JP3624390B2 - Railway track abnormality detection method and abnormality detection apparatus - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a railway track abnormality detection method and detection apparatus, and more particularly to a railway track abnormality detection method and detection apparatus using a neural network.
[0002]
[Prior art]
In railcars, when abnormalities such as track misalignment, floating sleepers, rail top and bottom irregularities, and rail wavy wear occur on railroad tracks (hereinafter simply referred to as tracks), not only does the ride feel worse, but also the extent of the abnormalities. Depending on the situation, it may lead to vehicle derailment accidents, etc., and in order to detect abnormalities in the track, level measurement and vibration measurement using a vehicle equipped with a measuring device, so-called Maya vehicles, or visual inspection by track maintenance workers, etc. have been made. It was.
[0003]
However, the measurement by Maya vehicles and track maintenance workers not only has the problem that it takes a lot of time for the measurement, but also the measurement data is only recorded on the magnetic tape or the record book, so the analysis is also necessary. There was a problem that it took a lot of time.
[0004]
However, in the present situation, a railway vehicle is traveling at a high speed and a large number of personnel are transported at a time. Therefore, a method or an apparatus for quickly detecting an abnormality in a track is desired. For this reason, research in response to such requests has been conducted.
[0005]
For example, Japanese Patent Laid-Open No. 7-23502 discloses at least one motion sensor provided in each of at least two vehicles and a plurality of first motion data output in time series from these motion sensors. A conversion means for converting the second fluctuation data of the distance series, a correction means for aligning the second fluctuation data with data from a predetermined starting point, and adding each of the corrected second fluctuation data to calculate the second fluctuation data. There has been proposed a vehicle shake cause analysis data collecting apparatus characterized by comprising a calculation means for converting to 3 shake data and an output means for outputting at least the result of the third shake data.
[0006]
According to this vehicle shake cause analysis data collection device, the waveform resulting from the shake is analyzed, so that the obtained waveform is caused by a trajectory abnormality or other vehicle passing or noise etc. It seems that it can be determined whether it is caused by the cause. Then, by processing the discriminated waveform using an appropriate threshold value, it can be discriminated whether or not the oscillation is due to a trajectory error.
[0007]
However, in the apparatus according to the proposal of Japanese Patent Laid-Open No. 7-23502, even if it is possible to distinguish between normal and abnormal, in order to determine the type of abnormality, it is appropriate for each combination of abnormal type and vibration data. It is necessary to select a threshold value. For this reason, it is not impossible to discriminate various kinds of abnormalities in the trajectory by the apparatus, but there is a problem that a very complicated operation is necessary for the realization.
[0008]
For example, when a vibration waveform as shown in FIG. 4 is obtained as a result of the measurement, floating sleepers are generated in the portions with single underline in the figure, and rail wavy wear is in the portions with double underline. However, with a simple threshold process, there is a possibility that a normal part is determined to be abnormal, and it is also impossible to determine floating sleepers and rail wavy wear.
[0009]
[Problems to be solved by the invention]
The present invention has been made in view of the problems of the prior art, and an object thereof is to provide a railway track abnormality detection method and an abnormality detection device that can detect a track abnormality in real time by a simple operation.
[0010]
[Means for Solving the Problems]
The first aspect of the railway track abnormality detection method of the present invention is a railway track abnormality detection method using a neural network,
The procedure to detect the vibration of the car body ,
A procedure of transmitting and separating the detected vibration data through the first low-pass filter and the second low-pass filter having a higher frequency for transmitting the first low-pass filter;
An absolute value of vibration data separated through the first low-pass filter is calculated to calculate a moving average, the moving average is used as first feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the second feature data;
An absolute value of vibration data transmitted through the second low-pass filter is calculated to calculate a moving average thereof, the moving average is used as third feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the fourth feature data;
A procedure for calculating certainty data regarding normality and various abnormalities by inputting the first, second, third and fourth feature data to the neural network, respectively.
Characterized in that it contains and.
