Deterioration determination device, learning device, and deterioration determination method
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- MITSUBISHI ELECTRIC CORP
- Filing Date
- 2024-04-01
- Publication Date
- 2026-05-12
AI Technical Summary
Existing methods for determining equipment deterioration in railway vehicles are inadequate when speed or vibration fluctuations due to deterioration are small relative to vehicle speed or vibration, leading to inaccurate abnormality detection.
A deterioration determination device that generates vibration and running data, calculates frequency distributions from feature and travel data combinations, and compares these distributions to reference data to accurately assess equipment condition.
Enhances the accuracy of detecting equipment deterioration by analyzing frequency distributions of combined vibration and travel data, improving detection beyond speed and vibration magnitude-based methods.
Abstract
Description
Deterioration determination device, learning device, and deterioration determination method
[0001] The present disclosure relates to a deterioration determination device, a learning device, and a deterioration determination method.
[0002] Equipment mounted on railway vehicles vibrates when the railway vehicle is traveling. If a component of the equipment deteriorates, for example, the vibration of the equipment may increase, potentially affecting the operation of the equipment. Therefore, it is common to determine whether the component of the equipment has deteriorated based on the vibration of the equipment. An example of this type of deterioration determination device is disclosed in Patent Document 1. The abnormality detection device disclosed in Patent Document 1 determines that an abnormality has occurred in the tachometer generator and the acceleration sensor when a state in which the traveling speed is equal to or greater than a first threshold and the vibration is less than a second threshold, or a state in which the traveling speed is less than a third threshold and the vibration is equal to or greater than a fourth threshold, continues for a predetermined period of time or longer.
[0003] JP 2013-127403 A
[0004] The abnormality detection method performed by the abnormality detection device disclosed in Patent Document 1 cannot detect abnormalities when the speed fluctuation due to equipment deterioration is small relative to the vehicle speed, or when the vibration fluctuation due to equipment deterioration is small relative to the vibration.
[0005] The present disclosure has been made in consideration of the above circumstances, and aims to provide a deterioration determination device, a learning device, and a deterioration determination method that can accurately determine whether or not a target device has deteriorated.
[0006] To achieve the above object, the deterioration determination device of the present disclosure includes a vibration data generation unit, a feature data generation unit, a running data generation unit, a combination data generation unit, a frequency distribution calculation unit, and a deterioration determination unit. The vibration data generation unit generates vibration data by sampling, at a first sampling frequency, a sensor signal output by a vibration sensor that measures the vibration of a target device mounted on a railway vehicle. The feature data generation unit generates feature data for each unit time, numerically indicating features of the vibration of the target device that change depending on whether or not the target device has deteriorated, from the vibration data over a unit time longer than the reciprocal of the first sampling frequency. The running data generation unit generates running data indicating at least one type of physical quantity for each unit time from measured or estimated values of a physical quantity that changes due to the running of the railway vehicle, other than vibration. The combination data generation unit generates combination data by associating feature data and running data corresponding to the same unit time. The frequency distribution calculation unit calculates a frequency distribution indicating the frequency of combination data corresponding to each combination of feature data ranges and driving data ranges from combination data spanning a target period of determination including multiple unit times, using multiple feature data ranges indicating ranges of different feature data values and multiple driving data ranges indicating ranges of different driving data values. The deterioration determination unit determines whether the target device has deteriorated from the frequency distribution.
[0007] The deterioration determination device according to the present disclosure calculates a frequency distribution indicating the frequency of combination data corresponding to each combination of feature data ranges and travel data ranges from combination data obtained by associating feature data and travel data corresponding to the same unit time, and determines whether or not a target device has deteriorated from the frequency distribution. Because the presence or absence of deterioration is determined based on the frequency distribution, it is possible to determine whether or not the target device has deteriorated more accurately than methods that determine whether or not an abnormality is present based only on the speed and magnitude of vibration.
[0008] FIG. 1 is a diagram showing an example of mounting of target equipment, which is the subject of deterioration determination by the deterioration determination device according to embodiment 1, on a railway vehicle. FIG. 2 is a diagram showing an example of the mounting position of a vibration sensor in embodiment 1. FIG. 3 is a block diagram of the deterioration determination device according to embodiment 1. FIG. 4 is a diagram showing a hardware configuration of the deterioration determination device according to embodiment 1. FIG. 5 is a flowchart showing an example of deterioration determination processing performed by the deterioration determination device according to embodiment 1. FIG. 6 is a diagram showing an example of feature data and running data generated by the deterioration determination device according to embodiment 1. FIG. 7 is a diagram showing an example of a target histogram generated by the deterioration determination device according to embodiment 1. FIG. 8 is a block diagram of the deterioration determination device according to embodiment 2. FIG. 9 is a flowchart showing an example of deterioration determination processing performed by the deterioration determination device according to embodiment 2. FIG. 10 is a diagram showing an example of a target histogram generated by the deterioration determination device according to embodiment 2.
[0009] Hereinafter, a deterioration determination device, a learning device, and a deterioration determination method according to embodiments of the present disclosure will be described in detail with reference to the drawings, in which the same or equivalent parts are designated by the same reference numerals.
[0010] (Embodiment 1) A deterioration determination device 1 according to embodiment 1 will be described using as an example a deterioration determination device that determines whether or not an electric motor that is mounted on a railway vehicle and generates propulsive power for the railway vehicle has deteriorated. The railway vehicle 100 shown in Fig. 1 includes one or more vehicles 71, a current collector 72 that acquires power supplied from a substation via a power supply line, and a bogie 81 that is movable on rails and supports each vehicle 71.
[0011] 1 , the X-axis direction indicates the width direction of the vehicle 71. The Y-axis direction indicates the direction of travel of the railway vehicle 100. The Z-axis is perpendicular to both the X-axis and the Y-axis. When the railway vehicle 100 is positioned horizontally, the Z-axis direction indicates the vertical direction.
