JP7334753B2 - Rail subsidence prediction method, prediction device, rail repair time determination method, and rail subsidence abnormality time determination method - Google Patents

Description

TECHNICAL FIELD The present invention relates to a rail subsidence prediction method, a prediction device, a rail repair time determination method, and a rail subsidence abnormality time determination method for predicting the rail subsidence amount of a rail on which a truck for transporting objects to be transported travels. .

In a steelworks, a slab continuously cast in a continuous casting facility may be transported to a hot rolling facility by a slab transport vehicle running on rails. Power is supplied to this slab carrier from a trolley wire through a power supply pantograph.

On the other hand, as an abnormality monitoring system for a contact wire and a power supply pantograph used in a mobile machine that receives power supply from a contact wire through a power supply pantograph, for example, the system disclosed in Patent Document 1 is conventionally known.
In the abnormality monitoring system for the contact wire and the power supply pantograph disclosed in Patent Document 1, the relative distance between the mobile machine and the contact wire support frame is measured by a laser rangefinder while the mobile machine is running in the contact wire deviation detection unit. Contact wire deflection can be detected based on the measured distance. Further, in the abnormality monitoring system, the load measuring unit can measure the pantograph load acting on the power feeding pantograph. In the abnormality monitoring system, the determination unit can determine the contact wire deflection and the pantograph load, and can determine that there is an abnormality when the measured values thereof are out of a predetermined range, for example.

Further, as a contact wire display device that accurately displays the position of a contact wire with respect to a reference on a track on which an electric train runs, there is conventionally known, for example, the device disclosed in Patent Document 2.
The trolley wire display device disclosed in Patent Literature 2 displays height and deviation based on reference values obtained from rails that constitute a track on which a train travels, among point cloud data obtained by three-dimensionally measuring an object. A display reference setting unit is provided for setting a display reference point that serves as a reference for displaying each point of the trolley wire whose position has been measured. Further, the trolley wire display device obtains a first point corresponding to the display reference point and a second point corresponding to each point on the trolley line on the track center line, and obtains each second point from the first point. and based on the second distance from the starting point to the first point and the first distance when displaying each point on the trolley line, from the starting point to each second point is provided with a distance measuring unit for calculating a third distance of . Further, the trolley wire display device includes a display control section for creating display data for displaying the height and deviation of each point on the trolley line corresponding to the second point based on the third distance.

JP-A-2017-13761 International Publication No. 2018/087931

However, the conventional contact wire and power supply pantograph abnormality monitoring system disclosed in Patent Document 1 and the conventional contact wire display device disclosed in Patent Document 2 have the following problems.
That is, in the case of the abnormality monitoring system for the contact wire and the power feeding pantograph disclosed in Patent Document 1, the contact wire deviation is detected, the pantograph load is measured, and an abnormality is detected when the measured values are out of a predetermined range. However, it is not possible to detect the amount of subsidence of the rail when the truck (mobile machine) is running on the rail. The amount of subduction could not be predicted.
Further, in the case of the contact wire display device disclosed in Patent Document 2, the position of the contact wire relative to the reference on the track on which the electric train travels can be displayed with high accuracy. ) could not detect the amount of rail sinking when the truck was running on the rail, and could not predict the amount of rail sinking after a predetermined period of time had passed since the truck was running.

On the other hand, even if the amount of rail subsidence is measured while the bogie is running and the amount of rail subsidence is predicted after a predetermined period of time has passed since the bogie is running, the amount of rail subsidence is caused by factors other than changes over time (for example, slab weight, change in bogie speed, etc.), when setting a threshold for the predicted rail subsidence and using it as a criterion for rail repair, the threshold is also affected by the rail subsidence due to fluctuation factors other than aging. It was difficult to set the threshold because it was necessary to set it in consideration of variations.

Therefore, the present invention has been made to solve these problems. To provide a rail subsidence amount prediction method, a prediction device, a rail repair time determination method, and a rail subsidence abnormality time determination method capable of appropriately predicting the predicted value of .

In order to solve the above problems, a method for predicting the amount of rail subsidence according to one aspect of the present invention is a method for predicting the amount of rail subsidence for predicting the amount of subsidence of a rail on which a carriage for transporting an object to be transported travels. The amount of rail subsidence at a specific position when the bogie is running due to past secular change and the amount of change in the influence factor that affects the past secular change during a predetermined period from when the bogie is running. Input data is the amount of change in the amount of rail subsidence at a specific position after the passage of a predetermined period of time, and outputs a predicted value of the rail subsidence amount at a specific position after the passage of a predetermined period of time from the bogie running due to past secular changes for this input data. A model for predicting the amount of rail subsidence over time generated by machine learning using multiple learning data as data. Input the fluctuation amount of the rail subsidence amount at a specific position after the lapse of a predetermined period from when the bogie is running due to aging based on the amount of change in the influence factors that affect the amount of rail subsidence and aging over a predetermined period from when the bogie is running. The gist of the present invention is to predict a predicted value of a rail subsidence amount at a specific position after a predetermined period of time has passed since the bogie travels due to secular change.

Further, a rail repair timing determination method according to another aspect of the present invention is an amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie travels due to aging, which is predicted by the above-described method of predicting the amount of rail subsidence. The gist is to determine the rail repair timing based on the predicted value of

Further, a rail subsidence amount prediction apparatus according to another aspect of the present invention is a rail subsidence amount prediction apparatus for predicting the rail subsidence amount on which a carriage that conveys an object to be conveyed travels. Based on the amount of rail subsidence at a specific position when the bogie is running due to aging, and the amount of change in the influence factors that affect the past aging over a predetermined period from when the bogie is running due to aging after the lapse of a predetermined period Input data is the amount of change in the amount of rail subsidence at a specific position, and output data is a predicted value of the amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie traveled due to past secular changes in this input data. A predictive model for the amount of rail subsidence over time generated by machine learning the data for learning, and the amount of rail subsidence at a specific position during bogie running due to secular change excluding the influence of fluctuation factors other than secular change and the secular change. Based on the amount of change in the predetermined period from when the bogie is running on the influencing factor that affects the bogie due to aging. The gist of the present invention is that the vehicle is provided with an aging rail subsidence amount prediction unit that predicts a predicted value of the rail subsidence amount at a specific position after a predetermined period of time has elapsed from the time of running.

Further, a rail subsidence amount prediction method according to another aspect of the present invention is a rail subsidence amount prediction method for predicting the rail subsidence amount on which a carriage for conveying an object to be conveyed travels, wherein the carriage travels. Machine learning machine learning Prediction of rail subsidence amount y (t, p) due to aging change at the prediction target point when the bogie is running and aging change factor x at the future prediction target time The gist is to input the value xf and predict the amount of rail subsidence y(t+k, p) due to secular change at the prediction target point at a future prediction target time.

Further, a method for determining an abnormal timing of rail subsidence according to another aspect of the present invention is the amount of rail subsidence due to secular changes at a prediction target point at a future prediction target time, according to the above-described method for predicting the amount of rail subsidence. Prediction of y(t+k, p) is performed for a plurality of prediction target time points, and a prediction target time point at which the prediction result is equal to or greater than a predetermined threshold set in advance is determined to be an abnormal rail subsidence time. do.

A rail subsidence amount prediction apparatus according to another aspect of the present invention is a rail subsidence amount prediction apparatus for predicting a rail subsidence amount on which a carriage for conveying an object to be conveyed travels, wherein the carriage travels. Machine learning machine learning Prediction of rail subsidence amount y (t, p) due to aging change at the prediction target point when the bogie is running and aging change factor x at the future prediction target time A secular change rail subsidence amount prediction unit that inputs a value xf and predicts the rail subsidence amount y(t+k, p) due to secular change at the prediction target point at a future prediction target point. and

According to the rail subsidence amount prediction method, the prediction device, the rail repair time determination method, and the rail subsidence abnormality time determination method according to the present invention, the bogie due to secular change excluding the influence of fluctuation factors other than secular change Rail subsidence prediction method, predicting device, rail repair timing determination method, and rail subsidence abnormality timing capable of appropriately predicting a predicted value of rail subsidence at a specific position after a predetermined period of time has passed since running can provide a determination method for

1 is a schematic configuration diagram of a state in which a bogie is running on a rail to which a method for predicting a rail subsidence amount according to an embodiment of the present invention is applied; FIG. FIG. 2 is a diagram for explaining the configuration of a device for measuring the amount of rail subsidence when a truck is running on the rail in FIG. 1 ; 1 is a block diagram of a schematic configuration of a rail subduction amount prediction device according to an embodiment of the present invention; FIG. FIG. 4 is a flow chart for explaining a processing flow of an arithmetic processing unit of an arithmetic unit of the rail subsidence amount prediction device shown in FIG. 3; FIG. A graph showing an example of the predicted rail subsidence amount at a specific position from when the bogie is running (at the current time) until the judgment result exceeds a predetermined threshold due to aging predicted by the rail subsidence amount prediction unit. be.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings. The embodiments shown below exemplify apparatuses and methods for embodying the technical idea of the present invention. It is not intended to be specific to the following embodiments. Also, the drawings are schematic. Therefore, it should be noted that the relationship, ratio, etc. between the thickness and the planar dimensions are different from the actual ones, and the drawings include portions where the relationship and ratio of the dimensions are different from each other.

