WO2016098773A1 - 鉄道車両状態監視装置 - Google Patents
鉄道車両状態監視装置 Download PDFInfo
- Publication number
- WO2016098773A1 WO2016098773A1 PCT/JP2015/085085 JP2015085085W WO2016098773A1 WO 2016098773 A1 WO2016098773 A1 WO 2016098773A1 JP 2015085085 W JP2015085085 W JP 2015085085W WO 2016098773 A1 WO2016098773 A1 WO 2016098773A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- vehicle
- railway vehicle
- wheel
- carriage
- railway
- Prior art date
Links
- 238000012806 monitoring device Methods 0.000 title claims abstract description 21
- 230000005856 abnormality Effects 0.000 claims abstract description 196
- 238000001514 detection method Methods 0.000 claims abstract description 28
- 238000010801 machine learning Methods 0.000 claims abstract description 20
- 230000002159 abnormal effect Effects 0.000 claims description 45
- 238000012544 monitoring process Methods 0.000 claims description 21
- 238000004088 simulation Methods 0.000 claims description 10
- 230000008520 organization Effects 0.000 claims description 5
- 230000014509 gene expression Effects 0.000 claims description 3
- 230000015572 biosynthetic process Effects 0.000 description 20
- 238000004364 calculation method Methods 0.000 description 15
- 238000000034 method Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 12
- 238000009940 knitting Methods 0.000 description 11
- 230000007246 mechanism Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000009434 installation Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61F—RAIL VEHICLE SUSPENSIONS, e.g. UNDERFRAMES, BOGIES OR ARRANGEMENTS OF WHEEL AXLES; RAIL VEHICLES FOR USE ON TRACKS OF DIFFERENT WIDTH; PREVENTING DERAILING OF RAIL VEHICLES; WHEEL GUARDS, OBSTRUCTION REMOVERS OR THE LIKE FOR RAIL VEHICLES
- B61F5/00—Constructional details of bogies; Connections between bogies and vehicle underframes; Arrangements or devices for adjusting or allowing self-adjustment of wheel axles or bogies when rounding curves
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M17/00—Testing of vehicles
- G01M17/08—Railway vehicles
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61F—RAIL VEHICLE SUSPENSIONS, e.g. UNDERFRAMES, BOGIES OR ARRANGEMENTS OF WHEEL AXLES; RAIL VEHICLES FOR USE ON TRACKS OF DIFFERENT WIDTH; PREVENTING DERAILING OF RAIL VEHICLES; WHEEL GUARDS, OBSTRUCTION REMOVERS OR THE LIKE FOR RAIL VEHICLES
- B61F5/00—Constructional details of bogies; Connections between bogies and vehicle underframes; Arrangements or devices for adjusting or allowing self-adjustment of wheel axles or bogies when rounding curves
- B61F5/02—Arrangements permitting limited transverse relative movements between vehicle underframe or bolster and bogie; Connections between underframes and bogies
- B61F5/22—Guiding of the vehicle underframes with respect to the bogies
- B61F5/24—Means for damping or minimising the canting, skewing, pitching, or plunging movements of the underframes
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61K—AUXILIARY EQUIPMENT SPECIALLY ADAPTED FOR RAILWAYS, NOT OTHERWISE PROVIDED FOR
- B61K13/00—Other auxiliaries or accessories for railways
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B61—RAILWAYS
- B61L—GUIDING RAILWAY TRAFFIC; ENSURING THE SAFETY OF RAILWAY TRAFFIC
- B61L1/00—Devices along the route controlled by interaction with the vehicle or train
- B61L1/02—Electric devices associated with track, e.g. rail contacts
- B61L1/06—Electric devices associated with track, e.g. rail contacts actuated by deformation of rail; actuated by vibration in rail
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01G—WEIGHING
- G01G19/00—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups
- G01G19/02—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for weighing wheeled or rolling bodies, e.g. vehicles
- G01G19/04—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for weighing wheeled or rolling bodies, e.g. vehicles for weighing railway vehicles
- G01G19/045—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for weighing wheeled or rolling bodies, e.g. vehicles for weighing railway vehicles for weighing railway vehicles in motion
- G01G19/047—Weighing apparatus or methods adapted for special purposes not provided for in the preceding groups for weighing wheeled or rolling bodies, e.g. vehicles for weighing railway vehicles for weighing railway vehicles in motion using electrical weight-sensitive devices
Definitions
- the present invention detects railway vehicle information represented by wheel weights and the like of wheels of a railway vehicle traveling on a track, and determines a vehicle state such as the presence or absence of abnormality of the railway vehicle based on the detected vehicle information.
- the present invention relates to a state monitoring device.
- the present invention relates to a railway vehicle state monitoring apparatus that can easily determine a parameter for determining a vehicle state without requiring a lot of labor.
- Patent Documents 1 to 3 are methods for comparing an evaluation index with a predetermined threshold value and detecting an abnormality of the railway vehicle according to the magnitude.
- the threshold value for detecting this abnormality can change as appropriate depending on the structure of the railway vehicle, loading conditions, traveling conditions, and the like. For this reason, in order to detect an abnormality with high accuracy, it is necessary to determine a large number of threshold values for each of these conditions, which requires a lot of time and labor.
- the present invention has been made to solve such problems of the prior art, and can be easily determined without requiring a great deal of effort in adjusting the parameters for determining the vehicle state such as the presence or absence of abnormality of the railway vehicle. It is an object of the present invention to provide a railway vehicle state monitoring device capable of performing the above.
- a railway vehicle state monitoring apparatus includes a detection device that detects vehicle information represented by wheel weights of wheels included in a railway vehicle traveling on a track, and the detected vehicle information. And a determination device including a classifier that outputs a vehicle state such as the presence or absence of abnormality of the railway vehicle. Then, the classifier outputs the known vehicle state when the vehicle information is input, using the vehicle information and the vehicle state of the railway vehicle whose vehicle state is known as teacher data. It is generated using machine learning.
- teacher data (input vehicle information and output vehicle state) are represented by vehicle information and vehicle state represented by wheel weights of wheels of a railway vehicle whose vehicle state such as presence or absence of vehicle abnormality is known.
- the classifier is generated by machine learning.
- the vehicle information is detected by the detection device, and the detected vehicle information is input to the classifier, so that the vehicle state as the determination result is output from the classifier.
- it is only necessary to generate a classifier by machine learning using known data without requiring much time and labor for determining a threshold value as in the prior art. Can be determined.
- the vehicle information detected by the detection device and input to the classifier is represented by wheel weights of wheels included in the railway vehicle, and the classifier is configured as the vehicle state.
- the presence / absence of abnormality and the type of abnormality are output.
- the railway vehicle includes a pair of carriages having two pairs of left and right wheels on the front and rear
- vehicle information detected by the detection device includes the following equations (1) and (2 )
- the vehicle status output from the classifier is the presence / absence and type of abnormality of the railway vehicle.
- P1 is the wheel weight of the front right wheel of the front carriage
- P2 is the wheel weight of the front left wheel of the front carriage
- P3 is the wheel weight of the right rear wheel of the front carriage
- P4 is the front wheel weight
- P5 is the wheel weight of the left front wheel of the rear cart
- P6 is the wheel weight of the front left wheel of the rear cart
- P7 is the rear wheel weight of the rear cart.
- P8 indicates the wheel weight of the rear left wheel of the rear carriage.
- the primary spring abnormality is an abnormality of a primary spring (shaft spring) provided on the carriage, and can be exemplified by breakage of a coil spring provided on the carriage.
- the primary spring is provided for each wheel, and when the primary spring of any wheel becomes abnormal, the weight of the carriage on the wheel provided with the primary spring is the front and back of the wheel in the carriage. Hang on the wheel located in the left and right direction. For example, when the primary spring of the front right wheel of the carriage becomes abnormal, the weight of the carriage that is applied to the front right wheel is applied to the front left wheel and the rear right wheel.
- the primary spring abnormality of the front-side carriage can be detected by evaluating the difference between the two.
