Determination device, determination system, determination method, and program

Abstract

To detect an overweight vehicle.SOLUTION: A determination device includes: a feature quantity calculation unit that calculates a feature quantity dependent on magnitude of a peak wave of time-series data of observed values representing displacements of a bridge in a vehicle traveling direction at a position on the bridge where sensors are disposed, a vehicle determination unit that determines a weight of a vehicle, which has passed the bridge, using a determination model for use in determining the weight of the vehicle on the basis of the feature quantity; an overweight vehicle determination unit that, when the feature quantity is larger than a threshold, determines that the vehicle having passed the bridge is an overweight vehicle; a model learning unit that trains the determination model in learning processing on the basis of feature quantities calculated when plural test vehicles whose weights are already known pass the bridge; and an outlier learning unit that discriminates the feature quantities which are calculated in the learning processing when the respective test vehicles have passed the bridge, as a group of normal values, detects a border between the normal values and outliers, and designates the detected border as a threshold.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to a determination device, a determination system, a determination method, and a program. [Background technology]

[0002] There is a known system that estimates the weight of a vehicle passing over a bridge based on the displacement of the bridge (see, for example, Patent Document 1). The relationship between the displacement of the bridge and the weight of the vehicle differs for each bridge. For this reason, such a system must generate a conversion model in advance through learning, which represents the correspondence relationship between the displacement of the bridge and the weight of the vehicle. For example, such a system generates a conversion model by acquiring the displacement of the bridge when a test vehicle, whose weight is known in advance, passes over the bridge.

[0003] In such a system, it is desirable to generate a transformation model by having a test vehicle travel alone and pass over a bridge multiple times at a set time. Also, a method has been proposed in which a camera is installed on the bridge to identify passing vehicles and detect the time when the test vehicle passes over the bridge (for example, Patent Document 2).

[0004] However, in order to allow a test vehicle to travel alone across a bridge multiple times at a set time, traffic restrictions had to be implemented or the test had to be conducted at night when there was less traffic.Furthermore, when cameras were installed on the bridge, which is a public road, to detect the passage of the test vehicle, it was necessary to take into consideration the personal information of people captured on camera.

[0005] It is also said that bridge deterioration is caused by overloaded large vehicles, which account for about 0.3% of all traffic. In particular, overweight vehicles that exceed the bridge's design load have a significant impact on bridge deterioration. Therefore, detecting overweight vehicles passing over bridges is extremely important in understanding bridge deterioration. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] International Publication No. 2020 / 031405 [Patent Document 2] Patent No. 6890258 Summary of the Invention [Problem to be solved by the invention]

[0007] The present invention has been made in view of the above, Traffic restrictions will be put in place to allow test vehicles to run on bridges, or at night when traffic is low. Test vehicle By running the vehicle on the bridge, a conversion model is generated that accurately represents the relationship between the bridge displacement and the vehicle weight. Even without 、 An overweight vehicle passed over the bridge. At low cost The present invention provides a determination device, a determination system, a determination method, and a program capable of detecting the above. [Means for solving the problem]

[0008] In order to solve the above-mentioned problems and achieve the object, a determination device according to the present invention is provided that detects, from a sensor device installed on a bridge, a position on the bridge where the sensor device is installed. Before an acquisition unit that acquires time-series data of observed values ​​that represent the displacement of the bridge; a feature calculation unit that calculates feature values ​​according to the magnitude of peak waveforms in the time-series data of observed values; car a vehicle determination unit that determines the weight of the vehicle that has passed over the bridge using a determination model that determines the weight of the vehicle; an overheavy vehicle determination unit that determines that the vehicle that has passed over the bridge is an overheavy vehicle when the feature amount is greater than a threshold value; and a learning process that is executed in advance, The vehicle is not overweight. The system includes a model learning unit that trains the judgment model based on the feature values ​​calculated when each of a plurality of test vehicles of known weights passes over the bridge, and an outlier learning unit that, in the learning process, sets the feature values ​​calculated when each of the plurality of test vehicles passes over the bridge as a group of normal values, detects the boundary between the normal values ​​and outliers that are larger than the normal values, and sets the detected boundary as the threshold value. [Effects of the Invention]

[0009] According to the present invention, Traffic restrictions will be put in place to allow test vehicles to run on bridges, or at night when traffic is low. Test vehicle By running the vehicle on the bridge, a conversion model is generated that accurately represents the relationship between the bridge displacement and the vehicle weight. Even without 、 Overweight vehicles At low cost It can be detected. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing the configuration of a determination system. [Figure 2] FIG. 2 is a diagram showing the arrangement of sensor devices when the bridge is viewed from the side. [Figure 3] FIG. 3 is a diagram showing the arrangement of sensor devices when the bridge is viewed from above. [Figure 4] FIG. 4 is a diagram illustrating a first example of the functional configuration of the information processing device according to the first embodiment. [Figure 5] FIG. 5 is a diagram showing an example of time-series data of observed values ​​detected by the sensor device when a small vehicle, a medium-sized vehicle, and a large vehicle pass by. [Figure 6] FIG. 6 is a diagram showing a processing flow of the information processing device during the learning processing. [Figure 7] FIG. 7 is a diagram showing an example of normal value ranges and thresholds of feature quantities. [Figure 8] FIG. 8 is a diagram illustrating a second example of the functional configuration of the information processing device. [Figure 9] FIG. 9 is a graph showing time on the horizontal axis and observed values ​​on the vertical axis. [Figure 10] FIG. 10 is a diagram illustrating a third example of the functional configuration of the information processing device. [Figure 11] FIG. 11 is a diagram showing a plurality of sensor devices arranged on a bridge. [Figure 12] FIG. 12 is a diagram illustrating the functional configuration of the information processing device according to the second embodiment. [Figure 13] FIG. 13 is a diagram illustrating the functional configuration of an information processing device according to the third embodiment. [Figure 14] FIG. 14 is a diagram showing changes over time in the range of normal values ​​of feature amounts. [Figure 15]FIG. 15 is a diagram illustrating a hardware configuration of an information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, an embodiment will be described with reference to the drawings.