FIG. 1 shows a schematic configuration of a truck running on a rail to which a method for predicting the amount of rail subsidence according to an embodiment of the present invention is applied.
A truck 1 shown in FIG. 1 conveys a slab S as an object to be conveyed, which is continuously cast in a continuous casting facility, to a hot rolling facility (not shown), and runs on rails 12 .
This trolley 1 comprises a trolley body 2 on which slabs S are loaded, and on both sides of the trolley body 2, a plurality of pairs of wheels 3 (only one pair is shown in FIG. 1) running on paired rails 12. The axle 3a of each wheel 3 is connected to a speed reducer 4, the rotation shaft 5a of one motor 5 is connected to the speed reducer 4, and the rotation of the motor 5 is reduced by the speed reducer 4 to rotate the wheels 3. Each wheel 3 , speed reducer 4 and motor 5 are supported by the carriage body 2 via suspensions 11 . The rails 12 are laid on paving stones 13 that have been pounded and hardened.

Electric power is supplied to the truck 1 from a contact wire (not shown) via a power supply pantograph (not shown). It is fixed to the tip of the horizontal portion 15 extending from the vicinity of the upper end.
A trolley position measuring device 6 for measuring the vertical position P1 of the trolley wire support frame 16 is installed on the side surface of the trolley main body 2 of the trolley 1 . As the trolley position measuring device 6, for example, a two-dimensional laser rangefinder is used.

A speed reducer position measuring device 7 for measuring the distance to the speed reducer 4 and for measuring the vertical position P2 of the speed reducer 4 is installed on the lower surface of the truck body 2 . A laser rangefinder, for example, is used as the speed reducer position measuring device 7 .
The trolley position measuring device 6 and the speed reducer position measuring device 7 are connected to a rail sinking amount measuring device 8 . A vertical position P1 of the contact wire support frame 16 measured by the trolley position measuring device 6 and a vertical position P1 of the speed reducer 4 measured by the speed reducer position measuring device 7 are stored in the rail sinking amount measuring device 8. P2 is input.

Also, as shown in FIG. 2, one of the wheels 3 of the bogie 1 is measured for the number of revolutions of the wheel 3 from a predetermined reference time to the time (t) when the bogie is running, and the number of revolutions and the running time are calculated. A traveling distance calculation sensor 17 is provided for calculating the traveling distance of the truck 1 from a predetermined reference time to the traveling time (t) of the truck from t. The traveling distance calculation sensor 17 is connected to the truck position management device 19 . The truck position management device 19 receives information on the traveling distance of the truck 1 from a predetermined reference time to the truck traveling time (t), which is calculated by the traveling distance calculation sensor 17 .

Also, in the vicinity of the rail 12, as shown in FIG. 2, a plurality (n pieces) of carriage position detection devices 18 ₁ to 18 ₁ are arranged along the rail 12 at predetermined intervals. Each of the truck position detection devices 18 ₁ to 18 ₁ detects the passage of the truck 1 when the truck 1 traveling on the rail 12 passes through the respective detection positions of the truck position detection devices 18 ₁ to 18 ₁ . . Each truck position detection device 18 ₁ to 18 ₁ is connected to a truck position management device 19 . The carriage passage information from each of the carriage position detection devices 18 ₁ to 18 ₁ is input to the carriage position management device 19 .

Then, the carriage position management device 19 receives travel distance information of the carriage 1 from the travel distance calculation sensor ₁₇ to _the time (t) when the carriage travels from a predetermined reference time, The position of the truck (p) at the time when the truck is running (t) is calculated from the truck passage information from each of them.
Further, the truck position management device 19 is connected to the rail sinking amount measuring device 8 described above. The truck position (p) at the time of truck travel (t) calculated by the truck position management device 19 is input to the rail sinking amount measuring device 8 .

The rail sinking amount measuring device 8 measures the vertical position P1 of the contact wire support frame 16 measured by the trolley position measuring device 6, and the vertical position P2 of the speed reducer 4 measured by the speed reducer position measuring device 7. The upper surface position P3 of the rail 12 is calculated from . Then, the rail sinking amount measuring device 8 calculates the sinking amount from the normal state of the calculated upper surface position P3 of the rail 12 . The rail sinking amount measuring device 8 combines the calculated sinking amount with the information on the truck position (p) when the truck is traveling (t) input from the truck position management device 19, is transmitted as a rail subsidence amount Z(t, p) at a specific position (p) of the rail subsidence amount to a rail subsidence amount prediction device 20 described later.

A slab weight measuring device 9 for measuring the weight of the slab S is installed on the carriage body 2 of the carriage 1 . Further, the motor 5 is provided with a truck speed measuring device 10 that measures the speed of the truck 1 and thus the speed change (acceleration) of the truck 1 by measuring the rotation speed of the rotating shaft 5a of the motor 5. The weight of the slab S measured by the slab weight measuring device 9 is transmitted to the rail subsidence amount predicting device 20, which will be described later, as the slab weight n ₁ (t), which is a variable factor other than the secular change during running of the bogie. be. In addition, the speed change (acceleration) of the bogie 1 measured by the bogie speed measuring device 10 is also used as the bogie speed change amount n ₂ (t), which is a variation factor other than the secular change while the bogie is running, and is used as a predictor of the rail subsidence amount. 20.

The reason why the slab weight n ₁ (t) and the bogie speed variation n ₂ (t) are selected in this embodiment as fluctuation factors other than aging while the bogie is running is as follows. The sinking of the rail 12 due to fluctuation factors other than aging is caused by the force applied to the rail 12. One of the factors for the force applied to the rail 12 is the slab weight n ₁ (t). Another reason is that if the rail 12 is inclined, the acceleration/deceleration force of the truck 1 is also applied to the rail 12 in the vertical direction, so the truck speed change amount n ₂ (t) is selected.
The sinking of the rails 12 due to aging is caused by the wear of the rails 12 and the expansion of the gaps between the paving stones 13 under the rails 12. Since the gap between the paving stones 13 is generated by the application of force and the movement of the paving stones 13 due to rain, the amount of rain x1 and the production amount x2 of the slab S, which is closely related to the running of the bogie 1, are the factors that affect the aging. selected.

Next, the rail subduction amount prediction device 20 shown in FIG. Or connected wirelessly.
This rail subsidence amount prediction device 20 predicts the rail subsidence amount at a specific position (p) after a predetermined period (k) has elapsed from the bogie running time (t) due to secular change excluding the effects of fluctuation factors other than secular change. is to predict the predicted value y(t+k, p) of .
The rail subduction amount prediction device 20 is, as shown in FIG. The computing device 21, the input device 44, the storage device 45, and the output device 46 are not limited to this connection mode, and may be connected wirelessly, or may be connected in a combination of wired and wireless. good.

The input device 44 is, for example, a keyboard, a pen tablet, a touch pad, a mouse, or any other input interface capable of detecting the operation of the administrator of this system. ing. Also, the position (p) of the truck 1 on the rail 12 transmitted from the truck traveling control device (not shown) and the specific position (p) at the time of truck traveling (t) transmitted from the rail sinking amount measuring device 8 A rail sinking amount Z(t, p) is input. Also, the slab weight n ₁ (t), which is a variable factor other than the secular change during the running of the bogie (t), transmitted from the slab weight measuring device 9 is input. Furthermore, the bogie speed change amount n ₂ (t), which is a variation factor other than the secular change during the bogie running (t), transmitted from the bogie speed measuring device 10 is input. In addition, the input device 44 accepts operations related to various processing instructions to the arithmetic device 21 .

The storage device 45 is, for example, a hard disk drive, a semiconductor drive, an optical drive, or the like, and is a device that stores necessary information in this system. The storage device 45 stores, for example, input data of fluctuation factors (slab weight n ₁ (t) and bogie speed change amount n ₂ (t)) other than the secular change during past bogie running (t), and past data for this input data. Amount _of rail subsidence (rail subsidence due to slab _weight A plurality of learning data are stored in which output data are the sinking amount h ₁ (t, p) and the rail sinking amount h ₂ (t, p) based on the bogie speed change amount. In addition, the storage device 45 stores the rail subsidence amount y(t, p) at a specific position (p) when the bogie travels (t) due to past secular change, and influence factors (rainfall x1 and At a specific position (p) after a predetermined period (k) has passed from the time (t) when the truck is running due to aging based on the amount of change in the production amount x2) of the slab S during the predetermined period (k) from when the truck is running (t) Input data of fluctuation amount of rail subsidence amount and prediction value y ( It stores a plurality of learning data with t+k, p) as output data. In addition, the storage device 45 stores information such as a generated variation factor subsidence amount calculation model (slab weight subsidence calculation model) and a generated secular change rail subsidence amount prediction model, which will be described later.

