CN106809251B - The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system - Google Patents

The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system Download PDF

Info

Publication number
CN106809251B
CN106809251B CN201510872938.7A CN201510872938A CN106809251B CN 106809251 B CN106809251 B CN 106809251B CN 201510872938 A CN201510872938 A CN 201510872938A CN 106809251 B CN106809251 B CN 106809251B
Authority
CN
China
Prior art keywords
data
suspendability
track
input
sample
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201510872938.7A
Other languages
Chinese (zh)
Other versions
CN106809251A (en
Inventor
龙志强
窦峰山
戴春辉
范成鑫
侯圣杰
文艳辉
黄号凯
吴峻
周文武
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
BEIJING HOLDING MAGNETIC SUSPENSION TECHN DEVELOPMENT Co Ltd
National University of Defense Technology
Original Assignee
BEIJING HOLDING MAGNETIC SUSPENSION TECHN DEVELOPMENT Co Ltd
National University of Defense Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by BEIJING HOLDING MAGNETIC SUSPENSION TECHN DEVELOPMENT Co Ltd, National University of Defense Technology filed Critical BEIJING HOLDING MAGNETIC SUSPENSION TECHN DEVELOPMENT Co Ltd
Priority to CN201510872938.7A priority Critical patent/CN106809251B/en
Publication of CN106809251A publication Critical patent/CN106809251A/en
Application granted granted Critical
Publication of CN106809251B publication Critical patent/CN106809251B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B61—RAILWAYS
    • B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
    • B61L25/00—Recording or indicating positions or identities of vehicles or vehicle trains or setting of track apparatus
    • B61L25/02—Indicating or recording positions or identities of vehicles or vehicle trains
    • B61L25/021—Measuring and recording of train speed
    • 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
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B61—RAILWAYS
    • B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
    • B61L23/00—Control, warning, or like safety means along the route or between vehicles or vehicle trains
    • B61L23/04—Control, warning, or like safety means along the route or between vehicles or vehicle trains for monitoring the mechanical state of the route
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B61—RAILWAYS
    • B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
    • B61L25/00—Recording or indicating positions or identities of vehicles or vehicle trains or setting of track apparatus
    • B61L25/02—Indicating or recording positions or identities of vehicles or vehicle trains
    • B61L25/028—Determination of vehicle position and orientation within a train consist, e.g. serialisation

Abstract

The present invention provides a kind of monitoring of magnetic suspension train rail irregularity and prediction techniques, device and system.The monitoring method includes: the first running state data and the first suspendability data for obtaining magnetic suspension train when running in actual track；The certain tracks irregularity type of the track irregularity point is determined according to the first train position information；The first track irregularity degree corresponding with first speed of service, the first suspendability data is searched from the suspendability data of each track irregularity degree of certain tracks irregularity type and each corresponding relationship of the speed of service.Track irregularity degree of the track irregularity point in different time sections on magnetic suspension train rail can be monitored by this method, so as to realize the monitoring to magnetic suspension train rail irregularity, and then can guarantee the safety of magnetic suspension train operation.