[0012]
A second aspect of the railway track abnormality detection method of the present invention is a railway track abnormality detection method using a neural network,
The procedure to detect the vibration of the car body ,
A procedure of transmitting and separating the detected vibration data through the first low-pass filter and the second low-pass filter having a higher frequency for transmitting the first low-pass filter;
An absolute value of vibration data separated through the first low-pass filter is calculated to calculate a moving average, the moving average is used as first feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the second feature data;
An absolute value of vibration data transmitted through the second low-pass filter is calculated to calculate a moving average thereof, the moving average is used as third feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the fourth feature data;
A procedure for inputting first, second, third and fourth feature data to a neural network, respectively, and calculating certainty data regarding normality and various abnormalities;
And a procedure for performing a maximum value selection process on the certainty data and determining the presence / absence of a railroad track abnormality and the type of abnormality .
[0014]
In the first and second aspects of the railroad track abnormality detection method of the present invention, the vibration data is, for example, vertical vibration data of a vehicle body axle box, and the vertical vibration data and the floating sleepers and / or rails are used. Wavy wear is detected or determined.
[0015]
In the first aspect and the second aspect of the railway track abnormality detection method of the present invention, it is preferable that the vibration data is associated with a travel distance.
[0017]
On the other hand, the first aspect of the railway track abnormality detection device of the present invention includes vibration detection means provided at an appropriate position of the vehicle body for detecting the vibration of the vehicle body,
A separation unit in which a first low-pass filter through which vibration data detected by the vibration detection unit is transmitted and a second low-pass filter having a higher transmission frequency than the first low-pass filter are arranged in parallel ;
Statistical processing means;
A neural network ,
The statistical processing means is
1) Calculate the moving average by calculating the absolute value of the vibration data that has been transmitted through the first low-pass filter, calculate the moving average as the first feature data, and calculate the moving standard deviation , The moving standard deviation as the second feature data,
2) Calculate the moving average by calculating the absolute value of the vibration data separated through the second low-pass filter, use the moving average as the third feature data, and calculate the moving standard deviation , The moving standard deviation as the fourth feature data,
The neural network is characterized in that the first, second, third, and fourth feature data is input, and reliability data relating to normality and various abnormalities is output .
[0018]
According to a second aspect of the railroad track abnormality detection device of the present invention, the vibration detection means provided at an appropriate position of the vehicle body for detecting the vibration of the vehicle body,
A separation unit in which a first low-pass filter through which vibration data detected by the vibration detection unit is transmitted and a second low-pass filter having a higher transmission frequency than the first low-pass filter are arranged in parallel ;
Statistical processing means;
A neural network;
A determination means ,
The statistical processing means is
1) Calculate the moving average by calculating the absolute value of the vibration data that has been transmitted through the first low-pass filter, calculate the moving average as the first feature data, and calculate the moving standard deviation , The moving standard deviation as the second feature data,
2) Calculate the moving average by calculating the absolute value of the vibration data separated through the second low-pass filter, use the moving average as the third feature data, and calculate the moving standard deviation , The moving standard deviation as the fourth feature data,
The neural network receives first, second, third and fourth feature data as input and outputs certainty data regarding normality and various abnormalities,
The determination means performs a maximum value selection process on the certainty factor data to determine the presence / absence of a trajectory and the type of abnormality .
[0019]
In the first aspect and the second aspect of the railway track abnormality detection device of the present invention, it is preferable that the travel distance is measured by the travel distance measuring means, and the vibration data is associated with the travel distance.
[0020]
In the first aspect and the second aspect of the railway track abnormality detection device of the present invention, the detection unit of the vibration detection means is, for example, an acceleration sensor.
[0021]
Furthermore, in the first and second aspects of the railroad track abnormality detection device of the present invention, the vibration detection means is provided at an appropriate position of the vehicle body axle box, and the railroad track abnormality detection device of the present invention is provided . In the second aspect, floating sleepers and / or rail corrugated wear is determined by the determination means.
[0022]
[Action]
The vehicle body vibration data detected by the vibration detection means is separated for each vibration data including a component related to a specific trajectory abnormality. Each separated vibration data is subjected to statistical processing and converted into feature data for a certainty factor data calculation means. The certainty factor data calculating means calculates certainty factor data relating to a specific trajectory abnormality by processing the feature data. Then, the presence / absence of a trajectory abnormality and its type are determined by, for example, performing a maximum value selection process on the certainty data.
[0023]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the present invention will be described based on an embodiment with reference to the accompanying drawings, but the present invention is not limited to the embodiment.