In addition, the output device 46 outputs the output data, for example, after a predetermined period of time (k) has elapsed from the time (t) when the truck traveled due to aging, which is the output data of the determination result output unit 42, which was determined to be a rail abnormality, or when it has been changed. A predicted value y (any one of t+k to t+nk, p) of the rail subsidence amount at a specific position (p) after the lapse of a predetermined period (any of 2k to nk) is output. In addition, the output device 46 predicts the amount of rail subsidence at a specific position (p) from the time (t) when the bogie is running due to the secular change predicted by the secular change rail subsidence amount prediction unit 40 until it is determined that there is a rail abnormality. It also outputs the value y(t+k to t+nk, p). The output device 46 can display a screen based on output data and signals by including any display such as a liquid crystal display or an organic display.

The arithmetic device 21 also includes a RAM 22 , a ROM 23 , and an arithmetic processing section 31 . The ROM 23 contains a slab weight subsidence amount calculation model generation program 24, a bogie speed change subduction amount calculation model generation program 25, a slab weight subduction amount calculation program 26, a bogie speed change subduction amount calculation program 27, and a secular change subduction amount. A calculation program 28, a secular change amount prediction model generation program 29, and a secular change amount prediction program 30 are stored. Moreover, the arithmetic processing unit 31 has an arithmetic processing function and is connected to the RAM 22 and the ROM 23 via the bus 43 . The RAM 22 , ROM 23 and arithmetic processing section 31 are also connected to an input device 44 , a storage device 45 and an output device 46 via a bus 43 .

The arithmetic processing unit 31 includes, as functional blocks, a variation factor sagging amount calculation model generation unit 32, a variation factor rail sagging amount calculation unit 35, an aging rail sagging amount calculation unit 38, and an aging rail sagging amount prediction model generation unit. A section 39 , an aging rail subsidence amount prediction section 40 , a determination section 41 , and a determination result output section 42 are provided.
Here, the variation factor subsidence amount calculation model generation section 32 includes a slab weight subduction amount calculation model generation section 33 and a bogie speed change subduction amount calculation model generation section 34 .

When the slab weight sinking amount calculation model generation unit 33 receives an instruction to generate a calculation model from the input device 44, the slab weight n ₁ (t) of the past bogie running time (t) stored in the storage device 45 is calculated. A plurality of learning data are acquired with the input data and the rail subduction amount h ₁ (t, p) at a specific position (p) when the bogie is running due to the past slab weight n ₁ (t) for this input data as output data. do. Then, the slab weight subsidence amount calculation model generation unit 33 executes the slab weight subsidence amount calculation model generation program 24 to machine-learn the plurality of learning data, thereby forming the variation factor subsidence amount calculation model. Generate a subduction amount calculation model.

In addition, when receiving an instruction to generate a calculation model from the input device 44, the bogie speed change subsidence amount calculation model generation unit 34 calculates the bogie speed change amount of the past bogie running time (t) stored in the storage device 45. Let n ₂ (t) be input data and rail subsidence amount h ₂ (t, p) at a specific position (p) during bogie travel based on past bogie speed change amount n ₂ (t) with respect to this input data as output data Get multiple training data Then, the bogie speed change subsidence amount calculation model generation unit 34 executes the bogie speed change subduction amount calculation model generation program 25 and machine-learns the plurality of learning data to construct a variation factor subduction amount calculation model. A bogie speed change subduction amount calculation model is generated.

Next, the variation factor rail subsidence amount calculation unit 35 includes a slab weight subsidence amount calculation unit 36 and a bogie speed change subsidence amount calculation unit 37 .
When receiving a calculation instruction from the input device 44, the slab weight subsidence amount calculation unit 36 executes the slab weight subsidence amount calculation program 26 to The slab weight n ₁ (t), which is measured by the slab weight measuring device 9 and input to the input device 44, is input to the weight sinking amount calculation model. A rail subsidence amount h ₁ (t, p) at a specific position (p) when the bogie is running (t) due to the slab weight, which is a variable factor other than aging, is calculated.

Here, the rail subsidence amount h ₁ (t, p) at a specific position (p) when the bogie is traveling due to the slab weight ( _t ) is the amount of rail depression (t ) at a specific position (p) of rail subduction h ₁ (t, p) is fh ₁ (p, n ₁ (t)), it can be expressed as the following equation (1) .
_h1 (t,p)= _fh1 (p, _n1 (t))= _A1 (p) _n1 (t)+ _B1 (p) (1)
Here, A ₁ (p) and B ₁ (p) are constants for each position (p) of the truck 1 on the rail 12 .

Further, when receiving a calculation instruction from the input device 44, the bogie speed change subsidence amount calculation unit 37 executes the bogie speed change subsidence amount calculation program 27, and the bogie speed change subsidence amount calculation model generation unit 34 In the bogie speed variation subsidence amount calculation model generated in , the bogie speed change amount n ₂ (t) is input to calculate the rail subsidence amount h ₂ (t, p) at a specific position (p) when the truck is running (t) based on the amount of truck speed variation that constitutes a variable factor other than aging.

Here, the amount of rail subsidence h ₂ (t, p) at a specific position (p) when the bogie is running (t) due to a change in bogie speed depends on the amount of bogie speed change with respect to the change in the bogie speed change amount n ₂ (t). Assuming that fh ₁ (p, n ₂ (t)) is the relational expression of the fluctuation amount of the rail subduction amount h ₂ (t, p) at a specific position (p) when the truck is traveling (t), the following (2) can be expressed as
_h2 (t,p)= _fh2 (p, _n2 (t))= _A2 (p) _n2 (t)+ _B2 (p) (2)
Here, A ₂ (p) and B ₂ (p) are constants for each position (p) of the truck 1 on the rail 12 .

In addition, when receiving a calculation instruction from the input device 44, the aging rail subsidence amount calculation unit 38 executes the aging change amount calculation program 28, and the rail subduction amount measurement device 8 calculates and inputs The amount of rail subsidence Z(t, p) at a specific position (p) when the bogie is running (t) input to the device 44 and the fluctuation factors other than the secular change calculated by the slab weight subsidence amount calculation unit 36 are calculated. The amount of rail subsidence h ₁ (t, p) at a specific position (p) when the bogie is running (t) due to the weight of the slab, and the fluctuation factor other than the secular change calculated by the bogie speed change subsidence amount calculation unit 37 Based on the rail subduction amount h ₂ (t, p) at a specific position (p) when the bogie is running (t) due to the amount of bogie speed change, the influence of fluctuation factors other than aging is calculated based on the following equation (3). A rail subsidence amount y(t, p) at a specific position (p) when the bogie is running (t) due to secular change excluding .

y(t, p)=Z(t, p)−(h ₁ (t, p)+h ₂ (t, p)) (3)
That is, the secular change rail subsidence amount calculation unit 38 calculates the rail subsidence amount (h ₁ (t, p)+h ₂ (t, p)) is excluded to calculate the rail subsidence amount y(t, p) at the specific position (p) during bogie running (t) due to aging.

In addition, when receiving an instruction to generate a prediction model from the input device 44, the secular change rail subsidence amount prediction model generation unit 39 specifies the bogie running time (t) due to the past secular change stored in the storage device 45. A predetermined period (k ) based on the amount of change in the rail subduction amount at a specific position (p) after a predetermined period of time (k) from the time when the bogie was running (t) due to aging. A plurality of learning data are obtained in which the predicted value y(t+k, p) of the rail subsidence amount at a specific position (p) after the elapse of a predetermined period (k) from the time (t) when the truck is running is obtained as output data.

Then, the aging rail subsidence amount prediction model generation unit 39 executes the aging change amount prediction model generation program 29, performs machine learning on the acquired plurality of learning data, and generates the aging change rail subsidence amount prediction model. Generate.
The slab weight subsidence amount calculation model generated by the slab weight subsidence amount calculation model generation unit 33, the bogie speed change subsidence amount calculation model generated by the bogie speed change subduction amount calculation model generation unit 34, and the aging The aging rail subsidence amount prediction model generated by the changing rail subsidence amount prediction model generation unit 39 is a model constructed by multiple regression analysis or a neural network.

In addition, the aging rail subsidence amount prediction unit 40 executes the aging change rail subsidence amount prediction program 30 to generate the aging rail subsidence amount prediction model generated by the aging rail subsidence amount prediction model generation unit 39. , rail subsidence amount y(t, p). In addition, the secular change rail subsidence amount prediction unit 40 adds to the secular change rail subsidence amount prediction model the influence factors that affect the secular change (rain amount x1 and production amount of slab S x2) from the bogie running time (t). Input the amount of change in the amount of rail subsidence at a specific position (p) after the passage of a predetermined period (k) from the time when the bogie is running (t) due to aging based on the amount of change in the predetermined period (k). A predicted value y(t+k, p) of the rail subsidence amount at a specific position (p) after a predetermined period of time (k) has elapsed since the truck travels (t) is predicted.

Here, the predicted value y(t+k, p) of the rail subsidence amount at a specific position (p) after the lapse of a predetermined period (k) from the time when the bogie is running due to aging (t) is The amount of rail subduction y (t, p) at a specific position (p) of (t) and the integral term of the following equation (4) are factors that affect aging (rainfall x1 and slab S production x2 ) based on the amount of change in the predetermined period (k) from the time (t) when the bogie is running (t). As a quantity, it can be expressed by the following formula (4). Here, F (t, k, p) is the integral term of the formula (4), and is based on the amount of change in the factor that affects aging over a predetermined period (k) from when the bogie is running (t). It is the amount of change in the rail subsidence amount at a specific position (p) after a predetermined period of time (k) has passed since the time when the bogie was running (t) due to aging.