Description

The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system
Technical field
The present invention relates to the monitoring of magnetic suspension train field more particularly to a kind of track irregularity of magnetic suspension train and in advance Survey methods, devices and systems.
Background technique
Magnetic suspension train is the support and guiding control that vehicle is realized with non-contacting electromagnetic force.In the rail of magnetic suspension train In road construction, largely use the structures such as elevated bridge, due to bridge base formation the problems such as, build one it is completely smooth The phenomenon that train rail is difficult to realize, and a variety of track irregularities such as height, level are constantly present in train rail.And Track irregularity is the main contributor for causing magnetic suspension train to vibrate, and serious track irregularity can not only cause magnetic suspension train High vibration, increase active force, result even in magnetic suspension train and the direct phase of track Collision, jeopardizes traffic safety.
Therefore, in order to ensure the safe and stable operation of magnetic suspension train, seek the prison to magnetic suspension train rail irregularity Survey method is this field technical problem urgently to be resolved.
Summary of the invention
In view of this, the present invention provides a kind of monitoring of magnetic suspension train rail irregularity and prediction technique, device and System.
In order to solve the above-mentioned technical problem, present invention employs following technical solutions:
A kind of monitoring method of magnetic suspension train rail irregularity, comprising:
Obtain first running state data and first suspendability data of the magnetic suspension train when running in actual track； First running state data includes the first train position information and first speed of service；It include track in the actual track Irregularity point；
The certain tracks irregularity type of the track irregularity point is determined according to the first train position information；
From each of the suspendability data of each track irregularity degree of certain tracks irregularity type and the speed of service The first track irregularity journey corresponding with first speed of service, the first suspendability data is searched in a corresponding relationship Degree.
A kind of prediction technique of magnetic suspension train rail irregularity, comprising:
The method according to above embodiment obtains track of the track irregularity point at T different moments of past not Evenness；Wherein, T >=2, and T is integer；
It is pre- using time-based prediction model according to the T different moments corresponding track irregularity degree in the past Survey track irregularity degree when the following predetermined time.
A kind of prediction technique of magnetic suspension train rail irregularity, comprising:
Obtain the suspendability data that magnetic suspension train passed through actual track irregularity point within the n moment of past；Wherein, N >=2, and T is integer；
According to the n different moments corresponding suspendability data in the past, predicted using time-based prediction model Suspendability data when the following predetermined time；
The relationship for searching suspendability data and track irregularity degree obtains and the suspendability in the following predetermined time The corresponding track irregularity degree of data.
A kind of monitoring and forecasting system of magnetic suspension train rail irregularity, comprising:
Positioning-speed-measuring device passes through the actual track for monitoring magnetic suspension train in actual track in operational process On track irregularity point the first running state data；First running state data include the first train position information and First speed of service；
Suspendability data monitoring device, for monitor magnetic suspension train in actual track in operational process by described First suspendability data of the track irregularity point in actual track；The first suspendability data include levitation gap width At least one of value and levitating current amplitude；
Processor, for executing method described in any of the above-described embodiment.
A kind of monitoring device of magnetic suspension train rail irregularity, comprising:
First acquisition unit, for obtain magnetic suspension train in actual track run when the first running state data and First suspendability data；First running state data includes the first train position information and first speed of service；It is described It include track irregularity point in actual track；
First determination unit, for determining the specific rail of the track irregularity point according to the first train position information Road irregularity type；
First searching unit, the suspendability number for each track irregularity degree from certain tracks irregularity type According to searching corresponding with first speed of service, the first suspendability data the in each corresponding relationship with the speed of service One track irregularity degree.
A kind of prediction meanss of magnetic suspension train rail irregularity, comprising:
Third acquiring unit obtains track irregularity point in mistake for the monitoring method according to any of the above-described embodiment Go track irregularity degree when T different moments；Wherein, T >=2, and T is integer；
First predicting unit, for according to the T different moments corresponding track irregularity degree in the past, using being based on The prediction model of time predicts track irregularity degree when the following predetermined time.
A kind of prediction meanss of magnetic suspension train rail irregularity, comprising:
4th acquiring unit, for obtaining magnetic suspension train within the n moment of past by actual track irregularity point Suspendability data；Wherein, n >=2, and T is integer；
Second predicting unit is used for according to the n different moments corresponding suspendability data in the past, when using being based on Between suspendability data of prediction model when predicting the following predetermined time；
Third searching unit obtains pre- with future for searching the relationship of suspendability data and track irregularity degree If the corresponding track irregularity degree of suspendability data in the moment.
Compared to the prior art, the invention has the following advantages:
As seen through the above technical solutions, in the monitoring method of magnetic suspension train rail irregularity provided by the invention, in advance First obtain each of suspendability data under each track irregularity degree under different track irregularity types and the speed of service A corresponding relationship.Wherein, the positioning-speed-measuring device being mounted on magnetic suspension train can obtain magnetic suspension train in actual track The first train position information and first speed of service when upper operation, since track irregularity point position in orbit is fixed Constant, and the corresponding track irregularity type of track irregularity point in the actual track can also be known in advance, so, It can determine the track irregularity type of track irregularity point by the first train position information.Again due to a track irregularity Each track irregularity degree corresponds to the different corresponding relationships of suspendability data and the speed of service under type, therefore, from determination The track irregularity type under the corresponding suspendability data of each track irregularity degree, each correspondence of the speed of service The first track irregularity journey corresponding with first speed of service, the first suspendability data can be found in relationship Degree.Therefore, in this way, track of the track irregularity point in different time sections on magnetic suspension train rail can be monitored Irregularity degree so as to realize the monitoring to magnetic suspension train rail irregularity, and then can guarantee that magnetic suspension train is transported Capable safety.
Detailed description of the invention
In order to which a specific embodiment of the invention is expressly understood, used when a specific embodiment of the invention is described below The attached drawing arrived makees a brief description.It should be evident that these attached drawings are only section Example of the invention, those skilled in the art Under the premise of not making the creative labor, other attached drawings can also be obtained.
Fig. 1 is the relevant technologies middle orbit in random triangular irregularity schematic diagram；
Fig. 2 is the relevant technologies middle orbit faulting of slab ends schematic diagram；
Fig. 3 is the relevant technologies middle orbit long wave periodicity irregularity schematic diagram；
Fig. 4 be under different random triangular irregularity degree, train running speed and levitation gap fluctuation amplitude it Between relationship；
Fig. 5 be under different random triangular irregularity degree, train running speed and levitating current fluctuation amplitude it Between relationship；
Fig. 6 be speed be 25m/s when, the relationship between different size faulting of slab ends value and levitation gap fluctuation amplitude；
Fig. 7 be speed be 25m/s when, the relationship between different size faulting of slab ends value and levitating current fluctuation amplitude；
Fig. 8 is Stochastic track irregularity simulation schematic diagram；
Fig. 9 is that the two sections of section of track seam crossings simulation schematic diagram of faulting of slab ends occur；
Figure 10 is track irregularity amplitude awIn random irregularities track when respectively 2.5mm and 3.5mm, different speeds With the relation curve between levitation gap fluctuation amplitude；
Figure 11 is monitoring and the forecasting system frame of magnetic suspension train rail irregularity degree provided in an embodiment of the present invention Figure；
Figure 12 is the flow diagram for the magnetic suspension train rail irregularity monitoring method that the embodiment of the present invention one provides；
Figure 13 be it is provided by Embodiment 2 of the present invention to the following predetermined time when magnetic suspension train rail irregularity it is pre- The flow diagram of survey method；
Figure 14 is the specific implementation flow diagram of the step S132 in the embodiment of the present invention two；
Figure 15 is recurrent composite BP neural network model structure schematic diagram in the related technology；
Figure 16 is the prediction technique flow diagram for the magnetic suspension train rail irregularity that the embodiment of the present invention three provides；
Figure 17 is the specific implementation flow diagram for the step S172 that the embodiment of the present invention three provides；
Figure 18 is the monitoring device structural representation for the magnetic suspension train rail irregularity degree that the embodiment of the present invention four provides Figure；
Figure 19 is the prediction meanss structural schematic diagram for the magnetic suspension train rail irregularity that the embodiment of the present invention five provides；
Figure 20 is the prediction meanss structural schematic diagram for the magnetic suspension train rail irregularity that the embodiment of the present invention six provides.
Specific embodiment
To keep goal of the invention of the invention, technological means and technical effect clearer, complete, with reference to the accompanying drawing to this The specific embodiment of invention is described.
Before introducing a specific embodiment of the invention, the track irregularity type of magnetic suspension train is introduced first.
Track irregularity type includes height, horizontal and rail to three classes.It is surveyed by the precision to existing magnetic suspension line track Amount, line track irregularity mainly appear on along road transport row height direction, mainly include shortwave randomness irregularity, two sections The faulting of slab ends and long wave periodicity irregularity that section of track seam crossing occurs.
Fig. 1 shows track in random triangular irregularity schematic diagram.Wherein, lwFor track irregularity length, as showing Example, lwUsually 2.4 meters.awFor track irregularity amplitude.Wherein, track irregularity amplitude awFor track highest point or minimum point With the maximum height difference between track normal level.In embodiments of the present invention, track irregularity amplitude awIt can be used to indicate that Track irregularity degree.
Fig. 2 is track faulting of slab ends schematic diagram.Faulting of slab ends is mainly the difference of height that two sections of sections of track as caused by track seam occur aw, wherein difference of height awFor faulting of slab ends value.In embodiments of the present invention, faulting of slab ends value awIt can also indicate track irregularity degree.
Fig. 3 is the long wave periodicity irregularity schematic diagram of track.Such track irregularity can be approximated to be sinusoidal song Line, irregularity amplitude awAs shown in figure 3, it is the highest point of track irregularity point or minimum point between track normal level Maximum height difference.
Different types of track irregularity type is to the suspendability data during magnetic suspension train operation in order to obtain Influence, present invention is alternatively directed to suspendability data of the different rail smooth types under multiple and different track irregularity degree with Relationship between train running speed has done emulation experiment.
It is influenced compared to Short wave irregularity and faulting of slab ends, the operation of track long wave irregularity type centering low speed magnetic suspension train Influence is smaller, so, the influence of long wave moment irregularity is ignored when to Stochastic track irregularity type analysis.The present invention is implemented Example faulting of slab ends both track irregularity types that only example goes out that shortwave randomness irregularity and two sections of section of track seam crossings occur it is imitative True experiment.It should be noted that in embodiments of the present invention, suspendability data include levitation gap fluctuation amplitude and the electricity that suspends Flow at least one of fluctuation amplitude.
As an example, the vehicle system parameter in emulation experiment model is derived from the emulation experiment of the embodiment of the present invention The magnetic suspension train for the CMS04 type that Beijing Holding Magnetic Suspension Techn Development Co., Ltd. and the National University of Defense technology develop.It needs It is bright, in the emulation experiment of the embodiment of the present invention, the magnetic suspension train of other models can also be used, when the other models of selection Magnetic suspension train when, related data will adjust, but emulation experiment is identical.
Referring first to shortwave random triangular irregularity track in different track irregularity degree, train running speed with Simulation result between suspendability data.