[0024]
Here, in order to facilitate understanding of the following description, a neural network will be briefly described first.
[0025]
FIG. 1 shows the basic structure of a neural network, which is composed of an input layer, an intermediate layer, and an output layer. Each layer is composed of a plurality of processing units, and each processing unit is coupled between the next layer and the previous layer with a weighting factor determined by learning. Each processing unit has an adder for summing the inputs from the previous layer and a threshold function having a threshold determined by learning. In this neural network learning, an input pattern (teacher input) and a target output pattern (teacher output) pair are presented. Immediately after this presentation, the weights and thresholds are adjusted so that the difference between the network output and the target output is reduced. In learning, a learning set which is a set of pairs of input patterns and target patterns is used, and this is repeatedly presented to the network. When this learning is completed, a network operation test is performed.
[0026]
In this learning, there are a forward propagation step and a back propagation step executed after that. Both the forward propagation step and the back propagation step are executed each time a pattern is presented during learning. The forward propagation step begins with the presentation of the input pattern to the input layer of the network and continues while the activity level calculation is forward propagated through the intermediate layer. All the processing units in each layer (indicated by ◯ in FIG. 1) obtain the sum of the inputs and calculate the output using a threshold function. The output layer of the unit then outputs the network.
[0027]
A comparison is made between the network output pattern and the target output pattern, and the backpropagation step is initiated when there is a difference. In the back propagation step, the unit threshold value and weight change of each layer are calculated, and this is started from the output layer and is traced back to the intermediate layer in order. In this back-propagation step, the network adjusts the weights and thresholds so that the observed differences are reduced.
[0028]
Since such learning is performed, the output pattern from the network basically basically matches the target output pattern.
[0029]
Specifically, this neural network is configured by storing a program corresponding to the arithmetic processing in a computer.
[0030]
The railway track abnormality detection method and detection apparatus according to the present invention will be described below based on the understanding of the neural network.
[0031]
The railway track abnormality detection device (hereinafter simply referred to as an abnormality detection device) according to the present invention is shown schematically in FIG. 2 and in a block diagram in FIG. 3, and this abnormality detection device A detects the vibration of the vehicle body T. Vibration detection means 10, separation means 20 for separating vibration data obtained by vibration detection means 10 into vibration data containing components relating to a specific abnormality, and statistical processing means 30 for statistically processing the separated data The certainty degree data calculation means 40, the determination means 50, and the travel distance measurement means 60 are provided as main parts.
[0032]
Specifically, the vibration detecting means 10 is a vibration sensor 11. The vibration sensor 11 is attached to a sensor body 12, for example, an acceleration sensor 1 (vehicle acceleration acceleration sensor) 12A that detects vibration acceleration of the vehicle body T at an appropriate position on the vehicle body T floor, and an appropriate position on the carriage. An acceleration sensor 2 (trolley vibration acceleration sensor) 12B which is attached and detects vibration acceleration of the carriage, and an acceleration sensor 3 (axle box vibration acceleration sensor which is attached to an appropriate position of the wheel axle box and detects vibration acceleration of the axle box. ) 12C and an amplifier 13 for amplifying a signal from the sensor main body 12. The acceleration sensors 12A, 12B, and 12C are, for example, sensors that can detect six-axis acceleration.
[0033]
Specifically, the separating unit 20 is configured such that a main part is constituted by a plurality of low-pass filters arranged in parallel. For example, when separating a component related to rail wavy wear and a component related to floating sleepers that are greatly related to vertical vibrations that greatly affect the ride comfort of the passenger, the rail wavy wear has a constant wavelength of about 1.2 to 1.5 m. Whereas the unevenness of the cycle is continuous, the deformation of the rail due to the floating sleepers extends to the range of several before and after the poorly supported sleepers, so that if the first low-pass filter 21 is, for example, 30 Hz, the rail It is possible to separate a component related to wavy wear and a component related to floating sleepers. Further, the component relating to the rail wave wear can be separated from the high frequency noise during traveling at a high speed, for example, traveling at 300 Km / h, by using 300 Hz as the second low-pass filter 22.
