CN111651840A - Method for detecting arch state on track slab based on deep learning technology - Google Patents
Method for detecting arch state on track slab based on deep learning technology Download PDFInfo
- Publication number
- CN111651840A CN111651840A CN202010275388.1A CN202010275388A CN111651840A CN 111651840 A CN111651840 A CN 111651840A CN 202010275388 A CN202010275388 A CN 202010275388A CN 111651840 A CN111651840 A CN 111651840A
- Authority
- CN
- China
- Prior art keywords
- track slab
- track
- slab
- data
- mortar
- 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 50
- 238000013135 deep learning Methods 0.000 title claims abstract description 17
- 238000005516 engineering process Methods 0.000 title claims abstract description 17
- 238000006073 displacement reaction Methods 0.000 claims abstract description 48
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 239000004570 mortar (masonry) Substances 0.000 claims description 58
- 229910000831 Steel Inorganic materials 0.000 claims description 34
- 239000010959 steel Substances 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 23
- 238000013016 damping Methods 0.000 claims description 21
- 239000011800 void material Substances 0.000 claims description 18
- 230000000694 effects Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 241001499740 Plantago alpina Species 0.000 claims description 10
- 238000004088 simulation Methods 0.000 claims description 10
- 238000012805 post-processing Methods 0.000 claims description 6
- 201000010099 disease Diseases 0.000 claims description 5
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 5
- 238000005452 bending Methods 0.000 claims description 4
- 238000009795 derivation Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 8
- 230000008569 process Effects 0.000 abstract description 7
- 238000004364 calculation method Methods 0.000 abstract description 6
- 238000012544 monitoring process Methods 0.000 description 10
- 238000007667 floating Methods 0.000 description 9
- 230000008878 coupling Effects 0.000 description 7
- 238000010168 coupling process Methods 0.000 description 7
- 238000005859 coupling reaction Methods 0.000 description 7
- 230000007547 defect Effects 0.000 description 4
- 239000004576 sand Substances 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000003780 insertion Methods 0.000 description 2
- 230000037431 insertion Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 239000000725 suspension Substances 0.000 description 2
- 238000011179 visual inspection Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000010426 asphalt Substances 0.000 description 1
- 239000004568 cement Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61K—AUXILIARY EQUIPMENT SPECIALLY ADAPTED FOR RAILWAYS, NOT OTHERWISE PROVIDED FOR
- B61K9/00—Railway vehicle profile gauges; Detecting or indicating overheating of components; Apparatus on locomotives or cars to indicate bad track sections; General design of track recording vehicles
- B61K9/08—Measuring installations for surveying permanent way
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Machines For Laying And Maintaining Railways (AREA)
Abstract
The invention belongs to the technical field of track slab arch-up detection, and particularly relates to a track slab arch-up state detection method based on a deep learning technology. The method of the invention converts the classification problem into a time sequence classification problem, namely inputting the time sequence classification problem into a series of track slab displacement data signal segments and outputting the time sequence classification problem into a state classification corresponding to the track slab; aiming at the characteristics of the identification task of the arch state on the track slab, the identification process is divided into two stages: a characteristic extraction stage and a classification stage; considering that the noise of the original signal is more, particularly the signal including the track slab irregularity runs through all the time, the electroencephalogram signal is represented by multiple types of features and divided into two stages to simplify the calculation amount of the classification process and improve the calculation speed.
Description
Technical Field
The invention belongs to the technical field of track slab arch-up detection, and particularly relates to a track slab arch-up state detection method based on a deep learning technology.
Background
The novel high-speed track structure CRTS II type slab ballastless track is developed by studying and innovating after the introduction of the German Bo lattice slab ballastless track. The Jingjin high-speed railway is the first high-speed railway with the speed per hour of 300-350 km/h in China, the total length of a high-speed line is 1318km, the high-speed railway is the highest-standard high-speed railway with the longest construction mileage, the largest investment and the highest standard after the establishment of new China, ballastless tracks are laid on the whole line, and the CRTS II type plate ballastless track technology is adopted. At present, CRTS II plate-type ballastless tracks are already used in a plurality of passenger special lines such as Jingjin, Jinghu, HuHangzhou and the like. The CRTS II type plate ballastless track on the roadbed mainly comprises components such as a steel rail, a fastener system, a track plate, a Cement Asphalt Mortar (CA Mortar for short) layer, a concrete supporting layer and the like. As the CRTS II type plate ballastless track is influenced by factors such as train impact, environment temperature and the like, various diseases inevitably occur to the structure below the ballastless track, wherein the occurrence of gap of the plate ballastless track and the upward arching of the track plate are typical diseases.
At present, the field detection and maintenance of the open joint and the upper arch of the slab ballastless track mainly adopts methods of visual inspection, steel ruler insertion measurement, field slab uncovering and track inspection. The visual inspection and steel ruler insertion measurement method has the defects that the accurate distribution conditions of the middle local open joints and the open joints cannot be detected; the plate uncovering method has the defects that the plate uncovering method is only suitable for building a railway, and has high cost and low efficiency; the rail inspection vehicle has the defects of high manufacturing cost, detection only in the empty window period of train operation and incapability of monitoring the real-time state.
Patent CN201910620162.8 discloses a track slab arching distributed monitoring system and a monitoring method, which detects an arching angle of a track slab in real time, and filters an angle value detected when a train passes through by processing the arching angle, so as to retain the angle value measured when the track slab is static; the track slab arching distributed monitoring system and the monitoring method have the following defects:
1) the existing algorithm can only realize the measurement of the angle value measured when the road plate is static, and the application scene is single;
2) when a train passes by, the data at the moment can be directly filtered by the existing algorithm, and the acquired original data is missing, so that the later data analysis is not facilitated;
3) the track slab arching distributed monitoring method and system adopt node output to process data, and have the disadvantages of high circuit power, high power consumption and reduced node service life.
