convolution filter number is 16；The above training process is repeated,
Obtain the prediction model of all decomposition layers；
(1)-E. carries out transfer learning to the prediction model that step (1)-D is obtained in real time；Specially in train actual motion
Process summarizes, and records from T0The military service performance parameter of all parts for the pantograph that moment starts, wherein T0After component maintenance
Moment；T in component operational processiMoment obtains the health index parameters H at the momenti, and it is discrete small using being greatly overlapped
Wave conversion algorithm is decomposed into S layers of wavelet coefficientAnd in i≤X1When carried out using master mould it is remaining
Service life rolls iteration prediction, and in i > X1When, in s layers of wavelet coefficient withIt is defeated
Enter, withThe parameter of s layers of prediction model is refreshed for output, and carries out remaining life using the model after refreshing
Roll iteration prediction；
In the specific implementation, it records in train actual moving process from ToThe military service of each component of the pantograph that moment starts
Performance parameter；Wherein ToAt the time of for after component maintenance, time interval is 1 day；T in component operational processiMoment, by this
The military service performance parameter at moment is input in VAE the and MLP model that training is completed in step 1, obtains health indicator Hi, and make
8 layers of wavelet coefficient are decomposed into MODWT algorithmDecomposing the wavelet function used is " morlet "；
In i≤5, remaining life is carried out using master mould and rolls iteration prediction；When i > 5, in s layers of wavelet coefficient, withTo input,The parameter of s layers of prediction model is refreshed for input；
(1)-F. carries out remaining life using the model after step (1)-E study and rolls iteration prediction；Specially for every
One subsequence, willIt is input in s layers of prediction model, obtains wavelet coefficient prediction
ValueThen wavelet coefficient predicted value is incorporated into s layers of wavelet coefficient sequence, and deletes first small echo in timing
Coefficient obtainsThen it is changed commanders each layer of wavelet coefficient weight using the big overlapping discrete wavelet transformer of inverse pole
Structure is time seriesAnd it takes thereinFor one-step prediction value；Further, willS layers of prediction model are input to, wavelet coefficient predicted value is obtainedThen
By wavelet coefficient predicted valueIt is incorporated into s layers of wavelet coefficient sequence, and deletes first wavelet coefficient in timing, obtain
?Then using inverse pole it is big overlapping discrete wavelet transformer change commanders each layer wavelet coefficient reconstruct
For time seriesAnd it takes thereinFor two-staged prediction value；It repeats the above steps,
Until obtaining advanced multi-step prediction value；And the stop condition to repeat the above steps isWherein h-2 is TiThe remaining service life L at momenti；
It in the specific implementation,, will for s layers for each subsequenceIt is input to
In s layers of prediction model, wavelet coefficient predicted value is obtainedWavelet coefficient predicted value is incorporated into s layers of wavelet systems number sequence
In column, and first wavelet coefficient in timing is deleted, obtainedThen using inverse pole it is big be overlapped from
It dissipates wavelet transformation and each layer of wavelet coefficient is reconstructed into time seriesIt takes thereinFor
One-step prediction value.Further willS layers of prediction model are inputted, wavelet coefficient predicted value is obtainedWavelet coefficient predicted value is incorporated into s layers of wavelet coefficient sequence, and deletes first wavelet coefficient in timing,
It obtainsThen it is changed commanders each layer of wavelet coefficient weight using the big overlapping discrete wavelet transformer of inverse pole
Structure is time seriesIt takes thereinFor two-staged prediction value.Above step is repeated to obtain
Take advanced multi-step prediction value.The multi-Step Iterations calculate stop condition beThen take
H-2 is TiThe remaining service life L at momenti；
The military service performance parameter timing of each component of pantograph is brought into above method.Pantograph is calculated using above method
Sledge system, rising bow system, Four-connecting-rod hinge system and damping system remaining service life
(2) it establishes from component residue service life to the mapping model of pantograph residue service life；Specially use by
The accelerated aging tests of pantograph entirety obtain all parts of pantograph and the remaining service life of entirety；Then it uses
Hallbreg-Peck model calculates the equivalent remaining military service longevity corresponding to the obtained remaining service life of accelerated aging tests
Life；
In the specific implementation, using each component of accelerated aging tests acquisition pantograph of pantograph entirety and remaining for entirety
Remaining service life.It is calculated corresponding to the obtained remaining service life of accelerated aging tests using Hallbreg-Peck model
Equivalent residue service life；Using the equivalent remaining service life of pantograph each section of accelerated aging tests acquisition as input,
The equivalent remaining service life whole using pantograph uses Gauss as output, training support vector regression SVR model, the model
Core, nuclear parameter 1.0,1.0；In the T of train operationiMoment, by each component residue service life of the pantograph being calculatedIt is input in the SVR model of training completion；Obtain pantograph system residue service lifeWherein subscript p is
The number of each component of pantograph；