In the emulation experiment, magnetic suspension train operation speed v takes 5~30m/s, track irregularity degree awTake 1~ 6mm.Wherein, Fig. 4 is shown under different random triangular irregularity degree, and train running speed and levitation gap fluctuate width Relationship between value.Fig. 5 is shown under different random triangular irregularity degree, train running speed and levitating current wave Relationship between dynamic amplitude.As can be seen that triangle track irregularity is to suspension system gap and electric current from Fig. 4 and Fig. 5 It influences with irregularity amplitude awIncrease with running velocity and increase.In order to guarantee the safety of this aerotrain operation, Gap fluctuations amplitude, that is, levitation gap the threshold value allowed is set, when train is under a certain speed in operational process, if levitation gap More than the levitation gap threshold value of the setting, then it is assumed that train is run there are security risk at such speeds, needs to reduce speed.Make For example, in the analogue system, the levitation gap threshold value of setting can be 2.5mm.In track irregularity amplitude awFor 6mm, When speed is 30m/s, levitation gap fluctuation amplitude is 2.8mm.The levitation gap fluctuation amplitude has been more than the levitation gap of setting Threshold value, illustrating train, there are security risks under the speed of service.
It is described below when track irregularity type is faulting of slab ends, the relationship of faulting of slab ends value and suspension system gap, electric current.
In emulation experiment, in the case where different size faulting of slab ends value, selection train speed is 0~30m/s.Pass through emulation As a result it obtains, the fluctuation of the levitation gap fluctuation amplitude and levitating current fluctuation amplitude of magnetic suspension train is influenced by train speed It is smaller.Fig. 6 show speed be 25m/s when, the relationship between different size faulting of slab ends value and levitation gap fluctuation amplitude, Fig. 7 shows Gone out speed be 25m/s when, the relationship between different size faulting of slab ends value and levitating current fluctuation amplitude.From fig. 6 it can be seen that The fluctuation amplitude and faulting of slab ends value a of levitation gapwIt is equal, as long as faulting of slab ends value awThe thing that train and track bump against will be caused greater than 8mm Therefore.Moreover, influence of the faulting of slab ends value to levitating current fluctuation amplitude is also bigger, this easily causes electromagnet overcurrent protection.In phase Under same travel speed, this is a kind of maximum track irregularity type of security risk.I.e. train is by square wave type distortion When track seam crossing, corresponding levitation gap and electromagnet current will generate big ups and downs, in some instances it may even be possible to exceed magnetic suspension Threshold value set by train normally travel.
In addition, by theoretical calculation and experiment test obtain in certain speed, under different irregularity degree suspend between The family of curves of gap and levitating current is it is found that random triangular irregularity type and faulting of slab ends irregularity type have significant difference, same In the case where one faulting of slab ends value, levitation gap and current fluctuation amplitude are influenced smaller or even can be ignored by speed.And compared to Irregularity degree in random triangular irregularity type, influence of the faulting of slab ends value to levitation gap and current fluctuation amplitude are bigger.
In order to obtain accurately every kind of track irregularity type different track irregularity degree under, suspendability data with There is shortwave randomness injustice by artificially adjusting wire guides come analog orbit in relationship between train running speed, the present invention There is faulting of slab ends and long wave periodicity irregularity situation in suitable, two sections of section of track seam crossings, driving actual operation test of going forward side by side, from And obtain above-mentioned family of curves.
Wherein, the simulation of shortwave randomness irregularity is specific as follows: artificial to lift within the scope of 2.4m in route flat segments Orbit altitude at high sleeper, to simulate shortwave randomness irregularity.Analog orbit random irregularities schematic diagram is as shown in Figure 8.Its In, according to demand, track irregularity amplitude awDifferent values can be taken, such as can be 2.5mm, 3.5mm.
The simulation that faulting of slab ends occur in two sections of section of track seam crossings is specific as follows: wherein will raise carry out mould by one end track in seam crossing It is quasi-, as shown in Figure 9.Equally, according to demand, faulting of slab ends value awDifferent value can also be taken, such as can be 1mm, 1.5mm, 2.5mm.
The simulation of track long wave irregularity is specific as follows: in route flat segments, within the scope of 10m, thinking to raise in proportion Orbit altitude at sleeper, to simulate this irregularity, as shown in Figure 10.Wherein, track irregularity amplitude can also take difference Value, such as can be 4mm.
After adjusting track irregularity test segment, control magnetic suspension train with friction speed by adjusting a certain rail The track irregularity test segment of road irregularity amplitude passes through the track not using suspendability data monitoring device monitoring train The suspendability data of smooth test segment, such as levitation gap fluctuation amplitude and levitating current fluctuation amplitude.In this way, can obtain The relation curve between speed and suspendability data under to a certain track irregularity degree.Using same experiment side Method, the relation curve between speed and suspendability data under available other track irregularity degree.Figure 10 example goes out Track irregularity amplitude awIn random irregularities track when respectively 2.5mm and 3.5mm, different speeds and levitation gap are fluctuated Relation curve between amplitude.
It should be noted that above-mentioned relation curve is obtained by being fitted, correcting to lot of experimental data.In song In line fitting, makeover process, the experimental data for deviating considerably from most of experimental data is removed, so that being intended by experimental data It closes, the curve that amendment obtains is able to reflect truth.So that obtained curve and simulation curve coincide substantially.
Above-mentioned relation curve can be used as train suspendability data and operating rate under each track irregularity degree Each corresponding relationship.Further, it is also possible to the function expression of every curve be obtained according to above-mentioned curve, at this point it is possible to should Curvilinear function expression formula is each corresponding as train suspendability data and the operating rate under each track irregularity degree Relationship.
In addition it is also possible to be track irregularity degree standard scale by above-mentioned relation Curve transform.Wherein, track irregularity journey Spending the data in standard scale is the experimental data or fitting data on relation curve.In the row side of track irregularity degree standard scale To in column direction, a direction is train running speed, another direction is suspendability data, and ranks crosspoint is fortune The corresponding track irregularity degree of scanning frequency degree, suspendability data.As an example, table 1 be according to levitation gap fluctuation amplitude with Train speed judges that the track irregularity degree standard scale of track irregularity degree, table 2 are according to levitating current fluctuation amplitude The track irregularity degree standard scale of track irregularity degree is judged with train operation rate.
Table 1
Table 2
It should be noted that suspendability data under each track irregularity degree of same track irregularity type with Relation curve between the speed of service can become a family of curves.Each curve in family of curves described above, family of curves Function expression and track irregularity degree standard scale can indicate each track of same track irregularity type not Train suspendability data under evenness and the relationship between operating rate.
When knowing track irregularity type, train suspendability data and operating rate, respective carter can use not The relationship of train suspendability data and operating rate under each track irregularity degree of smooth type, knows and the train The corresponding track irregularity degree of suspendability data, the speed of service, so as to realize the prison to track irregularity degree It surveys.
It should be noted that in embodiments of the present invention, it is in advance that each track of each track irregularity type is uneven It stores along the relationship of train suspendability data and operating rate under degree in the processor, to facilitate subsequent processor to look into It looks for.
In order to realize monitoring and prediction to magnetic suspension train rail irregularity degree, the embodiment of the invention provides A kind of monitoring and forecasting system of magnetic suspension train rail irregularity degree.
Figure 11 is monitoring and the forecasting system frame of magnetic suspension train rail irregularity degree provided in an embodiment of the present invention Figure.As shown in figure 11, the monitoring and forecasting system include: positioning-speed-measuring device 111, levitation gap sensor 112, levitating current Sensor 113, processor 114, wherein processor 114 can integrate on magnetic suspension train, also can be set in ground maintenance On machine.When processor 114 is integrated on magnetic suspension train, positioning-speed-measuring device 111, levitation gap sensor 112, suspend electricity Flow sensor 113 can be communicated to connect by CAN communication bus and processor 114 respectively.When the setting of processor 114 is tieed up on ground On shield machine, positioning-speed-measuring device 111, levitation gap sensor 112, levitating current sensor 113 can pass through wireless office respectively Domain net and processor 114, which are realized, to be communicated to connect.In this way, positioning-speed-measuring device 111, levitation gap sensor 112, levitating current pass Sensor 113 can be by the data transmission respectively monitored to processor 114, so that processor 114 is carried out according to data are received Analysis processing, to obtain the track irregularity degree of the track irregularity point of train process.
In embodiments of the present invention, locating test device 111, levitation gap sensor 112, levitating current sensor 113 Before train operation, it is pre-installed on train.
Wherein, positioning-speed-measuring device 111 is used to monitor first fortune of the magnetic suspension train in actual track in operational process Row status data；First running state data includes the first train position information and first speed of service；Wherein, practical rail Road is track of the magnetic suspension train in actual moving process.There are track irregularity points in the actual track.
Levitation gap sensor 112 passes through the reality for monitoring magnetic suspension train in actual track in operational process First levitation gap fluctuation amplitude of the track irregularity point on track；
Levitating current sensor 113 passes through the reality for monitoring magnetic suspension train in actual track in operational process First levitating current fluctuation amplitude of the track irregularity point on track；
Processor 114 is for executing magnetic suspension train rail irregularity degree monitoring method described in following any embodiments And/or the prediction technique of magnetic suspension train rail irregularity degree.
As another embodiment of the present invention, monitoring described above and forecasting system can also include magnetic suspension train vehicle Data logger 115 is carried, which is pre-installed on train before train operation.It needs to illustrate It is to pass through CAN bus when processor 114 integrates ON TRAINS, between vehicle mounted data recorder 115 and processor 114 and realize Communication connection passes through between vehicle mounted data recorder 115 and processor 114 when processor 114 is arranged on ground maintenance machine WLAN realizes communication connection.
In the process of running, positioning-speed-measuring device 111, levitation gap sensor 112, levitating current sensor 113 will supervise The data measured are transferred in vehicle mounted data recorder 115 by CAN bus.In this way, vehicle mounted data recorder 115 is i.e. recordable Fortune train operation state number in positioning-speed-measuring device 111, levitation gap sensor 112, levitating current sensor 113 According to, levitation gap fluctuation amplitude, levitating current fluctuation amplitude, thus for processor 114 carry out Data Management Analysis data are provided Source.
When processor 114 is arranged when wireless local area is online, vehicle mounted data recorder 115 can lead to when train is put in storage It crosses the equipment such as WLAN and the data of its record is sent to processor 114, monitored so that processor 114 can be got The train operation state data and suspendability data arrived.In addition, what vehicle mounted data recorder 115 can also be recorded in real time Data are sent to processor 114.
It should be noted that levitation gap sensor 112 and levitating current sensor 113 in the embodiment of the present invention can be with It is referred to as suspendability data monitoring device, and two devices are only the examples of suspendability data monitoring device, in fact, As the extension of the embodiment of the present invention, suspendability data monitoring device can also include that other suspendability data monitorings fill It sets, no longer enumerated here.
The magnetic suspension train rail irregularity monitoring provided based on the above embodiment and forecasting system, the present invention provides magnetic The specific embodiment of aerotrain track irregularity monitoring method.Referring specifically to embodiment one.
Embodiment one
Figure 12 is the flow diagram for the magnetic suspension train rail irregularity monitoring method that the embodiment of the present invention one provides, such as Shown in Figure 12, method includes the following steps:
The first running state data of S121, positioning-speed-measuring device monitoring magnetic suspension train when being run in actual track, The first of track irregularity point of the suspendability data monitoring device monitoring magnetic suspension train when being run in actual track simultaneously Suspendability data:
It should be noted that including track irregularity point in actual track in embodiments of the present invention.Moreover, the track The irregularity type of irregularity point can be known before monitoring.In addition, may include that multiple tracks are uneven in actual track Along point.
In order to facilitate monitoring, controls magnetic suspension train and travel at the uniform speed in actual track according to certain speed.In the present invention In embodiment, it can control magnetic suspension train and travel at the uniform speed in actual track according to first speed of service.Moreover, for safety For the sake of, it avoids in track irregularity monitoring process, suspendability data of the train in actual track when driving are more than to allow Suspendability data threshold, first speed of service that magnetic suspension train travels in actual track are generally low speed.