[0012] 1 is a diagram showing the configuration of a determination system 10 according to an embodiment. The determination system 10 detects that a vehicle has passed over a bridge and the weight of the vehicle that has passed over the bridge. Furthermore, the determination system 10 determines whether the vehicle that has passed over the bridge is an overweight vehicle.

[0013] Here, an overweight vehicle is a vehicle that is heavier than a predetermined value, i.e., exceeds the design load of the bridge, for example, exceeds 25 tons.

[0014] The determination system 10 includes a sensor device 20, a terminal device 22, and an information processing device 24.

[0015] The sensor device 20 is installed at a predetermined target portion of the bridge. The sensor device 20 detects an observation value that represents displacement in the direction of travel or in a direction perpendicular to the bridge at the target portion of the bridge where the sensor device 20 is installed. In this embodiment, the sensor device 20 uses an optical scale to measure the amount of expansion and contraction in the direction of travel at the target portion of the bridge. The amount of expansion and contraction is, for example, a change in distance of several nanometers to several hundred nanometers between two points that are several tens of centimeters apart.

[0016] Note that the sensor device 20 may detect other physical quantities instead of the expansion / contraction amount in the traveling direction, as long as it can detect observed values ​​representing displacement in the traveling direction or the vertical direction. For example, the sensor device 20 may be a strain meter that detects strain in the traveling direction in a target portion of a bridge. Also, for example, the sensor device 20 may be an accelerometer that detects the magnitude of expansion / contraction acceleration in the traveling direction in a target portion of the bridge, or the magnitude of expansion / contraction acceleration in the vertical direction in a target portion of the bridge. Furthermore, in the case of a bridge that has a thickness in the vertical direction, such as a box bridge, the sensor device 20 may detect the amount of expansion / contraction in the vertical direction in the target portion.

[0017] The sensor device 20 continuously detects observation values ​​representing displacement in the travel direction or vertical direction of a target portion of the bridge at predetermined time intervals. For example, the sensor device 20 detects observation values ​​every few milliseconds. The sensor device 20 transmits a group of the detected observation values ​​to the terminal device 22.

[0018] The terminal device 22 is an edge computer connected to the sensor device 20 via a network. For example, the terminal device 22 is connected by wire via a LAN (Local Area Network).

[0019] The terminal device 22 acquires a group of detected observation values ​​from the sensor device 20. Then, the terminal device 22 generates time-series data of the observation values, which is a data group in which each observation value included in the group of observation values ​​is associated with a detection time.

[0020] The terminal device 22 transmits time-series data of the observed values ​​of the sensor device 20 to the information processing device 24 via an external network. The external network may be wired, wireless, or a combination of wired and wireless. The external network may be, for example, a LAN, a PAN (Personal Area Network), a WAN (Wide Area Network), or a combination of a PAN, a LAN, and a WAN. The external network may also include a cellular communication line such as LTE (Long Term Evolution).

[0021] The information processing device 24 is a computer such as a server device connectable to an external network. The information processing device 24 may be a single computer or may be configured by multiple computers like a cloud system.

[0022] The information processing device 24 receives time-series data of observed values ​​representing displacement in the traveling direction or vertical direction of a target portion of the bridge from the terminal device 22 via an external network. The information processing device 24 identifies the weight of a vehicle that has passed over the bridge based on the received time-series data of observed values ​​of the sensor device 20. Furthermore, the information processing device 24 determines whether the vehicle that has passed over the bridge is an overweight vehicle.

[0023] The information processing device 24 may be integrated with the terminal device 22. In this case, the information processing device 24 acquires time-series data of observed values ​​representing displacement in the traveling direction or vertical direction of the target portion of the bridge from the terminal device 22 without going through an external network.

[0024] Fig. 2 is a diagram showing the arrangement of the sensor devices 20 when the bridge is viewed from the side, and Fig. 3 is a diagram showing the arrangement of the plurality of sensor devices 20 when the bridge is viewed from above.

[0025] The sensor device 20 is attached, for example, closer to the end of the bridge than the center in the travel direction, and is provided approximately in the center of the lane in the width direction of the bridge.

[0026] The sensor device 20 is attached, for example, to the underside of the bridge, near the abutment. This allows workers to easily attach the sensor device 20 to the bridge even after the bridge is completed. The sensor device 20 may be attached at any position in the travel direction of the bridge. For example, part of the sensor device 20 may be attached to the center of the bridge in the travel direction, although this may make installation more difficult for workers. Furthermore, part of the sensor device 20 may be attached to the side of the bridge. In this case, part of the sensor device 20 can measure the amount of expansion and contraction of the bridge in the vertical direction.

[0027] 4 is a diagram showing a first example of the functional configuration of the information processing device 24. The information processing device 24 includes an acquisition unit 62, a storage unit 64, a feature calculation unit 66, a vehicle determination unit 68, an overweight vehicle determination unit 70, a time identification unit 72, an output unit 74, a model learning unit 76, and an outlier learning unit 78.

[0028] The acquisition unit 62 acquires time-series data of observed values ​​representing displacement of the bridge in the vehicle's traveling direction at the position where the sensor device 20 is installed from the sensor device 20 installed on the bridge via the terminal device 22. In this embodiment, the observed values ​​are the amount of expansion and contraction of the target portion in the traveling direction.