Disclosure of Invention
Aiming at the technical problem, the invention provides a method for detecting the arch state on a track slab based on a deep learning technology. And converting the classification problem into a time series classification problem, namely inputting the time series classification problem into a series of track slab displacement data signal segments and outputting the time series classification problem into the state classification corresponding to the track slab.
The invention is realized by the following technical scheme:
a method for detecting the arch state on a track slab based on a deep learning technology comprises the following steps:
(1) the traditional vehicle-plate type track dynamics model is improved into a vehicle-plate type track dynamics model under the condition of considering the CA mortar void effect;
(2) constructing a database for deep learning of the classified neural network: outputting track slab displacement simulation data under different voiding degrees by setting different parameters by using the vehicle-slab track dynamics model under the voiding effect of the CA mortar obtained in the step (1), and after training for multiple times to obtain enough simulation data, adding labels to the simulation data according to disease types to form a database, wherein the database comprises a training set for inputting a classification neural network and a test set for inputting the classification neural network;
(3) training a classification network by using the database constructed in the step (2): the classified network structure comprises four layers of networks, wherein the first layer of network is an input layer, the second layer of network is a BilSTM, the third layer of network is a full-connection layer, and the fourth layer of network is a softmax layer;
(4) and (3) carrying out arch state classification on the track slab according to the classification network result: taking the track slab displacement data as the network input quantity of the trained classification network, and sorting the original track slab displacement data into a characteristic sequence to be input into a BilSTM network; and after circulation of the BilSTM layer, inputting the circular shape into a softmax layer to judge the type of the arch state on the track slab, and giving a final result of judgment of the arch state on the track slab.
Further, in the step (4), the original track slab displacement data is sorted into a feature sequence, and the specific method is as follows:
(1) preprocessing a track slab displacement signal: expressing track slab displacement signals by using various types of characteristics, normalizing time sequence signals to a (0,1) range, segmenting original track slab displacement signals, and extracting signal characteristics, wherein the signal characteristic extraction is to calculate data characteristics in small segments of data obtained by segmentation and then rearrange the data characteristics into a group of data to form characteristic data; after signal characteristics are extracted, connecting the characteristic data of each section in sequence to form a final characteristic data group;
the data features in the signal feature extraction process comprise a maximum value, a minimum value, an average value, a peak-to-peak value, a rectified mean value, a variance, a standard deviation, a kurtosis, a root mean square, a form factor, a peak factor, a kurtosis factor, a pulse factor and a margin factor;
(2) and calculating corresponding result relevance weight for each feature of the obtained feature data set by utilizing a Relief algorithm, and eliminating the features with lower weight according to a set weight threshold value to obtain a final trained feature sequence.
Further, step (4) is followed by step (5): and performing post-processing according to the obtained final result of the judgment of the arch-up state of the track slab, wherein the post-processing is to judge whether the track slab is arched up by combining other judgment programs, and the classification result can correct obvious identification errors, further optimize the identification effect and perform early warning work.
Further, in the step (1), the detailed form of the vibration type coordinate differential equation set of the vehicle-plate type track dynamics model under the consideration of CA mortar void action is as follows:
in the formula: t isn(t) introducing a free beam orthogonal function system (X) after a Ritz method is adopted in a vertical vibration differential equation of the track slabnNMS is the mode order of the track plate, and NMS generalized coordinates T are selectedn(t)；
Is Tn(t) a second derivative; esIsBending rigidity of the track slab; m issβ being the unit length mass of the track slabnIs a constant; m is0Finger m0A plurality of discrete units; csqIs the distributed damping of the CA mortar at q; xpIs a free beam orthogonal function system { X }nMeaning of, p is similar to n, the range is the same as n, XnIs a free beam orthogonal function system, X, of the traditional track slabpThe free beam orthogonal function system of the improved track slab is provided; t isp(t) and XpIn the same way, TnP of (t) is like n, takenThe value range is the same as n, the band n represents the generalized coordinate of the traditional track slab, and the band p represents the generalized coordinate of the improved track slab; ksqThe distributed rigidity of the CA mortar at the position q; f corresponds to the concrete supporting layer, zf(x, t) is the vibration displacement (m), z of the concrete supporting layerf(xqAnd t) is the vibration displacement of the q CA mortar discrete unit concrete supporting layer;
is zf(xqT) obtaining a derivation, and representing the vibration speed of the q CA mortar discrete unit concrete supporting layer; l issIs the length of a single track slab; n is0The number of fasteners of a steel rail on one rail plate; subscript s corresponds to the track slab; x is the number ofiThe method is characterized by comprising the following steps of (1) establishing a coordinate system of an original coordinate system on the basis of a steel rail model, wherein N is the number of fasteners in a plate type track length range L; x is the number ofqIs xiTime q ═ i; cpiIs in a coordinate system xiDamping of the under-rail pad layer when in position; NM is the modal order of the steel rail; kpiIs in a coordinate system xiThe stiffness of the under-rail pad in position; y isp(x(j-1)×n0+i)qp(t) expressing the vertical vibration displacement of the steel rail by applying the regular vibration mode function of the simply supported beam by a Ritz method
In the formula, k is p and x has a subscript of (j-1) n0+ i wherein n0The number of fasteners of the steel rail on one rail plate is equal to the number of fasteners of the steel rail on the other rail plate.