(3) environment and load condition parametric prediction model are established；Specially it is by y, the parameter definition of the m month
Dy,m；Parameter is divided into 12 months by annual prediction model individually to be modeled；The training dataset of model is joining the m month for history
NumberWherein NyFor historical years quantity；History parameters are reconstructed into input matrixAnd output matrixAnd
With ImTo input,For output, training m month parameter shot and long term Memory Neural Networks LSTM prediction model, model it is hidden
Number containing layer is 50；Then it carries out rolling iteration prediction using the model that training is completed, to obtain the prediction result of several steps；
For ambient condition parameter and load condition parameter, prediction model is established using above-mentioned process respectively；
(4) all parts and environment/load parameter during whole remaining service life for establishing pantograph influence
Quantitative model, so that prediction obtains the remaining service life of each component of pantograph and pantograph entirety；Specially transported in train
T at the time of rowi, remaining service life is Li, calculate from TiTo Ti+LiBetween all months ambient condition parameter prediction value and
Load condition parameter prediction value；Then pass through the influence quantizating index of state parameter during method of weighting calculating remaining life；
In the specific implementation, in the T of train operationiMoment, remaining service life are Li, calculate from TiTo Ti+LiBetween
All months；It is calculated using the state parameter prediction model of foundation from the state in months all during remaining service life and is joined
Number predicted value；By the influence quantizating index of state parameter during method of weighting calculating remaining life, calculation formula isWherein M is all months that remaining life includes；For m-th month during remaining service life
The predicted state data of part, wherein the status data less than one month was calculated with one month；wmIt, should for the weight in m-th of month
Weight is related to advanced prediction step number, and weight is reduced with prediction step number；It predicts that the shorter predicted value accuracy of step number is higher, assigns
Higher weights are given, the long predicted value accuracy of prediction step number is lower, assigns lower weight, the weight of difference prediction step number is such as
Shown in table 1:
The weight of the different prediction step numbers of table 1
| Predict step number | 1 step | 2 steps | 3 steps or more |
| Weight | 1 | 0.9 | 0.8 |
In train operation TiMoment, by each component of the pantograph being calculated and entirety remaining service life bring into
Upper calculation method obtains the shadow of wind speed in each component residue service life of pantograph, rainfall, snowfall and effective passenger capacity
Ring quantizating indexWithThe whole remaining military service of train pantograph
Wind speed in service life, rainfall, snowfall and effective passenger capacity influence quantizating indexWherein subscript q is
Environment/load condition parameter number；
S4. the remaining service life of the step S3 each component of pantograph predicted and pantograph entirety is repaired
Just；The historical data of pantograph actual moving process is specially used, will predict remaining service life, ambient condition parameter and is born
The real surplus service life that state parameter is carried as input, train pantograph is used as output, for all parts of pantograph
And pantograph is whole, and tree-shaped Gaussian process TGP model is respectively trained；The all parts for the pantograph that step (1) is obtained
Predict the prediction that the prediction residue service life for the pantograph entirety that remaining service life, step (2) obtain, step (3) obtain
Environment and load condition parameter input training after the completion of tree-shaped Gaussian process TGP model, to obtain each of pantograph
The amendment residue service life prediction result of a component and pantograph entirety；
In the specific implementation, in the T of train operationiMoment, by the surplus of each component of the pantograph being calculated and entirety
Remaining service lifeAndEnvironment/load condition parameter of each component of the pantograph being calculated and entiretyAndBring into training completion by
The TGP model of each component of pantograph and entirety obtains the amendment residue service life of pantograph component and entirety
AndThe amendment residue service life of acquisition is probability distribution；The coefficient initial value of TGP model is selected by cross validation
It is fixed；The probability distribution that TGP model generates can describe the uncertainty of remaining service life prediction；
S5. the correction result obtained according to step S4, the military service performance dynamic monitoring for obtaining final train pantograph are commented
Estimate result；
S6. the military service performance dynamic monitoring assessment result of the step S5 final train pantograph obtained is visualized
It shows, to realize that the dynamic and visual of the military service performance of train pantograph is shown；It is specially carried out using following steps visual
Change and show:
1) using time, the health status of pantograph, environment and load condition parameter and remaining service life as axis,
It is to visualize content with all parts of pantograph and the remaining service life of pantograph entirety, realizes that train is arbitrarily being transported
The remaining service life at row moment, all parts and the pantograph entirety of pantograph visualizes；
2) maintenance decision of the pantograph all parts of train is visualized；Specially using step carry out by
The maintenance decision of all parts of pantograph:
2) the holistic health correlation and accidental damage cost of -1. calculating each components of pantograph；
2) -2. pareto analysis method is used, finds out Pareto optimality face；The corresponding Pareto optimality face is whole
Body health correlation is most strong and damages the highest component of cost；Then the component is divided into the first important component；
2) after -3. the first important components of removing, step 2) -2 is repeated, until all components are divided and finish；
2) -4. for the component after division importance, using following Rulemaking maintenance decision:
For the first important component, the decision residue service life of component takes 0.5% quantile of service life probability；And every
It is a repair journey at the end of, if component predicts that remaining service life is less than next length for repairing journey, carried out at the end of repairing journey
Maintenance；
For the second important component, the decision residue service life of component takes 2.5% quantile of service life probability, and surplus
Remaining service life is replaced the previous moon；
For the remainder in addition to the first important component and the second important component, the decision residue service life of component
5% quantile of service life probability is taken, and is replaced in the first half of the month of remaining service life；
3) the whole maintenance decision of the pantograph of train is visualized；To train by electricity described in step 3)
The whole maintenance decision of bow is visualized, specially to the remaining service life probability distribution of train pantograph and maintenance
Time is visualized；And take 0.5% quantile of the remaining service life probability distribution of the amendment of pantograph as decision
Remaining service life carries out depot repair to pantograph in decision residue service life the first two months.
It is illustrated in figure 2 the functional block diagram of system of the invention: realizing a kind of intelligence the present invention also provides a kind of
Can train the pantograph dynamic monitoring of military service performance and appraisal procedure system, including pantograph state parameter detection module, by electricity
Bend remaining service life prediction module and maintenance visualization model；Pantograph state parameter detection module, pantograph residue are on active service
Life prediction module and maintenance visualization model are sequentially connected in series；Pantograph state parameter detection module is used to detect the shape of pantograph
State parameter simultaneously uploads pantograph residue service life prediction module；Pantograph residue service life prediction module is used for pantograph
All parts and pantograph integrally carry out life prediction, and prediction result is uploaded into maintenance visualization model；Maintenance visualization
Module is for visualizing the bimetry of pantograph, generating maintenance measures data and being visualized.
This intelligent train pantograph military service performance dynamic monitoring provided by the invention and appraisal procedure and its system, will be by
Pantograph is decomposed into sledge system, rising bow system, Four-connecting-rod hinge system and damper system, and is on active service to these components
Performance parameter dynamic monitoring；Compared to the monitoring modular for only monitoring pantograph pan military service performance, which can be adopted
Collect performance parameter of being more comprehensively on active service；Military service Performance Evaluation Model disclosed in the present application is by the military service of pantograph multiple portions
Performance is merged, to obtain more reasonably remaining service life prediction result.
The inherent laws that pantograph military service performance parameter is excavated using depth network self-adapting are realized to each portion of train
The high-precision forecast of part and pantograph residue service life；Using VAE-MLP model realization to pantograph each component with
And whole health status depth dimensionality reduction and strong monotonic property health indicator calculates；VAE model can extract the military service of pantograph
Service state is mapped to the feature space positioned at circle coordinates system by the changing rule of state, extracts the quadrant angle of feature as strong
Health state feature；MLP realizes the accurate adjustment of health status index, the monotonicity of health status index is improved, to meet Mechatronic Systems
Dull degradation characteristics.
Prediction using MODWT-LSTM model realization to each component residue service life of pantograph；By means of MODWT
The side-effect-free feature of algorithm and the strong nonlinearity capability of fitting of LSTM are realized and are on active service to each component residue of flashlight Nangong
Service life is effectively predicted；Each component residue service life is merged using SVR model, is realized whole to pantograph remaining
The adaptive prediction of service life.
It realizes and the prediction of environment during remaining service life/load condition parameter is quantified to model, establish to residue
The uncertain correction model of service life；The annual cycles of use state parameter realize environment/load condition parameter
High-precision forecast；Since precision of prediction increases with the increase of prediction step number, quantify remaining be on active service using decaying weighted model
Environment/load condition parameter influence power during service life, weighting coefficient increase with the increase of prediction step number；Due to residue
The intrinsic uncertainty of service life, the uncertainty calculated after considering environment/load condition parameter influence using TGP model are surplus
Remaining service life.