- index of the front trolley represented by Formula (1) is used as vehicle information (vehicle information input into a classifier) detected by a detection device.
- the reason for using the primary spring abnormality index of the rear carriage represented by the formula (2) is the same reason.
- the secondary spring abnormality is an abnormality of the secondary spring provided in the carriage, and for example, an abnormality in the supply and exhaust of the air spring provided in the carriage can be exemplified.
- the secondary springs are provided on the left and right sides of the front and rear carriages, in other words, on the front and rear, left and right sides of the railway vehicle. If the secondary springs of any truck become abnormal and cannot support the weight of the vehicle body, The weight of the vehicle body applied to the wheel on the side where the secondary spring is provided is applied to the wheel in the vicinity of the secondary spring located in the front-rear direction and the left-right direction of the secondary spring.
- the weight of the vehicle body on the wheel on the right side of the front carriage (a pair of front and rear wheels) is the wheel located on the left side of the front carriage. (A pair of front and rear wheels) and a wheel (a pair of front and rear wheels) located on the right side of the rear carriage.
- the influence extends to the wheels in the vicinity of the secondary spring located in the front-rear direction or the left-right direction with respect to the secondary spring in which the abnormality has occurred.
- the total wheel weights of the wheels located on the right side of the front carriage and the left side of the rear carriage (P1 + P3 + P6 + P8) and the wheels of the wheels located on the left side of the front carriage and the right side of the rear carriage.
- index of a railway vehicle represented by Formula (3) is used as vehicle information (vehicle information input into a classifier) detected by a detection device.
- the primary spring abnormality index of the front carriage As described above, in the preferable configuration described above, as the vehicle information to be detected (vehicle information input to the classifier), the primary spring abnormality index of the front carriage, the primary spring abnormality index of the rear carriage, and the railway vehicle Since the secondary spring abnormality index is used, it can be expected that the determination accuracy of the presence / absence and abnormality type of the vehicle abnormality (abnormality related to the primary spring or the secondary spring) output from the classifier as the determination result is increased.
- the primary spring abnormality index for the front carriage, the primary spring abnormality index for the rear carriage, and the secondary spring abnormality index for the railway vehicle are as follows from the formulas (1) to (3). It can be calculated by detecting the weight.
- the wheel load of a wheel can be detected by installing a wheel load sensor using a strain gauge or a load cell on a track as described in Patent Document 3, for example.
- the primary spring abnormality index of the front and rear carts and the secondary spring abnormality index of the railway vehicle described above serve as an index for detecting an abnormality related to the primary spring and the secondary spring, and at the same time, the railway vehicle. It can also be considered as an index showing individual differences. That is, it may be used as an index for distinguishing one railway vehicle from other railway vehicles, in other words, an index for determining vehicle composition.
- the vehicle information to be detected includes a primary spring abnormality index of the front carriage, a primary spring abnormality index of the rear carriage, and the railway vehicle While the secondary spring abnormality index is used, the vehicle state that is the determination result is set as the vehicle composition.
- the railway vehicle includes a pair of carriages having two pairs of left and right wheels on the front and rear, and vehicle information detected by the detection device is expressed by the above formulas (1) and (2).
- the vehicle state output from the classifier is a vehicle organization of the railway vehicle.
- the vehicle information (the primary spring abnormality index of the front carriage of the railway vehicle and the primary spring abnormality of the rear carriage) detected by the detection device without receiving information relating to vehicle formation from the outside.
- the vehicle composition can be determined using only the index and the secondary spring abnormality index of the railway vehicle).
- a classifier first classifier
- a classification that outputs the vehicle organization as in the present preferred configuration
- the presence / absence and type of abnormality of the railway vehicle and the vehicle organization of the railway vehicle can be determined at the same time. As a result, whether the determined abnormality of the railway vehicle and the type of abnormality are related to what kind of vehicle formation, information relating to the vehicle formation is received from the outside and linked to the determination result such as the presence or absence of abnormality of the railway vehicle. There is no need to attach it and it can be easily specified.
- a normal railway vehicle (a railway vehicle having no abnormality) is used during machine learning.
- the train data of a normal railway vehicle includes vehicle information of the normal railway vehicle (primary spring abnormality index of the front carriage of the railway vehicle, primary spring abnormality index of the rear carriage, and secondary spring abnormality index of the railway vehicle. ) Is actually detectable, so it is easy to prepare.
- the vehicle information calculated from the vehicle information detected for the normal railway vehicle for example, by numerical simulation using general-purpose mechanism analysis software.
- the classifier uses, as the vehicle information of the teacher data, a primary spring abnormality index of the front carriage actually detected by the detection device for a normal railway vehicle, and 1 of the rear carriage.
- a secondary spring abnormality index, a secondary spring abnormality index of the railway vehicle, a primary spring abnormality index of the front carriage of the normal railway vehicle, a primary spring abnormality index of the rear carriage, and 2 of the railway vehicle A primary spring abnormality index of the front carriage of the abnormal railway vehicle, a primary spring abnormality index of the rear carriage, and a secondary spring abnormality index of the railway vehicle calculated by a numerical simulation from a secondary spring abnormality index are used.
- the normal railway vehicle uses the presence / absence and type of abnormality of the known railway vehicle and the abnormal railway vehicle, the normal railway vehicle and When the vehicle information of the abnormal railway vehicle is input, to output a vehicle state of the normal rail vehicle and said abnormal railcar, it is generated using the machine learning.
- the detection device includes a sensor provided on the track.
- the detection device may include a sensor provided in the railway vehicle. Examples of the sensor provided on the track include the above-described wheel load sensor described in Patent Document 3, and examples of the sensor provided on the railway vehicle include the sensors described in Patent Documents 1 and 2 described above.
- classifier various configurations such as a support vector machine and a neural network can be adopted as long as they can be generated using machine learning.
- a naive Bayes classification having an advantage that the mechanism is simple and the calculation speed is high.
- a vessel is preferably used.
- FIG. 1 is a block diagram illustrating a classifier included in a determination device included in the railway vehicle state monitoring device according to the first embodiment of the present invention.
- FIG. 2 is an explanatory diagram for explaining a classifier generation method and a classifier determination method shown in FIG.
- FIG. 3 shows the determination result obtained when the vehicle information of the abnormal railway vehicle in the determination target data is input to the classifier shown in FIG.
- FIG. 4 shows a determination result obtained when the vehicle information of a normal railway vehicle among the determination target data is input to the classifier shown in FIG.
- FIG. 5 is a block diagram illustrating a classifier included in the determination device included in the railway vehicle state monitoring apparatus according to the second embodiment of the present invention.
- FIG. 6 shows the determination result obtained when the determination target data is input to the classifier shown in FIG.
- FIG. 7 is a schematic diagram showing a schematic configuration of the railway vehicle state monitoring apparatus according to the first and second embodiments of the present invention.
- monitoring apparatus a railway vehicle state monitoring apparatus (hereinafter, appropriately referred to as “monitoring apparatus”) according to an embodiment of the present invention will be described with reference to the accompanying drawings as appropriate. First, the overall configuration will be described.
- FIG. 7 is a schematic diagram showing a schematic configuration of the monitoring apparatus according to the embodiment of the present invention.
- the monitoring apparatus according to the first embodiment and the monitoring apparatus according to the second embodiment are included in the determination apparatus. Only the classifier generation method and the determination method (determination contents) by the classifier are different, and both have a common overall configuration as shown in FIG.
- the monitoring device 100 according to the present embodiment includes a detection device 1 that detects vehicle information represented by wheel load of wheels 31 included in a railway vehicle 3 that travels on a track, and detected vehicle information. Is input, and a determination device 2 including a classifier 21 that outputs a vehicle state such as the presence or absence of abnormality of the railway vehicle 3 is provided.
- the detection device 1 included in the monitoring device 100 is attached to the left and right rails R constituting the track, and includes a wheel load sensor 11 for measuring the wheel load of the wheels 31 included in the railcar 3, And an arithmetic unit 12 connected to the heavy sensor 11.
- a wheel load sensor 11 for measuring the wheel load of the wheels 31 included in the railcar 3
- an arithmetic unit 12 connected to the heavy sensor 11.