Further, in the step (2), when constructing the database, selecting 104 meters, that is, 16 track slabs as the length of the vehicle-slab track dynamics model under the consideration of the CA mortar void effect, wherein the vehicle speed is 300 kilometers per hour, selecting the displacement of the track slabs as sample data when a train passes through one track slab, and when the sampling interval is 10-4When s is needed, different sample data are obtained by setting different emptying lengths, CA mortar emptying conditions with the longitudinal lengths of 0 meter, 0.325 meter, 0.65 meter, 1.3 meter and 1.95 meter are set to obtain samples with five emptying conditions, and then, for each sample, the data are acquiredThe number of samples obtained for each of the blanking conditions is sufficiently large to obtain 64 sets of samples.
The invention has the beneficial technical effects that:
(1) the method for detecting the arch state on the track slab based on the deep learning technology converts the classification problem into a time series classification problem, namely, the classification problem is input into a series of track slab displacement data signal segments and output into the state classification corresponding to the track slab.
(2) The invention provides a track slab arch state detection method based on a deep learning technology, aiming at the characteristics of a track slab arch state identification task, the identification process is divided into two stages: a feature extraction stage and a classification stage. Considering that the noise of the original signal is more, especially the signal including the track slab irregularity runs through all the time, the calculation resource needed by the signal training model after the preprocessing is directly used is very large, and the type of the learned information is single, so that the calculation amount of the classification process can be simplified by using various types of features to represent the electroencephalogram signal in two stages, and the calculation speed is improved.
(3) In the method provided by the invention, the data acquisition adopts track slab monitoring equipment laid along the railway, the track slab monitoring equipment has the problems of large environmental interference, irregular installation and the like, and the requirements on the aspects of data accuracy, stability and the like are more strict; in the method provided by the invention, the artificial intelligence technology is utilized, the interference signals such as track irregularity and the like are added during the training of the classification network, and the classification network can continuously learn according to the updating of the data set, so that the high accuracy and the high robustness are achieved in the aspect of identifying the abnormal arch state on the track slab in the displacement signals of the track slab with poor signal quality and large interference, the subsequent labor cost is greatly reduced, and the period for generating the final conclusion is shortened.
Drawings
FIG. 1 is a conventional model of high speed railway vehicle-slab track dynamics in an embodiment of the present invention;
FIG. 2 is a diagram illustrating a stress relationship between a steel rail and a conventional model according to an embodiment of the present invention;
FIG. 3 is a diagram illustrating a stress analysis of a track slab model according to a conventional model in an embodiment of the present invention;
FIG. 4 is a high-speed railway vehicle-plate type track dynamics improved model (i.e. vehicle-plate type track dynamics model under consideration of CA mortar void action) in the embodiment of the invention;
FIG. 5 is a force analysis of an improved track slab model according to an embodiment of the present invention;
FIG. 6 is a BiLSTM neural network structure according to an embodiment of the present invention;
FIG. 7 is a diagram of a classification neural network structure according to an embodiment of the present invention;
FIG. 8 is a feature weight arrangement after a Relief algorithm in an embodiment of the present invention;
FIG. 9 is a feature sequence with low weight removed according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
On the contrary, the invention is intended to cover alternatives, modifications, equivalents and alternatives which may be included within the spirit and scope of the invention as defined by the appended claims. Furthermore, in the following detailed description of the present invention, certain specific details are set forth in order to provide a better understanding of the present invention. It will be apparent to one skilled in the art that the present invention may be practiced without these specific details.
Aiming at the technical problems of the track slab arching distributed monitoring system and monitoring method in the prior art, the invention provides a track slab arching state detection method based on a deep learning technology, which comprises the following steps:
(1) the traditional vehicle-plate type track dynamics model is improved into a vehicle-plate type track dynamics model under the condition of considering the CA mortar void effect;
the traditional vehicle-plate track dynamics model of the high-speed railway is provided by Zhai Ming professor of southwest transportation university in vehicle-track coupling dynamics, and is called the traditional model for short. The conventional model is a vertical model of the vehicle-plate type track coupling system as shown in fig. 1, which is established by using the vehicle-track coupling dynamics principle and integrating a vehicle system and a plate type track system.
The conventional model has the following assumptions:
(1) as the CA mortar under the track slab mainly plays a role in supporting, and the vertical vibration of the ballastless track is mainly influenced by the void of the CA mortar, only the vibration in the vertical direction of the vehicle-slab track coupling system is considered.
(2) The vehicle system and the plate-type track system are symmetrical to the central line of the track, and only half of structural research of the vehicle track coupling system is considered in the calculation process.
(3) The vehicle system is a multi-rigid-body model with a two-series spring damping system and comprises a vehicle body, two bogies and four wheel pairs. A primary suspension system is arranged between the bogie and the wheel pair, and a secondary suspension system is arranged between the vehicle body and the bogie. Considering the ups and downs of the vehicle body and the bogie and the vibration of the point head, the ups and downs of the wheel pair, and the total ten degrees of freedom of the vehicle system.
(4) The steel rail is a simply supported beam with limited length on the basis of discrete elastic point support, and the elasticity and damping of the base plate under the rail and the fastener are respectively determined by the elastic coefficient KpAnd damping coefficient CpAnd (4) showing.
(5) The track slab is a finite-length free beam supported on continuously distributed linear springs and linear dampers, and the distributed rigidity and the distributed damping of the CA mortar respectively use elastic coefficients ksAnd damping coefficient csAnd (4) showing.
(6) The coupling relationship between the vehicle system and the plate-type track system is realized through wheel-rail interaction force, and a classic Hertz nonlinear elastic contact model is adopted.
The physical quantities represented by the symbols in the model are shown in Table 1
TABLE 1 physical quantities represented by symbols in the conventional model
The model equation of motion includes:
(1) vehicle body sinking and floating movement
zcIs used for sinking and floating of the vehicle body,
is zcAfter derivation, the sinking and floating speed of the vehicle body is represented,
representing the sinking and floating acceleration of the vehicle body;
and zt1Similarly, where subscript t1 denotes the front truck, t2 denotes the rear truck.