[0012] At least one of the vehicles 71 is an electric vehicle. Fig. 1 shows the vehicle 71, which is an electric vehicle. The current collector 72 is, for example, a pantograph that acquires power through an overhead line, which is a power supply line, or a current collector shoe that acquires power through a third rail, which is a power supply line. Fig. 1 shows the current collector 72, which is a pantograph, installed on the roof of the vehicle 71. A power converter 73 is installed under the floor of the electric vehicle 71, which converts power supplied from the current collector 72 and supplies the converted power to other on-board equipment, such as an electric motor 91.
[0013] Each car 71 is provided with two bogies 81. Only one of the bogies 81 is shown in Fig. 1. Two electric motors 91 driven by power output from the power conversion device 73 are attached to each of the bogies 81. In other words, each car 71 is provided with four electric motors 91. The four electric motors 91 attached to the same car 71 are driven by receiving power supplied from the same power conversion device 73.
[0014] 2 , which is a view of the bogie 81 viewed from above in the vertical direction, the bogie 81 includes a joint 83 connected to the shaft of each electric motor 91, a gear device 84 that transmits the rotational force transmitted from the electric motor 91 via the joint 83 to an axle 85, the axle 85, and wheels 82 attached to both ends of the axle 85. When each electric motor 91 is driven by receiving power from the power conversion device 73, the shaft of each electric motor 91 rotates, and the rotational force of the shaft is transmitted to the axle 85 via the joint 83 and the gear device 84. Then, as the axle 85 rotates, the wheels 82 attached to both ends of the axle 85 rotate, thereby generating propulsive force for the railway vehicle 100. The same applies to the other bogie 81 not shown in FIG. 1 .
[0015] Because the electric motor 91 is attached to the bogie 81, the electric motor 91 vibrates when the railway vehicle 100 is traveling. When the electric motor 91 deteriorates, changes occur in vibration characteristics such as vibration amplitude, vibration frequency, and periodic change in vibration amplitude. Therefore, the deterioration determination device 1 determines whether the electric motor 91 has deteriorated based on a combination of feature data indicating the vibration characteristics of the electric motor 91, which is the target device, and traveling data based on physical quantities that change as the railway vehicle 100 travels.
[0016] As shown in Fig. 3 , the deterioration determination device 1 acquires a sensor signal from a vibration sensor 92 and acquires railway vehicle speed data from a speed sensor 93. To avoid complicating the diagram, one vibration sensor 92 and one speed sensor 93 are shown in Fig. 3 , but a vibration sensor 92 and a speed sensor 93 are provided for each electric motor 91. The deterioration determination device 1 determines whether or not each electric motor 91 has deteriorated, based on the sensor signal acquired from each vibration sensor 92 and the speed of the railway vehicle 100 acquired from each speed sensor 93.
[0017] 2, the vibration sensor 92 is provided at a position adjacent to and spaced apart from the shaft of each electric motor 91. The sensor signal output by the vibration sensor 92 is an analog signal whose amplitude changes depending on the magnitude of the vibration of the object to be measured.
[0018] The speed sensor 93 is attached, for example, to the end of the shaft of the electric motor 91 on the non-drive side. The electric motor 91 has a PG (Pulse Generator) that outputs a sensor signal whose value changes with the rotation of the shaft. The speed sensor 93 detects the number of rotations per unit time of the electric motor 91 from the sensor signal output by the PG, calculates the speed of the railway vehicle 100 from the detected number of rotations, and outputs speed data that indicates the speed of the railway vehicle 100.
[0019] 3 includes a vibration data generation unit 11 that generates vibration data from a sensor signal acquired from a vibration sensor 92, a feature data generation unit 12 that generates feature data for each unit time that numerically indicates the vibration features of the target equipment from the vibration data, and a running data generation unit 13 that generates running data that indicates the speed of the railway vehicle 100 for each unit time from speed data of the railway vehicle 100 acquired from a speed sensor 93. The deterioration determination device 1 also includes a combination data generation unit 14 that generates combination data by associating feature data and running data that correspond to the same unit time, a frequency distribution calculation unit 15 that uses a feature data range that indicates a range of values of the feature data and a running data range that indicates a range of values of the running data to calculate, for each combination of a feature data range and a running data range, a frequency distribution that indicates the frequency of combination data that corresponds to the combination, and a deterioration determination unit 16 that determines whether or not the electric motor 91 is deteriorated from the frequency distribution.
[0020] The hardware configuration of the deterioration determination device 1 having the above configuration is shown in Fig. 4. The deterioration determination device 1 includes a processor 61, a memory 62, and an interface 63. The processor 61, the memory 62, and the interface 63 are connected to one another via a bus 60. The processor 61 includes transistors, other electronic circuits, etc., and is considered to be a circuit or a processor circuit.
[0021] The functions of the deterioration determination device 1 are realized by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in the memory 62. The processor 61 reads and executes the programs stored in the memory 62, thereby realizing the functions of the above-mentioned components. That is, the memory 62 stores programs for executing the processing of the deterioration determination device 1.
[0022] The memory 62 includes, for example, non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read-Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable and Programmable Read-Only Memory), magnetic disk, flexible disk, optical disk, compact disk, mini disk, DVD (Digital Versatile Disc), etc.
[0023] The deterioration determination device 1 is connected to a vibration sensor 92, a speed sensor 93, etc. via an interface 63. The interface 63 has an interface module that complies with one or more standards depending on the connected device.
[0024] 4, the deterioration determination device 1 has one processor 61 and one memory 62, but the deterioration determination device 1 may have a plurality of processors 61 and a plurality of memories 62. In this case, the plurality of processors 61 and the plurality of memories 62 may work together to execute each function of the deterioration determination device 1.
[0025] When the railway vehicle 100 starts operation, the deterioration determination device 1 having the above configuration starts the deterioration determination process shown in Fig. 5. As an example, when the deterioration determination device 1 receives a powering command instructing the railway vehicle 100 to accelerate from a master controller provided in the driver's cab (not shown), the deterioration determination device 1 starts the deterioration determination process shown in Fig. 5.
[0026] The vibration data generator 11 generates vibration data by sampling the sensor signal output by the vibration sensor 92 at a first sampling frequency (step S11). As an example, the vibration data generator 11 generates the vibration data shown in graph A of Fig. 6 by sampling the sensor signal output by the vibration sensor 92 at a first sampling frequency of 1 kHz or higher. The horizontal axis of Fig. 6 represents time. The vertical axis of graph A of Fig. 6 represents the magnitude of vibration.