In embodiments of the present invention, suspendability data detection device may include levitation gap sensor and levitating current At least one of sensor.Correspondingly, the first suspendability data include levitation gap fluctuation amplitude and levitating current fluctuation At least one of amplitude.
S122, processor obtain the first running state data and the first suspendability data:
The first running state data monitored is transferred to processor, suspension by CAN bus by positioning-speed-measuring device The the first suspendability data monitored are transferred to processor by CAN bus by energy data monitoring device, to make processor Get the first running state data and the first suspendability data.
S123, processor determine the certain tracks irregularity class of the track irregularity point according to the first train position information Type；
It should be noted that since position of the track irregularity point in actual track is fixed and invariable, and track The irregularity type of irregularity point is known in advance, so, processor can determine track according to the first train position information The track irregularity type of irregularity point.
S124, processor are from suspendability data and fortune under each track irregularity degree of certain tracks irregularity type The first track corresponding with first speed of service, the first suspendability data is searched in each corresponding relationship of scanning frequency degree Irregularity degree.
It should be noted that under each track irregularity degree of different track irregularity types, suspendability data and Each corresponding relationship of the speed of service is stored in advance in the processor.Moreover, the corresponding relationship can be bent for relation curve, relationship Line function expression formula or track irregularity degree standard scale.
Processor is by the speed of service and suspension in first speed of service, the first suspendability data and each corresponding relationship Performance data compares, to find first speed of service and the corresponding first track irregularity degree of the first suspendability data.
By above step according to the first suspendability data under first speed of service of train and first speed of service The track irregularity degree of the track irregularity point on magnetic suspension train rail can be monitored.So as to being magnetic suspension train Safe operation provides safety guarantee.
In order to calculate corresponding suspendability data, magnetic suspension described above when train is run under other speed The monitoring method of train rail irregularity can further include following steps:
S125, processor obtain the suspension of the first track irregularity degree according to the first track irregularity degree First corresponding relationship of performance data and the speed of service:
Due to the relationship of a track irregularity degree corresponding a suspendability data and the speed of service, so, according to First track irregularity degree obtained above can obtain the suspendability data and fortune of the first track irregularity degree First corresponding relationship of scanning frequency degree.It should be noted that uneven according to each track under track irregularity type described above It can be by different expression ways, the first corresponding relationship along the relationship between the corresponding speed of service of degree and suspendability data There can also be different expression ways, specifically, the first corresponding relationship can be the pass of the speed of service and suspendability data Be data conversion on curve, the relation curve function expression or the relation curve at the track irregularity degree it is corresponding Track irregularity degree standard scale.
S126, processor search magnetic suspension train from first corresponding relationship and pass through the rail with second speed of service Corresponding second magnetic suspension performance data when road irregularity point:
It should be noted that second speed of service can be the speed of arbitrary size.However, logical in order to calculate train high speed With the presence or absence of security risk when crossing the track irregularity point, second speed of service is generally the higher speed of magnetic suspension train operation Degree, more specifically, second speed of service are greater than first speed of service.
S127, processor judge whether the second magnetic suspension performance data is more than magnetic suspension performance data threshold value, if It is to execute step S128；
For the sake of security, for the operation conditions of magnetic suspension train, the magnetic suspension performance data of a permission can be set Threshold value be considered as train when practical suspendability data when train operation are less than the threshold value of magnetic suspension performance data Operating status safety, practical suspendability data when train operation are more than the threshold value of magnetic suspension performance data, are considered as The operating status of train is dangerous, needs to take train safe handling measure.
S128, processor carry out limiting operation or to track injustice when passing through the track irregularity point to magnetic suspension train It is repaired along putting.
The above are the monitoring methods for the magnetic suspension train rail irregularity that the embodiment of the present invention one provides.The monitoring method is It is carried out by the running state data and suspendability data that monitor the magnetic suspension train run in actual track.Therefore, The monitoring method of track irregularity provided by the invention no longer rests on laboratory stage, but can be practical next real with incorporation engineering Now to the monitoring of track irregularity, therefore, this method can provide data safety guarantee for the actual motion of magnetic suspension train.
Since the monitoring method of magnetic suspension column track irregularity provided in an embodiment of the present invention can be examined so that incorporation engineering is practical Consider, therefore, in order to guarantee that the safety of train operation, monitoring method described above can be supervised daily or every fixed time period Survey the track irregularity degree of the track irregularity point in an actual track.
In this way, control train low speed operation, control train passes through before the train operation daily or every fixed time period The orbital segment of the track irregularity point of actual track, while the operating status in the operational process is monitored using data monitoring device Data and suspendability data, then from the speed of service of each track irregularity degree of each track irregularity type and outstanding In relationship between floating performance data, speed of service when low speed operation and the corresponding track of suspendability data are found not Evenness.According to the track irregularity degree found and the corresponding speed of service of track irregularity degree and suspension The relationship of performance data passes through the suspendability data of track irregularity point when can calculate the current high-speed cruising of train.The survey Suspendability data when obtained high-speed cruising provide decision-making foundation for the operation and maintenance of current train.Specifically:
If the suspendability when high-speed cruising of measuring and calculating by the suspendability data of track irregularity point more than operation Data threshold then controls magnetic suspension train in the track irregularity point and carries out limiting operation.If the suspendability of high-speed cruising When data are seriously more than suspendability data threshold, then stop same day train operation, or carry out to the orbital segment for being unsatisfactory for requiring Maintenance.
It should be noted that for monitor track irregularity when, for convenience monitor, control train on the track of all fronts or With the operation of a certain low speed constant speed on a certain orbital segment.
The monitoring method of the magnetic suspension train rail irregularity degree provided based on the above embodiment, the embodiment of the present invention is also The prediction technique of magnetic suspension train rail irregularity when providing to following certain moment.Referring specifically to embodiment two.
Embodiment two
It should be noted that since orbital forcing is in slow downward trend, different moments in the long-time service of track Track irregularity in a time series with increasing trend.But this track irregularity changes over time degree Nonlinear change, accordingly, it is difficult to the track irregularity degree at the method prediction following a certain moment with curve matching.
In order to solve this problem, the embodiment of the invention provides a kind of track irregularity journeys using multiple moment in the past Degree is predicted using magnetic suspension train rail irregularity degree of the time-based prediction algorithm to following certain moment.
Figure 13 be it is provided by Embodiment 2 of the present invention to the following predetermined time when magnetic suspension train rail irregularity it is pre- The flow diagram of survey method.As shown in figure 13, method includes the following steps:
The track irregularity degree of track irregularity point when S131, T different moments of acquisition past, wherein T >=2, and T For integer:
Rail when T different moments in the past is obtained using track irregularity degree monitoring method described in above-described embodiment one The track irregularity degree of road irregularity point.It should be noted that in the T different moments, between each adjacent two moment Duration may be the same or different.
S132, according to the T different moments corresponding track irregularity degree in the past, utilize time-based prediction mould Type predicts track irregularity degree when the following predetermined time.
In embodiments of the present invention, time-based prediction model may include: classical time series predicting model, Kalman filter prediction model, Grey Theory Forecast model, artificial nerve network model (being referred to as BP network model) With any one in recursion synthesis artificial nerve network model (being referred to as recurrent composite BP neural network model).
Wherein, recurrent composite BP neural network model by BP network model output node layer excitation function (sigmond letter Number) it is changed to linear function g (x)=x, and introduce input layer and be directly connected to weigh to output layer, to overcome the full of BP network model And on the other hand property increases lateral connection power, and using orderly in learning algorithm between hidden layer and output node layer Partial derivative replaces conventional partial derivative, with reflect network internal respectively measure between timing recurrence relation.This recurrent composite BP neural network Model is more suitable modeling and prediction to the time series with certainty increasing or decreasing trend.
The specific implementation for illustrating step S132 by taking recursion synthesizes artificial nerve network model as an example below, referring specifically to Figure 14.
It should be noted that structural schematic diagram such as Figure 15 of recursion synthesis artificial neuron model used in the embodiment of the present invention It is shown comprising input layer, hidden layer and output layer, structure are N-M-F structure, i.e. input layer, hidden layer and output layer pair The number of nodes answered is respectively N, M, F, wherein N, M, F are positive integer.As an example of recurrent composite BP neural network model, The structure of the recurrent composite BP neural network model is 9-5-1 structure, i.e. input layer, hidden layer and the corresponding number of nodes difference of output layer It is 9,5,1.
As shown in figure 14, according to T different moments of past corresponding track irregularity degree, artificial mind is synthesized using recursion The method of track irregularity degree when predicting the following predetermined time through network model the following steps are included:
S1321, the corresponding track irregularity degree of T different moments in the past is sorted according to chronological order, Form track irregularity degree time series:
As an example, the track irregularity degree time series formed can be expressed as follows: S (T)=[s (1), s (2), s (3),...,s(T)]；From in above-mentioned expression as can be seen that time series in time sequential value be different moments under track not Evenness.
In embodiments of the present invention, T=N+F+Q-1 is set；Wherein, N, F, Q are positive integer, and the physical significance of N, F It is identical as the physical significance of N and F described above, it is respectively the input layer of recurrent composite BP neural network model and the section of output layer Point, Q are the quantity of the input and output sample pair of subsequent divided.
S1322, the time sequential value in track irregularity degree time series is normalized, after being normalized Track irregularity degree time series；
It, can be according to following formula (1) to each track irregularity degree time series as an example of the invention In time sequential value be normalized；
Wherein, sminAnd smaxMinimum value and maximum value in respectively track irregularity degree time series S (T)；
Track irregularity degree when s (i) is i-th of moment；
For normalization after i-th of moment when track irregularity degree s (i).
As an example, the track irregularity degree time series after normalization is expressed as:
S1323, the structure of the input layer of artificial nerve network model, hidden layer and output layer is synthesized according to recursion by normalizing T time sequential value in track irregularity degree time series after change is divided into Q to input and output sample pair:
Division Q in input and output sample pair, each pair of input and output sample is to including N+F time sequential value, preceding N A time sequential value is the time sequential value in input sample, and rear F time sequential value is output sample.
Further, since track irregularity degree is in be slowly increased trend in the long-time service of track, this just illustrates track Irregularity also increases trend with certain certainty in addition to randomness, i.e. time factor has track irregularity degree It is certain to influence, so, time factor is also used as to the value of the input sample of recurrent composite BP neural network model.So in the present invention In the input sample of embodiment in addition to include track irregularity degree, that is, time sequential value other than, further include in the input sample At the time of time sequential value constitutes the output sample of input and output sample pair.Due to the recurrent composite BP neural network used in the present invention In model, output layer has F node, it is assumed that the last one moment in input sample is T, the then output for including in input sample It is N+1, N+2 at the time of sample ..., N+F.It is subsequent in input sample in input sample at the time of the output sample for including F moment.
As an example, the Q of division of the embodiment of the present invention is to input and output sample to as shown in table 3.
Table 3
When the structure of recurrent composite BP neural network model is 9-5-1 structure, the Q that the embodiment of the present invention divides is to input and output Sample is to as shown in table 4.