Predetermined period from when the bogie is running (t) due to secular change based on the amount of change during the predetermined period (k) from when the bogie is running (t) in the influencing factors (rainfall x1 and production volume of slab S x2) that affect aging (k) The amount of change in the rail subsidence amount at the specific position (p) after the lapse of time is input from the host computer 47 to the secular change rail subsidence amount prediction model. The host computer 47 stores the rainfall x1(t) when the truck is running (t) calculated from the annual average rainfall, the rainfall x1(t+k) after a predetermined period of time (k) has passed from the time when the truck is running (t), Rainfall x1 (2k) after a predetermined period of time (k) has passed from time (t), rainfall x1 (t+(n-1 )k), and the amount of rain x1(t+nk) after a predetermined period of time (nk) has elapsed from the time when the truck is running (t) is input in advance.

In addition, the host computer 47 stores the production amount x2(t) of the slab S when the truck is running (t) calculated from the production plan of the slab S, S production amount x2(t+k), Slab S production amount x2(t+2k) after a predetermined period of time (2k) has elapsed from the time (t) when the truck is running, and a predetermined period ((n- 1) Input in advance x2(t+(n−1)k) of slab S production after k) and 2(t+nk) of slab S after a predetermined period of time (nk) has passed from the time when the truck is running (t). It is

Then, the host computer 47 calculates the amount of change in rainfall (x1(t+k)-x1(t)) after the lapse of a predetermined period (k) from the time when the bogie is running (t) calculated from the annual average rainfall, and the amount of rainfall of the slab S The amount of change (x2(t+k)-x2(t)) in the production amount of the slab S after the lapse of a predetermined period (k) from the time (t) when the truck travels calculated from the production plan, and the amount of change in the amount of rainfall (x1( t+k)-x1(t)) and the amount of change in the production volume of the slab S (x2(t+k)-x2(t)), which is calculated based on the change over time, for a predetermined period (k) from the time when the truck is running (t) The post-elapsed rail subsidence variation amount (integral term of formula (4)) at the specific position (p) is sent to the secular change rail subsidence amount prediction unit 40 and input to the secular change rail subsidence amount prediction model. do. Here, the amount of change in rainfall (x1(t+k)-x1(t)) after a predetermined period of time (k) has passed since the bogie is running (t) calculated from the annual average rainfall is calculated from the bogie running (t). It means the cumulative amount of rainfall after a predetermined period (k) has passed. In addition, the amount of change in the production amount of the slab S (x2(t+k)-x2(t)) after the lapse of a predetermined period (k) from the time (t) when the slab S is running is calculated from the production plan of the slab S. It means the production amount of the difference between the production amount of the slabs S after the lapse of a predetermined period (k) from (t) and the production amount of the slabs when the truck is running (t).

Note that the aging rail subsidence amount prediction unit 40 executes the aging change rail subsidence amount prediction program 30 to generate the aging rail subsidence amount prediction model generated by the aging rail subsidence amount prediction model generation unit 39. , rail subsidence amount y(t, p). In addition, the secular change rail subsidence amount prediction unit 40 adds to the secular change rail subsidence amount prediction model the influence factors that affect the secular change (rain amount x1 and production amount of slab S x2) from the bogie running time (t). Input the amount of change in the amount of rail subsidence at a specific position (p) after the passage of a predetermined period (k) from the time when the bogie is running (t) due to aging based on the amount of change in the predetermined period (k). A predicted value y(t+k, p) of the rail subsidence amount at a specific position (p) after a predetermined period of time (k) has elapsed since the truck travels (t) is predicted.

Further, the determination unit 41 determines that the predicted value y(t+k, p) of the rail subsidence amount at the specific position (p) after the lapse of a predetermined period (k) from the bogie running time (t) due to secular change is a predetermined threshold value. (For example, 20 mm, which is the maximum value in the past record) or more.
Then, the determination unit 41 determines that the predicted value y(t+k, p) of the rail subsidence amount at the specific position (p) after the lapse of the predetermined period (k) from the bogie running time (t) due to aging is the above-mentioned predetermined value. is not equal to or greater than the threshold, the predetermined period (k) is changed (predetermined period (k)→predetermined period (2k)).

In addition, the secular change rail subsidence amount prediction unit 40 predicts the predicted value y(t+2k, p) of the rail subsidence amount due to secular change at the specific position (p) after the lapse of the changed predetermined period (2k). again. At this time, the secular change rail subsidence amount prediction unit 40 adds the secular change rail subsidence amount prediction model to a specific position ( Input the rail sinking amount y(t+k, p) at p). In addition, the aging rail subsidence amount prediction unit 40 adds, to the aging rail subsidence amount prediction model, influence factors (rainfall x1 and production amount of slab S x2) affecting aging change from a predetermined period (t+k) to a predetermined By inputting the variation amount of the rail subduction amount at the specific position (p) from the predetermined period (t+k) to the predetermined period (t+2k) due to aging based on the amount of change up to the period (t+2k), the changed bogie running time A predicted value y(t+2k, p) of rail subsidence due to secular change at a specific position (p) after a predetermined period (2k) has elapsed from (t) is predicted.

Here, the predicted value y(t+2k, p) of the rail subsidence amount at a specific position (p) after the elapse of a predetermined period (2k) from the bogie running time (t) due to secular change is given by the following equation (5): can be expressed as
y(t+2k, p)=y(t+k,p)+F(t+k,2k,P) (5)
In addition, the predicted value y (t + 2k to t + nk, p) of the rail subsidence amount at a specific position (p) after a predetermined period (3k to nk) has passed since the time when the bogie was running (t) due to aging is obtained by the following equation. can be expressed as
y(t+3k,p)=y(t+2k,p)+F(t+2k,3k,p)
↓
y(t+nk, p)=y(t+(n−1)k,p)+F(t+(n−1)k,nk,P)

Then, the determination unit 41 and the secular change rail subsidence amount prediction unit 40 change and predict the predicted value for a predetermined period until the predicted value y (t+2k to t+nk, p) becomes the above-described predetermined threshold value or more. The prediction and determination of are repeated.
In addition, the determination result output unit 42 outputs the rail at a specific position (p) after a predetermined period (k) or a changed predetermined period (any of 2k to nk) has elapsed since the bogie traveled (t) due to aging. When the predicted value y (any one of t+k to t+nk, p) of the amount of subduction is equal to or greater than a predetermined threshold value, it is determined that the rail is abnormal.
The determination result of the determination result output unit 42 is output to the output device 46 .

In addition, the prediction value of the rail subsidence amount at a specific position (p) from the time when the bogie is running (current time) (t) due to the secular change predicted by the secular change rail subsidence amount prediction unit 40 until it is determined that there is a rail abnormality. y(t+k to t+nk, p) is also output to the output device 46 . FIG. 5 shows an example of the predicted value of the rail subsidence amount at a specific position from when the bogie is running due to aging (current time) until it is determined that there is a rail abnormality.

Next, the flow of processing of the arithmetic processing unit 31 showing the method of predicting the amount of rail subsidence according to one embodiment of the present invention will be described with reference to FIG.
First, as shown in FIG. 4, in step S1, the variation factor sinking amount calculation model generating section 32 of the arithmetic processing section 31 receives a calculation model generation instruction from the input device 44, and calculates the variation factor sinking amount. Generate a computational model.

Specifically, when the slab weight subsidence amount calculation model generation unit 33 of the variation factor subsidence amount calculation model generation unit 32 receives an instruction to generate a calculation model from the input device 44, the slab weight subsidence amount calculation is performed. By executing the model generation program 24, the slab weight n ₁ (t) of the past bogie running time (t) stored in the storage device 45 is input data and the bogie based on the past slab weight n ₁ (t) for this input data. A plurality of learning data are acquired with the rail subduction amount h ₁ (t, p) at a specific position (p) during running as output data. Then, the slab weight sinking amount calculation model generation unit 33 performs machine learning on the plurality of pieces of learning data to generate a slab weight sinking amount calculation model constituting a variation factor sinking amount calculation model.

In addition, when receiving an instruction to generate a calculation model from the input device 44, the bogie speed change subduction amount calculation model generation unit 34 of the variation factor subduction amount calculation model generation unit 32 generates a bogie speed change subduction amount calculation model. By executing the program 25, the amount of change in bogie speed n 2 (t) at the time when the past truck traveled (t) stored in the storage device 45 is used as input data and the amount of change in past bogie speed _{n 2 (} t) for this input data. A plurality of learning data are acquired with the rail sinking amount h ₂ (t, p) at a specific position (p) when the bogie is traveling by using as output data. Then, the bogie speed change subsidence amount calculation model generation unit 34 performs machine learning on the plurality of learning data to generate a bogie speed change subsidence amount calculation model that configures the variation factor subsidence amount calculation model.