- the calculation unit 12 calculates each abnormality index described later based on the wheel weight measured by the wheel weight sensor 11 and transmitted from the wheel weight sensor 11. Specifically, for example, the calculation unit 12 stores calculation formulas (formulas (1) to (3)) for each abnormality index, and substitutes the wheel load transmitted from the wheel load sensor 11 into the calculation formula.
- PC personal computer
- the calculating part 12 outputs each calculated abnormality parameter
- the determination device 2 included in the monitoring device 100 is installed with a program that functions as a classifier 21 that is generated using machine learning and outputs a vehicle state according to input vehicle information.
- PC a program that functions as a classifier 21 that is generated using machine learning and outputs a vehicle state according to input vehicle information.
- the calculation unit 12 of the detection device 1 and the determination device 2 are separated, but the present invention is not limited to this, and the calculation unit 12 and the determination device 2 are separated. It is also possible to configure using a single PC in which a program that fulfills both functions is installed.
- monitoring device 100 According to the first and second embodiments of the present invention will be described in order.
- FIG. 1 is a block diagram illustrating a classifier included in a determination device included in the monitoring device according to the first embodiment of the present invention.
- FIG. 1A is a block diagram showing how a classifier is generated using machine learning
- FIG. 1B is a block diagram showing how a vehicle state is determined using the generated classifier.
- the monitoring device 100 according to the first embodiment includes a first step in which the detection device 1 detects vehicle information represented by wheel weights of wheels included in a railway vehicle traveling on a track, and a determination device for the detected vehicle information. And the second step of outputting the vehicle state such as the presence / absence of an abnormality of the railway vehicle (railway vehicle whose vehicle state is unknown) from the classifier 21 (FIG. 1B). )reference).
- the monitoring device 100 is intended for a railway vehicle including a pair of carriages having two pairs of left and right wheels on the front and rear.
- vehicle information detected in the first step is represented by the following equations (1) and (2) as shown in FIG.
- the primary spring abnormality index of the front trolley of the vehicle, the primary spring abnormality index of the rear trolley, and the secondary spring abnormality index of the railway vehicle represented by the following expression (3) are used.
- P1 is the wheel weight of the front right wheel of the front carriage
- P2 is the wheel weight of the front left wheel of the front carriage
- P3 is the wheel weight of the right rear wheel of the front carriage
- P4 is the front wheel weight
- P5 is the wheel weight of the left front wheel of the rear cart
- P6 is the wheel weight of the front left wheel of the rear cart
- P7 is the rear wheel weight of the rear cart.
- each abnormality index represented by the above formulas (1) to (3) is calculated by the calculation unit 12 included in the detection device 1.
- the vehicle state output from the classifier 21 in the second step is the presence / absence and type of abnormality of the railway vehicle, as shown in FIG.
- the classifier 21 of the first embodiment is a naive Bayes classifier as will be described later, the probability that the railway vehicle is normal and various abnormalities in the railway vehicle (FIG. 1 (b ), For the sake of convenience, only the abnormality A and the abnormality B are described) and the probability of occurrence) is calculated.
- the determination apparatus 2 outputs the vehicle state with the highest probability among the probabilities that various types of abnormalities output from the classifier 21 have occurred, as a final determination result.
- the classifier 21 used in the second step includes vehicle information on railway vehicles whose vehicle state is known as teacher data (a combination of input vehicle information and output vehicle state), and The vehicle state is generated using machine learning so that the vehicle state is output when vehicle information is input.
- teacher data about a normal railway vehicle a railway vehicle having no abnormality
- teacher data about an abnormal railway vehicle are required.
- the train data of a normal railway vehicle includes vehicle information of the normal railway vehicle (primary spring abnormality index of the front carriage of the railway vehicle, primary spring abnormality index of the rear carriage, and secondary spring abnormality index of the railway vehicle. ) Is actually detectable, so it is easy to prepare.
- the classifier 21 used in the second step uses the primary spring abnormality index of the front carriage, the primary spring abnormality index of the rear carriage, which is actually detected for a normal railway vehicle, as the vehicle information of the teacher data, and Calculated by numerical simulation from the secondary spring abnormality index of the railway vehicle, the primary spring abnormality index of the front carriage of a normal railway vehicle, the primary spring abnormality index of the rear carriage, and the secondary spring abnormality index of the railway vehicle.
- the primary spring abnormality index of the front carriage of the abnormal vehicle, the primary spring abnormality index of the rear carriage, and the secondary spring abnormality index of the railway vehicle are used.
- normal railway vehicle and abnormal railway So as to output both the vehicle condition is preferably generated using a machine learning.
- FIG. 2 is an explanatory diagram for explaining a generation method of the classifier 21 and a determination method by the classifier 21.
- FIG. 2A by inputting teacher data to the classifier 21, a frequency distribution of vehicle information is formed for each vehicle state.
- FIG. 2 (a) the frequency distribution of the primary spring abnormality index of the front carriage formed for each vehicle state (normal, abnormality A, abnormality B) is shown.
- a similar frequency distribution is formed for other vehicle information (primary spring abnormality index of the rear carriage, secondary spring abnormality index of the railway vehicle).
- primary spring abnormality index of the rear carriage secondary spring abnormality index of the railway vehicle.
- FIG. 2 (a) for convenience, there are three cases where the vehicle state is normal, abnormality A, and abnormality B, but in practice, the number of frequencies according to the number of types of abnormality is shown. A distribution is formed.
- a normal distribution (probability density distribution) is formed based on the frequency distribution formed as described above.
- the normal distribution of the primary spring abnormality index of the front carriage formed for each vehicle state (normal, abnormality A, abnormality B) is shown superimposed.
- a similar normal distribution is formed and stored for other vehicle information (primary spring abnormality index of the rear carriage, secondary spring abnormality index of the railway vehicle) (see FIG. 2C).
- the classifier 21 is generated.
- the abnormal rail vehicle (abnormal A and abnormal B) is also actually detected. It is possible to form a frequency distribution of the vehicle information, and to form a normal distribution (probability density distribution) shown in FIG. 2B based on this frequency distribution.
- the vehicle information detected for the normal railway vehicle is used for the abnormal railway vehicle. It is conceivable to calculate by numerical simulation. Specifically, for example, the normal distribution (probability density distribution) of the vehicle information about the abnormal railway vehicle (the abnormality A, in the example shown in FIG.
- the normal distribution (probability density distribution) (see FIG. 2B) formed using the vehicle information actually detected for the normal railway vehicle is averaged without changing its standard deviation ⁇ . Only the value ⁇ is shifted by the amount of change obtained in the above (3), and the shifted normal distribution is calculated as the normal distribution of the vehicle information about the abnormal railway vehicle. This is a calculation based on the assumption that the normal distribution of vehicle information for normal railway vehicles and the normal distribution of vehicle information for abnormal rail vehicles differ from each other in mean value ⁇ , but the standard deviation ⁇ will be equivalent. Is the method.
- the vehicle information (primary spring abnormality index of the front carriage, the rear carriage) detected in the first step is added to the classifier 21 generated as described above.
- (Secondary spring abnormality index of railway vehicle, secondary spring abnormality index of railway vehicle) is input (the part indicated as “detected value” in FIG. 2C indicates each vehicle information that is input).
- the classifier 21 calculates the probability of each vehicle state according to the value of each input vehicle information. In the example shown in FIG. 2 (c), the probability that the railway vehicle is normal, the probability that abnormality A has occurred in the railway vehicle, and the railway vehicle according to the input value of the primary spring abnormality index of the front carriage.
- the probabilities that the abnormality B has occurred are calculated as n 1 , a 1 , and b 1 , respectively. Further, the probability that the railway vehicle is normal, the probability that the railway vehicle has an abnormality A, and the probability that the railway vehicle has an abnormality B according to the value of the primary spring abnormality index of the input rear carriage. the calculated as n 2, a 2, b 2, respectively. Furthermore, the probability that the railway vehicle is normal, the probability that the railway vehicle has an abnormality A, and the probability that the railway vehicle has an abnormality B are determined in accordance with the value of the secondary spring abnormality index of the railway vehicle that is input. Calculated as n 3 , a 3 , b 3 .