Claims (10)
1. a kind of intelligence train pantograph military service performance dynamic monitoring and appraisal procedure, include the following steps:
S1. the working parameters of train pantograph are obtained；
S2. according to the working parameters of the step S1 train pantograph obtained, the strong of all parts of train pantograph is extracted
Kang Zhibiao timing；
S3. the healthy timing indicator obtained according to step S2, prediction obtain the residue of each component of pantograph and pantograph entirety
Service life；
S4. the remaining service life of the step S3 each component of pantograph predicted and pantograph entirety is modified；
S5. the correction result obtained according to step S4 obtains the military service performance dynamic monitoring assessment knot of final train pantograph
Fruit.
2. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 1 and appraisal procedure, feature
It is to further include following steps:
S6. visualization exhibition is carried out to the military service performance dynamic monitoring assessment result of the step S5 final train pantograph obtained
Show, to realize that the dynamic and visual of the military service performance of train pantograph is shown.
3. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 1 and appraisal procedure, feature
It is the working parameters of train pantograph described in step S1, specially obtains working parameters using following steps:
A. train pantograph system is divided into sledge system, rising bow system, Four-connecting-rod hinge system and damper system；
B. it is directed to pantograph pan system, obtains the wearing valve, disalignment amount and pantograph tilt quantity of pantograph pan
As military service performance parameter；
For rising bow system, the pressure value of pantograph cylinder pressure is obtained as military service performance parameter；
For Four-connecting-rod hinge system, the transverse direction of each bearing and the vibration signal of longitudinal direction are obtained when rising bow as military service
It can parameter；
For damper system, the vibration signal of damper is as military service performance parameter when obtaining rising bow；
C. according to the real-time positioning information of train, the meteorological data in train location is obtained as ambient condition parameter；
D. according to the real-time passenger capacity information of train, effective passenger loading data of train is obtained as load condition parameter.
4. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 3 and appraisal procedure, feature
It is to extract the health indicator timing of all parts of train pantograph described in step S2, is specially extracted using following steps
The health indicator timing of all parts of train pantograph:
A. all parts of pantograph are obtained from T by accelerated aging tests0Moment is to TNThe military service performance parameter at moment；It is shown
T0Moment is original state moment, TNMoment is retired state moment, T0Moment is to TNThe operating status moment between moment is fixed
Justice is Ti；The all parts of the pantograph include sledge system, rising bow system, Four-connecting-rod hinge system and damper system
System；
B. by Hallbreg-Peck model be calculated equivalent between accelerated aging tests time and actual run time plus
Fast scale parameter AF；Specially using following formula as Hallbreg-Peck model:
AF=(RHa/RH0)2×exp{(Ef/K)×[(1/T0)-(1/Ta)]}
RH in formulaaFor the relative humidity of accelerated aging tests；RH0For the relative humidity in actual motion；TaFor accelerated aging tests
In temperature；T0For the temperature in actual motion；K is Boltzmann constant；EfIt can and be obtained by experiment for fault activation；exp
() is exponential function；
C. after the completion of accelerated aging tests, using the injury experiment data training variation autocoding of all parts of pantograph
Device VAE obtains high-order feature, and calculates the quadrant angle of each high-order feature；Specially the input of variation autocoder VAE and
Output is the status data of all parts of pantograph, the expression formula of prior distribution are as follows:
Beta is Beta distribution in formula, and Uni is to be uniformly distributed；The high-order that VAE is generated after training is special
Sign is approximate to obey above-mentioned prior distribution；Running component working parameters are input in trained VAE model and are obtained
To high-order feature, it is (r, θ) which, which is obtained coordinate into circle coordinates system,；Then override feature is radially-wide
Degree, so that the coordinate of feature is reduced to θ；
D. accurate adjustment is carried out using sequence of differences of the multilayer perceptron MLP model to the obtained quadrant angle of step c, obtains pantograph
All parts are in TiThe health indicator H at momenti, to extract the health indicator timing of all parts of train pantograph.
5. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 4 and appraisal procedure, feature
It is that prediction described in step S3 obtains the remaining service life of each component of pantograph and pantograph entirety, specially using such as
Lower step predicts remaining service life:
(1) decomposition prediction model of the health indicator of all parts of pantograph is established；It is specially established and is divided using following steps
Solve prediction model:
(1)-A. obtains the health indicator timing of all parts of pantograph；The all parts of the pantograph include slide plate system
System, rising bow system, Four-connecting-rod hinge system and damper system；
(1)-B. is divided the health indicator timing of all parts of pantograph using greatly overlapping Algorithms of Discrete Wavelet Transform MODWT
Solution is each wavelet coefficient layer；The s layers of wavelet coefficient wherein decomposed are
(1) it is IN that-C., which takes input number, and output number is OUT, constructs the input matrix I of s layers of wavelet coefficient prediction modelsWith
Output matrix Os；
(1)-D. is with IsFor input, OsFor output, the full convolutional neural networks prediction model of non-extraction of s layers of wavelet coefficient of training；
(1)-E. carries out transfer learning to the prediction model that step (1)-D is obtained in real time；
(1)-F. carries out remaining life using the model after step (1)-E study and rolls iteration prediction；
(2) it establishes from component residue service life to the mapping model of pantograph residue service life；
(3) environment and load condition parametric prediction model are established；
(4) all parts and environment/load parameter during whole remaining service life for establishing pantograph influence quantization
Model, so that prediction obtains the equivalent remaining service life of each component of pantograph and pantograph entirety.
6. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 5 and appraisal procedure, feature
It is that the real-time prediction model obtained to step (1)-D described in step (1)-E carries out transfer learning, specially in train reality
Operational process summarizes, and records from T0The military service performance parameter of all parts for the pantograph that moment starts, wherein T0For component maintenance
At the time of afterwards；T in component operational processiMoment obtains the health index parameters H at the momenti, and using greatly be overlapped from
Scattered Wavelet Transformation Algorithm is decomposed into S layers of wavelet coefficientAnd in i≤X1When carried out using master mould
Remaining life rolls iteration prediction, and in i > X1When, in s layers of wavelet coefficient with
To input, withThe parameter of s layers of prediction model is refreshed for output, and carries out residue using the model after refreshing
Service life rolls iteration prediction；
Remaining life described in step (1)-F rolls iteration prediction, will specially for each subsequenceIt is input in s layers of prediction model, obtains wavelet coefficient predicted valueThen will
Wavelet coefficient predicted value is incorporated into s layers of wavelet coefficient sequence, and deletes first wavelet coefficient in timing, is obtainedThen the time is reconstructed into using change commanders each layer of wavelet coefficient of the big overlapping discrete wavelet transformer of inverse pole
SequenceAnd it takes thereinFor one-step prediction value；Further, willS layers of prediction model are input to, wavelet coefficient predicted value is obtainedThen
By wavelet coefficient predicted valueIt is incorporated into s layers of wavelet coefficient sequence, and deletes first wavelet coefficient in timing, obtain
?Then using inverse pole it is big overlapping discrete wavelet transformer change commanders each layer wavelet coefficient reconstruct
For time seriesAnd it takes thereinFor two-staged prediction value；It repeats the above steps,
Until obtaining advanced multi-step prediction value；And the stop condition to repeat the above steps isWherein h-2 is TiThe remaining service life L at momenti。
7. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 6 and appraisal procedure, feature
Be foundation described in step (2) from component residue service life to the mapping model of pantograph residue service life, specially
The all parts of pantograph and the remaining service life of entirety are obtained using the accelerated aging tests of pantograph entirety；Then make
The equivalent remaining military service longevity corresponding to the obtained remaining service life of accelerated aging tests is calculated with Hallbreg-Peck model
Life；
Environment and load condition parametric prediction model are established described in step (3), specially determine y, the parameter of the m month
Justice is Dy,m；Parameter is divided into 12 months by annual prediction model individually to be modeled；The training dataset of model is the m of history
Month parameterWherein NyFor historical years quantity；History parameters are reconstructed into input matrixAnd output matrixAnd
With ImTo input,For output, the shot and long term Memory Neural Networks LSTM prediction model of training m month parameter；Then it uses
The model that training is completed carries out rolling iteration prediction, to obtain the prediction result of several steps；For ambient condition parameter and bear
State parameter is carried, prediction model is established using above-mentioned process respectively；
The all parts that pantograph is established described in step (4) and environment/load ginseng during whole remaining service life
Number influences quantitative model, specially T at the time of train operationi, remaining service life is Li, calculate from TiTo Ti+LiBetween
The ambient condition parameter prediction value and load condition parameter prediction value in all months；Then remaining life is calculated by method of weighting
The influence quantizating index of period state parameter.
8. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 7 and appraisal procedure, feature
Be described in step S4 to the remaining service life of the step S3 each component of pantograph predicted and pantograph entirety into
Row amendment, specially uses the historical data of pantograph actual moving process, will predict remaining service life, ambient condition parameter
Real surplus service life with load condition parameter as input, train pantograph is used as output, for each of pantograph
Component and pantograph are whole, and tree-shaped Gaussian process TGP model is respectively trained；The each portion for the pantograph that step (1) is obtained
The prediction residue service life for the pantograph entirety that the prediction residue service life of part, step (2) obtain, step (3) obtain
Tree-shaped Gaussian process TGP model after the completion of environment and load condition parameter the input training of prediction, to obtain pantograph
All parts and pantograph entirety amendment residue service life prediction result.
9. a kind of intelligent train pantograph military service performance dynamic monitoring according to claim 8 and appraisal procedure, feature
It is to carry out the military service performance dynamic monitoring assessment result of the step S5 final train pantograph obtained described in step S6
It visualizes, is specially visualized using following steps:
1) using time, the health status of pantograph, environment and load condition parameter and remaining service life as axis, with by
The remaining service life of all parts and the pantograph entirety of pantograph is to visualize content, realizes train in any operation
It carves, the remaining service life of all parts and the pantograph entirety of pantograph visualizes；
2) maintenance decision of all parts of the pantograph of train is visualized；
3) the whole maintenance decision of the pantograph of train is visualized；
The maintenance decision of all parts of the pantograph of train is visualized described in step 2), specially using step
The maintenance decision of the rapid all parts for carrying out pantograph:
2) the holistic health correlation and accidental damage cost of -1. calculating each components of pantograph；
2) -2. pareto analysis method is used, finds out Pareto optimality face；The corresponding Pareto optimality face is whole strong
Health correlation is most strong and damages the highest component of cost；Then the component is divided into the first important component；
2) after -3. the first important components of removing, step 2) -2 is repeated, until all components are divided and finish；
2) -4. for the component after division importance, using following Rulemaking maintenance decision:
For the first important component, the decision residue service life of component takes 0.5% quantile of service life probability；And it is repaired each
At the end of journey, if component predicts that remaining service life is less than next length for repairing journey, safeguarded at the end of repairing journey；
For the second important component, the decision residue service life of component takes 2.5% quantile of service life probability, and takes in residue
The labour service life is replaced the previous moon；
For the remainder in addition to the first important component and the second important component, the decision residue service life of component takes the longevity
5% quantile of probability is ordered, and is replaced in the first half of the month of remaining service life；
The whole maintenance decision of the pantograph of train is visualized described in step 3), specially to train pantograph
Remaining service life probability distribution and maintenance time visualized；And take the amendment residue service life of pantograph general
0.5% quantile of rate distribution as decision residue service life, decision residue service life the first two months to pantograph into
Row depot repair.
10. it is a kind of realize described in one of claim 1~9 it is a kind of intelligence the military service performance dynamic monitoring of train pantograph and assessment
The system of method, it is characterised in that including pantograph state parameter detection module, pantograph residue service life prediction module and
Safeguard visualization model；Pantograph state parameter detection module, pantograph residue service life prediction module and maintenance visualization
Module is sequentially connected in series；Pantograph state parameter detection module is used to detect the state parameter of pantograph and uploads pantograph residue clothes
Use as a servant life prediction module；Pantograph residue service life prediction module be used for it is whole to all parts and pantograph of pantograph into
Row life prediction, and prediction result is uploaded into maintenance visualization model；Safeguard that visualization model is used for the prediction longevity to pantograph