[0027] 5 , the feature data generating unit 12 generates, from the vibration data over a unit time longer than the reciprocal of the first sampling frequency, feature data for each unit time that numerically indicates the vibration feature of the electric motor 91, which changes depending on whether or not the target device, the electric motor 91, has deteriorated (step S12). In detail, the feature data generating unit 12 generates, from the vibration data over the unit time, feature data that includes at least one of the effective value of the acceleration of the electric motor 91 over the unit time, the amplitude of the envelope of the vibration data, the period of the envelope of the vibration data, and the value of the amplitude spectrum at a frequency corresponding to the natural frequency of the electric motor 91.
[0028] As an example, the feature data generator 12 calculates an effective value of the acceleration of the electric motor 91 from vibration data over a unit time Δt, as shown in graph B of Fig. 6. The unit time is a time longer than the reciprocal of the first sampling frequency, for example, a time that is 100 times or more the reciprocal of the first sampling frequency. Specifically, the unit time is, for example, 0.1 seconds or more and 2 seconds or less. From a plurality of pieces of vibration data over the unit time Δt from time T1 to time T2 shown in graph A of Fig. 6, an effective value of acceleration, which is an example of feature data and is indicated by a plot point in graph B, is obtained.
[0029] As shown in FIG. 5 , the traveling data generation unit 13 generates traveling data indicating the value of at least one type of physical quantity per unit time from measured or estimated values of physical quantities other than vibrations that change due to the traveling of the railway vehicle 100 (step S13). In detail, the traveling data generation unit 13 acquires measured values from a physical quantity sensor that is different from the vibration sensor 92 and measures physical quantities that change due to the traveling of the railway vehicle 100. Specifically, the traveling data generation unit 13 acquires speed data indicating the speed of the railway vehicle 100 from the speed sensor 93. In graph C of FIG. 6 , the speed data is indicated by plotted points. As shown in graph C, the sampling period Δt' of the speed data may differ from the unit time Δt. The sampling period Δt' of the measured or estimated values of the physical quantities acquired by the traveling data generation unit 13 may be equal to or shorter than the unit time Δt, or may be longer than the unit time Δt.
[0030] The traveling data generation unit 13 interpolates the acquired measured values or estimated values of the physical quantities to generate traveling data having a period of unit time Δt, regardless of the sampling period Δt' of the measured values or estimated values of the physical quantities. In detail, as shown in graph C of FIG. 6 , when the sampling period Δt' of the measured values or estimated values of the physical quantities is shorter than the unit time Δt, the traveling data generation unit 13 performs downsampling to generate traveling data. When the sampling period Δt' of the measured values or estimated values of the physical quantities is longer than the unit time Δt, the traveling data generation unit 13 performs upsampling to generate traveling data. When the sampling period Δt' of the measured values or estimated values of the physical quantities matches the unit time Δt, the traveling data generation unit 13 uses the measured values or estimated values of the physical quantities as traveling data.
[0031] Specifically, the traveling data generation unit 13 resamples the speed data acquired from the speed sensor 93 at a second sampling frequency corresponding to the reciprocal of the unit time Δt, thereby generating traveling data as shown in graph D. Resampling in the traveling data generation unit 13 makes it possible to match the sampling frequency of the traveling data with the sampling frequency of the feature data.
[0032] As described above, since the second sampling frequency is lower than the first sampling frequency, the data volume of the feature data and the driving data is smaller than that of the vibration data, which enables the deterioration determination device 1 to perform processing at a higher speed.
[0033] As shown in Fig. 5, the combination data generation unit 14 generates combination data by associating feature data and running data corresponding to the same unit time (step S14). As shown in graphs B and D in Fig. 6, feature data and running data indicated by plot points in unit time from time T1 to time T2 are associated with each other. As an example, the combination data is array data whose elements are the feature data and running data. Specifically, the combination data generation unit 14 generates combination data that is array data whose elements are the effective acceleration value and the speed of the railway vehicle 100 corresponding to the same unit time.
[0034] The frequency distribution calculation unit 15 uses a plurality of feature data ranges indicating ranges of different feature data values and a plurality of running data ranges indicating ranges of different running data values to calculate a frequency distribution indicating the frequency of combination data corresponding to each combination of feature data ranges and running data ranges from the combination data spanning a target discrimination period including a plurality of unit times Δt (step S15). The target discrimination period is, for example, an arbitrarily determined period such as a period during which the railway vehicle 100 makes a predetermined number of round trips along a predetermined route, or a predetermined time since the start of operation of the railway vehicle 100.
[0035] The feature data range is arbitrarily determined according to the range of possible values of the feature data, for example, the effective acceleration value of the electric motor 91, and the accuracy required for the deterioration determination. The running data range is arbitrarily determined according to the range of possible values of the running data, for example, the speed of the railway vehicle 100, and the accuracy required for the deterioration determination.
[0036] As an example, the frequency distribution calculation unit 15 generates an object histogram, which is a two-dimensional histogram showing the frequency distribution with the units of the feature data and the units of the running data as axes, as shown in Fig. 7. The horizontal axis of the object histogram shown in Fig. 7 indicates the speed (unit: km / h) of the railway vehicle 100, which is an example of running data. The vertical axis of the object histogram shown in Fig. 7 indicates the effective acceleration value (unit: m / s) of the electric motor 91, which is an example of feature data. ２ ) is shown.
[0037] In the target histogram shown in FIG. 7, the driving data range is divided into 2.5 km / h intervals, and the feature data range is 2.5 m / s ２ The frequency of combination data corresponding to the combination of the feature data range and the driving data range is indicated by color.
[0038] 5, the deterioration determining unit 16 determines whether or not the target device, the electric motor 91, has deteriorated from the frequency distribution acquired from the frequency distribution calculating unit 15 (step S16). In detail, the deterioration determining unit 16 determines whether or not the target device, the electric motor 91, has deteriorated by comparing the frequency distribution with a reference distribution that indicates the frequency distribution when the target device, the electric motor 91, has not deteriorated. It is assumed that the deterioration determining unit 16 stores information about the reference distribution in advance.