[0029] Each of the multiple observation values ​​included in the time series data of observation values ​​is associated with a detection time. For example, the time series data of observation values ​​may include multiple observation values ​​and multiple time data that are associated one-to-one with the multiple observation values. Each of the multiple time data represents the time at which the corresponding observation value was detected. The time series data of observation values ​​has a predetermined sample interval, which is the time interval between samples. Therefore, the time series data of observation values ​​may include multiple observation values ​​and time data of the first observation value. In this case, the detection time of each of the multiple observation values ​​is calculated based on the time data of the first observation value, the order of the corresponding observation values, and the sample time interval.

[0030] The storage unit 64 stores the time series data of the observed values ​​acquired by the acquisition unit 62 .

[0031] The feature calculation unit 66 extracts the time series data of the observed values ​​stored in the storage unit 64 for each predetermined time unit. Then, the feature calculation unit 66 calculates a feature corresponding to the magnitude of the peak waveform in the time series data of the observed values ​​for each extracted time unit.

[0032] The time unit of the extracted time-series data of observed values ​​is at least longer than the time from when a change in the expansion / contraction amount in the traveling direction of the portion of the bridge being measured due to the passage of a vehicle begins to when the change ends. Note that the feature amount calculation unit 66 may overlap two adjacent time units in time. For example, the feature amount calculation unit 66 may calculate the feature amount for each time unit by overlapping the latter half of a first time unit with the former half of a second time unit following the first time unit.

[0033] The feature calculation unit 66 detects either an upwardly convex waveform or a downwardly convex waveform included in the time-series data as a peak waveform. In this embodiment, when detecting an upwardly convex peak waveform, the upper value is deemed to be greater than the lower value, and when detecting a downwardly convex peak waveform, the lower value is deemed to be greater than the upper value.

[0034] For example, the feature amount calculation unit 66 includes a moving average unit 80, a difference calculation unit 82, and a peak value detection unit 84.

[0035] The moving average unit 80 calculates time series data of moving average values ​​by performing a moving average using a preset time window on the time series data of observed values. The time window may be represented by the number of consecutive observed values ​​for which the moving average is calculated. The time window may also be represented by time. When the time window is represented by time, the moving average unit 80 calculates the moving average value using the number of observed values ​​obtained by dividing the time window by the sample interval.

[0036] The moving average unit 80 may calculate time series data of moving average values ​​by taking a simple moving average of the time series data of observed values. That is, the moving average unit 80 may add up all the number of observed values ​​included in the time window, for example, N, and divide the sum by N to calculate the moving average value.

[0037] Furthermore, the moving average unit 80 may calculate time series data of moving average values ​​by performing a weighted moving average on the time series data of observed values. That is, the moving average unit 80 may multiply each of the number of observed values ​​included in the time window by a predetermined weight, add up all of the N weighted observed values, and divide the sum by N to calculate the moving average value.

[0038] The moving average unit 80 may also calculate time series data of moving average values ​​by taking a moving average of the time series data of the observed values ​​using an FIR (Finite Impulse Response) filter.The moving average unit 80 may also calculate time series data of moving average values ​​by performing a convolution operation on the time series data of the observed values ​​with time series window function data of a time length corresponding to the time window.

[0039] The difference calculation unit 82 acquires time series data of the observed values ​​and time series data of the moving average values ​​calculated by the moving average unit 80. The difference calculation unit 82 calculates the difference between the observed values ​​and the moving average values ​​for each sample, thereby calculating time series data of difference values ​​that represent the differences between the observed values ​​and the moving average values. For example, the difference calculation unit 82 calculates the time series data of difference values ​​by subtracting the moving average values ​​from the observed values.

[0040] The peak value detection unit 84 detects peak waveforms in the time-series data of difference values. Furthermore, the peak value detection unit 84 detects peak values, which are difference values ​​at peak points of the detected peak waveforms. For example, the peak value detection unit 84 detects, as peak waveforms, continuous ranges greater than a predetermined value in the time-series data of difference values. Then, the peak value detection unit 84 detects, as peak points, points at which the maximum observed value is obtained within the continuous range greater than the predetermined value.

[0041] The feature calculation unit 66 outputs the calculated peak value as a feature for each time unit. If the feature calculation unit 66 cannot detect a peak waveform, it may output the feature as 0, a negative value, or a predetermined value.

[0042] The vehicle determination unit 68 acquires feature amounts for each unit time from the feature amount calculation unit 66. The vehicle determination unit 68 determines the weight of a vehicle that has passed over a bridge for each unit time using a determination model that determines vehicle weight based on the feature amounts. The determination model is trained in advance by the model learning unit 76 in the learning process.

[0043] The vehicle determination unit 68 may simply determine whether or not a vehicle has passed over a bridge. For example, the vehicle determination unit 68 may determine that a vehicle has not passed over a bridge if a peak waveform is not detected or if the feature amount is smaller than the passing determination value.

[0044] The overheavy vehicle determination unit 70 acquires the feature amount from the feature amount calculation unit 66 for each unit time. If the feature amount is greater than a threshold, the overheavy vehicle determination unit 70 determines that the vehicle passing through the bridge is an overheavy vehicle. The threshold is detected in advance by the outlier learning unit 78 in the learning process.

[0045] The time identifying unit 72 identifies the time when the vehicle passed over the bridge when the weight of the vehicle is determined by the vehicle determining unit 68. For example, the time identifying unit 72 identifies the time of the peak point in the peak waveform.