Next, in step S2, when receiving a calculation instruction from the input device 44, the variation factor rail subsidence amount calculator 35 calculates the rail subsidence amount at a specific position during bogie travel due to variation factors other than aging. .
Specifically, the slab weight subsidence amount calculation unit 36 of the variation factor rail subsidence amount calculation unit 35 executes the slab weight subsidence amount calculation program 26 when receiving a calculation instruction from the input device 44. , the slab weight subsidence amount calculation model generated by the slab weight subsidence amount calculation model generation unit 33 is measured by the slab weight measurement device 9 and input to the input device 44, other than the secular change during the truck running (t) Input the slab weight n ₁ (t) that is a factor of variation in , and the amount of rail subsidence h ₁ (t, p ) is calculated.

Further, when receiving a calculation instruction from the input device 44, the bogie speed change subsidence amount calculation unit 37 of the fluctuation factor rail subsidence amount calculation unit 35 executes the bogie speed change subsidence amount calculation program 27 to The bogie speed change subsidence amount calculation model generated by the speed change subduction amount calculation model generation unit 34 is measured by the bogie speed measurement device 10 and input to the input device 44, other than the secular change during the bogie running (t). Input the amount of change in bogie speed n ₂ (t) that constitutes the variation factor of , and the amount of rail subsidence h ₂ at a specific position (p) when the bogie is traveling (t) due to the amount of change in bogie speed that constitutes a variation factor other than aging. Calculate (t, p).

Next, in step S3, when receiving a calculation instruction from the input device 44, the aging rail subsidence amount calculation unit 38 executes the aging change amount calculation program 28, and the rail subduction amount measuring device 8 Rail subsidence amount Z(t, p) at a specific position (p) when the bogie is running (t), which is calculated and input to the input device 44, and other than the secular change calculated by the slab weight subsidence amount calculation unit 36 Rail subsidence amount h ₁ (t, p) at a specific position (p) when the bogie is running (t) due to the slab weight that constitutes a variation factor of From the rail subsidence amount h ₂ (t, p) at a specific position (p) when the bogie is running (t) due to the bogie speed change amount that constitutes a fluctuation factor, based on the above-mentioned equation (3), A rail subsidence amount y(t, p) at a specific position (p) when the bogie is running (t) due to secular change excluding the influence of fluctuation factors is calculated.

Next, in step S4, the aging rail subsidence amount prediction model generation unit 39 generates a secular change rail subsidence amount prediction model when receiving a prediction model generation instruction from the input device 44. FIG.
More specifically, when receiving an instruction to generate a prediction model from the input device 44, the secular change rail subsidence amount prediction model generation unit 39 executes the secular change amount prediction model generation program 29 and stores it. The amount of rail subsidence y(t, p) at a specific position (p) when the bogie is running (t) due to the past secular change stored in the device 45 and the influencing factors that affect the past secular change (rainfall x1 and slab Rail at a specific position (p) after a predetermined period of time (k) has elapsed from the time when the bogie is running (t) due to aging based on the amount of change in the production amount of S x2) during the predetermined period (k) from when the bogie is running (t) The amount of variation in the amount of rail subsidence is used as input data, and a predicted value y ( A plurality of learning data are acquired with t+k, p) as output data.
Then, the secular change rail subsidence amount prediction model generation unit 39 performs machine learning on the acquired plurality of learning data to generate the secular change rail subsidence amount prediction model.

Next, in step S5, the secular change rail subsidence amount prediction unit 40 calculates a specific position (p ), the predicted value y(t+k, p) of the amount of rail subsidence is predicted.
Specifically, when receiving a prediction instruction from the input device 44, the aging rail subsidence amount prediction unit 40 executes the aging change rail subsidence amount prediction program 30 to generate an aging rail subsidence amount prediction model. The aging rail subsidence amount prediction model generated by the generation unit 39 is added to the bogie running time (t ), input the rail sinking amount y(t, p) at the specific position (p). In addition, the secular change rail subsidence amount prediction unit 40 adds to the secular change rail subsidence amount prediction model the influence factors that affect the secular change (rain amount x1 and production amount of slab S x2) from the bogie running time (t). The amount of change in the rail subsidence amount at a specific position (p) after a predetermined period (k) has passed from the bogie running (t) due to secular change based on the amount of change in the predetermined period (k) is input. As a result, the predicted value y(t + k , p) are predicted.

Here, when the bogie runs (t) due to secular change based on the amount of change in the influencing factors (rainfall x1 and production volume of slab S x2) that affect the secular change during a predetermined period (k) from when the bogie runs (t) The amount of change in the amount of rail subsidence at a specific position (p) after the elapse of a predetermined period (k) from , is input from the host computer 47 to the secular change in rail subsidence amount prediction model, as described above.

Next, in step S6, the determination unit 41 calculates a predicted value y(t+k , p) is greater than or equal to a predetermined threshold value (for example, 20 mm, which is the maximum value of past performance). When the determination result in step 6 is YES, step 8 is performed, and when the determination result in step 6 is NO, step 7 is performed.
In step 7, the predicted value y(t+k, p) of the rail subsidence amount at a specific position (p) after the lapse of a predetermined period (k) from the bogie running time (t) due to aging is calculated as the above-mentioned predetermined value. When the determination result is not equal to or greater than the threshold, the determination unit 41 changes the predetermined period (k) to the predetermined period (2k), and returns to step S5.

Then, in step S5 after returning, the secular change rail subsidence amount prediction unit 40 predicts the rail subsidence amount at a specific position (p) after a lapse of a predetermined period (2k) from the time when the bogie is running (t) due to secular change. Predict y(t+2k,p) again. At this time, the secular change rail subsidence amount prediction unit 40 adds the secular change rail subsidence amount prediction model to a specific position ( Input the rail sinking amount y(t+k, p) at p). In addition, the aging rail subsidence amount prediction unit 40 adds, to the aging rail subsidence amount prediction model, influence factors (rainfall x1 and production amount of slab S x2) affecting aging change from a predetermined period (t+k) to a predetermined By inputting the variation amount of the rail subduction amount at the specific position (p) from the predetermined period (t+k) to the predetermined period (t+2k) due to aging based on the amount of change up to the period (t+2k), the changed bogie running time A predicted value y(t+2k, p) of rail subsidence due to secular change at a specific position (p) after a predetermined period (2k) has elapsed from (t) is predicted.

Then, the process proceeds to step S6, and the determination unit 41 obtains a predicted value y(t +2k,p) is equal to or greater than the predetermined threshold.
Then, the determining unit 41 and the secular change rail subsidence amount predicting unit 40 change the predetermined period in step S7 until the predicted value y (t+2k to t+nk, p) is determined to be equal to or greater than the predetermined threshold value. , the prediction of the predicted value in step S5 and the determination in step S6 are repeated.

Then, in step S8, the determination result output unit 42 outputs a specific position ( If the predicted value y (any of t+k to t+nk, p) of the rail subsidence amount in p) is determined to be equal to or greater than a predetermined threshold value, the rail is determined to be abnormal.
Then, the determination result of the determination result output unit 42 is output to the output device 46 . In addition, the prediction value of the rail subsidence amount at a specific position (p) from the time when the bogie is running (current time) (t) due to the secular change predicted by the secular change rail subsidence amount prediction unit 40 until it is determined that there is a rail abnormality. y(t+k to t+nk, p) is also output to the output device 46 .

FIG. 5 shows an example of the predicted value of the rail subsidence amount at a specific position from when the bogie is running due to aging (current time) until it is determined that there is a rail abnormality.
The output device 46 outputs the determination result of the determination result output unit 42 and the time from the time (current time) (t) when the truck is running due to the aging predicted by the aging rail subsidence amount prediction unit 40 until it is determined that the rail is abnormal. A predicted value y (t+k to t+nk, p) of the amount of rail subsidence at a specific position (p) is output. As a result, the operator can know the specific position (p) of the rail 12 that is considered to be rail abnormal and the timing of the rail abnormality (any of t to t+nk), so that repair can be performed before the rail abnormality occurs for a certain period of time. You will be able to formulate a plan.

As described above, according to the rail subsidence amount prediction method and the rail subsidence amount prediction apparatus according to the present embodiment, the secular change rail subsidence amount prediction model generated by performing machine learning on a plurality of learning data has variations other than secular change. Amount of rail subduction y (t, p) at a specific position (p) when the bogie is running (t) due to secular changes excluding the effects of factors, and factors that affect secular changes (rainfall x1 and slab S production x2) The amount of rail subsidence at a specific position (p) after a predetermined period (k) has passed since the bogie is running (t) due to aging based on the amount of change in the predetermined period (k) from when the bogie is running (t) By inputting the amount of variation, predict the predicted value y (t + k, p) of the rail subsidence amount at a specific position (p) after a predetermined period (k) has passed since the time when the bogie was running (t) due to aging ( Step S5, secular change rail sinking amount prediction unit 40).

As a result, the predicted value y(t + k , p) can be reasonably predicted. Therefore, since the effects of fluctuation factors other than aging are excluded, it is possible to easily set the threshold value of the predicted value, which is the criterion for judging rail abnormalities, and to appropriately predict the timing of rail abnormalities. be able to. As a result, it is possible to appropriately determine the timing of formulating a repair plan based on the timing of the rail abnormality, thereby avoiding the possibility of excessive repairs.