[0039] As an example, the deterioration determination unit 16 is assumed to previously store a reference histogram, which is a two-dimensional histogram showing a frequency distribution with axes representing units of feature data and units of running data when the electric motor 91 is normal, in other words, when there is no deterioration of the electric motor 91. The reference histogram may be determined from feature data and running data obtained by test running, simulation, etc. of the railway vehicle 100. The deterioration determination unit 16 determines whether or not there is deterioration of the electric motor 91 by comparing the target histogram obtained from the frequency distribution calculation unit 15 with the reference histogram.
[0040] Specifically, the deterioration determination unit 16 determines whether the target device, the electric motor 91, is deteriorated by performing pattern matching between image data representing the target histogram and image data representing the reference histogram to determine the degree of similarity. As an example, the deterioration determination unit 16 performs pattern matching using DP (Dynamic Planning) matching. If the degree of similarity is equal to or greater than a threshold, for example, 80%, and it is determined that the image data representing the target histogram and the image data representing the reference histogram match, it can be determined that the electric motor 91 is not deteriorated. If the degree of similarity is less than the threshold, and it is determined that the image data representing the target histogram and the image data representing the reference histogram do not match, it can be determined that the electric motor 91 is deteriorated.
[0041] As shown in FIG. 5, when the process of step S16 is completed, the deterioration determination device 1 ends the deterioration determination process.
[0042] As described above, the deterioration determination device 1 according to the first embodiment determines whether or not the target device has deteriorated based on the frequency distribution of the combination data generated by associating the feature data with the travel data. This makes it possible to determine whether or not the target device has deteriorated more accurately than a method of determining whether or not an abnormality has occurred based only on the speed and the magnitude of vibration.
[0043] Since the driving data generation unit 13 generates driving data by resampling, it is possible to generate driving data having the same sampling frequency as the feature data from the measurement values of existing sensors operating at independent timing.
[0044] By setting the second sampling frequency sufficiently lower than the first sampling frequency, the number of data items to be processed by the deterioration determination device 1 is reduced, thereby enabling the processing speed of the deterioration determination device 1 to be increased and the storage capacity to be reduced.
[0045] (Embodiment 2) The degradation determination method is not limited to the above-described example. In particular, the data used to determine degradation of the target equipment is not limited to the above-described example, and may be any data as long as it is feature data indicating vibration characteristics and running data indicating at least one type of physical quantity that changes as the railway vehicle 100 runs. Furthermore, the reference distribution used to determine degradation may be determined from the frequency distribution calculated by the frequency distribution calculation unit 15 during a period in which the target equipment is considered to be normal. The degradation determination device 2 according to embodiment 2 shown in FIG. 8 will be described below, focusing on the differences from embodiment 1.
[0046] The deterioration determination device 2 acquires a sensor signal from a vibration sensor 92, acquires railway vehicle speed data from a speed sensor 93, and acquires current data indicating the value of current flowing through the electric motor 91 from a current sensor 94. The hardware configuration of the deterioration determination device 2 is similar to that of the deterioration determination device 1 according to the first embodiment. The deterioration determination device 2 is connected to the vibration sensor 92, the speed sensor 93, and the current sensor 94 via an interface 63.
[0047] The current sensor 94 has, for example, a current transformer (CT) provided on a bus bar connecting the power conversion device 73 and the electric motor 91, and measures the value of the current flowing between the power conversion device 73 and the electric motor 91. The current sensor 94 outputs current data indicating the measured value of the current to the deterioration determination device 2.
[0048] When the railway vehicle 100 starts operation, the deterioration determination device 2 starts the deterioration determination process shown in Fig. 9. The processes of steps S11 and S12 are similar to the processes of steps S11 and S12 performed by the deterioration determination device 1 according to the first embodiment shown in Fig. 5.
[0049] In step S13, the traveling data generation unit 13 samples the speed data acquired from the speed sensor 93 at a second sampling frequency corresponding to the reciprocal of the unit time Δt. The traveling data generation unit 13 samples the current data acquired from the current sensor 94 at the second sampling frequency. The traveling data generation unit 13 outputs the traveling data including the speed data and current data sampled at the second sampling frequency to the combination data generation unit 14.
[0050] In step S14, the combination data generation unit 14 generates combination data, which is array data whose elements are the effective acceleration value corresponding to the same unit time, the speed of the railway vehicle 100, and the current flowing through the electric motor 91, and outputs the combination data to the frequency distribution calculation unit 15.
[0051] In step S15, the frequency distribution calculation unit 15 generates an object histogram, which is a three-dimensional histogram showing a frequency distribution with the units of the feature data and the units of the running data as axes, as shown in Fig. 10. The axes of the object histogram shown in Fig. 10 are the speed (unit: km / h) of the railway vehicle 100, which is an example of running data, the current (unit: A) flowing through the electric motor 91, which is another example of running data, and the effective acceleration value (unit: m / s) of the electric motor 91, which is an example of feature data. ２ ) is shown.
[0052] In the target histogram shown in FIG. 10, the speed range of the railway vehicle 100, which is an example of the travel data range, is divided into 2.5 km / h intervals, the current range flowing through the electric motor 91, which is another example of the travel data range, is divided into 5 A intervals, and the acceleration effective value range, which is an example of the feature data range, is divided into 2.5 m / s ２ The frequency of combination data corresponding to a combination of a feature data range and a running data range, specifically, a combination of an acceleration effective value range, a speed range of the railway vehicle 100, and a current range of the electric motor 91, is indicated by color.
[0053] The deterioration determination unit 16 uses the frequency distribution calculated by the frequency distribution calculation unit 15 as a reference distribution from the combined data obtained by associating the characteristic data based on the vibration data during the normal period from the start of operation of the target equipment, the electric motor 91, until the period during which the electric motor 91 can be considered normal has elapsed with the driving data based on the measurement values during the normal period.