[0046] The output unit 74 outputs to an external device the vehicle weight determined by the vehicle determination unit 68 for each unit time, the presence or absence of an overheavy vehicle passing for each unit time determined by the overheavy vehicle determination unit 70, and the time at which the vehicle passed over the bridge identified by the time identification unit 72. For example, the output unit 74 displays this information on a display device. Furthermore, for example, the output unit 74 may transmit this information to a terminal device 22 held by a manager or the like.

[0047] In a learning process executed prior to the vehicle weight determination process, the model learning unit 76 trains a determination model used in the vehicle determination unit 68 based on the features calculated when each of multiple test vehicles of known weights passes over a bridge.

[0048] For example, in the learning process, the model learning unit 76 acquires feature amounts when multiple test vehicles, whose passing times and weights are known, pass over a bridge from the feature amount calculation unit 66. Then, the model learning unit 76 trains the judgment model so that when the feature amounts when each of the multiple test vehicles passes over a bridge are input to the judgment model, the actual weight of the test vehicle is output.

[0049] In the learning process, the outlier learning unit 78 classifies the feature quantities calculated when each of the multiple test vehicles passes over a bridge into a group of normal values, and detects the boundary between the normal values ​​and outliers larger than the normal values. The outlier learning unit 78 then sets the detected boundary as a threshold in the overheavy vehicle determination unit 70.

[0050] FIG. 5 is a diagram showing an example of time-series data of observed values ​​detected by the sensor device 20 when a small vehicle, a medium-sized vehicle, and a large vehicle pass by.

[0051] The heavier the vehicle, the larger the peak value of the peak waveform represented in the time-series data of the observed values ​​becomes for the amount of bridge expansion and contraction, as shown in Fig. 5. Therefore, in this embodiment, the determination model detects peak waveforms that appear in the time-series data when a vehicle passes over a bridge, and determines the weight of the vehicle based on the peak value of the detected peak waveform.

[0052] For example, vehicles are classified into small, medium, and large vehicles depending on their weight. In this case, the determination model includes, for example, a first determination value and a second determination value. In this case, the determination model determines a small vehicle if the feature amount is smaller than the first determination value, a medium vehicle if the feature amount is equal to or greater than the first determination value but smaller than the second determination value, and a large vehicle if the feature amount is equal to or greater than the second determination value.

[0053] The determination model may also be a function that represents the relationship between the feature amount and the vehicle weight. In this case, the vehicle determination unit 68 calculates the vehicle weight by substituting the acquired feature amount into the argument of the determination model.

[0054] Here, an overheavy vehicle is a vehicle that is heavier than a predetermined weight, and is heavier than a large vehicle. Therefore, the threshold value for determining whether an overheavy vehicle has passed is greater than the second determination value for determining whether a vehicle is a large vehicle.

[0055] 6 is a diagram showing the flow of processing by the information processing device 24 during the learning process. Prior to the process of determining the vehicle weight, the information processing device 24 executes the learning process according to the flow shown in FIG.

[0056] First, in S101, the model learning unit 76 trains a judgment model based on feature amounts calculated when multiple test vehicles of known weights each pass a bridge. For example, the model learning unit 76 acquires feature amounts when multiple test vehicles with known passing times and weights pass a bridge. Then, the model learning unit 76 trains the judgment model so that when the feature amounts calculated when each of the multiple test vehicles passes a bridge are input to the judgment model, the actual weight is output.

[0057] Note that each of the multiple test vehicles is not an overweight vehicle, but a vehicle of a normal weight that is expected to travel across a bridge. The multiple test vehicles may also include vehicles of different weights. The test vehicles may pass over the bridge at any time. The model learning unit 76 may also change the period for executing the learning process depending on the number of times the test vehicles travel.

[0058] Next, in S102, the outlier learning unit 78 groups the plurality of feature amounts acquired in S101 into a group of normal values. Then, the outlier learning unit 78 uses a predetermined outlier detection algorithm to detect the boundary between a normal value and an outlier larger than the normal value.

[0059] Next, in S103, the model learning unit 76 sets the determination model obtained in S101 in the vehicle determination unit 68. Furthermore, the outlier learning unit 78 sets the boundary obtained in S102 as a threshold in the overheavy vehicle determination unit 70. When the processing of S103 ends, the information processing device 24 ends the learning processing and proceeds to processing for detecting the weight of a vehicle passing over a bridge.

[0060] FIG. 7 is a diagram showing an example of normal value ranges and thresholds of feature quantities.

[0061] For example, the feature amount when a normal vehicle, not an overheavy vehicle, passes over a bridge falls within a certain range of values, as shown in Fig. 7. Furthermore, the feature amount when an overheavy vehicle passes over a bridge falls outside the range of values ​​when a normal vehicle passes over a bridge, and is also larger than the range of values ​​when a normal vehicle passes over a bridge.

[0062] Therefore, the outlier learning unit 78 classifies the multiple feature amounts obtained when multiple test vehicles pass over the bridge into a group of normal values, and detects the boundary using an outlier detection algorithm, thereby obtaining threshold values ​​for the feature amounts obtained when a normal vehicle passes over the bridge and the feature amounts obtained when an overweight vehicle passes over the bridge.

[0063] For example, the outlier learning unit 78 detects the boundary using a one-class support vector machine (SVM) algorithm. Note that the kernel function of the one-class SVM is, for example, rbf, linear, poly, sigmoid, or precomputed.

[0064] The outlier learning unit 78 may use other outlier detection algorithms such as K-approximation and LOF (Local Outlier Factor) instead of the One Class SVM algorithm.