- the classifier 21 has a probability P N (equation (4) in FIG. 2C) that the railway vehicle is normal, and a probability that an abnormality A occurs in the railway vehicle.
- P A (Equation (5) in FIG. 2C) and probability P B (Equation (6) in FIG. 2C) that an abnormality B has occurred in the railway vehicle are calculated.
- the determination device 2 outputs the vehicle state having the highest probability among the probabilities P N , P A , and P B calculated by the classifier 21 as a final determination result.
- Vehicle information primary spring of the front bogie
- An abnormality index, a primary spring abnormality index of a rear carriage, and a secondary spring abnormality index of a railway vehicle, and a classifier 21 is generated by machine learning using the detected vehicle information.
- the vehicle state was determined by inputting the detected vehicle information.
- An abnormality in which a leveling valve connected to an air spring provided on the inner track side of the front carriage fails and the exhaust operation remains performed (abbreviated as “front inner track exhaust”)
- the leveling valve connected to the air spring provided on the outer gauge side of the front carriage fails and remains exhausted (abbreviated as “front outer gauge exhaust”).
- FIG. 3 shows a determination result obtained when the vehicle information of the abnormal rail vehicle among the determination target data is input to the classifier 21.
- the type of abnormality output as the determination result completely matched the assumed (simulated) type of abnormality.
- FIG. 4 shows a determination result obtained when vehicle information of a normal railway vehicle among the determination target data is input to the classifier 21. As shown in FIG. 4, most of the determination results are determined to be normal. Thereby, according to the monitoring apparatus 100 which concerns on 1st Embodiment, it turns out that it can determine with a comparatively sufficient precision also about a normal railway vehicle.
- FIG. 5 is a block diagram illustrating a classifier included in the determination device included in the monitoring device according to the second embodiment of the present invention.
- FIG. 1A is a block diagram showing how a classifier is generated using machine learning
- FIG. 1B is a block diagram showing how a vehicle state is determined using the generated classifier.
- the monitoring apparatus 100 according to the second embodiment also includes a first step of detecting vehicle information represented by wheel weights of wheels included in a railway vehicle traveling on a track by the detection device 1. Then, the detected vehicle information is input to the classifier 21 included in the determination device 2, and the second step of outputting the vehicle state of the railway vehicle (railway vehicle whose vehicle state is unknown) is executed from the classifier 21 (FIG.
- the second embodiment is different from the first embodiment only in that the vehicle state output from the classifier 21 in the second step is not the presence / absence of abnormality and the type of abnormality, but the train formation of the railway vehicle. More specifically, the classifier 21 calculates the probability that the railway vehicle to be determined has each vehicle formation, and the determination device 2 has the highest probability among the vehicle formation probabilities output from the classifier 21. Is output as a final determination result. Since it is not necessary to determine the presence or absence of abnormality and the type of abnormality, in the second embodiment, it is only necessary to prepare teacher data for normal railway vehicles as teacher data, and teacher data for abnormal rail vehicles is not necessarily required. Absent.
- 10 types of vehicle formation of X series (a formation, b formation, c formation, d formation, e formation, f formation, g formation, h formation, i formation, j formation) and 11 types of vehicle formation of k series (k (Knitting, l-knitting, m-knitting, n-knitting, o-knitting, p-knitting, q-knitting, r-knitting, s-knitting, t-knitting, u-knitting) vehicle information when passing through the curved section of the track under the following conditions (Primary spring abnormality index of the front carriage, primary spring abnormality index of the rear carriage, secondary spring abnormality index of the railway vehicle) is detected, and the classifier 21 is detected by machine learning using the detected vehicle information.
- k series k (Knitting, l-knitting, m-knitting, n-knitting, o
- the vehicle state was determined by inputting the detected vehicle information to the generated classifier 21.
- Teacher data Vehicle information detected for normal railway vehicles that passed the curve section (for 10 days)