Life is visualized, and is generated maintenance measures data and is visualized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910676732.5A CN110361180B (en) | 2019-07-25 | 2019-07-25 | Intelligent train pantograph service performance dynamic monitoring and evaluating method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910676732.5A CN110361180B (en) | 2019-07-25 | 2019-07-25 | Intelligent train pantograph service performance dynamic monitoring and evaluating method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110361180A true CN110361180A (en) | 2019-10-22 |
| CN110361180B CN110361180B (en) | 2021-01-26 |
Family
ID=68221611
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910676732.5A Active CN110361180B (en) | 2019-07-25 | 2019-07-25 | Intelligent train pantograph service performance dynamic monitoring and evaluating method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110361180B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110849627A (en) * | 2019-11-27 | 2020-02-28 | 哈尔滨理工大学 | Width migration learning network and rolling bearing fault diagnosis method based on same |
| CN111597759A (en) * | 2020-05-18 | 2020-08-28 | 中车永济电机有限公司 | Construction method of residual service life prediction model of IGBT (insulated Gate Bipolar translator) of converter device |
| CN112115643A (en) * | 2020-09-15 | 2020-12-22 | 中南大学 | Smart train service life non-invasive prediction method |
| CN112541228A (en) * | 2020-12-10 | 2021-03-23 | 重庆交通大学 | Pantograph active control method capable of memorizing network prediction of contact force duration |
| CN113032985A (en) * | 2021-03-11 | 2021-06-25 | 北京必创科技股份有限公司 | Intelligent service life assessment method and device for wireless sensing equipment |
| CN114001850A (en) * | 2021-10-25 | 2022-02-01 | 南京地铁建设有限责任公司 | Pantograph pressure detection method and system |
| CN115723578A (en) * | 2022-12-29 | 2023-03-03 | 湖南行必达网联科技有限公司 | Pantograph control method, device, equipment and working machine |
| CN116400623A (en) * | 2023-04-06 | 2023-07-07 | 南京星河世纪信息技术有限公司 | Intelligent monitoring system for high-voltage equipment |
| CN118133947A (en) * | 2024-05-07 | 2024-06-04 | 广东艾林克能源装备有限公司 | AI processing method and system for air compressor energy system |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090277287A1 (en) * | 2008-05-06 | 2009-11-12 | Chartered Semiconductor Manufacturing, Ltd. | Method for performing a shelf lifetime acceleration test |
| CN101865962A (en) * | 2010-06-30 | 2010-10-20 | 福州大学 | Pantograph electrical life prediction and analysis system |
| JP2013130460A (en) * | 2011-12-21 | 2013-07-04 | Railway Technical Research Institute | Light measurement apparatus |
| CN103528624A (en) * | 2013-10-18 | 2014-01-22 | 中国航空工业集团公司北京长城计量测试技术研究所 | Optical fiber type comprehensive on-line real-time pantograph detection and control system |
| CN204576779U (en) * | 2015-02-16 | 2015-08-19 | 苏州华兴致远电子科技有限公司 | A kind of bow net operating state monitoring system |
| CN105468866A (en) * | 2015-12-15 | 2016-04-06 | 长春工业大学 | Method for predicting remaining life of LED driving power of railway vehicles |
| JP2018052301A (en) * | 2016-09-29 | 2018-04-05 | 東日本旅客鉄道株式会社 | Power supply system for electric vehicle |
| CN108680890A (en) * | 2018-08-23 | 2018-10-19 | 重庆市计量质量检测研究院 | Intelligent electric energy meter life characteristics detection method |
-
2019
- 2019-07-25 CN CN201910676732.5A patent/CN110361180B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090277287A1 (en) * | 2008-05-06 | 2009-11-12 | Chartered Semiconductor Manufacturing, Ltd. | Method for performing a shelf lifetime acceleration test |
| CN101865962A (en) * | 2010-06-30 | 2010-10-20 | 福州大学 | Pantograph electrical life prediction and analysis system |
| JP2013130460A (en) * | 2011-12-21 | 2013-07-04 | Railway Technical Research Institute | Light measurement apparatus |
| CN103528624A (en) * | 2013-10-18 | 2014-01-22 | 中国航空工业集团公司北京长城计量测试技术研究所 | Optical fiber type comprehensive on-line real-time pantograph detection and control system |
| CN204576779U (en) * | 2015-02-16 | 2015-08-19 | 苏州华兴致远电子科技有限公司 | A kind of bow net operating state monitoring system |
| CN105468866A (en) * | 2015-12-15 | 2016-04-06 | 长春工业大学 | Method for predicting remaining life of LED driving power of railway vehicles |
| JP2018052301A (en) * | 2016-09-29 | 2018-04-05 | 東日本旅客鉄道株式会社 | Power supply system for electric vehicle |
| CN108680890A (en) * | 2018-08-23 | 2018-10-19 | 重庆市计量质量检测研究院 | Intelligent electric energy meter life characteristics detection method |
Non-Patent Citations (2)
| Title |
|---|
| XU XIA 等: "A Routing Protocol for Multisink Wireless Sensor Networks in Underground Coalmine Tunnels", 《SENSORS》 * |
| 衣磊: "PLC控制器应力加速试验", 《建筑机械技术与管理》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110849627A (en) * | 2019-11-27 | 2020-02-28 | 哈尔滨理工大学 | Width migration learning network and rolling bearing fault diagnosis method based on same |