[0065] The determination system 10 according to the present embodiment as described above can detect the passage of an overheavy vehicle during the learning process, even without having the test vehicle pass over the bridge. When a test vehicle passes over a bridge, the administrator must restrict traffic or must do so at night or other times when there is less traffic. However, the determination system 10 according to the present embodiment does not require the test vehicle to pass over the bridge, so it can detect the passage of an overheavy vehicle at low cost and without placing a burden on the bridge.

[0066] Fig. 8 is a diagram showing a second example of the functional configuration of the information processing device 24. The feature amount calculation unit 66 may have the configuration shown in Fig. 8 instead of the configuration shown in Fig. 4. That is, the feature amount calculation unit 66 may have a configuration including a moving average unit 80, a peak point detection unit 86, and an area calculation unit 88.

[0067] The moving average unit 80 executes the same processing as that of the configuration shown in FIG.

[0068] The peak point detection unit 86 detects peak points of peak waveforms in the time-series data of observed values. For example, the peak point detection unit 86 may detect, as a peak waveform, a continuous range of the time-series data of observed values ​​that is greater than a predetermined value. Then, the peak point detection unit 86 detects, as a peak point, a point at which the maximum observed value is obtained within the continuous range that is greater than the predetermined value.

[0069] In a graph showing time on the horizontal axis and observed values ​​on the vertical axis, the area calculation unit 88 calculates the area of ​​a triangle connecting the peak point, the closest point before the peak where the time series data of the observed values ​​intersects with the time series data of the moving average value, and the closest point after the peak where the time series data of the observed values ​​intersects with the time series data of the moving average value.The area calculation unit 88 then outputs the calculated area as a feature.

[0070] FIG. 9 is a graph showing time on the horizontal axis and observed values ​​on the vertical axis.

[0071] The area calculation unit 88 calculates the area of ​​a triangle on a graph as shown in Fig. 9 for the peak waveform. The triangle has three vertices: a peak point, a first point, and a second point. The first point is the closest point in time to the peak point among the points where the time series data of the observed value and the time series data of the moving average intersect. The second point is the closest point in time to the peak point among the points where the time series data of the observed value and the time series data of the moving average intersect.

[0072] The area of ​​such a triangle represents the amount of expansion and contraction of the bridge when a vehicle passes over it. When a vehicle passes over it at a relatively slow speed, the time period for the expansion and contraction to fluctuate is long, but the amplitude of the expansion and contraction is small. On the other hand, when a vehicle passes over it at a relatively high speed, the amplitude of the expansion and contraction to fluctuate is large, but the time period for the expansion and contraction to fluctuate is short.

[0073] Therefore, by calculating the area of ​​such a triangle as a feature, the feature calculation unit 66 can stably detect whether a vehicle is passing and its weight, regardless of the vehicle's passing speed. However, if multiple vehicles pass over a bridge at short intervals, there is a possibility that the triangular waveforms will overlap. Therefore, by calculating the peak value as a feature instead of the triangular area as in the configuration shown in Figure 4, even if multiple vehicles pass over a bridge at short intervals, the peak waveforms will be separated in time, making it possible to accurately detect whether a vehicle has passed and its weight for each vehicle.

[0074] Fig. 10 is a diagram showing a third example of the functional configuration of the information processing device 24. As shown in Fig. 10, the feature calculation unit 66 may have a moving average unit 80, a difference calculation unit 82, a peak point detection unit 86, and an area calculation unit 88.

[0075] The moving average unit 80 and the difference calculation unit 82 perform the same processing as in the configuration shown in FIG.

[0076] The peak point detection unit 86 detects peak points of the peak waveform in the time-series data of difference values. For example, the peak point detection unit 86 may detect a continuous range of the time-series data of difference values ​​that is greater than a predetermined value as the peak waveform. Then, the peak point detection unit 86 detects the point at which the largest difference value is obtained within the continuous range that is greater than the predetermined value as the peak point.

[0077] In a graph showing time on the horizontal axis and difference values ​​on the vertical axis, the area calculation unit 88 calculates the area of ​​a triangle connecting the peak point, the closest point before the peak where the time series data of the difference values ​​intersects with the time series data of the moving average value, and the closest point after the peak where the time series data of the difference values ​​intersects with the time series data of the moving average value.The area calculation unit 88 then outputs the calculated area as a feature.

[0078] The area of ​​the triangle in the third example also represents the amount of expansion and contraction of the bridge when a vehicle passes over it, as in the second example. Therefore, by calculating the area of ​​such a triangle as a feature, the feature calculation unit 66 can stably detect whether a vehicle is passing and the weight of the vehicle, regardless of the passing speed of the vehicle.

[0079] FIG. 11 is a diagram showing an example of time-series data of observed values ​​detected by each of the plurality of sensor devices 20. As shown in FIG.

[0080] The determination system 10 may include a plurality of sensor devices 20. Each of the plurality of sensor devices 20 is provided at approximately the center of a lane in the width direction of the bridge. Furthermore, the plurality of sensor devices 20 are provided in the same position in the traveling direction, lined up in a straight line in the width direction. If the bridge has a plurality of lanes, the plurality of sensor devices 20 are provided for each lane. For example, if the bridge has four lanes, four sensor devices 20 are provided for any of the four lanes. Furthermore, the plurality of sensor devices 20 detect observation values ​​at synchronized timing.

[0081] When the determination system 10 includes multiple sensor devices 20, the feature calculation unit 66 of the information processing device 24 calculates feature values ​​for each of the multiple sensor devices 20. The vehicle determination unit 68 then identifies the largest maximum feature value among the feature values ​​of the multiple sensor devices 20 and determines the weight of the vehicle based on the identified maximum feature value. Furthermore, the overheavy vehicle determination unit 70 determines that the vehicle passing through the bridge is an overheavy vehicle if the identified maximum feature value is greater than a threshold value.