- -Judgment target data vehicle information detected for normal railway vehicles that have passed the curve section (for 20 days after acquiring teacher data)
- FIG. 6 shows a determination result obtained when the determination target data is input to the classifier 21. As shown in FIG. 6, most of the determination results are the same as the actual vehicle formation. Thereby, according to the monitoring apparatus 100 which concerns on 2nd Embodiment, it turns out that it can determine comparatively accurately also about vehicle organization.
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Train Traffic Observation, Control, And Security (AREA)
- Electric Propulsion And Braking For Vehicles (AREA)
- Traffic Control Systems (AREA)
Abstract
Description
しかしながら、上記の方法で、走行中の鉄道車両の異常を検出しようとすると、全ての鉄道車両にセンサを取り付けなければならず、センサの保守・点検等に手間が掛かる。このため、鉄道車両の異常を容易に検出できないと共に多額の費用が必要になるという問題がある。
斯かる方法によれば、鉄道車両毎にセンサを取り付ける場合に比べて鉄道車両の異常を容易に且つ安価に検出することができる。
前側の台車の1次ばね異常指標=(P1+P4)-(P2+P3)・・・(1)
後側の台車の1次ばね異常指標=(P5+P8)-(P6+P7)・・・(2)
鉄道車両の2次ばね異常指標=(P1+P3+P6+P8)-(P2+P4+P5+P7)・・・(3)
ただし、P1は前側の台車の前方右側の車輪の輪重を、P2は前側の台車の前方左側の車輪の輪重を、P3は前側の台車の後方右側の車輪の輪重を、P4は前側の台車の後方左側の車輪の輪重を、P5は後側の台車の前方右側の車輪の輪重を、P6は後側の台車の前方左側の車輪の輪重を、P7は後側の台車の後方右側の車輪の輪重を、P8は後側の台車の後方左側の車輪の輪重を示す。
1次ばねは車輪毎に設けられており、いずれかの車輪の1次ばねが異常になると、その1次ばねが設けられた車輪に掛かっていた台車の重量は、その台車におけるその車輪の前後方向及び左右方向に位置する車輪に掛かる。例えば、台車の前方右側の車輪の1次ばねが異常になると、前方右側の車輪に掛かっていた台車の重量は、前方左側の車輪及び後方右側の車輪に掛かる。このように、1次ばね異常が発生すると、その影響は、1次ばね異常が発生した車輪に対して前後方向及び左右方向に位置する車輪に及ぶことになる。
従って、前側の台車の前方右側及び後方左側に位置する車輪のそれぞれの輪重の合計(P1+P4)と、前側の台車の前方左側及び後方右側に位置する車輪のそれぞれの輪重の合計(P2+P3)との差を評価することにより、前側の台車の1次ばね異常を検出することができると考えられる。このため、上記の好ましい構成では、検出装置で検出する車両情報(分類器に入力する車両情報)として、式(1)で表わされる前側の台車の1次ばね異常指標を用いている。式(2)で表わされる後側の台車の1次ばね異常指標を用いているのも同様の理由である。
2次ばねは前後の台車の左右それぞれに、換言すれば鉄道車両の前後左右に設けられており、いずれかの台車の2次ばねが異常になって車体の重量を支持できなくなると、その台車の2次ばねが設けられた側の車輪に掛かっていた車体の重量は、その2次ばねの前後方向及び左右方向に位置する2次ばねの近傍の車輪に掛かることになる。例えば、前側の台車の右側の2次ばねが異常になると、前側の台車の右側に位置する車輪(前後1対の車輪)に掛かっていた車体の重量は、前側の台車の左側に位置する車輪(前後1対の車輪)及び後側の台車の右側に位置する車輪(前後1対の車輪)に掛かる。このように、2次ばね異常が発生すると、その影響は、異常が発生した2次ばねに対して前後方向又は左右方向に位置する2次ばねの近傍の車輪に及ぶことになる。
従って、前側の台車の右側及び後側の台車の左側に位置する車輪のそれぞれの輪重の合計(P1+P3+P6+P8)と、前側の台車の左側及び後側の台車の右側に位置する車輪のそれぞれの輪重の合計(P2+P4+P5+P7)との差を評価することにより、鉄道車両の2次ばね異常を検出することができると考えられる。このため、上記の好ましい構成では、検出装置で検出する車両情報(分類器に入力する車両情報)として、式(3)で表わされる鉄道車両の2次ばね異常指標を用いている。
なお、前側の台車の1次ばね異常指標、後側の台車の1次ばね異常指標及び鉄道車両の2次ばね異常指標は、式(1)~(3)から明らかなように、車輪の輪重を検出することで算出可能である。車輪の輪重は、例えば、特許文献3に記載のように、歪ゲージを用いた輪重センサやロードセルを軌道に設置することで検出可能である。
この場合、上記の好ましい構成と同様に、検出する車両情報(分類器に入力する車両情報)としては前側の台車の1次ばね異常指標、後側の台車の1次ばね異常指標及び鉄道車両の2次ばね異常指標を用いる一方、判定結果である車両状態を車両編成とすることになる。
すなわち、好ましくは、前記鉄道車両は、前後に左右2対の車輪を有する台車を前後に1対具備し、前記検出装置で検出する車両情報は、前述の式(1)及び式(2)でそれぞれ表わされる前記鉄道車両の前側の台車の1次ばね異常指標及び後側の台車の1次ばね異常指標と、前述の式(3)で表わされる前記鉄道車両の2次ばね異常指標であり、前記分類器から出力する車両状態は、前記鉄道車両の車両編成である。
正常な鉄道車両の教師データは、正常な鉄道車両の車両情報(鉄道車両の前側の台車の1次ばね異常指標及び後側の台車の1次ばね異常指標と、鉄道車両の2次ばね異常指標)を実際に検出可能であるため、用意することは容易である。これに対し、異常な鉄道車両の教師データは、異常な鉄道車両を用意すること自体が困難(しかも多数用意することは極めて困難)であるため、車両情報を実際に検出することが困難であり、用意することは容易ではない。
このため、異常な鉄道車両の教師データについては、その車両情報として、正常な鉄道車両について検出した車両情報から、例えば、汎用機構解析ソフトを用いた数値シミュレーションによって算出したものを用いることが好ましい。
すなわち、前記検出装置は、前記軌道に設けられたセンサを具備することが好ましい。しかしながら、前記検出装置は、前記鉄道車両に設けられたセンサを具備することも可能である。軌道に設けられたセンサとしては、前述した特許文献3に記載の輪重センサを、鉄道車両に設けられたセンサとしては、前述した特許文献1、2に記載のセンサを例示できる。
図7は、本発明の実施形態に係る監視装置の概略構成を示す模式図である。なお、後述する第1実施形態に係る監視装置及び第2の実施形態に係る監視装置(以下、これらを総称して、適宜「本実施形態に係る監視装置」という)は、判定装置が具備する分類器の生成方法及び分類器による判定方法(判定内容)が異なるだけであって、双方共に図7に示すような共通する全体構成を有する。
図7に示すように、本実施形態に係る監視装置100は、軌道上を走行する鉄道車両3が具備する車輪31の輪重で表わされる車両情報を検出する検出装置1と、検出した車両情報が入力され、鉄道車両3の異常の有無等の車両状態を出力する分類器21を具備する判定装置2とを備えている。
なお、図7に示す監視装置100では、検出装置1の演算部12と、判定装置2とが別体とされているが、本発明はこれに限るものではなく、演算部12と判定装置2の双方の機能を果たすプログラムをインストールした単一のPCを用いて構成することも可能である。
図1は、本発明の第1実施形態に係る監視装置が備える判定装置が具備する分類器を説明するブロック図である。図1(a)は機械学習を用いて分類器を生成する様子を示すブロック図であり、図1(b)は生成された分類器を用いて車両状態を判定する様子を示すブロック図である。
第1実施形態に係る監視装置100は、軌道上を走行する鉄道車両が具備する車輪の輪重等で表わされる車両情報を検出装置1で検出する第1ステップと、検出した車両情報を判定装置2が具備する分類器21に入力し、分類器21から鉄道車両(車両状態が未知である鉄道車両)の異常の有無等の車両状態を出力する第2ステップとを実行する(図1(b)参照)。
そして、前記第1ステップで検出する車両情報(分類器21に入力される車両情報)は、図1(b)に示すように、以下の式(1)及び式(2)でそれぞれ表わされる鉄道車両の前側の台車の1次ばね異常指標及び後側の台車の1次ばね異常指標と、以下の式(3)で表わされる鉄道車両の2次ばね異常指標とされている。
前側の台車の1次ばね異常指標=(P1+P4)-(P2+P3)・・・(1)
後側の台車の1次ばね異常指標=(P5+P8)-(P6+P7)・・・(2)
鉄道車両の2次ばね異常指標=(P1+P3+P6+P8)-(P2+P4+P5+P7)・・・(3)
ただし、P1は前側の台車の前方右側の車輪の輪重を、P2は前側の台車の前方左側の車輪の輪重を、P3は前側の台車の後方右側の車輪の輪重を、P4は前側の台車の後方左側の車輪の輪重を、P5は後側の台車の前方右側の車輪の輪重を、P6は後側の台車の前方左側の車輪の輪重を、P7は後側の台車の後方右側の車輪の輪重を、P8は後側の台車の後方左側の車輪の輪重を示す。
上記の式(1)~(3)で表わされる各異常指標は、前述のように、検出装置1が具備する演算部12によって算出される。
また、前記第2ステップで分類器21から出力する車両状態は、図1(b)に示すように、鉄道車両の異常の有無及び異常の種別である。具体的には、第1実施形態の分類器21は、後述のようにナイーブベイズ分類器とされているため、鉄道車両が正常である確率と、鉄道車両に各種別の異常(図1(b)では、便宜上、異常A、異常Bのみ記載している)が生じている確率とを算出する。そして、判定装置2は、分類器21から出力された各種別の異常が生じている確率のうち、最も確率の高い車両状態を最終的な判定結果として出力する。