That is, the predicted value of the rail subsidence amount at a specific position (p) after a predetermined period of time (k) has passed since the time when the bogie was running (t) due to aging, which is predicted by the rail subsidence amount prediction method according to the present embodiment. A rail repair timing determination method for determining the repair timing of the rail 3 can be provided by y(t+k, p).
If the rail subsidence amount is predicted without excluding the effects of fluctuation factors other than secular change, if the forecast is not made without removing the current fluctuation factors, the predicted value will fluctuate due to fluctuation factors other than secular change. It is difficult to judge and can lead to wrong decisions and excessive repairs.

In addition, by predicting the predicted value y(t+k, p) of the rail subsidence amount at a specific position (p) after a predetermined period (k) has passed from the bogie running time (t) due to secular change, it is possible to detect rail abnormalities. It becomes easy to determine the specific position (p). Therefore, it is possible to easily determine the portion to be repaired on the rail 12, and to efficiently perform the repair.
The sinking of the rail 12 is caused by the gap between the rail 12 and the pavement 13, and the deformation of the rail 12 due to the weight of the truck 1 during running. The gap is reduced by compacting the paving stones around the position.
The trolley 1 needs to be stopped for the stamping. It is practically impossible.

In addition, according to the rail subsidence amount prediction method and apparatus according to the present embodiment, when the bogie travels due to secular change excluding the effects of fluctuation factors other than secular change input to the secular change rail subsidence amount prediction model, The amount of rail subsidence y(t, p) at a specific position (p) of (t) is obtained by applying a variation factor subsidence amount calculation model generated by machine-learning a plurality of learning data to the measured bogie running time (t ), input the variation factors n ₁ (t) and n ₂ (t) other than aging, and determine the rail subsidence amount h ₁ (t, p ), h ₂ (t, p) is calculated (step S2, fluctuation factor rail subsidence amount calculation unit 35), and the measured rail subsidence amount Z(t, p) and the amount of rail subsidence h ₁ (t, p) and h ₂ (t, p) at a specific position (p) when the bogie is running (t) due to fluctuation factors other than aging. (Step S3, secular change rail sinking amount calculator 38).

As a result, the rail subsidence amount y(t, p) can be calculated appropriately and easily.
Further, according to the method and apparatus for predicting the amount of rail subsidence according to the present embodiment, the amount of rail subsidence at a specific position (p) after the lapse of a predetermined period (k) from the time (t) when the bogie is running due to secular change. It is determined whether or not the predicted value y(t, p) is greater than or equal to a predetermined threshold (step S6, determination unit 41). Accordingly, it is possible to appropriately determine whether or not the predicted value y(t, p) of the rail subsidence amount is equal to or greater than the predetermined threshold value.

Further, according to the method and apparatus for predicting the amount of rail subsidence according to the present embodiment, the amount of rail subsidence at a specific position (p) after the lapse of a predetermined period (k) from the time (t) when the bogie is running due to secular change. When the predicted value y(t, k) is not equal to or greater than the predetermined threshold value, the predetermined period (k) is changed, and the rail at the specific position (p) after the changed predetermined period (2k to nk) has passed. The predicted value y (t+2k to t+nk, p) of the subduction amount is predicted again, and the predetermined period is changed until the predicted value y (t+2k to t+nk, p) reaches a predetermined threshold value or more. , the prediction of the predicted value and the determination are repeated. As a result, it is possible to accurately and accurately determine when the rail abnormality occurs.

Although the embodiment of the present invention has been described above, the present invention is not limited to this and can be modified and improved in various ways.
For example, fluctuation factors other than aging are not limited to slab weight n1(t) and bogie speed variation n2(t). may be a change in
In this case, it is necessary to machine-learn a plurality of learning data for each variation factor other than aging to generate a variation factor rail sinking amount calculation model.

In addition, the influence factor that influences the secular change is not limited to the amount of rainfall x1 and the production amount x2 of the slab S, but may be, for example, the occurrence of an earthquake that affects the laying condition of paving stones.
In addition, the fluctuation factor subduction amount calculation model (slab weight subduction amount calculation model and bogie speed change subduction amount calculation model) and aging rail subduction amount prediction model are models constructed by multiple regression analysis or neural networks. However, it may be constructed by machine learning techniques other than multiple regression analysis or neural networks.
Also, the object to be transported by the carriage 1 may be something other than the slab S.
In addition, in the modification of the rail subsidence amount prediction method and prediction device according to the present embodiment, the prediction target point (specific position (p ))) and the past data of the aging factor x, which is a factor affecting the aging change, are machine-learned to A changed rail sinking amount prediction model may be generated.

Specifically, the secular change rail subsidence amount prediction model generation unit 39 calculates the rail subsidence amount y(t, p) due to secular change at the prediction target point (specific position (p)) during bogie travel (t). Past data and past data of the aging factor x, which is a factor that affects aging, are input data, and a prediction target point (specific position (p) ) is machine-learned on a plurality of learning data having past data of the rail subsidence amount y(t+k, p) due to secular change as output data to generate a secular change rail subsidence amount prediction model.

Then, in the secular change rail subsidence amount prediction unit 40, the rail subsidence amount y(t, p) due to secular change at the prediction target point (specific position (p)) when the bogie is running (t), and the future prediction target A prediction target point (specific position (p) ) due to secular change y(t+k, p) of the rail is predicted.

In contrast, in the rail subsidence amount prediction method and apparatus according to the above-described embodiment, the secular change rail subsidence amount prediction unit 40 uses the secular change rail subsidence amount prediction model as the bogie traveling due to secular change. Aging based on the amount of rail subsidence y(t, p) at a specific position (p) at time (t) and the amount of change in the influence factor (x) that affects the aging in a predetermined period from the time (t) when the bogie is running Input the variation F(t, k, p) of the rail subduction amount at a specific position (p) after the lapse of a predetermined period (k) from the time (t) when the truck travels due to change, A predicted value y(t+k, p) of the rail subsidence amount at a specific position (p) after the elapse of a predetermined period of time (t+k) is predicted.

That is, in the above-described embodiment, the aging rail subsidence amount prediction unit 40 provides the aging rail subsidence amount prediction model with the rail Whereas y(t+k, p) is predicted by inputting the variation amount F(t, k, p) of the subsidence amount, in this modified example, the secular change rail subsidence amount prediction unit 40 y(t+k, p) is predicted by inputting y(t, p) and the predicted value xf of the secular change factor x into the changed rail subsidence amount prediction model. According to this modification, since the predicted value xf of the secular change factor x is input, the amount of rail subsidence at a specific position during running of the bogie due to past secular changes and past secular changes that cannot be protected by the above-described embodiment. Based on the amount of change in the predetermined period from when the bogie is running on the influencing factors Factors other than determining the amount of change in the amount of rail subduction at a specific position after the passage of a predetermined period from when the bogie is running due to aging (earthquake, etc.) variation, etc.) can also be considered.
Here, in a modified example, the model for predicting the amount of rail subsidence due to aging is a model constructed by multiple regression analysis or a neural network.

In addition, the aging factor x correlates with the amount of rainfall x1 from the start of the subduction amount measurement, the total weight x2 of the passing objects (passing slab S) from the start of the subduction amount measurement, and the total weight of the passing conveyed objects x2. contains as influencing factors one or more of a certain factor x3 of
Here, the amount of rail subsidence y(t, p) due to aging of the prediction target point when the truck travels is the height of the truck 1 ( obtained by measuring the vertical position P1 of the contact wire support frame 16 measured by the trolley position measuring device 6 and the vertical position P2 of the speed reducer 4 measured by the speed reducer position measuring device 7). , from the actual measurement value Z(t, p) of the rail subduction amount when the bogie travels (t) at the prediction target point (specific position (p)), the prediction target point (specific position (p)) when the bogie travels (t ) is obtained by subtracting the fluctuation factor subsidence amount h(t, p), which is an estimated value of the rail subsidence amount due to fluctuation factors other than aging.

Also, the variation factor sinking amount h(t, p) is estimated based on a model constructed by multiple regression analysis or a neural network.
Further, the variation factor sinking amount h(t, p) is estimated based on the weight n1 of the object to be conveyed and/or the amount of change n2 in the speed of the truck when the truck travels through the prediction target point.
In the modified example, the determination unit 41 determines whether or not the rail subsidence amount y(t+k, p) due to secular change at the prediction target point at the future prediction target point is equal to or greater than a predetermined threshold value set in advance. .

Then, when the determining unit 41 determines that the amount of rail subsidence y(t+k, p) due to secular change at the prediction target point at the future prediction target point is not equal to or greater than a predetermined threshold value, the prediction target time is changed. Then, the rail subsidence amount y(t+k, p) due to the secular change of the prediction target point is predicted again, and it is determined that the rail subsidence amount y(t+k, p), which is the predicted value, is equal to or greater than the predetermined threshold value. Change, prediction, and determination are repeated until
When the determination result by the determination unit 41 is equal to or greater than a predetermined threshold value, it is determined that the rail is abnormal.

Further, in the method for determining the abnormal timing of rail subsidence according to the present invention, the rail subsidence amount y(t+k, p ) is performed for a plurality of prediction target time points, and a prediction target time point at which the prediction result is equal to or greater than a predetermined threshold value set in advance is determined to be a time when the rail sinking abnormality occurs.