(2) Vehicle body nodding motion
(3) Sinking and floating movement of front bogie
(4) Front bogie nodding motion
(5) Sinking and floating movement of rear bogie
(6) Rear bogie nodding motion
(7) Sinking and floating movement of the first wheel pair
(8) Sinking and floating movement of the second wheel pair
(9) The third wheel moves up and down
(10) Fourth wheel pair sinking and floating movement
In the formula, piAnd (t) is the vertical wheel-rail force (i is 1-4) of the single-side wheel.
The steel rail in the traditional model is regarded as a finite-length simply supported beam on a discrete elastic point supporting basis, and the stress analysis model is shown in figure 2. Wherein p isiThe force is wheel-rail force and moves forwards at a speed v along with the vehicle; frsi(i is 1-N) is steel rail fulcrum counter-force, and N is the number of fasteners in the plate track length range L; ox is a fixed coordinate system fixedly connected to the steel rail; and o 'x' is a moving coordinate system connected to the vehicle. The relationship between these two coordinates is:
x＝x′+x0+νt (11)
wherein x is0Fixed coordinates of the rear wheel at the starting moment; t is a time variable.
As can be seen from fig. 2, the differential equation of the vibration of the steel rail is:
wherein
In the formula, subscript "r" corresponds to steel rail, subscript "s" corresponds to track plate, and zr(xiT) and zs(xiT) respectively representing vertical displacement variables (m) of the steel rail and the track plate at the fastener;
and
respectively, the vertical velocity variables (m/s) of the rail and the rail plate at the fasteners.
Coordinates x of each wheelwj(j 1 to 4) are each:
coordinates of each fastener
xi＝ilp(15)
By adopting a Ritz method and applying a regular vibration mode function of the simply supported beam, the vertical vibration displacement of the steel rail can be expressed as:
wherein, NM is the mode order of the steel rail, and is generally 0.5N; vibration type of rail
Substituting formula (16) for formula (12) to obtain:
both sides of the above formula are multiplied by Yh(x) (h ═ 1,2,3,. and, NM), integrate x from 0 to L, and note the orthogonality of the modes:
is provided with
Depending on the nature of the function, equation (20) can be organized as:
because of the fact that
Therefore, equation (21) can be simplified as:
the method is a basic form of a steel rail vibration type coordinate second-order ordinary differential equation set.
Further, formula (16) is substituted for formula (13)
Then, the formula (24) becomes
This is the detailed form of the differential equation set of the vibration mode coordinate of the steel rail.
The slab track model simplifies the CA mortar into springs and dampers that are continuously distributed along the track slab, which is seen as a finite free beam supported on continuously distributed linear springs and linear dampers as shown in fig. 3.
The differential equation of the vertical vibration of the track slab is as follows:
in the formula, ks、csRespectively distributing rigidity (N/m/m) and damping (N.s/m/m) of a CA mortar layer under the track slab along the track direction; n is0The number of fasteners of the steel rail on one rail plate is increased.
Introducing a free beam orthogonal function system { X ] by a Ritz methodn1-NMS, selecting NMS generalized coordinates Tn(t) in which
In the formula, Cn、βnIs a constant. Cn、βnLsThe values of (A) are shown in Table 2.
TABLE 2 free Beam function coefficients
The vertical displacement of the track plate can be approximated as:
wherein NMS is the mode order of the track plate, and NMS is 0.5n0Substituting formula (29) for formula (27) to obtain:
both sides of the above formula are multiplied by Xp(x) (p 1-NMS), then L is in the whole length of the track platesIntegrate x and note the orthogonality of the modes
Is provided with
Depending on the nature of the function, equation (32) can be organized as:
because of the fact that
Therefore, equation (33) can be simplified as:
this is the dynamic equation of the track slab model.
Further, formula (29) is substituted for formula (25)
Then, equation (36) becomes:
this is the detailed form of the differential equation set of the track slab mode coordinate.
In the formula: in the formula: t isn(t) introducing a free beam orthogonal function system { X ] after a Ritz method is adopted in a track slab vertical vibration differential equation (27)n}(n＝1-NMS), NMS is the mode order of the track plate, and NMS generalized coordinates T are selectedn(t)；
Is Tn(t) a second derivative; esIsBending rigidity of the track slab; m issβ being the unit length mass of the track slabnIs a constant; m is0Finger m0A plurality of discrete units; csqIs the distributed damping of the CA mortar at q; xpIs a free beam orthogonal function system { X }nMeaning of, p is similar to n, the range is the same as n, XnIs a free beam orthogonal function system, X, of the traditional track slabpThe free beam orthogonal function system of the improved track slab is provided; t isp(t) and XpIn the same way, TnP of (t) is similar to n, the value range is the same as n, the n represents the generalized coordinate of the traditional track slab, and the p represents the generalized coordinate of the improved track slab; ksqThe distributed rigidity of the CA mortar at the position q; f corresponds to the concrete supporting layer, zf(x, t) is the vibration displacement (m), z of the concrete supporting layerf(xqAnd t) is the vibration displacement of the q CA mortar discrete unit concrete supporting layer;
is zf(xqT) obtaining a derivation, and representing the vibration speed of the q CA mortar discrete unit concrete supporting layer; l issIs the length of a single track slab; n is0The number of fasteners of a steel rail on one rail plate; subscript s corresponds to the track slab; x is the number ofiThe method is characterized by comprising the following steps of (1) establishing a coordinate system of an original coordinate system on the basis of a steel rail model, wherein N is the number of fasteners in a plate type track length range L; x is the number ofqIs xiTime q ═ i; cpiIs in a coordinate system xiDamping of the under-rail pad layer when in position; NM is the modal order of the steel rail; kpiIs in a coordinate system xiThe stiffness of the under-rail pad in position; y isp(x(j-1)×n0+i)qp(t) expressing the vertical vibration displacement of the steel rail by applying the regular vibration mode function of the simply supported beam by a Ritz methodIn the formula, k is p and x has a subscript of (j-1) n0+ i wherein n0The number of fasteners of a steel rail on one rail plate; xp(xi)Tp(t) is obtained by substituting formula (29) into (42), wherein formula (29) is obtained by introducing a free beam orthogonal function system { X ] by a Ritz methodn1-NMS, selecting NMS generalized coordinates Tn(t) in which
In the formula, Cn、βnIs a constant. Cn、βnLsThe values of (A) are shown in Table 2. The vertical displacement of the track plate can be approximated as:
in a ballastless track dynamics equation of a conventional vehicle-plate track dynamics model, CA mortar is simplified into springs and damping which are continuously distributed along a track plate, and acting force on the track plate is embodied in the form of distributed force in the dynamics equation, so that the void of a CA mortar layer with the length of the whole track plate can be simulated only by changing the rigidity damping value of the CA mortar under the bottom of the whole track plate, and an arbitrary range cannot be simulated. In the improved model provided by the invention, a CA mortar model is discretized in the longitudinal direction, distributed spring damping is replaced by the discrete spring damping, the supporting distribution force of the CA mortar on a track plate is converted into concentrated force, and a vehicle-plate type track dynamic model considering the CA mortar void effect is established, as shown in figure 4.