Table 4
S1324, by the Q of above-mentioned division at least partly sample of input and output sample centering to as training sample, will In input sample to each node of the input layer of recurrent composite BP neural network model in training sample, BP net is synthesized by recursion The operation of network model obtains the reality output of the model:
It should be noted that can by whole Q to input and output sample to the training as recurrent composite BP neural network model Sample, can also by Q to the part input and output sample in input and output sample to the instruction as recurrent composite BP neural network model Practice sample.When by part input and output sample to as training sample, q1 is to input and output sample to conduct before can choosing Training sample, wherein q1 < Q, and q1 are positive integer.
S1325, judge whether the reality output and the mean square error of target output reach preset requirement, if so, instruction White silk terminates, and executes step S1327, if not, executing step S1326:
In embodiments of the present invention, target output is to constitute input and output sample pair with the input sample in training sample Export sample.
S1326, adjust recurrent composite BP neural network model each node connection weight and threshold value, return to step S1324。
S1327, by the time series at nearest (N-F) a moment in the track irregularity degree time series after normalization In recurrent composite BP neural network model after value and the following F predetermined time input training, prediction obtains the following F predetermined time Track irregularity degree:
Specifically, by the time sequence at nearest (N-F) a moment in the track irregularity degree time series after normalization Train valueAnd F predetermined time T+1, T+2 of future ..., T+F is input to passing after training In the input layer for slippaging into BP network model, by the operation of the recurrent composite BP neural network model after training, recursion synthesis is obtained The reality output of BP network modelThe following F are calculated by formula (2) to preset Moment T+1, T+2 ..., track irregularity degree s'(T+1 when T+F), s'(T+2 ..., s'(T+F)；
Wherein, formula (2) is as follows:
S'(T+I) be following (T+I) a moment track irregularity degree；
It is the output at (T+I) a moment of recurrent composite BP neural network model output.
Track irregularity degree when the following predetermined time can be predicted by method shown in Figure 14, so as to for The safe operation of magnetic suspension train operation provides Data safeguard.
It should be noted that being that first sample data is normalized in the above-described embodiments, then carry out again defeated Enter to export the division of sample pair.In fact, can also first carry out input and output sample pair as another embodiment of the present invention It divides, then carries out the normalized of sample data, that is, first carry out step S1323, then execute step S1322.
In addition, when training sample to Q to the part input and output sample clock synchronization of input and output sample centering, can be by preceding q1 To input and output sample to as training sample, (Q-q1) is to input and output sample as test sample by after.The test specimens Whether this is correct for detecting the recurrent composite BP neural network model after training, to determine the recurrent composite BP neural network model after training Whether can be used for predicting the track irregularity degree of the following predetermined time.
As the specific embodiment of the present invention, in order to whether just detect the recurrent composite BP neural network model after training Really, before the track irregularity degree for predicting the following predetermined time, it can execute after model training and utilize test sample Detect the whether correct step of the recurrent composite BP neural network model after training.
Wherein, the whether correct specific implementation of recurrent composite BP neural network model after training is detected using test sample It is as follows:
The input sample of test sample is input in the recurrent composite BP neural network model after training, by model calculation, Obtain the reality output of input sample:
Judge whether the reality output and the mean square error of target output reach preset requirement, if it is, after training Recurrent composite BP neural network type it is correct, recurrent composite BP neural network model after can use the training carries out the following predetermined time The prediction of track irregularity degree；If not, the recurrent composite BP neural network model after training is incorrect, need to continue to the recursion Synthesis BP network model is trained:
Herein, target output is the output sample that input sample is corresponding in the test sample being input in model This.
Above-described embodiment based on the track irregularity degree that the measuring and calculating of embodiment one obtains second is that when predicting following default Track irregularity degree when quarter.In addition, track irregularity is exactly on most directly most important influence caused by train operation The exception of train suspendability data is caused, therefore the prediction of track irregularity degree can be converted into train suspendability The prediction of data fluctuations amplitude.Since orbital forcing is in slow downward trend in the long-time service of track, train is each Suspendability data fluctuations amplitude when being pinpointed with constant speed by track can also gradually increase, the fluctuation of suspendability data Amplitude is in a time series with increasing trend.Therefore, the forecasting problem of track irregularity degree be also just reduced to using Forecasting problem of the recurrent composite BP neural network to this time series.
Based on this, when being realized based on suspendability data fluctuations amplitude to the following predetermined time the present invention also provides one kind Magnetic suspension train rail irregularity prediction.Referring specifically to embodiment three.
Embodiment three
It should be noted that in embodiments of the present invention, suspendability data fluctuations amplitude includes levitation gap fluctuation width At least one of value and levitating current fluctuation amplitude.According to the relationship between train running speed and levitation gap fluctuation amplitude And/or the relationship between train running speed and levitating current fluctuation amplitude can determine track irregularity degree.
In addition, it is generally outstanding that track irregularity degree influences most apparent suspendability data in suspendability data Floating gap fluctuations amplitude.As an example, the embodiment of the present invention illustrates how to utilize using outstanding by taking levitation gap fluctuation amplitude as an example Floating performance data predicts the track irregularity degree of the following predetermined time.
Furthermore, it is contemplated that the incidence relation between levitating current and levitation gap can be used more to improve precision of prediction Variable prediction, therefore, the embodiment of the present invention is using levitation gap fluctuation amplitude and levitating current fluctuation amplitude as the ginseng of prediction Factor is examined, levitation gap fluctuation amplitude and levitating current fluctuation amplitude are used as the input sample of recurrent composite BP neural network model. Due to track irregularity degree in the long-time service of track in being slowly increased trend, this just illustrates track irregularity except having Outside randomness, also increasing trend with certain certainty, i.e. time factor also has a certain impact to gap fluctuations amplitude, therefore Time is also used as to the input sample of recurrent composite BP neural network model.
Figure 16 is the prediction technique flow diagram for the magnetic suspension train rail irregularity that the embodiment of the present invention three provides.Such as Shown in Figure 16, method includes the following steps:
S161, acquisition magnetic suspension train pass through the outstanding of actual track irregularity point at the n moment of past with pre-set velocity Floating performance data；
Suspendability data include levitation gap fluctuation amplitude and levitating current fluctuation amplitude.It should be noted that due to Suspendability data are not only related with track irregularity degree, also related with train running speed, so, in order to avoid train fortune Influence of the scanning frequency degree to suspendability data, in this step, it is uneven to pass through actual track for train when the n moment of past of acquisition Suspendability data along point are the data that train passes through actual track irregularity point with the operation of same pre-set velocity.
S162, according to the n different moments corresponding suspendability data in the past, utilize time-based prediction model Predict suspendability data when the following predetermined time.
It is identical with above-described embodiment two, time-based prediction model described in the embodiment of the present invention also may include: through Allusion quotation time series predicting model, Kalman filter prediction model, Grey Theory Forecast model, artificial nerve network model ( It is properly termed as BP network model) or recursion synthesis artificial nerve network model (being referred to as recurrent composite BP neural network model).
The specific implementation for illustrating step S162 by taking recursion synthesizes artificial nerve network model as an example below, referring specifically to Figure 17.
As shown in figure 17, step S162 specific implementation the following steps are included:
S1621, n different moments of past corresponding suspendability data are sorted according to chronological order, is generated outstanding Buoyancy energy data time series, wherein n is positive integer:
Specifically, n different moments of past corresponding levitation gap fluctuation amplitude is sorted according to chronological order, it is raw At levitation gap fluctuation amplitude time series Z (n), by n different moments of past corresponding levitating current fluctuation amplitude according to when Between sequencing sort, generate levitating current fluctuation amplitude I (n).
Wherein, Z (n)=[z (1), z (2) ..., z (n)]；I (n)=[i (1), i (2) ..., i (n)]
In embodiments of the present invention, n=N+F+Q-1 is set；Wherein, the physical significance of N, F are with N's described above and F Physical significance is identical, is respectively the input layer of recurrent composite BP neural network model and the number of nodes of output layer, and Q is subsequent divided Input and output sample is to quantity.
S1622, the time sequential value in suspendability data time series is normalized, after obtaining normalization Track irregularity degree time series:
Specifically, levitation gap fluctuation amplitude time series Z (n) is normalized according to above-mentioned formula (3), is obtained Levitation gap fluctuation amplitude time series Z ' (n) after to normalization, by levitating current fluctuation amplitude time series I (n) according to Above-mentioned formula (4) is normalized, levitation gap fluctuation amplitude time series I ' (n) after being normalized.
Formula (3) is as follows:
Wherein, zminAnd zmaxMinimum value and maximum value in respectively levitation gap fluctuation amplitude time series Z (T)；
Levitation gap fluctuation amplitude when z (i) is i-th of moment；
The value after levitation gap fluctuation amplitude z (i) normalization when for i-th of moment.
Levitation gap fluctuation amplitude time series after normalization
Formula (4) is as follows:
Wherein, iminAnd imaxMinimum value and maximum value in respectively levitating current fluctuation amplitude time series I (T)；
Levitating current fluctuation amplitude when i (i) is i-th of moment；
The value after levitating current fluctuation amplitude i (i) normalization when for i-th of moment；
Levitating current fluctuation amplitude time series after normalization
S1623, according to the input layer of recurrent composite BP neural network model and the number of nodes of output layer, when by suspendability data Between the data at n moment in sequence be divided into Q to input and output sample pair:
Division Q in input and output sample pair, each pair of input and output sample is to including N+F time sequential value, preceding N A time sequential value is the time sequential value in input sample, and rear F time sequential value is output sample.
Due in embodiments of the present invention, being used as and being predicted using levitation gap fluctuation amplitude and levitating current fluctuation amplitude The reference factor of track irregularity degree, so, include in the input sample of each pair of input and output sample pair [(N-F)/2] a outstanding Floating gap fluctuations amplitude and [(N-F)/2] a levitating current fluctuation amplitude, and exporting in sample only includes F levitation gap fluctuation Amplitude；Wherein, the quantity of the levitation gap fluctuation amplitude in input sample is the value that [(N-F)/2] rounds up, and is suspended The quantity of current fluctuation amplitude is that [(N-F)/2] is rounded downwards obtained value.
In embodiments of the present invention, n=N+F+Q-1 is set, as an example, the input and output sample that this step divides is to can With as shown in table 5.
Table 5
When the structure of recurrent composite BP neural network model is 9-5-1 structure, the Q that the embodiment of the present invention divides is to input and output Sample is to as shown in table 6.
Table 6
S1624, by the Q of above-mentioned division at least partly sample of input and output sample centering to as training sample, will In input sample to each node of the output layer of recurrent composite BP neural network model in training sample, BP net is synthesized by recursion The operation of network model obtains the reality output of the model:
It should be noted that can by whole Q to input and output sample to the training as recurrent composite BP neural network model Sample, can also by Q to the part input and output sample in input and output sample to the instruction as recurrent composite BP neural network model Practice sample.When by part input and output sample to as training sample, q1 is to input and output sample to conduct before can choosing Training sample, wherein q1 < Q, and q1 are positive integer.
It should be noted that in embodiments of the present invention, recurrent composite BP neural network model output variable is levitation gap wave Dynamic amplitude.
S1625, judge whether the reality output and the mean square error of target output reach preset requirement, if so, instruction White silk terminates, and executes step S1627, if not, executing step S1626:
In embodiments of the present invention, target output is to constitute input and output sample pair with the input sample in training sample Export sample.
S1626, adjust recurrent composite BP neural network model each node connection weight and threshold value, return to step S1624。
S1627, by the levitation gap fluctuation amplitude time series and levitating current fluctuation amplitude time series after normalization In nearest (N-F)/2 moment time sequential value and the following F predetermined time input it is trained after recurrent composite BP neural network The input layer of model, prediction obtain the suspendability data at the following default F moment:
Specifically, by the levitation gap fluctuation amplitude time series and levitating current fluctuation amplitude time series after normalization In nearest (N-F)/2 moment time sequential value z (n- (N-F)/2+1) ..., z (n), i (n- (N-F)/2+1) ..., i ..., (n) and future F predetermined time n+1, n+2 n+F be input to it is trained after recurrent composite BP neural network model input layer In, by the operation of the recurrent composite BP neural network model after training, obtain the output of recurrent composite BP neural network modelFollowing F predetermined time n+1, n+2 is calculated by formula (5) ..., n+F When levitation gap fluctuation amplitude z'(n+1), z'(n+2 ..., z'(n+F)；
Wherein, formula (5) is as follows:
Z'(n+I) be following (n+I) a moment track irregularity degree；
It is the output at (n+I) a moment of recurrent composite BP neural network model output.
S163, the relationship for searching suspendability data and track irregularity degree obtain presetting in F moment with following The corresponding track irregularity degree of suspendability data.
Since each track irregularity degree of the track irregularity type of a certain type corresponds to different suspendability numbers According to the relationship with irregularity degree, therefore, from the relationship of known suspendability data and track irregularity degree, Neng Goucha Find track irregularity degree corresponding with the default suspendability data at F moment of the above-mentioned future predicted.
The above are the prediction techniques of track irregularity degree described in the embodiment of the present invention three.In embodiments of the present invention, By the prediction for the suspendability data that the predictive conversion of the track irregularity degree of the following predetermined time is the following predetermined time, so Afterwards from the relationship of known suspendability data and track irregularity degree, the following predetermined time obtained with prediction is found The corresponding track irregularity degree of suspendability data.
Since under the premise of train running speed is certain and track irregularity type is certain, track irregularity degree is shadow The most important factor of suspendability data is rung, therefore, the suspendability data that prediction obtains the following predetermined time are also equivalent to Prediction has obtained the track irregularity degree of the following predetermined time.Therefore, embodiment three and embodiment two have the effect to play the same tune on different musical instruments Fruit.
It should be noted that in above-described embodiment three, it can also be only with a kind of suspendability data such as levitation gap Fluctuation amplitude or levitating current fluctuation amplitude are as the input in the input layer of recurrent composite BP neural network model, at this point, to future It is the pre- of the track irregularity degree of the following predetermined time in the prediction of the suspendability data of predetermined time and embodiment two Similar, difference is surveyed, is only that and the track irregularity degree in embodiment two is accordingly replaced with into suspendability data.? This, is not described in detail.
In addition, two type of levitation gap fluctuation amplitude and levitating current fluctuation amplitude is utilized in above-described embodiment two The suspendability data of type are predicted.In fact, the extension as the embodiment of the present invention, prediction provided in an embodiment of the present invention Method is not limited to above two suspendability data, can also be 3 kinds, 4 kinds or more suspendability data.Therefore, it is possible to Multiple embodiment is summarized are as follows: according to the suspendability data of m type to the suspendability number of one of seed type According to being predicted.Wherein, m is positive integer.At this point, every kind of suspension in the sample output of any pair of input and output sample centering The quantity of energy data is that [(N-F)/m] is a, wherein the quantity of some species of suspendability data takes [(N-F)/m] upwards Whole, the quantity of the suspendability data of other type is rounded downwards [(N-F)/m].
The monitoring method of a magnetic suspension train rail irregularity degree provided based on the above embodiment, the embodiment of the present invention Additionally provide a kind of monitoring device of magnetic suspension train rail irregularity.Referring specifically to example IV.
Example IV
Figure 18 is the monitoring device structural representation for the magnetic suspension train rail irregularity degree that the embodiment of the present invention four provides Figure.As shown in figure 18, which includes with lower unit:
First acquisition unit 181, for obtaining first operating status number of the magnetic suspension train when running in actual track According to the first suspendability data；First running state data includes the first train position information and first speed of service； It include track irregularity point in the actual track；
First determination unit 182, for determining the spy of the track irregularity point according to the first train position information Determine track irregularity type；
First searching unit 183, the suspension for each track irregularity degree from certain tracks irregularity type It is searched in each corresponding relationship of energy data and the speed of service corresponding with first speed of service, the first suspendability data The first track irregularity degree.
As a specific embodiment of the invention, monitoring device described above can also include:
Second acquisition unit 184 is uneven in first track for being obtained according to the first track irregularity degree Along the suspendability data of degree and the first corresponding relationship of the speed of service；
Second searching unit 185, for searching magnetic suspension train from first corresponding relationship with second speed of service Corresponding second magnetic suspension performance data when by the track irregularity point；
First judging unit 186, for judging whether the second magnetic suspension performance data is more than magnetic suspension performance data Threshold value carries out limiting operation when if so, passing through the track irregularity point to magnetic suspension train or clicks through to track irregularity Row maintenance.
The prediction technique of the two magnetic suspension train rail irregularities provided, the embodiment of the present invention also mention based on the above embodiment The prediction meanss for having supplied a kind of magnetic suspension train rail irregularity degree, referring specifically to embodiment five.
Embodiment five
Figure 19 is the prediction meanss structural schematic diagram for the magnetic suspension train rail irregularity that the embodiment of the present invention five provides.Such as Shown in Figure 19, which includes with lower unit:
It is a not in past T to obtain track irregularity point for the method according to embodiment one for third acquiring unit 191 Track irregularity degree when in the same time；Wherein, T >=2, and T is integer；
First predicting unit 192, for utilizing according to the T different moments corresponding track irregularity degree in the past Time-based prediction model predicts track irregularity degree when the following predetermined time.
As a specific embodiment of the invention, when the time-based prediction model is that recursion synthesizes artificial neuron Network model, the input layer of the recursion synthesis artificial nerve network model are N number of node, and output layer is F node；And N, F When being positive integer, the first predicting unit specifically includes following subelement,
First sorting subunit 1921, for by the T different moments corresponding track irregularity degree in the past according to Chronological order sequence, forms track irregularity degree time series；In the track irregularity degree time series when Between sequential value be different moments under track irregularity degree；T=N+F+Q-1, and Q is positive integer；
First normalization subelement 1922, for returning the time sequential value in track irregularity degree time series One changes, the track irregularity degree time series after being normalized；
First input and output sample sub-unit 1923, for synthesizing the input layer of artificial nerve network model according to recursion T time sequential value in the track irregularity degree time series after the normalization is divided into the number of nodes of output layer Q is to input and output sample pair, and for each pair of input and output sample to including N+F time sequential value, top n time sequential value is input Time sequential value in sample, rear F time sequential value are the time sequential value exported in sample；The input sample further includes At the time of constituting the output sample of input and output sample pair with the input sample；
First training subelement 1924, for the input sample in training sample to be input to recursion synthesis artificial neural network In the input layer of network model, the reality output of recursion artificial nerve network model is obtained；The training sample is the Q to input Export sample pair at least partly to input and output sample pair；
First judgment sub-unit 1925, for judging whether the mean square error that the reality output and target export reaches pre- If it is required that if so, training terminates, if not, the connection weight of each node of adjustment recursion synthesis artificial nerve network model With threshold value, returns and execute the input sample by training sample into recursion synthesis artificial nerve network model, passed The step of pushing away the reality output of artificial nerve network model；Wherein, the target output inputs to be formed with the input sample Export the output sample of sample pair；
First prediction subelement 1926, for will be nearest in the track irregularity degree time series after the normalization (N-F) recursion after the time sequential value at a moment and the following F predetermined time input are trained synthesizes artificial nerve network model In, prediction obtains the track irregularity degree of the following F predetermined time.
The prediction technique of the three magnetic suspension train rail irregularities provided, the embodiment of the present invention also mention based on the above embodiment The prediction meanss for having supplied a kind of magnetic suspension train rail irregularity degree, referring specifically to embodiment six.
Embodiment six
Figure 20 is the prediction meanss structural schematic diagram for the magnetic suspension train rail irregularity that the embodiment of the present invention six provides.Such as Shown in Figure 20, which includes with lower unit:
4th acquiring unit 201, for obtaining magnetic suspension train within the n moment of past by actual track irregularity point Suspendability data；Wherein, n >=2, and T is integer；
Second predicting unit 202, for utilizing base according to the n different moments corresponding suspendability data in the past Suspendability data when the prediction model of time predicts the following predetermined time；
Third searching unit 203 obtains and future for searching the relationship of suspendability data and track irregularity degree The corresponding track irregularity degree of suspendability data in predetermined time.
When the time-based prediction model is that recursion synthesizes artificial nerve network model, the artificial mind of recursion synthesis Input layer through network model is N number of node, and output layer is F node；And N, F are positive integer；And according to described n in the past The suspendability data of different moments corresponding m type, when predicting the following predetermined time using time-based prediction model Suspendability data, m be positive integer when, second predicting unit 202 include following subelement:
Second sorting subunit 2021, for pressing the suspendability data for passing by n different moments corresponding m type It sorts according to chronological order, generates suspendability data time series；Every kind of suspendability data generate corresponding one and hang Buoyancy energy data time series；
Second normalization subelement 2022, for the time sequential value in suspendability data time series to be carried out normalizing Change processing, the suspendability data time series after being normalized；
Second input and output sample to dividing subelement 2023, for according to the input layer of recurrent composite BP neural network model and The data at n moment in suspendability data time series are divided into Q to input and output sample by the number of nodes of output layer Right: for each pair of input and output sample to including N+F time sequential value, top n time sequential value is the time sequence in input sample Train value, rear F time sequential value are the time sequential value exported in sample；The input sample further includes and the input sample structure At input and output sample pair output sample at the time of；N=N+F+Q-1, and Q is positive integer；
Second training subelement 2024, for by the Q at least partly sample of input and output sample centering to conduct Input sample in training sample is input on each node of the input layer of recurrent composite BP neural network model by training sample, By the operation of recurrent composite BP neural network model, the reality output of the model is obtained:
Second judgment sub-unit 2025, for judging whether the mean square error that the reality output and target export reaches pre- If it is required that if so, training terminate, if not, adjustment recurrent composite BP neural network model each node connection weight and threshold value, It returns and executes in the input sample by training sample to each node of the output layer of recurrent composite BP neural network model, lead to The operation for crossing recurrent composite BP neural network model obtains the reality output of the model；Wherein, target output is and the input The output sample of sample formation input and output sample pair；
Second prediction subelement 2026, for nearest (N-F) in the suspendability data time series after normalizing/ The input layer of recurrent composite BP neural network model after the time sequential value at m moment and the following F predetermined time input training, in advance Measure the suspendability data of the following F predetermined time.
The above is only presently preferred embodiments of the present invention, is not intended to limit the present invention in any form.Though The right present invention is disclosed above in the preferred embodiment, and however, it is not intended to limit the invention.Anyone skilled in the art, Without departing from the scope of the technical proposal of the invention, all using the methods and technical content of the disclosure above to the technology of the present invention Scheme makes many possible changes and modifications or equivalent example modified to equivalent change.Therefore, all without departing from this hair The content of bright technical solution, according to the technical essence of the invention any simple modification made to the above embodiment, equivalent variations And modification, all of which are still within the scope of protection of the technical scheme of the invention.