[0034]
Specifically, the statistical processing means 30 includes an absolute value calculation processing unit 31, a moving average calculation processing unit 32, and a moving standard deviation calculation processing unit 33, and the absolute value calculation processing unit 31 calculates the absolute value of the vibration data. The moving average calculation processing unit 32 calculates the moving average, and the moving standard deviation calculation processing unit 33 calculates the moving standard deviation.
[0035]
The certainty degree data calculating means 40 is, for example, a neural network means 41, and vibration data in which a moving average or moving standard deviation is calculated by a statistical processing means 30 from vibration data relating to a specific trajectory abnormality separated by the separating means 20. That is, the feature data is input, and the feature data is subjected to neural network processing, thereby outputting the certainty data of normality and various abnormalities, for example, rail wave wear and floating sleepers.
[0036]
The learning of the neural network means 41 is performed through the following procedures: a network input / output determination procedure, a teacher data extraction procedure, and a weighting coefficient / threshold value calculation procedure.
[0037]
In the network input / output determination procedure, the number of processing units in the network input layer is determined and assigned. For the input layer, feature data having a strong correlation with the presence / absence of an abnormality to be detected is selected from the feature data and assigned as an input to each processing unit. On the other hand, for the output layer, each abnormality to be detected and the certainty of normality are assigned as the output of each processing unit.
[0038]
In the teacher data extraction procedure, data is extracted at locations where the influence of trajectory abnormalities such as rail wavy wear and floating sleepers is more prominent than the acquired vibration data. For example, when the teacher data is extracted from the data where the floating sleepers are generated, the teacher input pattern is subjected to statistical processing by the separating means 20 and the statistical processing means 30 on the vibration data at that point. The resulting feature data is used. The target output pattern uses normality and certainty of certain abnormalities (in this case, the certainty is 1 for floating sleepers and zero for others). A large number (tens to hundreds of patterns) of such teacher data is extracted.
[0039]
In the weight coefficient / threshold value calculation procedure, the weight / threshold value is adjusted by the method described above using the extracted teacher data as a learning set. At this time, the learning coefficient, which is the gain of the weight value / threshold change amount, and the number of processing units in the intermediate layer, the learning progress speed and the learning convergence value (the difference between the output from the network and the target output pattern no longer decreases). Time weighting factor / threshold), and affects the performance of the network after learning. Therefore, normally, the weighting coefficient / threshold value calculation procedure is repeatedly performed using the learning coefficient, the number of intermediate processing units, and the like as operation parameters.
[0040]
The determination unit 50 determines the presence / absence of a trajectory abnormality and the type of the trajectory abnormality by appropriately processing the certainty factor data calculated by the certainty factor data calculation unit 40, for example, a maximum value selection process.
[0041]
Specifically, the statistical processing means 30, the certainty degree data calculation means 40, and the determination means 50 are realized by storing a program corresponding to each of the means 30, 40, 50 in the computer C.
[0042]
Although the detailed description of the configuration of the travel distance measuring means 60 is omitted, it is the same as that conventionally used for measuring the travel distance of a railway vehicle. The measurement of the travel distance in the travel distance measuring means 60 is performed in synchronization with the vibration detection of the vibration detecting means 10. Further, in this embodiment, the travel distance is also measured by so-called P point information.
[0043]
Therefore, this determination result is transmitted to, for example, the central train command room by the output means 70, for example, the communication means 71.
[0044]
【Example】
Hereinafter, as an embodiment, a case will be described in which floating sleepers and rail corrugated wear that significantly affect passenger riding comfort are detected from vertical vibration data obtained by the vibration sensor 11 attached to the bogie axle box of the vehicle body T.
[0045]
In this embodiment, in the separation means 20, a 30 Hz low-pass filter is used as the first low-pass filter 21, and a 300 Hz low-pass filter is used as the second low-pass filter 22. In addition, here, the vibration data shown in FIG. 4 is used as the vibration data.