具体的には、機械学習の際、正常な鉄道車両(異常の無い鉄道車両)についての教師データと、異常な鉄道車両についての教師データとが必要である。
正常な鉄道車両の教師データは、正常な鉄道車両の車両情報(鉄道車両の前側の台車の1次ばね異常指標及び後側の台車の1次ばね異常指標と、鉄道車両の2次ばね異常指標)を実際に検出可能であるため、用意することは容易である。これに対し、異常な鉄道車両の教師データは、異常な鉄道車両を用意すること自体が困難(しかも多数用意することは極めて困難)であるため、車両情報を実際に検出することが困難であり、用意することは容易ではない。
このため、異常な鉄道車両の教師データについては、その車両情報として、正常な鉄道車両について検出した車両情報から数値シミュレーションによって算出したものを用いることが好ましい。
図2は、分類器21の生成方法及び分類器21による判定方法を説明する説明図である。
まず、図2(a)に示すように、分類器21に教師データを入力することによって、各車両状態毎に車両情報の頻度分布が形成される。なお、図2(a)に示す例では、各車両状態(正常、異常A、異常B)毎に形成された前側の台車の1次ばね異常指標の頻度分布を示しているが、実際には他の車両情報(後側の台車の1次ばね異常指標、鉄道車両の2次ばね異常指標)についても同様の頻度分布が形成される。また、図2(a)に示す例では、便宜上、車両状態が正常、異常A、異常Bの3つである場合を示しているが、実際には異常の種別の数に応じた数の頻度分布が形成される。
以上のようにして、分類器21は生成される。
しかしながら、前述のように、異常な鉄道車両についての車両情報を実際に検出することは困難であるため、異常な鉄道車両については、その車両情報を、正常な鉄道車両について検出した車両情報を用いた数値シミュレーションによって算出することが考えられる。具体的には、例えば、以下の(1)~(4)の手順で、異常な鉄道車両についての車両情報の正規分布(確率密度分布)(図2(b)に示す例では、異常A、異常Bの正規分布)を形成することが考えられる。
(1)正常な鉄道車両を想定して、汎用機構解析ソフト(例えば、シムパックジャパン(株)製マルチボディダイナミクス解析ツール「SIMPACK」)を用いた数値シミュレーションを実行し、車両情報(前側の台車の1次ばね異常指標、後側の台車の1次ばね異常指標及び鉄道車両の2次ばね異常指標)の数値計算結果を算出する。
(2)異常な鉄道車両を想定して、上記汎用機構解析ソフトを用いた数値シミュレーションを実行し、車両情報(前側の台車の1次ばね異常指標、後側の台車の1次ばね異常指標及び鉄道車両の2次ばね異常指標)の数値計算結果を算出する。この際、想定した異常の種別(図2(b)に示す例では、異常A、異常B)毎に数値計算結果を算出する。
(3)上記(1)及び(2)に基づき、正常な鉄道車両と異常な鉄道車両とで、車両情報がどの程度変化するか、その変化量を求める。すなわち、上記(1)で算出した異常な鉄道車両の数値計算結果から上記(2)で算出した正常な鉄道車両の数値計算結果を減算して、上記変化量を求める。
(4)前述のように、正常な鉄道車両について実際に検出した車両情報を用いて形成した正規分布(確率密度分布)(図2(b)参照)を、その標準偏差σは変えずに平均値μだけを上記(3)で求めた変化量分だけズラし、そのズラした正規分布を、異常な鉄道車両についての車両情報の正規分布として算出する。これは、正常な鉄道車両についての車両情報の正規分布と異常な鉄道車両についての車両情報の正規分布は、互いの平均値μは異なるものの、標準偏差σは同等であろうという仮定に基づく算出方法である。
そして、分類器21は、上記のようにして算出した確率に基づき、鉄道車両が正常である確率PN(図2(c)の式(4))、鉄道車両に異常Aが生じている確率PA(図2(c)の式(5))、鉄道車両に異常Bが生じている確率PB(図2(c)の式(6))を算出する。
最後に、判定装置2は、分類器21が算出した確率PN、PA、PBのうち最も確率の高い車両状態を最終的な判定結果として出力する。
X系列の5種の車両編成(a編成、b編成、c編成、d編成、e編成)について、以下の条件で、軌道の曲線区間を通過する際の車両情報(前側の台車の1次ばね異常指標、後側の台車の1次ばね異常指標、鉄道車両の2次ばね異常指標)を検出し、この検出した車両情報を用いた機械学習によって分類器21を生成し、生成した分類器21に検出した車両情報を入力することによって車両状態を判定した。
(A)対象とする曲線区間
・入口緩和曲線:長さ47m
・円曲線:長さ60.1m、半径251m、カント0.065m、スラック0.009m
・出口緩和曲線:長さ47m
・輪重センサ設置位置:円曲線の始点から15mの位置
(B)用いたデータ
・教師データ:上記曲線区間を通過した正常な鉄道車両について検出した車両情報(10日間分)、及び、該検出した正常な鉄道車両についての車両情報から数値シミュレーションによって算出した異常な鉄道車両の車両情報及びその異常の種別
・判定対象データ:上記曲線区間を通過した正常な鉄道車両について検出した車両情報(教師データ取得後の13日間分)、及び、該検出した正常な鉄道車両についての車両情報から汎用機構解析ソフトを用いた数値シミュレーションによって算出した異常な鉄道車両の車両情報
なお、教師データに含まれる異常の種別及び判定対象データによって判定される異常の種別のいずれについても、
(1)前側の台車の内軌側に設けられた空気ばねに接続されたレベリングバルブが故障して、排気動作を行ったままになる異常(「前内軌排気」と略称する)、
(2)前側の台車の外軌側に設けられた空気ばねに接続されたレベリングバルブが故障して、排気動作を行ったままになる(「前外軌排気」と略称する)、
(3)前側の台車の内軌側に設けられた空気ばねに接続されたレベリングバルブが故障して、給気動作を行ったままになる異常(「前内軌給気」と略称する)、
(4)前側の台車の外軌側に設けられた空気ばねに接続されたレベリングバルブが故障して、給気動作を行ったままになる異常(「前外軌給気」と略称する)、
(5)前側の台車の前方内軌側の軸ばねの折損(「1軸内軌折損」と略称する)、
(6)前側の台車の前方外軌側の軸ばねの折損(「1軸外軌折損」と略称する)
の6種類を異常の種別として想定した。
図3に示すように、判定結果として出力された異常の種別は、想定した(模擬した)異常の種別と完全に一致した。
これにより、第1実施形態に係る監視装置100によれば、異常な鉄道車両についてその異常の種別を精度良く判定可能であることがわかる。
図4に示すように、判定結果の大半は正常だと判定されている。
これにより、第1実施形態に係る監視装置100によれば、正常な鉄道車両についても比較的精度良く判定可能であることがわかる。
図5は、本発明の第2実施形態に係る監視装置が備える判定装置が具備する分類器を説明するブロック図である。図1(a)は機械学習を用いて分類器を生成する様子を示すブロック図であり、図1(b)は生成された分類器を用いて車両状態を判定する様子を示すブロック図である。
第2実施形態に係る監視装置100も、第1実施形態と同様に、軌道上を走行する鉄道車両が具備する車輪の輪重等で表わされる車両情報を検出装置1で検出する第1ステップと、検出した車両情報を判定装置2が具備する分類器21に入力し、分類器21から鉄道車両(車両状態が未知である鉄道車両)の車両状態を出力する第2ステップとを実行する(図5(b)参照)。
第2実施形態では、前記第2ステップで分類器21から出力する車両状態が、異常の有無及び異常の種別ではなく、鉄道車両の車両編成である点だけが第1実施形態と異なる。より具体的には、分類器21は、判定対象である鉄道車両が各車両編成である確率を算出し、判定装置2は、分類器21から出力された各車両編成の確率のうち、最も確率の高い車両編成を最終的な判定結果として出力する。異常の有無及び異常の種別を判定する必要が無いため、第2実施形態では、教師データとして正常な鉄道車両についての教師データを用意すれば良く、異常な鉄道車両についての教師データは必ずしも必要ではない。
X系列の10種の車両編成(a編成、b編成、c編成、d編成、e編成、f編成、g編成、h編成、i編成、j編成)及びY系列の11種の車両編成(k編成、l編成、m編成、n編成、o編成、p編成、q編成、r編成、s編成、t編成、u編成)について、以下の条件で、軌道の曲線区間を通過する際の車両情報(前側の台車の1次ばね異常指標、後側の台車の1次ばね異常指標、鉄道車両の2次ばね異常指標)を検出し、この検出した車両情報を用いた機械学習によって分類器21を生成し、生成した分類器21に検出した車両情報を入力することによって車両状態を判定した。
(A)対象とする曲線区間
・入口緩和曲線:長さ47m
・円曲線:長さ60.1m、半径251m、カント0.065m、スラック0.009m
・出口緩和曲線:長さ47m
・輪重センサ設置位置:円曲線の始点から15mの位置
(B)用いたデータ
・教師データ:上記曲線区間を通過した正常な鉄道車両について検出した車両情報(10日間分)
・判定対象データ:上記曲線区間を通過した正常な鉄道車両について検出した車両情報(教師データ取得後の20日間分)
図6に示すように、判定結果の大半は実際の車両編成と一致する判定結果が得られている。
これにより、第2実施形態に係る監視装置100によれば、車両編成についても比較的精度良く判定可能であることがわかる。
2・・・判定装置
3・・・鉄道車両
11・・・輪重センサ
12・・・演算部
21・・・分類器
31・・・車輪
100・・・鉄道車両状態監視装置
Claims (8)
- 軌道上を走行する鉄道車両が具備する車輪の輪重等で表わされる車両情報を検出する検出装置と、
前記検出した車両情報が入力され、前記鉄道車両の異常の有無等の車両状態を出力する分類器を具備する判定装置とを備え、
前記分類器は、教師データとして前記車両状態が既知である鉄道車両の前記車両情報及び前記車両状態を用いて、前記車両情報が入力されたときに前記既知の車両状態を出力するように、機械学習を用いて生成されていることを特徴とする鉄道車両状態監視装置。 - 前記検出装置で検出し前記分類器に入力される前記車両情報は、前記鉄道車両が具備する車輪の輪重で表わされ、
前記分類器は、前記車両状態として、前記鉄道車両の異常の有無及び異常の種別を出力することを特徴とする請求項1に記載の鉄道車両状態監視装置。 - 前記鉄道車両は、前後に左右2対の車輪を有する台車を前後に1対具備し、
前記検出装置で検出する車両情報は、以下の式(1)及び式(2)でそれぞれ表わされる前記鉄道車両の前側の台車の1次ばね異常指標及び後側の台車の1次ばね異常指標と、以下の式(3)で表わされる前記鉄道車両の2次ばね異常指標であり、
前記分類器から出力する車両状態は、前記鉄道車両の異常の有無及び異常の種別であることを特徴とする請求項1に記載の鉄道車両状態監視装置。
前側の台車の1次ばね異常指標=(P1+P4)-(P2+P3)・・・(1)
後側の台車の1次ばね異常指標=(P5+P8)-(P6+P7)・・・(2)