[0054] The normal period is, for example, the period from the start of operation, such as the time when the electric motor 91 is attached to the railway vehicle 100 and first operates, or the time when the railway vehicle 100 equipped with the electric motor 91 first starts commercial operation, until a predetermined period of time, such as 10 weeks or 3 months, has elapsed. The period is determined arbitrarily within a range in which it is possible to assume that no abnormalities due to aging will occur in the target equipment.
[0055] The deterioration determiner 16 has a timer (not shown) and, as shown in Fig. 9, calculates the elapsed time from the start of operation to the present time and determines whether it is within the normal period (step S21). If the elapsed time is within the normal period (step S21; Yes), the deterioration determiner 16 uses the frequency distribution calculated in step S15 as the reference distribution. Specifically, the deterioration determiner 16 stores the target histogram calculated in step S15 in a memory (not shown) as a reference histogram indicating the reference distribution (step S22).
[0056] When the process of step S22 is completed, the processes of steps S11 to S15 are repeated. As long as the current time is within the normal period (step S21; Yes), the process of step S22 is repeated. When step S22 is repeated, a new reference histogram is generated by adding the frequency distribution indicated by the target histogram newly calculated in step S15 to the frequency distribution indicated by the existing reference histogram.
[0057] If the elapsed time exceeds the normal period (step S21; No), the deterioration determination unit 16 determines whether or not the target equipment, the electric motor 91, has deteriorated by comparing the frequency distribution obtained in step S15 with the reference distribution stored in step S22 (step S16).
[0058] The deterioration determination unit 16 determines whether the target device, the electric motor 91, is deteriorated by performing pattern matching between image data representing a target histogram and image data representing a reference histogram to determine the similarity. Specifically, the deterioration determination unit 16 determines the similarity by performing pattern matching between two-dimensional image data of the target histogram and two-dimensional image data of the reference histogram on a surface parallel to a plane including two axes in FIG. 10 . The deterioration determination unit 16 determines the similarity between the target histogram and the reference histogram from the similarities determined for multiple surfaces. As an example, the deterioration determination unit 16 determines the similarity between the target histogram and the reference histogram from a simple average, weighted average, or the like of the similarities determined for multiple surfaces.
[0059] As described above, the deterioration determination device 2 according to the second embodiment determines whether or not a target device has deteriorated by comparing a target histogram, which is a three-dimensional histogram, with a reference histogram. This makes it possible to determine whether or not a target device has deteriorated more accurately than the deterioration determination device 1, which determines whether or not a target device has deteriorated by comparing a target histogram, which is a two-dimensional histogram, with a reference histogram.
[0060] The present disclosure is not limited to the above-described embodiment. The feature data and driving data used by the deterioration determination devices 1 and 2 for deterioration determination are not limited to the above-described examples. The feature data generator 12 generates, from the vibration data over a unit time, feature data including at least one of an effective value of the acceleration of the target device over a unit time, an amplitude of an envelope of the vibration data, a period of the envelope of the vibration data, and an amplitude spectrum value at a frequency corresponding to the natural frequency of the target device.
[0061] As an example, the feature data generating unit 12 may use a band pass filter (BPF) that includes the natural frequency of the target device to filter the vibration data, and calculate the effective acceleration value per unit time from the result.
[0062] As another example, the feature data generation unit 12 may detect envelope data by applying an LPF (Low Pass Filter) to the result of filtering the vibration data using a BPF including the natural frequency of the target device. The feature data generation unit 12 generates frequency domain data by performing an FFT (Fast Fourier Transform) on the envelope data. The feature data generation unit 12 detects peak values of the frequency domain data and obtains the amplitude of the envelope of the vibration data. Furthermore, the feature data generation unit 12 obtains the period of the envelope of the vibration data from the frequency at which the frequency domain data reaches its peak value.
[0063] As another example, the feature data generating unit 12 may generate frequency domain data by performing FFT on the vibration data acquired from the vibration data generating unit 11. The feature data generating unit 12 detects peak values of the frequency domain data and obtains amplitude spectrum values at frequencies corresponding to the natural frequency of the target device.
[0064] The measured or estimated value of the physical quantity acquired by the traveling data generation unit 13 may be any value that indicates a value of the physical quantity that changes as the railway vehicle 100 travels. As one example, the traveling data generation unit 13 may generate traveling data that indicates the temperature of the electric motor 91 from a measured value of a temperature sensor attached to the outer surface of the electric motor 91. As another example, the traveling data generation unit 13 may generate traveling data that indicates the speed of the railway vehicle 100 from the speed of the railway vehicle 100 acquired from an ATC (Automatic Train Control), a train information management system, or the like. As another example, the traveling data generation unit 13 may acquire an estimated value of the rotation speed of the electric motor 91 from a control device that controls the power conversion device 73, and generate traveling data that indicates the speed of the railway vehicle 100 from the rotation speed of the electric motor 91.
[0065] The deterioration determination process performed by the deterioration determination devices 1 and 2 is not limited to the above example. A modified example of the deterioration determination process, the outline of which is similar to the deterioration determination process performed by the deterioration determination device 1 shown in FIG. 5 , will be described below. As an example, the deterioration determination devices 1 and 2 may determine whether or not the target device has deteriorated using frequency distributions corresponding to unit times of different lengths. In detail, the feature data generator 12 generates feature data for each unit time for each of the different unit times. As an example, the feature data generator 12 generates feature data for each unit time for a unit time of 1 second and a unit time of 2 seconds.
[0066] The running data generating unit 13 generates running data for each unit time having a different length, for example, a unit time of 1 second and a unit time of 2 seconds.
[0067] The combination data generation unit 14 generates combination data for each unit time of different lengths. Specifically, the combination data generation unit 14 generates combination data by associating feature data and running data corresponding to the same unit time from feature data and running data for each unit time of one second. The combination data generation unit 14 generates combination data by associating feature data and running data corresponding to the same unit time from feature data and running data for each unit time of two seconds.
[0068] The frequency distribution calculation unit 15 calculates a frequency distribution from the combination data generated for each unit time of different lengths. Specifically, the frequency distribution calculation unit 15 generates an object histogram indicating the distribution of the frequency of combination data corresponding to a unit time of 1 second from the combination data generated from the feature data and driving data for each unit time of 1 second. The frequency distribution calculation unit 15 generates an object histogram indicating the distribution of the frequency of combination data corresponding to a unit time of 2 seconds from the combination data generated from the feature data and driving data for each unit time of 2 seconds.