In the vehicle-plate type track dynamic model considering the CA mortar void effect, provided by the invention, the CA mortar corresponding to one track plate is averagely divided into m0The CA mortar of each discrete unit is simplified into a spring and a damper which are concentrated to one point, and the elastic coefficients K are respectively usedsAnd damping coefficient CsThe longitudinal length of each unit is expressed as ls＝Ls/m0And under the condition that the CA mortar is not empty, the rigidity of the bearing surface is as follows:
ECAis the modulus of elasticity of the CA mortar; h isCAThe thickness of the CA mortar layer (CRTS II type plate ballastless track is 0.03m), and the vertical stiffness K of each discrete spring and the damping of the CA mortar model under the working condition of no voidsAnd damping coefficient CsRespectively as follows:
in the formula,
bearing surface stiffness, b0The overall width of the track slab (2.55 m for a CRTS type II track slab); c. CsDistribution of damping for CA mortar,. lsA longitudinal length for each discrete unit;
after the CA mortar model is discretized, the distributed force to the track slab is converted into concentrated force in the track slab vibration differential equation, as shown in fig. 5, in the improved track slab model, the track slab is regarded as a finite-length free beam on the basis of the discrete elastic point support, and the differential equation is:
wherein,
in the formula, the subscript f corresponds to a concrete supporting layer; z is a radical off(x, t) is the vibrational displacement (m) of the concrete support layer; fsfq(t)(q＝1～m0) The supporting force (N) of the q & ltth & gt CA mortar discrete unit to the track slab is shown; z is a radical ofs(xqT) is the vertical displacement variable of the track plate at q、KsqThe distributed rigidity of the CA mortar at q, CsqIs the distributed damping of the CA mortar at q;
wherein E issIsRepresenting the bending rigidity of the track slab; z is a radical ofs(x, t) vertical displacement variation of the track plate at the fastener; x, t are variables; m issThe mass of the track slab in unit length; m is0Is m0A plurality of discrete units; q is the q-th CA mortar discrete unit concrete supporting layer; (x-x)q) Is a differential variable; n is0The number of fasteners of a steel rail on one rail plate; frsj(t) the supporting force of the jth CA mortar discrete unit concrete supporting layer; (x-x)i) Is a differential variable.
Introducing a free beam orthogonal function system { X ] by a Ritz methodn1-NMS, selecting NMS generalized coordinates Tn(t)，{XnThe values are given in table 2.
Substituting equation (29) into equation (41), and multiplying both sides of the equation by Xp(x) (p 1-NMS), and then in the whole length range L of the track slabsInner pair x is integrated:
using the modal orthogonality and the nature of the function one can derive:
substituting formula (34) and formula (35) for formula (44):
this is the dynamic equation of the improved model of the track slab.
Further, formula (29) is substituted for formula (42):
substituting formula (37) and formula (46) for formula (45):
this is the detailed form of the improved track slab model mode shape coordinate differential equation set.
After the CA mortar is discretized, the distribution force of the CA mortar on the track plate is converted into a concentrated force, a train track coupling dynamic model considering the CA mortar void effect can be established, and the problem that the existing dynamic model cannot simulate CA mortar void under any working condition can be solved; because the modeling is based on matlab software modeling, the practical problems of low simulation efficiency, limited line length, short track irregularity wavelength and the like of a finite element model method can be solved.
(2) Constructing a database for deep learning of the classified neural network: outputting track slab displacement simulation data under different voiding degrees by setting different parameters by using the vehicle-slab track dynamics model under the voiding effect of the CA mortar obtained in the step (1), and after training for multiple times to obtain enough simulation data, adding labels to the simulation data according to disease types to form a database, wherein the database comprises a training set for inputting a classification neural network and a test set for inputting the classification neural network;
when constructing the database, in this embodiment, 104 meters, that is, 16 track slabs are selected as the length of the vehicle-slab track dynamics model under consideration of CA mortar void, the vehicle speed is 300 km/h, the displacement of the track slab is selected as sample data when a train passes through one track slab, and when the sampling interval is 10-4When s is carried out, different sample data are obtained by setting different void lengths, and CA mortar voids with longitudinal lengths of 0 meter, 0.325 meter, 0.65 meter, 1.3 meter and 1.95 meter are respectively arranged under the track slab to obtain samples under five conditions; then a sufficient number of samples are obtained for each of the conditions set for the void, obtaining 64 sets of samples.