鉄道車両の2次ばね異常指標=(P1+P3+P6+P8)-(P2+P4+P5+P7)・・・(3)
ただし、P1は前側の台車の前方右側の車輪の輪重を、P2は前側の台車の前方左側の車輪の輪重を、P3は前側の台車の後方右側の車輪の輪重を、P4は前側の台車の後方左側の車輪の輪重を、P5は後側の台車の前方右側の車輪の輪重を、P6は後側の台車の前方左側の車輪の輪重を、P7は後側の台車の後方右側の車輪の輪重を、P8は後側の台車の後方左側の車輪の輪重を示す。 - 前記鉄道車両は、前後に左右2対の車輪を有する台車を前後に1対具備し、
前記検出装置で検出する車両情報は、以下の式(1)及び式(2)でそれぞれ表わされる前記鉄道車両の前側の台車の1次ばね異常指標及び後側の台車の1次ばね異常指標と、以下の式(3)で表わされる前記鉄道車両の2次ばね異常指標であり、
前記分類器から出力する車両状態は、前記鉄道車両の車両編成であることを特徴とする請求項1に記載の鉄道車両状態監視装置。
前側の台車の1次ばね異常指標=(P1+P4)-(P2+P3)・・・(1)
後側の台車の1次ばね異常指標=(P5+P8)-(P6+P7)・・・(2)
鉄道車両の2次ばね異常指標=(P1+P3+P6+P8)-(P2+P4+P5+P7)・・・(3)
ただし、P1は前側の台車の前方右側の車輪の輪重を、P2は前側の台車の前方左側の車輪の輪重を、P3は前側の台車の後方右側の車輪の輪重を、P4は前側の台車の後方左側の車輪の輪重を、P5は後側の台車の前方右側の車輪の輪重を、P6は後側の台車の前方左側の車輪の輪重を、P7は後側の台車の後方右側の車輪の輪重を、P8は後側の台車の後方左側の車輪の輪重を示す。 - 前記分類器は、前記教師データの前記車両情報として、正常な鉄道車両について前記検出装置で実際に検出した前記前側の台車の1次ばね異常指標、前記後側の台車の1次ばね異常指標及び前記鉄道車両の2次ばね異常指標と、前記正常な鉄道車両の前記前側の台車の1次ばね異常指標、前記後側の台車の1次ばね異常指標及び前記鉄道車両の2次ばね異常指標から数値シミュレーションによって算出した異常な鉄道車両の前記前側の台車の1次ばね異常指標、前記後側の台車の1次ばね異常指標及び前記鉄道車両の2次ばね異常指標とを用いると共に、前記教師データの前記車両状態として、前記正常な鉄道車両及び前記異常な鉄道車両についての既知の異常の有無及び異常の種別を用い、前記正常な鉄道車両及び前記異常な鉄道車両の車両情報が入力されたときに、前記正常な鉄道車両及び前記異常な鉄道車両の車両状態を出力するように、機械学習を用いて生成されていることを特徴とする請求項3に記載の鉄道車両状態監視装置。
- 前記検出装置は、前記軌道に設けられたセンサを具備することを特徴とする請求項1から5の何れかに記載の鉄道車両状態監視装置。
- 前記検出装置は、前記鉄道車両に設けられたセンサを具備することを特徴とする請求項1から5の何れかに記載の鉄道車両状態監視装置。
- 前記分類器は、ナイーブベイズ分類器であることを特徴とする請求項1から7の何れかに記載の鉄道車両状態監視装置。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
AU2015364756A AU2015364756B2 (en) | 2014-12-15 | 2015-12-15 | Railway vehicle condition monitoring apparatus |
CN201580068521.3A CN107107927B (zh) | 2014-12-15 | 2015-12-15 | 铁道车辆状态监视装置 |
US15/534,064 US10379008B2 (en) | 2014-12-15 | 2015-12-15 | Railway vehicle condition monitoring apparatus |
EP15869978.5A EP3235705A4 (en) | 2014-12-15 | 2015-12-15 | Railway vehicle state monitoring device |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2014-253515 | 2014-12-15 | ||
JP2014253515A JP6509548B2 (ja) | 2014-12-15 | 2014-12-15 | 鉄道車両状態監視装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2016098773A1 true WO2016098773A1 (ja) | 2016-06-23 |
Family
ID=56126662
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/085085 WO2016098773A1 (ja) | 2014-12-15 | 2015-12-15 | 鉄道車両状態監視装置 |
Country Status (6)
Country | Link |
---|---|
US (1) | US10379008B2 (ja) |
EP (1) | EP3235705A4 (ja) |
JP (1) | JP6509548B2 (ja) |
CN (1) | CN107107927B (ja) |
AU (1) | AU2015364756B2 (ja) |
WO (1) | WO2016098773A1 (ja) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019138771A (ja) * | 2018-02-09 | 2019-08-22 | 日本製鉄株式会社 | 鉄道車両の状態監視装置 |
JP2021506677A (ja) * | 2018-08-01 | 2021-02-22 | 中▲車▼青▲島▼四方▲機車車▼輌股▲分▼有限公司Crrc Qingdao Sifang Co., Ltd. | 電磁横方向アクティブ制振システム、及びその制御方法と装置 |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20180065628A (ko) * | 2016-12-08 | 2018-06-18 | 재단법인대구경북과학기술원 | 철도차량이상판별장치 및 철도차량 이상 판별 방법 |
JP6889059B2 (ja) * | 2017-07-19 | 2021-06-18 | 株式会社東芝 | 情報処理装置、情報処理方法及びコンピュータプログラム |
US20190178754A1 (en) * | 2017-12-08 | 2019-06-13 | Jiaxing Broadsens Technology, Ltd. | Method and system for monitoring structural status of railcar draft gear |
KR102527319B1 (ko) * | 2018-01-29 | 2023-05-03 | 한국전자통신연구원 | 철도 차량의 부품 및 운영 환경 특징 벡터 정보 기반의 기계 학습 고장 진단 시스템 및 그 방법 |
JP7402127B2 (ja) | 2020-06-30 | 2023-12-20 | 株式会社日立製作所 | 軌道の保守要否判断方法、保守要否判断装置および車両 |
KR102673205B1 (ko) * | 2021-07-07 | 2024-06-11 | 한국철도기술연구원 | 대차 주행안전성 예측 장치 및 방법 |
GR1010518B (el) * | 2022-04-13 | 2023-08-10 | Αττικο Μετρο Α.Ε., | Διαγνωστικη μοναδα για αναρτησεις, τροχους και τροχιες σιδηροδρομικων οχηματων με μετρηση σηματων ταλαντωσης απο το εν λειτουργια οχημα |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1600351A1 (en) * | 2004-04-01 | 2005-11-30 | Heuristics GmbH | Method and system for detecting defects and hazardous conditions in passing rail vehicles |
US20090175509A1 (en) * | 2008-01-03 | 2009-07-09 | Apple Inc. | Personal computing device control using face detection and recognition |
JP2012058207A (ja) * | 2010-09-13 | 2012-03-22 | Railway Technical Research Institute | 鉄道車両の状態監視システムにおける健全性判定方法 |
JP5085175B2 (ja) * | 2007-03-30 | 2012-11-28 | 公益財団法人鉄道総合技術研究所 | 鉄道車両用懸架装置の動特性推定方法 |
JP5139163B2 (ja) * | 2008-06-06 | 2013-02-06 | 株式会社総合車両製作所 | 移動体の異常検出方法 |
JP2013511704A (ja) * | 2009-11-18 | 2013-04-04 | クノル−ブレムゼ ジステーメ フューア シーネンファールツォイゲ ゲゼルシャフト ミット ベシュレンクテル ハフツング | レール車両の、少なくとも一つの輪軸を有している台車の状態を監視するための方法 |
JP2013120100A (ja) * | 2011-12-06 | 2013-06-17 | Nippon Steel & Sumitomo Metal | 車両の異常検出方法 |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5085175A (ja) | 1973-11-29 | 1975-07-09 | ||
JPS5139163A (ja) | 1974-09-30 | 1976-04-01 | Suwa Seikosha Kk | Tokeiyomojiban |
PL1963161T3 (pl) * | 2005-12-23 | 2012-07-31 | Amsted Rail Co Inc | System monitorowania pociągów kolejowych |
JP2007256153A (ja) * | 2006-03-24 | 2007-10-04 | Hitachi Ltd | 鉄道車両台車異常検知システム |