[0069] The deterioration determination unit 16 determines whether or not the target device has deteriorated from a frequency distribution based on the combination data generated for each unit time of different lengths. By comparing the target histogram corresponding to a short unit time with the reference histogram, it becomes possible to detect deterioration of the target device that causes sudden vibration fluctuations. By comparing the target histogram corresponding to a long unit time with the reference histogram, it becomes possible to remove sudden vibration fluctuations due to disturbances and determine whether or not the target device has deteriorated.
[0070] The deterioration determination unit 16 may determine whether or not the target equipment has deteriorated using a plurality of reference distributions. As an example, the deterioration determination unit 16 may previously store a reference histogram when the target equipment is operating and a reference histogram when the target equipment is not operating. When the railway vehicle 100 is accelerating or decelerating, in other words, when a powering command or a braking command is input from a master controller provided in the driver's cab of the railway vehicle 100, the electric motor 91, which is the target equipment, is in an operating state. When the railway vehicle 100 is coasting, the electric motor 91, which is the target equipment, is in an inoperable state.
[0071] The deterioration determination devices 1 and 2 may determine whether or not the target device has deteriorated by using multiple frequency distributions corresponding to multiple feature data obtained from the same vibration data. Specifically, the feature data generator 12 uses multiple frequency filters with different passbands to generate feature data from the vibration data to which the frequency filters have been applied, for each frequency filter. For example, the feature data generator 12 uses a first frequency filter that is a BPF whose passband includes frequencies equal to or lower than 1 kHz and a second frequency filter that is a BPF whose passband includes frequencies higher than 1 kHz to generate feature data from the vibration data to which the frequency filters have been applied, for each frequency filter.
[0072] The combination data generation unit 14 generates combination data by associating each piece of feature data corresponding to a different frequency filter with the driving data. Specifically, the combination data generation unit 14 generates combination data by associating the feature data corresponding to a first frequency filter with the driving data, and generates combination data by associating the feature data corresponding to a second frequency filter with the driving data.
[0073] The frequency distribution calculation unit 15 generates a target histogram from combination data generated by associating feature data corresponding to a first frequency filter with driving data, and generates another target histogram from combination data generated by associating feature data corresponding to a second frequency filter with driving data.
[0074] The deterioration determination unit 16 determines whether the target device has deteriorated by comparing each target histogram with the reference histogram. For example, if it is determined that the similarity between the target histogram generated from combination data generated by associating feature data corresponding to the first frequency filter with driving data and the reference histogram is less than a threshold, it can be determined that deterioration has occurred in the rotating components of the electric motor 91. The passband of the frequency filter is determined depending on the natural frequency of the target device as a whole, the natural frequencies of the components of the target device, etc.
[0075] The feature data range and the driving data range may be divided into sections at equal intervals or at unequal intervals. In areas with low occurrence frequency, the length of the feature data range or the driving data range section may be longer than the other sections.
[0076] When the total number of frequencies shown in the target histogram differs from the total number of frequencies shown in the reference histogram, the degradation determination unit 16 performs normalization to match the total number of frequencies in one of the target histogram and the reference histogram to the total number of frequencies in the other, and then compares the target histogram with the reference histogram.
[0077] The threshold value of the similarity when the deterioration determiner 16 performs pattern matching between image data representing the target histogram and image data representing the reference histogram is not limited to the above example, and can be set arbitrarily.
[0078] The deterioration determining unit 16 may determine whether or not there is a sign of deterioration based on the similarity obtained as a result of pattern matching between image data representing the target histogram and image data representing the reference histogram. As an example, the deterioration determining unit 16 may determine that the target device has not deteriorated if the similarity is 90% or more, that the target device has a sign of deterioration if the similarity is 80% or more but less than 90%, and that the target device has deteriorated if the similarity is less than 80%.
[0079] The deterioration determiner 16 may determine whether or not the electric motor 91 has deteriorated by comparing the target histograms of each electric motor 91. In particular, the deterioration determiner 16 repeatedly performs pattern matching on two pieces of image data selected from the four pieces of image data, using image data representing the respective target histograms of the four electric motors 91 supplied from the same power conversion device 73, while changing the combination of the image data. If there is image data that does not match the other image data, it can be determined that deterioration has occurred in the electric motor 91 corresponding to the target histogram represented by that image data.
[0080] The deterioration determination devices 1 and 2 may generate a model for determining whether or not a target device has deteriorated by machine learning. In addition to the configuration of the deterioration determination device 1 according to the first embodiment, the deterioration determination device 3 shown in Fig. 11 includes a learning device 17 having: a learning unit 18 that learns by associating the combination data of feature data and traveling data with the frequency; and a model generation unit 19 that generates a deterioration determination model from the associations between the combination data of feature data and traveling data and the frequency learned by the learning unit 18.
[0081] The learning unit 18 associates the feature data and the driving data over the normal period with each other for each unit time and learns the association. Specifically, the learning unit 18 learns the combinations acquired from the combination data generating unit 14.
[0082] The model generation unit 19 generates a deterioration determination model that outputs whether or not the target device has deteriorated, using the feature data and driving data as input, from the learning results of the learning unit 18. As an example, the learning unit 18 uses as input data a data set including multiple pairs of feature data and driving data associated with each other for each unit time, in other words, a data set including multiple combinations of data, and generates a deterioration determination model that is a neural network model, using as result data a state in which no deterioration has occurred, since the input data was obtained during a normal period.
[0083] The deterioration determination unit 16 determines whether or not the target device has deteriorated by applying the feature data and driving data over the determination target period to the deterioration determination model generated by the model generation unit 19.
[0084] The learning device 17 may be a function of the deterioration determination device 3 or may be provided independently of the deterioration determination device 3 .
[0085] The deterioration determination device 1 - 3 may be mounted on the railway vehicle 100 as a whole, or only a part of the device may be mounted on the railway vehicle 100 .