(3) Training a classification network by using the database constructed in the step (2): the classified network structure comprises four layers of networks, wherein the first layer of network is an input layer, the second layer of network is a BilSTM, the third layer of network is a full-connection layer, and the fourth layer of network is a softmax layer; both first and second tier networks are commonly used to model context information in natural language processing tasks, where the full name of LSTM is Long Short-Term Memory, which is one of rnn (current Neural network). LSTM is well suited for modeling time series data, such as text data, due to its design features. BilSTM is an abbreviation of Bi-directional Long Short-Term Memory, and is formed by combining forward LSTM and backward LSTM, and the structure of the BilSTM is shown in figure 6;
(4) and (3) carrying out arch state classification on the track slab according to the classification network result: taking the track slab displacement data as the network input quantity of the trained classification network, and sorting the original track slab displacement data into a characteristic sequence to be input into a BilSTM network; and after circulation of the BilSTM layer, inputting the circular shape into a softmax layer to judge the type of the arch state on the track slab, and giving a final result of judgment of the arch state on the track slab.
The classification network structure is shown in fig. 7, and specifically, in this step, the rail plate displacement data as the network input quantity of the trained classification network may be collected by the method provided in patent CN201910620162.8, where the collection method specifically is to collect rail plate vibration displacement data by using the inclination angle sensing nodes installed on the rail plate, and then convert the rail plate vibration displacement data into rail plate inclination angle data for use in the present invention.
Specifically, in the step (4), the original track slab displacement data is sorted into a feature sequence, and the specific method is as follows:
(1) preprocessing a track slab displacement signal: the method comprises the steps of representing track slab displacement signals by using various types of characteristics, normalizing time sequence signals to a (0,1) range, segmenting original track slab displacement signals and extracting signal characteristics, wherein the original track slab displacement signals can be generally divided into three segments, if the signals are too long, the number of the segments can be increased, and the signal characteristic extraction is to calculate data characteristics in small segments of data obtained by segmentation and then rearrange the data characteristics into a group of data to form characteristic data; after signal characteristics are extracted, connecting the characteristic data of each section in sequence to form a final characteristic data group;
the data features in the signal feature extraction process comprise a maximum value, a minimum value, an average value, a peak-to-peak value, a rectified mean value, a variance, a standard deviation, a kurtosis, a root mean square, a form factor, a peak factor, a kurtosis factor, a pulse factor and a margin factor;
(2) and calculating a corresponding result correlation weight for each feature of the obtained feature data set by using a Relief algorithm, as shown in fig. 8, removing features with lower weights according to a set weight threshold, and then obtaining a final training feature sequence, as shown in fig. 9.
The Relief algorithm is a Feature weighting algorithm (Feature weighting algorithms), different weights are given to features according to the relevance of each Feature and category, and features with weights smaller than a certain threshold value are removed. The relevance of features and classes in the Relief algorithm is based on the discriminative power of features on close-range samples. Randomly selecting a sample R from a training set D by adopting a Relief algorithm, then searching a nearest neighbor sample H from samples similar to the R, wherein the nearest neighbor sample H is called NearHit, searching a nearest neighbor sample M from samples different from the R, wherein the nearest neighbor sample M is called NearMiss, and the samples NearHit and NearMiss are also samples in the training set D; the weight of each feature is then updated according to the following rules: if the distance between R and Near Hit on a feature is smaller than the distance between R and Near Miss, the feature is beneficial to distinguishing the nearest neighbors of the same class and different classes, and the weight of the feature is increased; conversely, if the distance between R and Near Hit is greater than the distance between R and Near miss, indicating that the feature has a negative effect on distinguishing between similar and dissimilar nearest neighbors, the weight of the feature is reduced. Repeating the above processes m times to obtain the average weight of each feature. The larger the weight of a feature is, the stronger the classification capability of the feature is, and conversely, the weaker the classification capability of the feature is. The running time of the Relief algorithm is increased linearly along with the increase of the sampling times m of the samples and the number N of the original features, so that the running efficiency is very high.
Further, step (4) is followed by step (5): and performing post-processing according to the obtained final result of the judgment of the arch-up state of the track slab, wherein the post-processing is to judge whether the track slab is arched up by combining other judgment programs, and the classification result can correct obvious identification errors, further optimize the identification effect and perform early warning work. The other judging program may be the judging program mentioned in patent CN201910620162.8, and judges whether the track slab is arched upward according to the comparison between the set inclination angle threshold and the collected inclination angle data.
Compared with the prior art, the method can realize that:
(1) events are converted into time series classification problems, and the application scenes are more diverse by utilizing new technology classification;
(2) the deep learning technology is adopted, so that data information loss is avoided, and the method can be conveniently expanded to other similar tasks;
(3) according to the difference of the collected data sets, more result types can be judged, so that the model has wider application range and more extensibility;
(4) and the module can be further more accurate and reliable by combining with other result judgment.