1 Carriage 2 Carriage Main Body 3 Wheel 3a Axle 4 Reduction Gear 5 Motor 5a Rotating Shaft 6 Trolley Position Measurement Device 7 Reduction Gear Position Measurement Device 8 Rail Subduction Amount Measurement Device 9 Slab Weight Measurement Device 10 Truck Speed Measurement Device 11 Suspension 12 Rail 13 paving stone 14 strut 15 horizontal part 16 trolley wire support stand 17 travel distance calculation sensor 18 ₁ to 18 _n bogie position detection device 19 bogie position management device 20 rail subduction amount prediction device 21 arithmetic device 22 RAN
23 ROMs
24 Slab Weight Subduction Amount Calculation Model Generation Program 25 Bogie Speed Change Subduction Amount Calculation Model Generation Program 26 Slab Weight Subduction Amount Calculation Program 27 Bogie Speed Change Subduction Amount Calculation Program 28 Secular Change Subduction Amount Calculation Program 29 Secular Change Sinking Subduction amount prediction model generation program 30 Aging subduction amount prediction program 31 Arithmetic processing unit 32 Variation factor subduction amount calculation model generation unit 33 Slab weight subduction amount calculation model generation unit 34 Bogie speed change subduction amount calculation model generation unit 35 Variation factor rail subsidence amount calculation unit 36 Slab weight subsidence amount calculation unit 37 Bogie speed change subsidence amount calculation unit 38 Aging rail subsidence amount calculation unit 39 Aging rail subsidence amount prediction model generation unit 40 Aging rail subsidence amount calculation unit Load amount prediction unit 41 Judgment unit 42 Judgment result output unit 43 Bus 44 Input device 45 Storage device 46 Output device 47 Host computer S Slab (conveying object)

Claims (35)