[0086] The target equipment that is the subject of the deterioration determination process of the deterioration determination device 1-3 is not limited to the electric motor 91, but may be any on-board equipment that vibrates due to the running of the railway vehicle 100. As an example, the deterioration determination device 1-3 may determine whether or not there is deterioration in the bogie 81, the wheels 82, the joints 83, the gear device 84, the axles 85, the power conversion device 73, etc.
[0087] The above hardware configuration and flowchart are merely examples and can be changed and modified as desired.
[0088] The hardware configuration of the deterioration determination devices 1-3 is not limited to the above-described example. As an example, the deterioration determination device 1 may be realized by a processing circuit 64 as shown in FIG. 12 . The processing circuit 64 is connected to a vibration sensor 92, a speed sensor 93, and the like via an interface circuit 65. When the processing circuit 64 is dedicated hardware, the processing circuit 64 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination thereof. Each component of the deterioration determination device 1 may be realized by a separate processing circuit 64, or each component of the deterioration determination device 1 may be realized by a common processing circuit 64. The same applies to the deterioration determination devices 2 and 3.
[0089] Some of the functions of the deterioration determination device 1 may be realized by dedicated hardware, and other functions may be realized by software or firmware. For example, the vibration data generation unit 11, the feature data generation unit 12, and the running data generation unit 13 may be realized by a processing circuit 64 shown in Fig. 12, and the combination data generation unit 14, the frequency distribution calculation unit 15, and the deterioration determination unit 16 may be realized by a processor 61 shown in Fig. 4 reading and executing programs stored in a memory 62. The same applies to the deterioration determination devices 2 and 3.
[0090] In the flowcharts shown in FIGS. 5 and 9, the vibration data generation process in step S11 and the running data generation process in step S13 may be performed in parallel.
[0091] The present disclosure allows various embodiments and modifications without departing from the broad spirit and scope of the present disclosure. Furthermore, the above-described embodiments are intended to illustrate the present disclosure and do not limit the scope of the present disclosure. That is, the scope of the present disclosure is defined by the claims, not the embodiments. Various modifications made within the scope of the claims and the meaning of equivalent disclosures are considered to be within the scope of the present disclosure.
[0092] 1, 2, 3 Deterioration determination device, 11 vibration data generation unit, 12 feature data generation unit, 13 running data generation unit, 14 combination data generation unit, 15 frequency distribution calculation unit, 16 deterioration determination unit, 17 learning device, 18 learning unit, 19 model generation unit, 60 bus, 61 processor, 62 memory, 63 interface, 64 processing circuit, 65 interface circuit, 71 vehicle, 72 current collector, 73 power conversion device, 81 bogie, 82 wheel, 83 joint, 84 gear device, 85 axle, 91 electric motor, 92 vibration sensor, 93 speed sensor, 94 current sensor, 100 railway vehicle.
Claims
1. A vibration data generation unit generates vibration data by sampling the sensor signal output by a vibration sensor that measures the vibration of target equipment mounted on a railway vehicle at a first sampling frequency, A feature data generation unit generates feature data for each unit time that numerically represents the vibration characteristics of the target device, which change depending on whether or not the target device has deteriorated, from the vibration data over a unit time longer than the reciprocal of the first sampling frequency, A running data generation unit generates running data that shows the value of at least one of the physical quantities for each unit time from measured or estimated values of physical quantities that change due to the running of the railway vehicle, which are different from the vibrations mentioned above. A combination data generation unit generates combination data by associating the feature data and the driving data corresponding to the same unit time, A frequency distribution calculation unit calculates a frequency distribution showing the frequency of the combination data corresponding to each combination of the feature data range and the driving data range from the combination data over a discrimination target period including a plurality of unit time periods, using a plurality of feature data ranges that show ranges of values of the feature data that are different from each other and a plurality of driving data ranges that show ranges of values of the driving data that are different from each other. A deterioration determination unit that determines whether or not the target device has deteriorated based on the frequency distribution, A deterioration detection device equipped with the following features.
2. The driving data generation unit generates the driving data by resampling the measured value or estimated value at a second sampling frequency corresponding to the reciprocal of the unit time. The deterioration determination device according to claim 1.
3. The feature data generation unit generates the feature data for each of the unit time periods of different lengths. The driving data generation unit generates the driving data for each of the unit time periods of different lengths. The combination data generation unit generates the combination data for each of the unit time periods of different lengths. The frequency distribution calculation unit obtains the frequency distribution from the combination data generated for each of the unit time periods of different lengths. The deterioration determination unit determines whether or not the target device has deteriorated based on the frequency distribution derived from the combination data generated for each of the unit time periods of different lengths. A deterioration determination device according to claim 1 or 2.
4. The feature data generation unit generates the feature data from the vibration data to which each frequency filter has been applied, using a plurality of frequency filters with different passbands. The combination data generation unit generates the combination data by associating the feature data with the driving data for each of the feature data corresponding to different frequency filters. The frequency distribution calculation unit obtains the frequency distribution from the combination data generated for each of the feature data corresponding to different frequency filters. The degradation determination unit determines whether or not the target device has deteriorated based on each of the frequency distributions obtained from the combination data generated for each of the feature data corresponding to different frequency filters. A deterioration determination device according to claim 1 or 2.
5. The deterioration determination unit determines whether or not the target equipment has deteriorated by comparing the frequency distribution obtained by the frequency distribution calculation unit with a reference distribution that shows the frequency distribution when the target equipment has not deteriorated. A deterioration determination device according to claim 1 or 2.
6. The frequency distribution calculation unit generates a target histogram, which is a two-dimensional histogram or a three-dimensional histogram showing the frequency distribution with the units of the feature data and the units of the driving data as axes. The deterioration discrimination unit determines whether or not the target equipment has deteriorated by comparing the target histogram with a reference histogram, which is a two-dimensional or three-dimensional histogram showing the reference distribution with the units of the feature data and the units of the driving data as axes. The deterioration detection device according to claim 5.
7. The degradation determination unit determines whether or not the target device has deteriorated by comparing the target histogram with the reference histogram when the target device is in operation or when the target device is not in operation. The deterioration determination device according to claim 6.