Claims (5)
1. A method for detecting the arch state on a track slab based on a deep learning technology is characterized by comprising the following steps:
(1) the traditional vehicle-plate type track dynamics model is improved into a vehicle-plate type track dynamics model under the condition of considering the CA mortar void effect;
(2) constructing a database for deep learning of the classified neural network: outputting track slab displacement simulation data under different voiding degrees by setting different parameters by using the vehicle-slab track dynamics model under the voiding effect of the CA mortar obtained in the step (1), and after training for multiple times to obtain enough simulation data, adding labels to the simulation data according to disease types to form a database, wherein the database comprises a training set for inputting a classification neural network and a test set for inputting the classification neural network;
(3) training a classification network by using the database constructed in the step (2): the classified network structure comprises four layers of networks, wherein the first layer of network is an input layer, the second layer of network is a BilSTM, the third layer of network is a full-connection layer, and the fourth layer of network is a softmax layer;
(4) and (3) carrying out arch state classification on the track slab according to the classification network result: taking the track slab displacement data as the network input quantity of the trained classification network, and sorting the original track slab displacement data into a characteristic sequence to be input into a BilSTM network; and after circulation of the BilSTM layer, inputting the circular shape into a softmax layer to judge the type of the arch state on the track slab, and giving a final result of judgment of the arch state on the track slab.
2. The method for detecting the camber state on the track slab based on the deep learning technology as claimed in claim 1, wherein in the step (4), the original track slab displacement data is sorted into a feature sequence, and the specific method comprises:
(1) preprocessing a track slab displacement signal: expressing track slab displacement signals by using various types of characteristics, normalizing time sequence signals to a (0,1) range, segmenting original track slab displacement signals, and extracting signal characteristics, wherein the signal characteristic extraction is to calculate data characteristics in small segments of data obtained by segmentation and then rearrange the data characteristics into a group of data to form characteristic data; after signal characteristics are extracted, connecting the characteristic data of each section in sequence to form a final characteristic data group;
the data features in the signal feature extraction process comprise a maximum value, a minimum value, an average value, a peak-to-peak value, a rectified mean value, a variance, a standard deviation, a kurtosis, a root mean square, a form factor, a peak factor, a kurtosis factor, a pulse factor and a margin factor;
(2) and calculating corresponding result relevance weight for each feature of the obtained feature data set by utilizing a Relief algorithm, and eliminating the features with lower weight according to a set weight threshold value to obtain a final trained feature sequence.
3. The method for detecting the arch state on the track slab based on the deep learning technology as claimed in claim 1, wherein the step (4) is followed by the step (5): and performing post-processing according to the obtained final result of the judgment of the arch-up state of the track slab, wherein the post-processing is to judge whether the track slab is arched up by combining other judgment programs, and the classification result can correct obvious identification errors, further optimize the identification effect and perform early warning work.
4. The method for detecting the arch state on the track slab based on the deep learning technology as claimed in claim 1, wherein in the step (1), the detailed form of the vibration type coordinate differential equation set of the vehicle-slab track dynamics model under consideration of CA mortar void is as follows:
in the formula: t isn(t) introducing a free beam orthogonal function system (X) after a Ritz method is adopted in a vertical vibration differential equation of the track slabnNMS is the mode order of the track plate, and NMS generalized coordinates T are selectedn(t)；
Is Tn(t) a second derivative; esIsBending rigidity of the track slab; m issβ being the unit length mass of the track slabnIs a constant; m is0Finger m0A plurality of discrete units; csqIs the distributed damping of the CA mortar at q; xpIs a free beam orthogonal function system { X }nMeaning of, p is similar to n, the range is the same as n, XnIs a free beam orthogonal function system, X, of the traditional track slabpThe free beam orthogonal function system of the improved track slab is provided; t isp(t) and XpIn the same way, TnP of (t) is similar to n, the value range is the same as n, the n represents the generalized coordinate of the traditional track slab, and the p represents the generalized coordinate of the improved track slab; ksqThe distributed rigidity of the CA mortar at the position q; f corresponds to the concrete supporting layer, zf(x, t) is the vibration displacement (m), z of the concrete supporting layerf(xqAnd t) is the vibration displacement of the q CA mortar discrete unit concrete supporting layer;
is zf(xqT) obtaining a derivation, and representing the vibration speed of the q CA mortar discrete unit concrete supporting layer; l issIs the length of a single track slab; n is0The number of fasteners of a steel rail on one rail plate; subscript s corresponds to the track slab; x is the number ofiThe method is characterized by comprising the following steps of (1) establishing a coordinate system of an original coordinate system on the basis of a steel rail model, wherein N is the number of fasteners in a plate type track length range L; x is the number ofqIs xiTime q ═ i; cpiIs in a coordinate system xiDamping of the under-rail pad layer when in position; NM is the modal order of the steel rail; kpiIs in a coordinate system xiThe stiffness of the under-rail pad in position; y isp(x(j-1)×n0+i)qp(t) expressing the vertical vibration displacement of the steel rail by applying the regular vibration mode function of the simply supported beam by a Ritz method
In the formula, k is p and x has a subscript of (j-1) n0+ i wherein n0The number of fasteners of the steel rail on one rail plate is equal to the number of fasteners of the steel rail on the other rail plate.