[0046]
A pre-processing The vibration data obtained from the vibration sensor 11 is input to the separating means 20 and then branched to be 30 Hz low-pass filter (first low-pass filter 21) and 300 Hz low-pass filter (second low-pass filter). 22) can be transmitted. The vibration data transmitted through the first low-pass filter 21 and the second low-pass filter 22 are subjected to statistical processing by the statistical processing means 30. In this statistical process, an absolute value is first calculated, and then the moving average and moving standard deviation are calculated using this absolute value as feature data. The characteristic data thus obtained is shown in FIG. In FIG. 5, (a) shows feature data 1, (b) shows feature data 2, (c) shows feature data 3, and (d) shows feature data 4. Here, the feature data 1 is obtained by calculating the moving average from the vibration data transmitted through the second low-pass filter 22, and the feature data 2 is obtained by calculating the moving standard deviation from the vibration data transmitted through the second low-pass filter 22. Yes, the feature data 3 is obtained by calculating the moving average from the vibration data transmitted through the first low-pass filter 21, and the feature data 4 is obtained by calculating the moving standard deviation from the vibration data transmitted through the first low-pass filter 21. .
[0047]
B abnormality determination processing Feature data 1 to feature data 4 obtained in the preprocessing are input to the input layer of the neural network means 41 shown in FIG. The neural network means 41 processes the inputted feature data 1 to feature data 4 and outputs the certainty data of normal, floating sleeper and rail wave wear from the output layer. FIG. 7 shows the certainty data output. In FIG. 7, (a) shows certainty data when normal, (b) shows certainty data when floating sleepers, and (c) shows certainty data when rail wavy wear. Each is shown.
[0048]
From FIG. 7, it can be seen that the certainty data of the floating sleepers is the highest for the portion with the single underline in FIG. 4 and the certainty data of the rail corrugated wear is the highest for the portion with the double underline. Therefore, according to the abnormality detection device A of this embodiment, floating sleepers and rail corrugated wear can be detected. That is, the type of abnormality occurring on the rail (track) can be determined by performing, for example, maximum value selection processing on the obtained certainty data.
[0049]
The present invention has been described based on the embodiments and examples. However, the present invention is not limited to the embodiments and examples, and various modifications can be made. For example, the feature data input to the certainty factor data calculation means 40 is not limited to vibration data, and information on the speed such as the number of rotations of the wheels can also be input as the feature data. By using the information on the speed as feature data, it is possible to grasp the synergistic action between the orbit abnormality and the speed.
[0050]
In the present embodiment and examples, the separating means 20, the statistical processing means 30, the certainty degree data calculating means 40, and the determining means 50 are also mounted on the vehicle body T. However, these means 20, 30, 40, 50 is installed in the central train command room, for example, and vibration data detected by the vibration detection means 10 is wirelessly input to the means 20, 30, 40, 50 installed in the central train command room. You may make it determine the presence or absence of abnormality.
[0051]
【The invention's effect】
As described above in detail, according to the present invention, it is possible to obtain an excellent effect that the presence / absence of a trajectory abnormality and the type of the abnormality can be detected or determined in real time by vibration data detected by a vibration detection means provided on the vehicle body. . This also has the secondary effect of improving the efficiency of track maintenance work and supporting safe operation of railway vehicles.
[Brief description of the drawings]
FIG. 1 is an explanatory diagram of a neural network.
FIG. 2 is a schematic view of a railway track abnormality detection apparatus according to the present invention.
FIG. 3 is a block diagram of an abnormality detection apparatus for railway tracks according to the present invention.
FIG. 4 is a graph showing an example of a vibration waveform obtained by a vibration sensor.
5 is a graph of a waveform obtained by processing the waveform shown in FIG. 4 and converting it into feature data.
FIG. 6 is an explanatory diagram of a neural network means used in the embodiment.
7 is a graph of certainty factor data calculated from the feature data shown in FIG.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 Vibration detection means 11 Vibration sensor 12 Sensor main body 13 Amplifier 20 Separation means 21 1st low pass filter 22 2nd low pass filter 30 Statistical processing means 31 Absolute value calculation processing part 32 Moving average calculation processing part 33 Moving standard deviation calculation processing part 40 Certainty Degree data calculation means 41 Neural network means 50 Determination means 60 Travel distance measurement means 70 Output means 71 Communication means A Abnormality detection device T Car body C Computer

Claims (11)

An abnormality detection method for railway tracks using a neural network,
The procedure to detect the vibration of the car body ,
A procedure of transmitting and separating the detected vibration data through the first low-pass filter and the second low-pass filter having a higher frequency for transmitting the first low-pass filter;
An absolute value of vibration data separated through the first low-pass filter is calculated to calculate a moving average, the moving average is used as first feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the second feature data;
An absolute value of vibration data transmitted through the second low-pass filter is calculated to calculate a moving average thereof, the moving average is used as third feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the fourth feature data;
A procedure for calculating certainty data regarding normality and various abnormalities by inputting the first, second, third and fourth feature data to the neural network, respectively.