8. The degradation determination unit determines whether or not the target device has deteriorated by performing pattern matching between the image data representing the target histogram and the image data representing the reference histogram. The deterioration determination device according to claim 6.
9. The vibration data generation unit generates vibration data by sampling the sensor signal output by the vibration sensor, which measures the vibration of the target device, which is an electric motor mounted on the railway vehicle and generates the propulsion force of the railway vehicle, at a first sampling frequency. The feature data generation unit generates feature data from the vibration data over the unit time, indicating the effective value of the acceleration of the electric motor over the unit time. The aforementioned running data generation unit generates the running data indicating the speed of the railway vehicle for each unit of time from the measured values of the speed sensor that measures the speed of the railway vehicle, The combination data generation unit generates the combination data by associating the feature data, which shows the effective value of the acceleration of the electric motor corresponding to the same unit time, with the running data, which shows the speed of the railway vehicle. The frequency distribution calculation unit generates the target histogram, which is a two-dimensional histogram showing the frequency distribution with the units of the speed of the railway vehicle and the units of the effective value of the acceleration of the electric motor as axes. The deterioration discrimination unit determines whether or not the target equipment has deteriorated by comparing the target histogram with the reference histogram, which is a two-dimensional histogram showing the reference distribution with the units of the speed of the railway vehicle and the units of the effective value of the acceleration of the electric motor as axes. The deterioration determination device according to claim 6.
10. The vibration data generation unit generates vibration data by sampling the sensor signal output by the vibration sensor, which measures the vibration of the target device, which is an electric motor mounted on the railway vehicle and generates the propulsion force of the railway vehicle, at a first sampling frequency. The feature data generation unit generates feature data from the vibration data over the unit time, indicating the effective value of the acceleration of the electric motor over the unit time. The running data generation unit generates the running data, which indicates the speed of the railway vehicle and the current flowing through the electric motor for each unit of time, from the measured values of a speed sensor that measures the speed of the railway vehicle and a current sensor that measures the current flowing through the electric motor. The combination data generation unit generates the combination data by associating the feature data, which shows the effective value of the acceleration of the electric motor corresponding to the same unit time, with the running data, which shows the speed of the railway vehicle and the current value flowing through the electric motor. The frequency distribution calculation unit generates the target histogram, which is a three-dimensional histogram showing the frequency distribution with the units of the speed of the railway vehicle, the units of the current flowing through the electric motor, and the units of the effective value of the acceleration of the electric motor as axes. The deterioration discrimination unit determines whether or not the target equipment has deteriorated by comparing the target histogram with the reference histogram, which is a three-dimensional histogram showing the reference distribution with the units of the speed of the railway vehicle, the units of the current flowing through the electric motor, and the units of the effective value of the acceleration of the electric motor as axes. The deterioration determination device according to claim 6.
11. The deterioration discrimination unit uses the frequency distribution calculated by the frequency distribution calculation unit as the reference distribution, obtained from the combination data obtained by associating the characteristic data based on the vibration data during the normal period from the start of operation of the target equipment until the period during which the target equipment can be considered normal has elapsed, with the driving data based on the measured values or estimated values of the physical quantities during the normal period. The deterioration detection device according to claim 5.
12. The feature data generation unit generates feature data from the vibration data in the unit time, including at least one of the effective value of the acceleration of the target device over the unit time, the amplitude of the envelope of the vibration data, the period of the envelope of the vibration data, and the amplitude spectral value at a frequency corresponding to the natural frequency of the target device. A deterioration determination device according to claim 1 or 2.
13. A learning unit acquires the characteristic data and driving data over the normal period from the start of operation of the target device until the period during which the target device can be considered normal has elapsed, and learns by associating the characteristic data and driving data corresponding to the same unit time. The system further comprises a model generation unit that generates a deterioration discrimination model that outputs whether or not the target equipment has deteriorated, based on the correspondence between the feature data and the driving data learned by the learning unit, and taking the feature data and the driving data as input. The deterioration determination unit determines whether or not the target equipment has deteriorated by applying the feature data and driving data over the determination period to the deterioration determination model generated by the model generation unit. A deterioration determination device according to claim 1 or 2.
14. A learning unit acquires, over the normal period from the start of operation of the target equipment mounted on a railway vehicle until the period during which the target equipment can be considered normal has elapsed, characteristic data for each unit time that numerically represents the characteristics of the vibration of the target equipment which change depending on whether or not the target equipment has deteriorated, generated from vibration data obtained by sampling the sensor signal output by a vibration sensor that measures the vibration of the target equipment at a first sampling frequency, and running data for each unit time that represents at least one type of physical quantity obtained from measured or estimated values of physical quantities that change due to the running of the railway vehicle, which are different from the vibration, and learns by associating the characteristic data and the running data corresponding to the same unit time. A model generation unit generates a deterioration discrimination model that outputs whether or not the target equipment has deteriorated, based on the correspondence between the feature data and the driving data learned by the learning unit, using the feature data and the driving data as inputs. A learning device equipped with the following features.
15. A deterioration detection method performed by a deterioration detection device that determines whether or not a target device installed on a railway vehicle has deteriorated, Vibration data is generated by sampling the sensor signal output by the vibration sensor that measures the vibration of the aforementioned target device at a first sampling frequency. From the vibration data over a unit time longer than the reciprocal of the first sampling frequency, characteristic data for each unit time is generated, which numerically represents the vibration characteristics of the target equipment that change depending on whether or not the target equipment has deteriorated. From measured or estimated values of physical quantities that change due to the movement of the railway vehicle, which are different from the vibrations mentioned above, running data is generated that shows the value of at least one of the physical quantities for each unit time. Combined data is generated by associating the feature data and the driving data corresponding to the same unit time. Using a plurality of feature data ranges that represent ranges of values for the feature data and a plurality of driving data ranges that represent ranges of values for the driving data, a frequency distribution is obtained from the combination data over a discrimination period including a plurality of unit time, for each combination of the feature data range and the driving data range, showing the frequency of the combination data corresponding to that combination. From the frequency distribution, it is determined whether or not the target device has deteriorated. Deterioration determination method.