5. The method according to claim 1, wherein in the step (2), when constructing the database, 104 meters, that is, 16 track slabs are selected as the length of the vehicle-slab track dynamics model considering CA mortar void, the vehicle speed is 300 km/h, the displacement of the track slab is selected as the sample data when the train passes through one track slab, and when the sampling interval is 10-4And s, obtaining different sample data by setting different emptying lengths, setting CA mortar emptying conditions with the longitudinal lengths of 0 meter, 0.325 meter, 0.65 meter, 1.3 meter and 1.95 meter to obtain samples of five emptying conditions, and then obtaining enough samples for each emptying setting condition to obtain 64 groups of samples.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010275388.1A CN111651840B (en) | 2020-04-09 | 2020-04-09 | Track slab arch state detection method based on deep learning technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010275388.1A CN111651840B (en) | 2020-04-09 | 2020-04-09 | Track slab arch state detection method based on deep learning technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111651840A true CN111651840A (en) | 2020-09-11 |
| CN111651840B CN111651840B (en) | 2024-04-02 |
Family
ID=72352449
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010275388.1A Active CN111651840B (en) | 2020-04-09 | 2020-04-09 | Track slab arch state detection method based on deep learning technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111651840B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113191051A (en) * | 2021-04-30 | 2021-07-30 | 西南交通大学 | Establishment method and application of nonlinear Hertz spring |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE10111957A1 (en) * | 2001-03-13 | 2002-09-19 | Baumann Theodor | High speed rail track in tubing tube matches tubing width to support spacing appropriate to track scale and support legs splayed out at set angle to bear centrally between tubing ring joins |
| CN101719183A (en) * | 2009-12-10 | 2010-06-02 | 北京交通大学 | Test simulation system for rail structures of high-speed railways and urban railway system |
| CN103678823A (en) * | 2013-12-25 | 2014-03-26 | 西南交通大学 | Plate-type ballastless track structure evaluation method based on CA mortar damage |
| CN105631124A (en) * | 2015-12-28 | 2016-06-01 | 北京交通大学 | Ballastless track damage analysis method performing combined solution based on definite element expansion and fatigue analysis |
| CN106650067A (en) * | 2016-12-12 | 2017-05-10 | 贵州理工学院 | Vehicle-track-bridge coupled vibrationfrequency domain model with middle and high frequency response output |
| CN107729649A (en) * | 2017-10-16 | 2018-02-23 | 西南交通大学 | A kind of method for calculating stock rail coupled system dynamics |
-
2020
- 2020-04-09 CN CN202010275388.1A patent/CN111651840B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE10111957A1 (en) * | 2001-03-13 | 2002-09-19 | Baumann Theodor | High speed rail track in tubing tube matches tubing width to support spacing appropriate to track scale and support legs splayed out at set angle to bear centrally between tubing ring joins |
| CN101719183A (en) * | 2009-12-10 | 2010-06-02 | 北京交通大学 | Test simulation system for rail structures of high-speed railways and urban railway system |
| CN103678823A (en) * | 2013-12-25 | 2014-03-26 | 西南交通大学 | Plate-type ballastless track structure evaluation method based on CA mortar damage |
| CN105631124A (en) * | 2015-12-28 | 2016-06-01 | 北京交通大学 | Ballastless track damage analysis method performing combined solution based on definite element expansion and fatigue analysis |
| CN106650067A (en) * | 2016-12-12 | 2017-05-10 | 贵州理工学院 | Vehicle-track-bridge coupled vibrationfrequency domain model with middle and high frequency response output |
| CN107729649A (en) * | 2017-10-16 | 2018-02-23 | 西南交通大学 | A kind of method for calculating stock rail coupled system dynamics |
Non-Patent Citations (3)
| Title |
|---|
| 向俊;赫丹;曾庆元;: "横向有限条与无砟轨道板段单元的车轨系统竖向振动分析法", 铁道学报, no. 04 * |
| 朱胜阳: "高速铁路无砟轨道结构伤损行为及其对动态性能的影响", 中国优秀博士学位论文全文数据库信息科技辑 * |
| 赵东田等: "板式无碴轨道CA砂浆的配制和性能", 天津大学学报, vol. 41, no. 7, pages 793 - 799 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113191051A (en) * | 2021-04-30 | 2021-07-30 | 西南交通大学 | Establishment method and application of nonlinear Hertz spring |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111651840B (en) | 2024-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111144039B (en) | Train dynamic weighing system and weighing method based on deep learning | |
| CN110704801B (en) | Bridge cluster structure operation safety intelligent monitoring and rapid detection complete method | |
| CN111256924B (en) | Intelligent monitoring method for expansion joint of large-span high-speed railway bridge | |
| Au et al. | Parameter identification of vehicles moving on continuous bridges | |
| JP5384166B2 (en) | Structural deformation detection system | |
| CN104598753B (en) | Bridge moving vehicle load recognition method based on Brakhage V method | |
| CN105923014B (en) | A kind of track transition Amplitude Estimation method based on evidential reasoning rule | |
| CN101266190A (en) | Normal traffic flow stayd-cable bridge mode parametric measurement device and method | |
| Yin et al. | Bridge damage identification under the moving vehicle loads based on the method of physics-guided deep neural networks | |
| CN110806324A (en) | Wheel polygon abrasion detection method based on rail displacement and data acquisition equipment | |
| CN116881819B (en) | Stay cable working state monitoring method based on isolated forest | |
| CN112329520B (en) | Truck bearing fault identification method based on generation countermeasure learning | |
| CN116052435B (en) | Urban road congestion influence range definition and road influence calculation method | |
| CN114882069A (en) | Taxi track abnormity detection method based on LSTM network and attention mechanism | |
| CN111651840A (en) | Method for detecting arch state on track slab based on deep learning technology | |
| Nguyen et al. | Damage detection of structural based on indirect vibration measurement results combined with Artificial Neural Network | |
| Zhang et al. | Structural Damage Identification System Suitable for Old Arch Bridge in Rural Regions: Random Forest Approach. | |
| Shen et al. | A boom damage prediction framework of wheeled cranes combining hybrid features of acceleration and Gaussian process regression | |
| CN117390551A (en) | Operation monitoring level high-speed railway bridge influence line tracking identification method | |
| CN112781551B (en) | Road surface condition monitoring system and monitoring method thereof | |
| CN114707205A (en) | Track foundation differential settlement identification method based on GRU neural network | |
| CN115455826A (en) | Asphalt pavement track data expansion method based on radial basis function neural network | |
| Cheema et al. | Bridge indirect monitoring using Uniform Manifold Approximation and Projection (UMAP) | |
| CN113420387B (en) | Migration diagnosis method and system for rolling bearing of compacting machine | |
| CN117709084B (en) | Contact force calculation method considering harshness of pantograph slide plate |
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 |