| CN111597759B (en) * | 2020-05-18 | 2022-04-26 | 中车永济电机有限公司 | Construction method of residual service life prediction model of IGBT (insulated Gate Bipolar translator) of converter device |
| CN111597759A (en) * | 2020-05-18 | 2020-08-28 | 中车永济电机有限公司 | Construction method of residual service life prediction model of IGBT (insulated Gate Bipolar translator) of converter device |
| CN112115643A (en) * | 2020-09-15 | 2020-12-22 | 中南大学 | Smart train service life non-invasive prediction method |
| CN112115643B (en) * | 2020-09-15 | 2022-06-10 | 中南大学 | Smart train service life non-invasive prediction method |
| CN112541228A (en) * | 2020-12-10 | 2021-03-23 | 重庆交通大学 | Pantograph active control method capable of memorizing network prediction of contact force duration |
| CN112541228B (en) * | 2020-12-10 | 2022-11-08 | 重庆交通大学 | Pantograph active control method for contact force duration memory network prediction |
| CN113032985A (en) * | 2021-03-11 | 2021-06-25 | 北京必创科技股份有限公司 | Intelligent service life assessment method and device for wireless sensing equipment |
| CN113032985B (en) * | 2021-03-11 | 2024-04-26 | 北京必创科技股份有限公司 | Intelligent evaluation method and device for service life of wireless sensing equipment |
| CN114001850A (en) * | 2021-10-25 | 2022-02-01 | 南京地铁建设有限责任公司 | Pantograph pressure detection method and system |
| CN115723578A (en) * | 2022-12-29 | 2023-03-03 | 湖南行必达网联科技有限公司 | Pantograph control method, device, equipment and working machine |
| CN116400623A (en) * | 2023-04-06 | 2023-07-07 | 南京星河世纪信息技术有限公司 | Intelligent monitoring system for high-voltage equipment |
| CN116400623B (en) * | 2023-04-06 | 2024-05-14 | 国网安徽省电力有限公司芜湖供电公司 | Intelligent monitoring system for high-voltage equipment |
| CN118133947A (en) * | 2024-05-07 | 2024-06-04 | 广东艾林克能源装备有限公司 | AI processing method and system for air compressor energy system |
| CN118133947B (en) * | 2024-05-07 | 2024-07-16 | 广东艾林克能源装备有限公司 | AI processing method and system for air compressor energy system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110361180B (en) | 2021-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110361180A (en) | Intelligent train pantograph military service performance dynamic monitoring and appraisal procedure and its system | |
| CN110376003A (en) | Intelligent whole train service life prediction technique and its system based on BIM | |
| CN103514366B (en) | Urban air quality concentration monitoring missing data recovering method | |
| Celik et al. | Generalized feed-forward based method for wind energy prediction | |
| CN106248381A (en) | A kind of rolling bearing life dynamic prediction method based on multiple features and phase space | |
| CN102169630A (en) | Quality control method of road continuous traffic flow data | |
| Pandit et al. | Data‐driven weather forecasting models performance comparison for improving offshore wind turbine availability and maintenance | |
| Zargarnezhad et al. | Predicting vehicle fuel consumption in energy distribution companies using ANNs | |
| Mathew et al. | Regression kernel for prognostics with support vector machines | |
| Deng et al. | Correlation model of deflection, vehicle load, and temperature for in‐service bridge using deep learning and structural health monitoring | |
| CN108694479A (en) | Consider the distribution network reliability prediction technique that weather influences time between overhaul | |
| Tamilselvan et al. | Optimization of wind turbines operation and maintenance using failure prognosis | |
| McMillan et al. | Flow forecasting for leakage burst prediction in water distribution systems using long short-term memory neural networks and Kalman filtering | |
| Zhao et al. | Rolling bearing remaining useful life prediction based on wiener process | |
| CN105205572A (en) | Photovoltaic power generation output forecasting method based on EMD (Empirical Mode Decomposition) and Elman algorithm | |
| CN105741184A (en) | Transformer state evaluation method and apparatus | |
| CN117269340A (en) | GIS equipment intelligent fault diagnosis method and system based on gas components | |
| Wang et al. | Prognosis-informed wind farm operation and maintenance for concurrent economic and environmental benefits | |
| Pang et al. | A New Fault Diagnosis Method for Quality Control of Electromagnet Based on T–S Fault Tree and Grey Relation | |
| CN115600695A (en) | Fault diagnosis method of metering equipment | |
| Wang et al. | Remaining Life Prediction for High-speed Rail Bearing Considering Hybrid Data-model-driven Approach | |
| KR20230123574A (en) | Transfer learning driven sequential forecasting and ventilation control of PM2.5 associated health risk levels in underground public facilities | |
| CN107124003A (en) | Wind power plant wind energy Forecasting Methodology and equipment | |
| Eshkevari et al. | AI-enabled indirect bridge strain sensing using field acceleration data | |
| Qian et al. | Research on deterioration evolution trend of primary loop piping in nuclear power plant based on fusion health index |
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 |