Claims (22)

1. a kind of monitoring method of magnetic suspension train rail irregularity characterized by comprising
Obtain first running state data and first suspendability data of the magnetic suspension train when running in actual track；It is described First running state data includes the first train position information and first speed of service；It include that track is uneven in the actual track Along point；
The certain tracks irregularity type of the track irregularity point is determined according to the first train position information；
From each right of the suspendability data of each track irregularity degree of certain tracks irregularity type and the speed of service It should be related to middle lookup the first track irregularity degree corresponding with first speed of service, the first suspendability data.
2. the method according to claim 1, wherein the magnetic suspension train is run specifically in actual track Are as follows: the magnetic suspension train is at the uniform velocity run in actual track with first speed of service.
3. the method according to claim 1, wherein each track from certain tracks irregularity type is not It searches in the suspendability data of evenness and each corresponding relationship of the speed of service and is hanged with first speed of service, first After the floating corresponding first track irregularity degree of performance data, further includes:
The suspendability data and fortune in the first track irregularity degree are obtained according to the first track irregularity degree First corresponding relationship of scanning frequency degree；
Magnetic suspension train is searched from first corresponding relationship with second speed of service by when the track irregularity point pairs The the second suspendability data answered；
Judge whether the second suspendability data are more than suspendability data threshold, if so, passing through to magnetic suspension train Limiting operation is carried out when the track irregularity point or track irregularity point is repaired.
4. the method according to claim 1, wherein each track of the certain tracks irregularity type is uneven Each corresponding relationship of suspendability data and the speed of service along degree is each different rails of certain tracks irregularity type The relation curves of suspendability data and the speed of service under road irregularity degree, relation curve function expression or by the pass Be data conversion on curve at track irregularity degree standard scale；
Wherein, the relation curve of each track irregularity degree corresponding suspendability data and the speed of service, the pass Be curve be by fitting experimental data, amendment obtain；
Wherein, in the line direction and column direction of track irregularity degree standard scale, a direction is movement velocity, other direction For suspendability data, ranks crosspoint is the speed of service, the corresponding track irregularity degree of suspendability data.
5. the method according to claim 1, wherein the suspendability data include levitation gap amplitude and hang At least one of floating current amplitude.
6. a kind of prediction technique of magnetic suspension train rail irregularity characterized by comprising
Method according to claim 1-5 obtains track of the track irregularity point at T different moments of past Irregularity degree；Wherein, T >=2, and T is integer；
According to the T different moments corresponding track irregularity degree in the past, not using the prediction of time-based prediction model Carry out track irregularity degree when predetermined time.
7. according to the method described in claim 6, it is characterized in that, the time-based prediction model includes: the classical time Sequential forecasting models, Kalman filter prediction model, Grey Theory Forecast model, artificial nerve network model or recursion synthesis Artificial nerve network model.
8. according to the method described in claim 6, it is characterized in that, the time-based prediction model is that recursion synthesis is artificial Neural network model, the input layer of the recursion synthesis artificial nerve network model are N number of node, and output layer is F node；And N, F is positive integer；
It is described according to the T different moments corresponding track irregularity degree in the past, it is pre- using time-based prediction model Track irregularity degree when the following predetermined time is surveyed, is specifically included:
The corresponding track irregularity degree of T different moments in the past is sorted according to chronological order, forms track not Evenness time series；Time sequential value in the track irregularity degree time series be track under different moments not Evenness；T=N+F+Q-1, and Q is positive integer；
Time sequential value in track irregularity degree time series is normalized, the track irregularity after being normalized Degree time series；
The input layer of artificial nerve network model and the number of nodes of output layer are synthesized by the track after the normalization according to recursion T time sequential value in irregularity degree time series is divided into Q to input and output sample pair, each pair of input and output sample pair Including N+F time sequential value, top n time sequential value is the time sequential value in input sample, and rear F time sequential value is Export the time sequential value in sample；The input sample further includes that the output of input and output sample pair is constituted with the input sample At the time of sample；
Input sample in training sample is input in the input layer of recursion synthesis artificial nerve network model, obtains recursion conjunction At the reality output of artificial nerve network model；The training sample is the Q at least partly right of input and output sample pair Input and output sample pair；
Judge whether the reality output and the mean square error of target output reach preset requirement, if so, training terminates, if No, it is described by training sample to return to execution for the connection weight and threshold value of each node of adjustment recursion synthesis artificial nerve network model Input sample in this is input in recursion synthesis artificial nerve network model, obtains recursion synthesis artificial nerve network model The step of reality output；Wherein, the target output is the output sample with input sample formation input and output sample pair；
By the time sequential value at nearest (N-F) a moment in the track irregularity degree time series after the normalization and not In recursion synthesis artificial nerve network model after carrying out F predetermined time input training, prediction obtains the following F predetermined time Track irregularity degree.
9. a kind of prediction technique of magnetic suspension train rail irregularity characterized by comprising
Obtain the suspendability that magnetic suspension train passed through actual track irregularity point within n different moments of past with pre-set velocity Data；Wherein, n >=2, and n is integer, the suspendability data pass through actual track for train with the operation of same pre-set velocity The data of irregularity point；The suspendability data and track irregularity degree and train running speed are related；
According to the n different moments corresponding suspendability data in the past, future is predicted using time-based prediction model Suspendability data when predetermined time；
Search the train running speed be the pre-set velocity when, the relationship of suspendability data and track irregularity degree, Obtain track irregularity degree corresponding with the suspendability data in the following predetermined time.
10. according to the method described in claim 9, it is characterized in that, the time-based prediction model includes: the classical time Sequential forecasting models, Kalman filter prediction model, Grey Theory Forecast model, artificial nerve network model or recursion synthesis Artificial nerve network model.
11. according to the method described in claim 10, it is characterized in that, the time-based prediction model is that recursion synthesizes people Artificial neural networks model, the input layer of the recursion synthesis artificial nerve network model are N number of node, and output layer is F node； And N, F are positive integer；
It is described according to the n different moments corresponding suspendability data in the past, utilize the prediction of time-based prediction model Suspendability data when the following predetermined time, specifically include:
According to the suspendability data of the corresponding m type of n different moments in the past, time-based prediction model is utilized Predict suspendability data when the following predetermined time, m is positive integer.
12. according to the method for claim 11, which is characterized in that if recursion synthesis artificial nerve network model is to pass BP network model is slippaged into, the suspendability data according to the corresponding m type of n different moments in the past utilize Time-based prediction model predicts suspendability data when the following predetermined time, specifically includes:
By n different moments of past, the suspendability data of corresponding m type sort according to chronological order, generate and suspend Performance data time series；Every kind of suspendability data generate a corresponding suspendability data time series；
Time sequential value in suspendability data time series is normalized, the suspendability after being normalized Data time series；
It, will be in suspendability data time series according to the number of nodes of the input layer of recurrent composite BP neural network model and output layer The data at n moment are divided into Q to input and output sample pair: each pair of input and output sample to include N+F time sequential value, it is preceding N number of time sequential value is the time sequential value in input sample, and rear F time sequential value is the time series exported in sample Value；The input sample further includes at the time of constituting the output sample of input and output sample pair with the input sample；N=N+F+Q- 1, and Q is positive integer；
By the Q at least partly sample of input and output sample centering to as training sample, by the input in training sample Sample is input on each node of the input layer of recurrent composite BP neural network model, passes through the fortune of recurrent composite BP neural network model It calculates, obtains the reality output of the model:
Judge whether the reality output and the mean square error of target output reach preset requirement, if so, training terminates, if It is no, the connection weight and threshold value of each node of recurrent composite BP neural network model are adjusted, return execution is described will be in training sample Input sample is input on each node of the output layer of recurrent composite BP neural network model, passes through recurrent composite BP neural network model Operation obtains the reality output of the model；Wherein, the target output is to form input and output sample pair with the input sample Output sample；
By the time sequential value and future F at the moment of nearest (N-F) in the suspendability data time series after normalization/m The input layer of recurrent composite BP neural network model after a predetermined time input training, prediction obtain the outstanding of the following F predetermined time Floating performance data.
13. according to the method for claim 11, which is characterized in that the suspendability data include levitation gap fluctuation width At least one of value and levitating current fluctuation amplitude.
14. a kind of monitoring and forecasting system of magnetic suspension train rail irregularity characterized by comprising
Positioning-speed-measuring device passes through in the actual track in operational process in actual track for monitoring magnetic suspension train First running state data of track irregularity point；First running state data includes the first train position information and first The speed of service；
Suspendability data monitoring device passes through the reality for monitoring magnetic suspension train in actual track in operational process First suspendability data of the track irregularity point on track；The first suspendability data include levitation gap amplitude and At least one of levitating current amplitude；
Processor, for executing the described in any item methods of claim 1-13.
15. system according to claim 14, which is characterized in that the system also includes:
Data logger, for recording first running state data and the first suspendability data；
The processor obtains first running state data and the first suspendability data from the data logger.
16. system according to claim 14, which is characterized in that the suspendability data monitoring device includes between suspending At least one of gap sensor and levitating current sensor.
17. a kind of monitoring device of magnetic suspension train rail irregularity characterized by comprising
First acquisition unit, for obtaining first running state data and first of the magnetic suspension train when running in actual track Suspendability data；First running state data includes the first train position information and first speed of service；The reality It include track irregularity point on track；
First determination unit, for determining the certain tracks of the track irregularity point according to the first train position information not Smooth type；
First searching unit, for each track irregularity degree from certain tracks irregularity type suspendability data and The first rail corresponding with first speed of service, the first suspendability data is searched in each corresponding relationship of the speed of service Road irregularity degree.
18. monitoring device according to claim 17, which is characterized in that further include:
Second acquisition unit, for being obtained according to the first track irregularity degree in the first track irregularity degree First corresponding relationship of suspendability data and the speed of service；
Second searching unit is used to search magnetic suspension train from first corresponding relationship with second speed of service described in Corresponding second suspendability data when track irregularity point；
First judging unit, for judging whether the second suspendability data are more than suspendability data threshold, if so, Limiting operation is carried out when passing through the track irregularity point to magnetic suspension train or track irregularity point is repaired.
19. a kind of prediction meanss of magnetic suspension train rail irregularity characterized by comprising
Third acquiring unit obtained track irregularity point in past T for method according to claim 1-5 Track irregularity degree when different moments；Wherein, T >=2, and T is integer；
First predicting unit is used for according to the T different moments corresponding track irregularity degree in the past, using based on the time Track irregularity degree of prediction model when predicting the following predetermined time.
20. prediction meanss according to claim 19, which is characterized in that the time-based prediction model is recursion conjunction At artificial nerve network model, the input layer of the recursion synthesis artificial nerve network model is N number of node, and output layer is F Node；And N, F are positive integer；
First predicting unit specifically includes:
First sorting subunit is used for the corresponding track irregularity degree of T different moments in the past according to time order and function Sequence sorts, and forms track irregularity degree time series；Time sequential value in the track irregularity degree time series For the track irregularity degree under different moments；T=N+F+Q-1, and Q is positive integer；
First normalization subelement is obtained for the time sequential value in track irregularity degree time series to be normalized Track irregularity degree time series after to normalization；
First input and output sample sub-unit, for synthesizing the input layer and output layer of artificial nerve network model according to recursion Number of nodes T time sequential value in the track irregularity degree time series after the normalization is divided into Q to input Sample pair is exported, for each pair of input and output sample to including N+F time sequential value, top n time sequential value is in input sample Time sequential value, rear F time sequential value be export sample in time sequential value；The input sample further includes defeated with this At the time of entering the output sample of sample composition input and output sample pair；
First training subelement, for the input sample in training sample to be input to recursion synthesis artificial nerve network model In input layer, the reality output of recursion synthesis artificial nerve network model is obtained；The training sample is the Q to input and output Sample pair at least partly to input and output sample pair；
First judgment sub-unit, for judging whether the reality output and the mean square error of target output reach preset requirement, If so, training terminates, if not, the connection weight and threshold value of each node of adjustment recursion synthesis artificial nerve network model, It returns to the execution input sample by training sample to be input in recursion synthesis artificial nerve network model, obtains recursion conjunction At artificial nerve network model reality output the step of；Wherein, the target output inputs to be formed with the input sample Export the output sample of sample pair；
First prediction subelement, for nearest (N-F) in the track irregularity degree time series after the normalization is a In recursion synthesis artificial nerve network model after the time sequential value at moment and the following F predetermined time input training, prediction Obtain the track irregularity degree of the following F predetermined time.
21. a kind of prediction meanss of magnetic suspension train rail irregularity characterized by comprising
4th acquiring unit passed through actual track within n different moments of past for obtaining magnetic suspension train with pre-set velocity The suspendability data of irregularity point；Wherein, n >=2, and n is integer, the suspendability data are train with same default speed Degree operation passes through the data of actual track irregularity point；The suspendability data and track irregularity degree and train are transported Scanning frequency degree is related；
Second predicting unit, for time-based according to n different moments corresponding suspendability data, the utilization in the past Prediction model predicts suspendability data when the following predetermined time；
When third searching unit for searching the train running speed is the pre-set velocity, suspendability data and track The relationship of irregularity degree obtains track irregularity degree corresponding with the suspendability data in the following predetermined time.
22. prediction meanss according to claim 21, which is characterized in that the prediction model is recurrent composite BP neural network mould Type, second predicting unit include:
Second sorting subunit, for the suspendability data of n different moments corresponding m type will to be pass by according to time elder generation Sequence sorts afterwards, generates suspendability data time series；Every kind of suspendability data generate a corresponding suspendability number According to time series；
Second normalization subelement, for the time sequential value in suspendability data time series to be normalized, Suspendability data time series after being normalized；
Second input and output sample is to dividing subelement, for according to the input layer of recurrent composite BP neural network model and output layer The data at n moment in suspendability data time series are divided into Q to input and output sample pair: each pair of defeated by number of nodes Enter to export sample to including N+F time sequential value, top n time sequential value is the time sequential value in input sample, and rear F is a Time sequential value is the time sequential value exported in sample；The input sample further includes constituting input and output with the input sample At the time of the output sample of sample pair；N=N+F+Q-1, and Q is positive integer；
Second training subelement, for by the Q at least partly sample of input and output sample centering to as training sample, Input sample in training sample is input on each node of the input layer of recurrent composite BP neural network model, is closed by recursion At the operation of BP network model, the reality output of the model is obtained:
Second judgment sub-unit, for judging whether the reality output and the mean square error of target output reach preset requirement, If so, training terminates, if not, the connection weight and threshold value of each node of adjustment recurrent composite BP neural network model, return are held On each node for the output layer that the row input sample by training sample is input to recurrent composite BP neural network model, pass through The operation of recurrent composite BP neural network model obtains the reality output of the model；Wherein, target output is and the input sample The output sample of this formation input and output sample pair；
Second prediction subelement, for nearest (N-F)/m moment in the suspendability data time series after normalizing Time sequential value and the following F predetermined time input it is trained after recurrent composite BP neural network model input layer, prediction obtains The suspendability data of the following F predetermined time.
CN201510872938.7A 2015-12-02 2015-12-02 The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system Active CN106809251B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510872938.7A CN106809251B (en) 2015-12-02 2015-12-02 The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510872938.7A CN106809251B (en) 2015-12-02 2015-12-02 The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system