[0082] When a vehicle passes over a bridge, the positions of the multiple sensor devices 20 on the bridge expand and contract in the direction of travel of the vehicle. Because the lane through which the vehicle passed sinks the furthest, the amplitude of the expansion and contraction amount for the lane through which the vehicle passed is the largest, and the amplitude of the expansion and contraction amount decreases as the lane becomes farther away from the lane through which the vehicle passed. Therefore, the information processing device 24 can determine which lane the vehicle passed over by comparing the amplitude of the expansion and contraction amount detected almost simultaneously for each lane.

[0083] Furthermore, the largest maximum feature value among the feature values ​​of the plurality of sensor devices 20 is a feature value calculated from time-series data of observation values ​​detected by the sensor device 20 installed in the lane through which the vehicle has passed. Therefore, by using the largest maximum feature value among the feature values ​​of the plurality of sensor devices 20, the information processing device 24 can accurately determine the weight of the vehicle and whether it is an overweight vehicle, regardless of which lane the vehicle has passed through.

[0084] (Second embodiment) Next, a determination system 10 according to a second embodiment will be described. The determination system 10 according to the second embodiment has substantially the same functions and configuration as the first embodiment. In the description of the second embodiment, devices and components having the same functions and configuration as the first embodiment are denoted by the same reference numerals, and detailed description will be omitted except for differences. The same applies to the third and subsequent embodiments.

[0085] FIG. 12 is a diagram showing the functional configuration of an information processing device 24 according to the second embodiment.

[0086] The information processing device 24 further includes a learning interval control unit 92. The learning interval control unit 92 generates an alert indicating that it is time to start the learning process. The learning interval control unit 92 provides the alert to, for example, the model learning unit 76. When the model learning unit 76 receives the alert from the learning interval control unit 92, it starts the learning process. The learning interval control unit 92 may also notify an administrator or the like by displaying the alert on, for example, a display device. This allows the administrator to, for example, prepare to run the test vehicle or provide the passing time and weight of the test vehicle to the model learning unit 76.

[0087] In this embodiment, the model learning unit 76 repeats the learning process at time intervals. For example, after executing a first learning process, the model learning unit 76 executes a second learning process at time intervals. In this case, the learning interval control unit 92 detects the number of times a peak waveform is detected by the feature calculation unit 66 after the first learning process, and if the number of times exceeds a preset value, issues an alert indicating that it is time to start the second learning process.

[0088] The strength of a bridge changes due to deterioration. Therefore, if time passes after the completion of the learning process, the accuracy of the information processing device 24 in determining the weight of a vehicle and whether it is a heavy vehicle or not may deteriorate. Bridge deterioration is also thought to depend on the volume of vehicle traffic. In other words, the more vehicles pass through the bridge, the faster the bridge deteriorates.

[0089] The learning interval control unit 92 issues an alert when the number of peak waveforms detected by the feature calculation unit 66 exceeds a preset value. The number of peak waveforms detected by the feature calculation unit 66 represents the number of times vehicles have passed. Therefore, the learning interval control unit 92 issues an alert when the number of times vehicles have passed exceeds a preset value, and can notify that it is time to start the next learning process. This allows the learning interval control unit 92 to start the next learning process before the judgment accuracy deteriorates due to bridge deterioration.

[0090] The determination system 10 according to the second embodiment as described above can determine the weight of a vehicle and whether or not it is an overweight vehicle with high accuracy over a long period of time.

[0091] (Third embodiment) Next, a determination system 10 according to a third embodiment will be described.

[0092] FIG. 13 is a diagram showing the functional configuration of an information processing device 24 according to the third embodiment.

[0093] The information processing device 24 further includes a deterioration level calculation unit 94. The deterioration level calculation unit 94 acquires the threshold value detected by the outlier learning unit 78 for each learning process, and calculates the deterioration level of the bridge based on the acquired threshold value.

[0094] In this embodiment, the model learning unit 76 repeats the learning process at time intervals. For example, after executing a first learning process, the model learning unit 76 executes a second learning process after a first period has elapsed. The deterioration level calculation unit 94 calculates the deterioration level of the bridge based on the amount of change between the threshold value detected in the first learning process and the threshold value detected in the second learning process. The deterioration level calculation unit 94 then outputs the calculated deterioration level. For example, the deterioration level calculation unit 94 displays the calculated deterioration level on a display device or the like to notify a manager or the like.

[0095] FIG. 14 is a diagram showing changes over time in the range of normal values ​​of feature amounts.

[0096] The strength of a bridge changes due to deterioration. For this reason, the amount of expansion and contraction of a bridge when a vehicle of the same weight passes over it becomes larger after the passage of time than when it is first constructed. For this reason, the range of variation in the feature values ​​acquired in the learning process, i.e., the range of normal values, becomes wider after the passage of time than when it is first constructed, as shown in FIG. 14. Accordingly, the threshold value detected by the outlier learning unit 78 also becomes larger after the passage of time than when it is first constructed.

[0097] The change in the threshold value over time can be said to represent the degree of deterioration of the bridge. Therefore, the deterioration level calculation unit 94 according to this embodiment can calculate the degree of deterioration of the bridge by comparing the threshold value calculated by the first learning process with the threshold value calculated by the second learning process.

[0098] For example, the deterioration level calculation unit 94 may calculate the ratio between the threshold value calculated by the first learning process and the threshold value calculated by the second learning process as the deterioration level, or may calculate the difference between the threshold value calculated by the first learning process and the threshold value calculated by the second learning process as the deterioration level. Furthermore, the deterioration level calculation unit 94 may use the first learning process as the learning process at the start of vehicle weight determination, for example, the learning process at the time a bridge is first constructed. This allows the deterioration level calculation unit 94 to calculate the deterioration level from the start of vehicle weight determination.