A rail sinking amount prediction method for predicting the rail sinking amount on which a truck for conveying an object travels, comprising:
A predetermined period of time has passed since the bogie was running due to aging, based on the amount of rail subsidence at a specific position when the bogie is running due to past secular changes, and the amount of change in the influence factor that affects the past secular change during a predetermined period from when the bogie is running. Input data is the amount of change in the amount of rail subsidence at a specific position later, and output data is the predicted value of the amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie was running due to past secular changes for this input data. A predictive model for the amount of rail subsidence over time generated by machine learning from multiple training data is applied to the amount of rail subsidence at a specific position during bogie travel due to aging, excluding the effects of fluctuation factors other than aging, and Input the amount of change in rail subsidence at a specific position after a predetermined period of time from when the bogie is running due to aging based on the amount of change in the factors that affect aging over a predetermined period from when the bogie is running. A method for predicting the amount of rail subsidence, characterized by predicting the predicted value of the amount of rail subsidence at a specific position after a predetermined period of time has elapsed since the bogie was running. The amount of rail subsidence at a specific position due to aging while the bogie is running, excluding the effects of fluctuation factors other than the above-mentioned aging, is obtained by using the past fluctuation factors other than the aging while the bogie is running as input data, and the past for this input data. Multiple learning data with the amount of rail subsidence at a specific position when the bogie is running due to fluctuation factors other than aging are output data. Calculate the amount of rail subsidence at a specific position while the bogie is running due to the variable factor other than the aging change by inputting the fluctuation factors other than the aging change, and compare the measured rail subsidence amount at the specific position while the bogie is running. 2. The method for predicting the amount of rail subsidence according to claim 1, wherein the amount of rail subsidence is calculated from the amount of rail subsidence at a specific position during running of the bogie caused by a variable factor other than aging. 3. The amount of rail subsidence according to claim 2 , wherein it is determined whether or not a predicted value of the amount of rail subsidence at a specific position after a lapse of a predetermined period from when the bogie is running due to aging is equal to or greater than a predetermined threshold value. Forecast method. When the predicted value of the rail subsidence amount at the specific position after the lapse of the predetermined period from the running of the bogie due to aging is a determination result that is not equal to or greater than a predetermined threshold value, the predetermined period is changed, and after the changed predetermined period has passed. predicting again the predicted value of the amount of rail subsidence at the specific position, and repeating the change, the prediction, and the determination until the predicted value reaches the predetermined threshold value or more. 3. The method for predicting the amount of rail subduction according to 3. Determining that the rail is abnormal when a predicted value of the rail subsidence amount at a specific position after a predetermined period of time has elapsed since the bogie traveled due to aging or after a changed predetermined period of time has elapsed is a predetermined threshold value or more. The method for predicting the amount of rail subduction according to claim 4. Variation factors other than the aging are the weight of the slab as the object to be conveyed and the speed change amount of the truck,
The variation factor subsidence amount calculation model uses the slab weight when the bogie has traveled in the past as input data, and the rail subduction amount at a specific position when the bogie travels due to the past slab weight for this input data as output data. A slab weight sinking amount calculation model generated by machine learning learning data and a bogie speed change amount during past bogie running are used as input data, and the past bogie speed change amount for this input data It is composed of a bogie speed change subsidence amount calculation model generated by machine learning a plurality of learning data whose output data is the amount of rail subsidence at a specific position,
Inputting the measured slab weight when the truck is running into the slab weight sinking amount calculation model, the rail sinking amount at a specific position when the truck is traveling due to the slab weight is calculated, and the bogie speed change sinking amount calculation model. , input the measured amount of change in bogie speed when the bogie is running, and calculate the amount of rail sinking at a specific position when the bogie is running based on the amount of change in bogie speed,
Measured amount of rail subsidence at a specific position while the bogie is running, rail subsidence at a specific position when the bogie is running due to the calculated slab weight, and rail subsidence at a specific position when the bogie is running due to the calculated amount of change in bogie speed 6. The amount of rail subsidence at a specific position due to aging while the bogie is running, excluding the influence of fluctuation factors other than the aging, is calculated from method of predicting the amount of rail subduction in Influencing factors affecting the aging are the production amount and rainfall of the slab,
The aging rail subsidence amount prediction model is based on the amount of rail subsidence at a specific position when the bogie is running due to past secular change, and the amount of change in the past slab production amount and rainfall during a predetermined period from when the bogie was running. The amount of change in the amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie was running due to a change is used as input data. Generated by machine learning a plurality of learning data with the predicted value of the amount as the output data,
In the model for predicting the amount of rail subsidence over time, the amount of rail subsidence at a specific position during bogie running due to secular change excluding the effects of fluctuation factors other than aging, and the amount of slab production and rainfall from when the bogie was running. By inputting the amount of change in the amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie was running due to aging based on the amount of change over a predetermined period, the rail at a specific position after a predetermined period has passed since the bogie was running due to aging. 7. The method for predicting the amount of rail subsidence according to claim 6 , wherein a predicted value of the amount of subsidence is predicted. 8. The rail subsidence amount prediction method according to any one of claims 1 to 7, wherein the secular change rail subsidence amount prediction model is a model constructed by multiple regression analysis or a neural network. . The rail subduction amount prediction method according to any one of claims 2 to 7 , wherein the variation factor subsidence amount calculation model is a model constructed by multiple regression analysis or a neural network. According to the predicted value of the amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie traveled due to aging, which is predicted by the method for predicting the amount of rail subsidence according to any one of claims 1 to 9, A rail repair timing determination method for determining the rail repair timing. A rail sinking amount predicting device for predicting the sinking amount of a rail on which a carriage for conveying an object to be conveyed runs,
A predetermined period of time has passed since the bogie was running due to aging, based on the amount of rail subsidence at a specific position when the bogie is running due to past secular changes, and the amount of change in the influence factor that affects the past secular change during a predetermined period from when the bogie is running. Input data is the amount of change in the amount of rail subsidence at a specific position later, and output data is the predicted value of the amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie was running due to past secular changes for this input data. A predictive model for the amount of rail subsidence over time generated by machine learning from multiple training data is applied to the amount of rail subsidence at a specific position during bogie travel due to aging, excluding the effects of fluctuation factors other than aging, and Input the amount of change in rail subsidence at a specific position after a predetermined period of time from when the bogie is running due to aging based on the amount of change in the factors that affect aging over a predetermined period from when the bogie is running. A rail subsidence amount prediction device, comprising: a secular change rail subsidence amount prediction unit that predicts a predicted value of the rail subsidence amount at a specific position after a predetermined period of time has elapsed since the bogie was running. Input data is variable factors other than aging while the bogie is running in the past, and multiple learning data are output data that is the amount of rail subsidence at a specific position while the bogie is running due to fluctuation factors other than the past aging for this input data. is input into a model for calculating the amount of subsidence due to variation factors generated by machine learning. a variation factor rail sinking amount calculation unit that calculates the amount of sinking;
Aging change excluding the effects of fluctuation factors other than aging from the measured amount of rail subsidence at a specific position while the bogie is running and the calculated amount of rail subsidence at a specific position while the bogie is running due to fluctuation factors other than aging. 12. The rail subsidence amount predicting device according to claim 11, further comprising a rail subsidence amount calculation unit that calculates a rail subsidence amount at a specific position during running of the bogie by means of a secular change rail subsidence amount calculation unit. 13. The rail according to claim 12 , further comprising a determination unit that determines whether or not a predicted value of the rail subsidence amount at a specific position after a predetermined period of time has passed since the truck traveled due to aging is equal to or greater than a predetermined threshold value. A device for predicting the amount of subduction. The determination unit changes the predetermined period when the predicted value of the rail subsidence amount at the specific position after the lapse of the predetermined period from the time when the bogie travels due to aging is not equal to or greater than a predetermined threshold, and changes the predetermined period. The changed rail subsidence amount prediction unit again predicts the predicted value of the rail subsidence amount due to secular change at the specific position after the elapse of the changed predetermined period, and the predicted value is determined to be equal to or greater than the predetermined threshold. 14. A prediction device for a rail subduction amount according to claim 13, wherein said change, prediction, and determination are repeated up to. When the determination result of the determination unit is that the predicted value of the rail subsidence amount at a specific position after a predetermined period of time has elapsed since the truck traveled due to aging or after a changed predetermined period of time has elapsed is a predetermined threshold value or more, 15. The rail subsidence amount prediction device according to claim 14, further comprising a determination result output unit for determining that the rail is abnormal. Variation factors other than the aging are the weight of the slab as the object to be conveyed and the speed change amount of the truck,
The variation factor subsidence amount calculation model uses the slab weight when the bogie has traveled in the past as input data, and the rail subduction amount at a specific position when the bogie travels due to the past slab weight for this input data as output data. A slab weight sinking amount calculation model generated by machine learning learning data and a bogie speed change amount during past bogie running are used as input data, and the past bogie speed change amount for this input data It is composed of a bogie speed change subsidence amount calculation model generated by machine learning a plurality of learning data whose output data is the amount of rail subsidence at a specific position,
The variation factor rail subsidence amount calculation unit inputs the measured slab weight when the bogie is traveling to the slab weight subduction amount calculation model, and calculates the rail subduction amount at a specific position when the bogie is traveling due to the slab weight. In addition, inputting the bogie speed variation measured when the bogie is running into the bogie speed change subsidence amount calculation model, and calculating the rail subsidence amount at a specific position when the bogie is running based on the bogie speed variation,
The aging rail subsidence amount calculation unit calculates the measured rail subsidence amount at a specific position while the bogie is traveling, the rail subsidence amount at a specific position while the bogie is traveling due to the calculated slab weight, and the calculated bogie speed change amount. A rail subsidence amount at a specific position when the bogie is running is calculated from the rail subsidence amount at the specific position when the bogie is running due to aging, excluding the effects of fluctuation factors other than the aging. 16. The rail subduction amount prediction device according to any one of 12 to 15. Influencing factors affecting the aging are the production amount and rainfall of the slab,
The aging rail subsidence amount prediction model is based on the amount of rail subsidence at a specific position when the bogie is running due to past secular change, and the amount of change in the past slab production amount and rainfall during a predetermined period from when the bogie was running. The amount of change in the amount of rail subsidence at a specific position after a predetermined period of time has passed since the bogie was running due to a change is used as input data. Generated by machine learning a plurality of learning data with the predicted value of the amount as the output data,
The secular change rail subsidence amount prediction unit adds to the secular change rail subsidence amount prediction model the rail subsidence amount at a specific position due to secular change excluding the influence of fluctuation factors other than secular change, and the slab Input the amount of change in rail subsidence at a specific position after a predetermined period of time from when the bogie is running due to aging based on the amount of change in the amount of production and rainfall during the predetermined period from when the bogie is running. 17. A predicting device for the amount of rail subsidence according to claim 16 , which predicts a predicted value of the amount of rail subsidence at a specific position after a predetermined period of time has elapsed. 18. The rail subsidence amount prediction device according to any one of claims 11 to 17, wherein the rail subsidence amount prediction model over time is a model constructed by multiple regression analysis or a neural network. . The rail subduction amount prediction device according to any one of claims 12 to 17 , wherein the variation factor subduction amount calculation model is a model constructed by multiple regression analysis or a neural network. A rail sinking amount prediction method for predicting the rail sinking amount on which a truck for conveying an object travels, comprising:
Data including past data of the rail subduction amount y(t, p) due to aging of the prediction target point during running of the bogie and past data of the aging change factor x, which is a factor affecting the aging change, A model for predicting the amount of rail subsidence over time generated by machine learning includes the amount of rail subsidence y(t, p) due to the change over time at the prediction target point when the bogie is running, and the aging change factor x at the future prediction target point. and predicting the amount of rail subsidence y(t+k, p) due to secular changes at a prediction target point at a future prediction target time. 21. The method for predicting the amount of rail subsidence according to claim 20, wherein the model for predicting the amount of rail subsidence over time is a model constructed by multiple regression analysis or a neural network. The aging factor x is selected from the rainfall amount x1 from the start of the subduction amount measurement, the total weight of the passing goods x2 from the start of the subduction amount measurement, and the factor x3 correlated with the total weight of the passing goods x2. 22. The method for predicting the amount of rail subduction according to claim 20 or 21, wherein one or more types are included as influencing factors. The amount of rail subsidence y(t, p) due to aging at the prediction target point when the truck travels is obtained by measuring the height of the truck when the truck travels through the prediction target point. Variation factor subsidence amount, which is an estimated value of the rail subsidence amount due to variation factors other than secular change while the bogie is running at the prediction target point, from the actual measured value Z (t, p) of the rail subsidence amount when the bogie is running at the point 23. The method for predicting the amount of rail subsidence according to any one of claims 20 to 22, wherein it is obtained by subtracting h(t, p). 24. The method of predicting rail subsidence according to claim 23, wherein the variation factor subsidence amount h(t, p) is estimated based on a model constructed by multiple regression analysis or a neural network. The variation factor sinking amount h(t, p) is estimated based on the weight n1 of the object to be conveyed and/or the amount of change n2 in the speed of the truck when the truck travels through the prediction target point. 25. A method for predicting a rail subduction amount according to Item 23 or 24. 26. It is determined whether or not the rail subduction amount y(t+k, p) due to secular change of the prediction target point at the future prediction target time is equal to or greater than a predetermined threshold set in advance. The method for predicting the amount of rail subduction according to any one of the above. When the determination result is that the amount of rail subsidence y(t+k, p) due to secular change at the prediction target point at the future prediction target point is not equal to or greater than a predetermined threshold value, the prediction target time is changed, and the prediction target point is changed. The rail subsidence amount y(t+k, p) due to secular change is predicted again, and the change is made until the rail subsidence amount y(t+k, p), which is a predicted value, becomes equal to or greater than the predetermined threshold value. 27. The method for predicting the amount of rail subsidence according to claim 26, wherein , prediction, and determination are repeated. 28. The method of predicting the amount of rail subsidence according to claim 27, wherein the rail is determined to be abnormal when the determination result is equal to or greater than the predetermined threshold. Prediction of rail subsidence amount y(t+k, p) due to secular change of a prediction target point at a future prediction target time by the rail subduction amount prediction method according to any one of claims 20 to 25. is performed for a plurality of prediction target time points, and a prediction target time point at which the prediction result is equal to or greater than a predetermined threshold value is determined as a time of an abnormal rail subsidence state. judgment method. A rail sinking amount predicting device for predicting the sinking amount of a rail on which a carriage for conveying an object to be conveyed runs,
Data including past data of the rail subduction amount y(t, p) due to aging of the prediction target point during running of the bogie and past data of the aging change factor x, which is a factor affecting the aging change, A model for predicting the amount of rail subsidence over time generated by machine learning includes the amount of rail subsidence y(t, p) due to the change over time at the prediction target point when the bogie is running, and the aging change factor x at the future prediction target point. and a forecasted value xf of and predicts the rail subsidence amount y(t+k, p) due to secular change at the prediction target point at a future prediction target time. A prediction device for the amount of rail subduction characterized by 31. The rail subsidence amount prediction device according to claim 30, wherein the secular change rail subsidence amount prediction model is a model constructed by multiple regression analysis or a neural network. The aging factor x is selected from the rainfall amount x1 from the start of the subduction amount measurement, the total weight of the passing goods x2 from the start of the subduction amount measurement, and the factor x3 correlated with the total weight of the passing goods x2. 32. The rail subduction amount prediction device according to claim 30 or 31, wherein one or more types are included as influencing factors. The amount of rail subsidence y(t, p) due to aging at the prediction target point when the truck travels is obtained by measuring the height of the truck when the truck travels through the prediction target point. Variation factor subsidence amount, which is an estimated value of the rail subsidence amount due to variation factors other than secular change while the bogie is running at the prediction target point, from the actual measured value Z (t, p) of the rail subsidence amount when the bogie is running at the point 33. The apparatus for predicting the amount of rail subsidence according to any one of claims 30 to 32, which is obtained by subtracting h(t, p). 34. The rail subsidence amount prediction device according to claim 33, wherein the variation factor subsidence amount h(t, p) is estimated based on a model constructed by multiple regression analysis or a neural network. The variation factor sinking amount h(t, p) is estimated based on the weight n1 of the object to be conveyed and/or the amount of change n2 in the speed of the truck when the truck travels through the prediction target point. 35. A rail subduction amount prediction device according to Item 33 or 34.

Images