And a method for detecting an abnormality in a railroad track. An abnormality detection method for railway tracks using a neural network,
The procedure to detect the vibration of the car body ,
A procedure of transmitting and separating the detected vibration data through the first low-pass filter and the second low-pass filter having a higher frequency for transmitting the first low-pass filter;
An absolute value of vibration data separated through the first low-pass filter is calculated to calculate a moving average, the moving average is used as first feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the second feature data;
An absolute value of vibration data transmitted through the second low-pass filter is calculated to calculate a moving average thereof, the moving average is used as third feature data, and a moving standard deviation is calculated. A procedure using the moving standard deviation as the fourth feature data;
A procedure for inputting first, second, third and fourth feature data to a neural network, respectively, and calculating certainty data regarding normality and various abnormalities;
A method for detecting an abnormality in a railway track, comprising : a step of performing a maximum value selection process on the certainty data and determining the presence / absence and type of abnormality of the railway track . Abnormality detecting method for railway track according to claim 1 or 2, wherein the vibration data is made to correspond with the distance traveled. The railway vibration according to claim 1 or 2 , wherein the vibration data is vertical vibration data of a vehicle body axle box, and floating sleepers and / or rail wavy wear is detected or determined using the vertical vibration data. Orbital abnormality detection method. Abnormality detecting method for railway track according to claim 1 or 2, wherein the vibration data is acceleration data. Vibration detecting means provided at an appropriate position of the vehicle body for detecting the vibration of the vehicle body;
A separation unit in which a first low-pass filter through which vibration data detected by the vibration detection unit is transmitted and a second low-pass filter having a higher transmission frequency than the first low-pass filter are arranged in parallel ;
Statistical processing means;
A neural network ,
The statistical processing means is
1) Calculate the moving average by calculating the absolute value of the vibration data that has been transmitted through the first low-pass filter, calculate the moving average as the first feature data, and calculate the moving standard deviation , The moving standard deviation as the second feature data,
2) Calculate the moving average by calculating the absolute value of the vibration data separated through the second low-pass filter, use the moving average as the third feature data, and calculate the moving standard deviation , The moving standard deviation as the fourth feature data,
The railroad track abnormality detection device , wherein the neural network receives the first, second, third and fourth feature data and outputs certainty data relating to normality and various abnormalities . Vibration detecting means provided at an appropriate position of the vehicle body for detecting the vibration of the vehicle body;
A separation unit in which a first low-pass filter through which vibration data detected by the vibration detection unit is transmitted and a second low-pass filter having a higher transmission frequency than the first low-pass filter are arranged in parallel ;
Statistical processing means;
A neural network;
A determination means ,
The statistical processing means is
1) Calculate the moving average by calculating the absolute value of the vibration data that has been transmitted through the first low-pass filter, calculate the moving average as the first feature data, and calculate the moving standard deviation , The moving standard deviation as the second feature data,
2) Calculate the moving average by calculating the absolute value of the vibration data separated through the second low-pass filter, use the moving average as the third feature data, and calculate the moving standard deviation , The moving standard deviation as the fourth feature data,
The neural network receives first, second, third and fourth feature data as input and outputs certainty data regarding normality and various abnormalities,
The railway track abnormality detection device , wherein the determination means performs a maximum value selection process on the certainty data and determines the presence / absence and type of abnormality of the track. The railway track abnormality detection device according to claim 6 or 7 , further comprising a travel distance measuring means, wherein the vibration data is associated with the travel distance. Abnormality detection apparatus railroad track according to claim 6 or 7 wherein the vibration detecting means and said Rukoto such provided at appropriate positions of the vehicle body axis box. The railroad track abnormality detection device according to claim 7, wherein floating sleepers and / or rail corrugated wear is determined by the determination means. The railroad track abnormality detection device according to claim 6 or 7 , wherein the detection unit of the vibration detection means is an acceleration sensor.

Images