[0099] The determination system 10 according to the third embodiment as described above can easily calculate the deterioration level of a bridge.

[0100] Fig. 15 is a diagram showing the hardware configuration of the information processing device 24. As an example, the information processing device 24 is realized by a device having the same hardware configuration as a general computer. The terminal device 22 may also have the same hardware configuration as that shown in Fig. 15. The information processing device 24 includes a CPU (Central Processing Unit) 301, an operation device 302, a display device 303, a main memory device 305, an auxiliary memory device 306, a communication device 307, and a bus 309. Each unit is connected via the bus 309.

[0101] The CPU 301 executes various processes in cooperation with various programs stored in advance in the auxiliary storage device 306, etc., using a predetermined area of ​​the main storage device 305 as a working area, and comprehensively controls the operation of each unit constituting the information processing device 24. The CPU 301 also operates the operation device 302, the display device 303, the communication device 307, etc. in cooperation with the programs.

[0102] The operation device 302 is an input device such as a touch panel, a mouse, or a keyboard, and receives information input by a user as an instruction signal, and outputs the instruction signal to the CPU 301 .

[0103] The display device 303 is a display unit such as an LCD (Liquid Crystal Display), etc. The display device 303 displays various information based on a display signal from the CPU 301.

[0104] The main storage device 305 is a volatile storage medium such as a Synchronous Dynamic Random Access Memory (SDRAM), etc. The main storage device 305 functions as a work area for the CPU 301.

[0105] The auxiliary storage device 306 is a rewritable storage device such as a semiconductor storage medium such as a flash memory, or a magnetically or optically recordable storage medium. The auxiliary storage device 306 stores programs used to control the information processing device 24.

[0106] The communication device 307 transmits and receives data to and from other devices. The communication device 307 may also transmit and receive data to and from a server or the like via a network.

[0107] The program executed by the information processing device 24 may be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. Alternatively, the program executed by the information processing device 24 may be provided in advance by being stored in a portable storage medium or the like.

[0108] The program executed by the information processing device 24 has a modular configuration including an acquisition module, a feature calculation module, a weight determination module, an overweight determination module, a time identification module, an output module, a model learning module, and an outlier learning module. The program may further include a learning interval control module and a deterioration level calculation module. The CPU 301 reads such program from a storage medium or the like and loads each of the above modules into the main memory device 305. By executing such program, the CPU 301 functions as the acquisition unit 62, the feature calculation unit 66, the vehicle determination unit 68, the overweight vehicle determination unit 70, the time identification unit 72, the output unit 74, the model learning unit 76, the outlier learning unit 78, the learning interval control unit 92, and the deterioration level calculation unit 94. By executing such program, the CPU 301 also causes the main memory device 305 or the auxiliary memory device 306 to function as the storage unit 64. In addition, some or all of the acquisition unit 62, feature calculation unit 66, vehicle determination unit 68, overweight vehicle determination unit 70, time identification unit 72, output unit 74, model learning unit 76, outlier learning unit 78, learning interval control unit 92 and deterioration level calculation unit 94 may be configured using hardware.

[0109] Although the embodiments of the present invention have been described above, these embodiments are presented as examples and are not intended to limit the scope of the invention. Various modifications can be made to the embodiments. [Explanation of symbols]

[0110] 10 Judgment System 20 Sensor device 22 Terminal equipment 24 Information processing equipment 62 Acquisition Department 64 Memory section 66 Feature calculation unit 68 Vehicle Judging Department 70 Overweight Vehicle Judgment Department 72 Time identification part 74 Output section 76 Model Learning Department 78 Outlier Learning Unit 80 Moving average part 82 Difference calculation unit 84 Peak value detector 86 Peak point detector 88 Area calculation part 92 Learning interval control unit 94 Deterioration degree calculation section

Claims

1. an acquisition unit that acquires, from a sensor device installed on a bridge, time series data of an observation value that represents a displacement of the bridge at a position on the bridge where the sensor device is installed; a feature calculation unit that calculates a feature according to the magnitude of a peak waveform in the time-series data of the observed value; a vehicle determination unit that determines the weight of the vehicle that has passed over the bridge using a determination model that determines the weight of the vehicle based on the feature amount; an overheavy vehicle determination unit that determines that the vehicle that has passed through the bridge is an overheavy vehicle when the feature amount is greater than a threshold value; a model learning unit that trains the determination model based on the feature amounts calculated when each of a plurality of test vehicles having known weights that are not the overheavy vehicle passes through the bridge in a learning process; an outlier learning unit that, in the learning process, classifies the feature amounts calculated when each of the plurality of test vehicles passes over the bridge into a group of normal values, detects a boundary between the normal value and an outlier greater than the normal value, and sets the detected boundary as the threshold value; A determination device comprising:

2. The sensor device detects, as the displacement of the bridge, the amount of expansion and contraction at the position where the sensor device is installed on the bridge. The determination device according to claim 1 .