DE102008028264B3 (de) * | 2008-06-13 | 2009-12-17 | Knorr-Bremse Systeme für Schienenfahrzeuge GmbH | Verfahren zur Überwachung wenigstens eines das Betriebsverhalten von Fahrzeugen oder Fahrzeugzügen beeinflussenden Systemparameters |
JP4388594B2 (ja) | 2009-06-05 | 2009-12-24 | 東海旅客鉄道株式会社 | 鉄道車両の異常検知装置 |
JP5306947B2 (ja) | 2009-09-03 | 2013-10-02 | 公益財団法人鉄道総合技術研究所 | 鉄道車両の状態監視システム |
US20140200951A1 (en) * | 2013-01-11 | 2014-07-17 | International Business Machines Corporation | Scalable rule logicalization for asset health prediction |
-
2014
- 2014-12-15 JP JP2014253515A patent/JP6509548B2/ja active Active
-
2015
- 2015-12-15 CN CN201580068521.3A patent/CN107107927B/zh active Active
- 2015-12-15 WO PCT/JP2015/085085 patent/WO2016098773A1/ja active Application Filing
- 2015-12-15 AU AU2015364756A patent/AU2015364756B2/en active Active
- 2015-12-15 EP EP15869978.5A patent/EP3235705A4/en not_active Withdrawn
- 2015-12-15 US US15/534,064 patent/US10379008B2/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1600351A1 (en) * | 2004-04-01 | 2005-11-30 | Heuristics GmbH | Method and system for detecting defects and hazardous conditions in passing rail vehicles |
JP5085175B2 (ja) * | 2007-03-30 | 2012-11-28 | 公益財団法人鉄道総合技術研究所 | 鉄道車両用懸架装置の動特性推定方法 |
US20090175509A1 (en) * | 2008-01-03 | 2009-07-09 | Apple Inc. | Personal computing device control using face detection and recognition |
JP5139163B2 (ja) * | 2008-06-06 | 2013-02-06 | 株式会社総合車両製作所 | 移動体の異常検出方法 |
JP2013511704A (ja) * | 2009-11-18 | 2013-04-04 | クノル−ブレムゼ ジステーメ フューア シーネンファールツォイゲ ゲゼルシャフト ミット ベシュレンクテル ハフツング | レール車両の、少なくとも一つの輪軸を有している台車の状態を監視するための方法 |
JP2012058207A (ja) * | 2010-09-13 | 2012-03-22 | Railway Technical Research Institute | 鉄道車両の状態監視システムにおける健全性判定方法 |
JP2013120100A (ja) * | 2011-12-06 | 2013-06-17 | Nippon Steel & Sumitomo Metal | 車両の異常検出方法 |
Non-Patent Citations (1)
Title |
---|
See also references of EP3235705A4 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2019138771A (ja) * | 2018-02-09 | 2019-08-22 | 日本製鉄株式会社 | 鉄道車両の状態監視装置 |
JP2021506677A (ja) * | 2018-08-01 | 2021-02-22 | 中▲車▼青▲島▼四方▲機車車▼輌股▲分▼有限公司Crrc Qingdao Sifang Co., Ltd. | 電磁横方向アクティブ制振システム、及びその制御方法と装置 |
US10960901B2 (en) | 2018-08-01 | 2021-03-30 | Crrc Qingdao Sifang Co., Ltd. | Electromagnetic transverse active damping system, and control method and apparatus therefor |
Also Published As
Publication number | Publication date |
---|---|
US10379008B2 (en) | 2019-08-13 |
CN107107927B (zh) | 2020-02-18 |
AU2015364756A1 (en) | 2017-06-22 |
US20170363517A1 (en) | 2017-12-21 |
CN107107927A (zh) | 2017-08-29 |
EP3235705A1 (en) | 2017-10-25 |
JP6509548B2 (ja) | 2019-05-08 |
AU2015364756B2 (en) | 2019-01-03 |
JP2016113029A (ja) | 2016-06-23 |
EP3235705A4 (en) | 2018-01-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2016098773A1 (ja) | 鉄道車両状態監視装置 | |
JP6096424B2 (ja) | 軌道状態監視装置 | |
JP5827598B2 (ja) | 故障確率算出装置及び故障確率算出方法並びに鉄道保守システム | |
JP5887651B2 (ja) | 車両の異常検出方法 | |
Dižo et al. | Vibration analysis of a coach with the wheel-flat due to suspension parameters changes | |
JP6747598B2 (ja) | 検査システム、検査方法、およびプログラム | |
Mızrak et al. | Determining effects of wagon mass and vehicle velocity on vertical vibrations of a rail vehicle moving with a constant acceleration on a bridge using experimental and numerical methods | |
Muñoz et al. | Estimation of lateral track irregularity using a Kalman filter. Experimental validation | |
KR20130013133A (ko) | 차축에 작용하는 외력을 이용한 차륜의 탈선 예측방법 | |
JP6722573B2 (ja) | 状態監視装置 | |
JP7024874B6 (ja) | 検査システム、検査方法、およびプログラム | |
JP6979898B2 (ja) | 鉄道車両の状態監視装置 | |
JP2020012733A (ja) | 軌道の状態評価方法及び評価装置 | |
JP2019177784A (ja) | 車両試験システム | |
JPWO2020008525A1 (ja) | 検査システム、検査方法、およびプログラム | |
Shahidi et al. | Estimation of bogie performance criteria through on-board condition monitoring | |
WO2020121943A1 (ja) | 検査システム、検査方法、およびプログラム | |
JP6770414B2 (ja) | 乗り心地測定方法及び装置 | |
JP6930609B2 (ja) | 脱線係数の推定方法、装置、及びプログラム | |
RU2601467C2 (ru) | Способ определения годности грузовых железнодорожных вагонов по величине зазоров в скользунах | |
JP7332152B2 (ja) | 学習モデル生成方法、学習モデル生成装置、及び学習プログラム | |
Yang et al. | A load estimation model for axle box bearings of high-speed trains based on vehicle dynamic response | |
CN106078358A (zh) | 基于列车轮对在线监测***的镟床测量数据运用方法 | |
Firlik et al. | Condition monitoring of a light rail vehicle—From concept to implementation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15869978 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15534064 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
ENP | Entry into the national phase |
Ref document number: 2015364756 Country of ref document: AU Date of ref document: 20151215 Kind code of ref document: A |
|
REEP | Request for entry into the european phase |
Ref document number: 2015869978 Country of ref document: EP |