Publications (2)

Publication Number Publication Date
CN106809251A CN106809251A (en) 2017-06-09
CN106809251B true CN106809251B (en) 2019-09-27

Family

ID=59105777

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510872938.7A Active CN106809251B (en) 2015-12-02 2015-12-02 The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system

Country Status (1)

Country Link
CN (1) CN106809251B (en)

Families Citing this family (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107433961A (en) * 2017-07-06 2017-12-05 湖南凌翔磁浮科技有限责任公司 A kind of Permanent Magnet and Electric magnetic-type magnetic-levitation train is crossing the suspension control algolithm of steps of track
WO2019010461A1 (en) * 2017-07-06 2019-01-10 Skytran, Inc. Path correction of a vehicle relative to projected magnetic flight path
CN108645451A (en) * 2018-07-12 2018-10-12 中铁磁浮科技（成都）有限公司 Middle speed magnetic suspension train on-line condition monitoring system
CN110323978B (en) * 2019-06-27 2021-01-05 中车青岛四方机车车辆股份有限公司 Linear generator control method and device
CN110766034A (en) * 2019-08-08 2020-02-07 广东毓秀科技有限公司 Method for predicting track deformation through big data
CN111444950A (en) * 2020-03-24 2020-07-24 湖南盛鼎科技发展有限责任公司 SVM track information and running state identification method based on maglev train data
CN113525098B (en) * 2020-04-20 2023-02-03 株洲中车时代电气股份有限公司 Suspension control method and device for magnetic suspension vehicle
TWI724888B (en) * 2020-05-05 2021-04-11 崑山科技大學 Deep learning proportional derivative control method for magnetic levitation system
CN115388815B (en) * 2022-07-08 2024-04-02 西南交通大学 Method and device for measuring irregularity of track functional part of magnetic suspension system in static mode
CN117068233B (en) * 2023-10-16 2023-12-15 江苏天佑路航认证检测有限公司 Magnetic levitation train safety detection system and method based on big data

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN2826375Y (en) * 2005-02-18 2006-10-11 杜虎荣 Cage type magnetic floating ball
CN101995322A (en) * 2010-10-18 2011-03-30 常州西南交通大学轨道交通研究院 Electromechanical coupling vibration test device for maglev train
JP5254676B2 (en) * 2008-06-17 2013-08-07 東海旅客鉄道株式会社 Trajectory error detector
CN103991463A (en) * 2014-04-11 2014-08-20 西南交通大学 Low-speed magnetic suspension track irregularity detection method based on two sensors

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN2826375Y (en) * 2005-02-18 2006-10-11 杜虎荣 Cage type magnetic floating ball
JP5254676B2 (en) * 2008-06-17 2013-08-07 東海旅客鉄道株式会社 Trajectory error detector
CN101995322A (en) * 2010-10-18 2011-03-30 常州西南交通大学轨道交通研究院 Electromechanical coupling vibration test device for maglev train
CN103991463A (en) * 2014-04-11 2014-08-20 西南交通大学 Low-speed magnetic suspension track irregularity detection method based on two sensors

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"磁浮列车轨道不平顺对悬浮状态的影响分析";龙志强;《2010 Chinese Control and Decision Conference》;20101231;942-947 *

Also Published As

Publication number Publication date
CN106809251A (en) 2017-06-09

Similar Documents

Publication Publication Date Title
CN106809251B (en) The monitoring of magnetic suspension train rail irregularity and prediction technique, device and system
Santoso et al. Power quality disturbance waveform recognition using wavelet-based neural classifier. II. Application
Liu et al. Data fusion for multi-source sensors using GA-PSO-BP neural network
CN102496280B (en) Method for obtaining road condition information in real time
CN110533229A (en) Orbital maintenance moment prediction technique and device
CN102646332A (en) Traffic state estimation device and method based on data fusion
Li et al. Dynamic simulation and characteristics analysis of traffic noise at roundabout and signalized intersections
CN109446920A (en) Urban track traffic passenger&#39;s degree of crowding detection method based on convolutional neural networks
CN109357747B (en) A kind of identification of online train and speed estimation method based on fiber-optic vibration signal
CN102122823A (en) Method for positioning oscillation disturbance source in power system
CN109885854A (en) The real-time forecasting system of Flutter Boundaries and prediction technique based on arma modeling
CN103235953B (en) A kind of method of optical fiber distributed perturbation sensor pattern recognition
Chen et al. Practical travel time prediction algorithms based on neural network and data fusion for urban expressway
CN108415884A (en) A kind of modal parameters real-time tracing method
CN103456163A (en) City expressway interchange traffic capacity and running status discrimination method and system
Chen et al. Adaptive ramp metering control for urban freeway using large-scale data
CN110474862A (en) A kind of network flow abnormal detecting method and device
CN107331163A (en) A kind of queue length computational methods and device
Hosseini et al. Traffic flow prediction using MI algorithm and considering noisy and data loss conditions: An application to Minnesota traffic flow prediction
CN105866633B (en) The replay method of transmission line malfunction current traveling wave waveform based on wave weight
Ma et al. An intelligent classification framework for complex PQDs using optimized KS-transform and multiple fusion CNN
CN104931838B (en) System damping on-line monitoring method and system based on traction load shock response
CN107132515A (en) A kind of point mark screening technique constrained based on multidimensional information
Gallo et al. An artificial neural network approach for spatially extending road traffic monitoring measures
CN106652443B (en) Method for predicting short-time traffic volume of expressway with similar longitudinal and transverse dimensions

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