3. The feature amount calculation unit a moving average unit that calculates time series data of a moving average value by performing a moving average on the time series data of the observed values ​​using a preset time window; a difference calculation unit that calculates time series data of difference values ​​that represent the difference between the observed value and the moving average value; a peak value detection unit that detects a peak value, which is a difference value of a peak point of a peak waveform in the time series data of the difference value, and outputs the detected peak value as the feature amount; The determination device according to claim 1 or 2, further comprising:

4. The feature amount calculation unit a moving average unit that calculates time series data of a moving average value by performing a moving average on the time series data of the observed values ​​using a preset time window; a peak point detection unit that detects peak points of a peak waveform in the time series data of the observed values; an area calculation unit that calculates an area of ​​a triangle connecting the peak point, a point before the peak point where the time series data of the observation value intersects with the time series data of the moving average value, and a point after the peak point where the time series data of the observation value intersects with the time series data of the moving average value in a graph that represents time on the horizontal axis and the observation value on the vertical axis, and outputs the calculated area as the feature amount; The determination device according to claim 1 or 2, further comprising:

5. The feature amount calculation unit a moving average unit that calculates time series data of a moving average value by performing a moving average on the time series data of the observed values ​​using a preset time window; a difference calculation unit that calculates time series data of difference values ​​that represent the difference between the observed value and the moving average value; a peak point detection unit that detects peak points of a peak waveform in the time series data of the difference values; an area calculation unit that calculates an area of ​​a triangle connecting the peak point, a point before the peak point where the time series data of the difference value intersects with the time series data of the moving average value, and a point after the peak point where the time series data of the difference value intersects with the time series data of the moving average value, in a graph that represents time on the horizontal axis and the difference value on the vertical axis, and outputs the calculated area as the feature amount; The determination device according to claim 1 or 2, further comprising:

6. The outlier learning unit detects the boundary using a One Class SVM (Support Vector Machine) algorithm. The determination device according to any one of claims 1 to 5.

7. the acquisition unit acquires time-series data of the observed values ​​from each of the plurality of sensor devices provided for each lane of the bridge; the feature calculation unit calculates the feature for each of the plurality of sensor devices; the vehicle determination unit identifies a maximum feature amount that is the largest among the feature amounts of each of the plurality of sensor devices, and determines a weight of the vehicle based on the identified maximum feature amount; The overheavy vehicle determination unit determines that the vehicle that has passed through the bridge is an overheavy vehicle when the maximum feature amount is greater than the threshold value. The determination device according to any one of claims 1 to 6.

8. A learning interval control unit is further provided, the model learning unit executes a first learning process and then executes a second learning process; The learning interval control unit detects the number of times the peak waveform is detected after the first learning process, and when the number of times exceeds a preset value, issues an alert indicating that it is time to start the second learning process. The determination device according to any one of claims 1 to 6.

9. A deterioration degree calculation unit is further provided, the model learning unit executes a first learning process, and then executes a second learning process after a first period has elapsed; The deterioration level calculation unit calculates the deterioration level of the bridge based on the amount of change between the threshold value detected in the first learning process and the threshold value detected in the second learning process. The determination device according to any one of claims 1 to 8.

10. a sensor device installed on a bridge; A determination device; Equipped with The determination device an acquisition unit that acquires, from the sensor device, time series data of observed values ​​that represent the displacement of the bridge at a position on the bridge where the sensor device is installed; a feature calculation unit that calculates a feature according to the magnitude of a peak waveform in the time-series data of the observed value; a vehicle determination unit that determines the weight of the vehicle that has passed over the bridge using a determination model that determines the weight of the vehicle based on the feature amount; an overheavy vehicle determination unit that determines that the vehicle that has passed through the bridge is an overheavy vehicle when the feature amount is greater than a threshold value; a model learning unit that trains the determination model based on the feature amounts calculated when each of a plurality of test vehicles having known weights that are not the overheavy vehicle passes through the bridge in a learning process; an outlier learning unit that, in the learning process, classifies the feature amounts calculated when each of the plurality of test vehicles passes over the bridge into a group of normal values, detects a boundary between the normal value and an outlier greater than the normal value, and sets the detected boundary as the threshold value; have Judging system.

11. an information processing device acquires, from a sensor device provided on a bridge, time series data of observed values ​​representing displacement of the bridge at a position on the bridge where the sensor device is provided; the information processing device calculates a feature amount corresponding to the magnitude of a peak waveform in the time-series data of the observed value; the information processing device determines a weight of the vehicle that has passed over the bridge using a determination model that determines a weight of the vehicle based on the feature amount; the information processing device determines that the vehicle that has passed through the bridge is an overweight vehicle when the feature amount is greater than a threshold value; the information processing device trains the judgment model based on the feature amounts calculated when each of a plurality of test vehicles having known weights that are not the overheavy vehicle passes through the bridge in a learning process; The information processing device, in the learning process, classifies the feature amounts calculated when each of the plurality of test vehicles passes over the bridge into a group of normal values, detects a boundary between the normal values ​​and outliers larger than the normal values, and sets the detected boundary as the threshold value. Judgment method.

12. A program for causing an information processing device to function as a determination device, The information processing device an acquisition unit that acquires, from a sensor device installed on a bridge, time series data of an observation value that represents a displacement of the bridge at a position on the bridge where the sensor device is installed; a feature calculation unit that calculates a feature according to the magnitude of a peak waveform in the time-series data of the observed value; a vehicle determination unit that determines the weight of the vehicle that has passed over the bridge using a determination model that determines the weight of the vehicle based on the feature amount; an overheavy vehicle determination unit that determines that the vehicle that has passed through the bridge is an overheavy vehicle when the feature amount is greater than a threshold value; a model learning unit that trains the determination model based on the feature amounts calculated when each of a plurality of test vehicles having known weights that are not the overheavy vehicle passes through the bridge in a learning process; an outlier learning unit that, in the learning process, classifies the feature amounts calculated when each of the plurality of test vehicles passes over the bridge into a group of normal values, detects a boundary between the normal value and an outlier greater than the normal value, and sets the detected boundary as the threshold value; A program that makes it work.

Images