Railway maintenance support system, railway maintenance support method

Abstract

To provide a railroad maintenance support system, and a railroad maintenance support method which can be formed at a lower cost, while considering relationship of maintenance of each facility and apparatus, with respect to a railroad in which two or more facilities and devices operate in association with each other.SOLUTION: In a railroad maintenance support system, a processor stored in a processing unit 113 estimates a deterioration prediction model 110 used for predicting deterioration of a railroad facility in which maintenance accuracy which is information about accuracy of maintenance of the railroad facility obtained by converting sensing data acquired from a sensor device of the railroad facility is set as an explanation variable for each of the plurality of railroad facilities, and stores the model in a storage unit (main storage (for example, a memory)) and an auxiliary storage (for example, an HDD)). In estimation of the deterioration prediction model, the processor performs processing which estimates the deterioration prediction model of a second railroad facility in which deterioration prediction of the deterioration prediction model of a first railroad facility already estimated and stored in the storage unit is included in the explanation variable, and outputs a whole maintenance cost due to the maintenance based on the deterioration prediction of the deterioration prediction model of each railroad facility.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a railway maintenance support system and a railway maintenance support method. [Background technology]

[0002] Social infrastructure such as electricity and railways has been improving year by year to enable people to live more prosperously. For example, when railways were first introduced, they consisted of locomotives and passenger cars. However, over time, small power units have been installed in passenger cars to allow for the provision of power, and lighting, air conditioning, and other equipment have been incorporated to improve passenger comfort. Tracks have also become increasingly equipped with numerous devices, including signals to ensure train safety and automatic train control (ATC) to limit speeds. However, as the number of devices increases, maintenance is inevitably required, increasing maintenance costs. To reduce maintenance costs, infrastructure operators are beginning to shift from traditional manual infrastructure inspection and repair to remote systems utilizing the Internet of Things (IoT), and from time-based maintenance (TBM) to condition-based maintenance (CBM) using sensor information. The following inventions are examples of methods for implementing CBM. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2017-16509 Summary of the Invention [Problem to be solved by the invention]

[0004] Patent Document 1 provides a means for collecting sensor information from equipment, estimating and predicting the state of deterioration and future deterioration, clarifying the relationship between the predicted state of deterioration and the occurrence of abnormal events in the equipment, and then estimating future costs.The means represented by Patent Document 1 is thought to be particularly effective for independently operating devices and equipment.

[0005] However, railways, in particular, are systems of systems that function normally only when multiple devices and equipment are involved, and an abnormality in one device can affect others. For example, for track maintenance, maintenance standards are set for track irregularities (gauge, alignment, elevation, level, and flatness). However, the probability of a malfunction is expected to increase when vehicles run on tracks with significant track irregularities. Despite railways being a system of systems, maintenance has traditionally been based on individual maintenance standards for each piece of equipment and device without considering the relationships between them. For example, there are the aforementioned track maintenance standards for tracks, vehicle maintenance standards for rolling stock, and similarly, overhead lines. This management approach often results in excessive maintenance because maintenance is performed without consideration of the condition of the other devices and equipment. In recent years, the Japanese railway industry, particularly in Japan, has seen a decline in railway ridership due to population decline. Railway companies are being called upon to reduce maintenance costs and increase corporate sustainability. However, the traditional maintenance approach based on individual maintenance standards for each piece of equipment and device is expected to reach its limits.

[0006] The objective of this invention is to derive a lower-cost maintenance method for railways, which is an example of a System of Systems, while taking into account the relationships between the maintenance of each piece of equipment and device, rather than the conventional maintenance method for each piece of equipment and device. [Means for solving the problem]

[0007] According to a first aspect of the present invention, there is provided a railway maintenance support system as described below. That is, the railway maintenance support system includes a processor and a storage unit. The processor estimates a deterioration prediction model for each of a plurality of railway facilities to be used to predict deterioration of the railway facilities, the explanatory variable being maintenance accuracy, which is information related to the accuracy of maintenance of the railway facilities obtained by converting sensing data acquired from sensor devices on the railway facilities, and stores the model in the storage unit. In estimating the deterioration prediction model, the processor performs processing to estimate a deterioration prediction model for a second railway facility that includes, as an explanatory variable, the deterioration prediction of a deterioration prediction model for a first railway facility that has already been estimated and stored in the storage unit. The processor outputs the total maintenance cost for maintenance based on the deterioration prediction of the deterioration prediction model for each railway facility.

[0008] According to a second aspect of the present invention, there is provided the following railway maintenance support system. That is, the railway maintenance support system includes a processor and a storage unit. The processor estimates a failure prediction model used to predict failures of railway equipment for each of a plurality of railway equipment and stores the model in the storage unit. In estimating the failure prediction model, the processor performs processing to estimate a failure prediction model for a second railway equipment that includes, as an explanatory variable, the failure prediction of the failure prediction model for a first railway equipment that has already been estimated and stored in the storage unit. The processor outputs the total maintenance cost for maintenance based on the failure predictions of the failure prediction model for each railway equipment.

[0009] According to a third aspect of the present invention, there is provided the following railway maintenance support method. That is, the railway maintenance support method is a method performed using a railway maintenance support system having a processor and a storage unit. The processor inputs maintenance accuracy, which is information regarding the accuracy of maintenance of railway equipment, and estimates a deterioration prediction model for each of multiple railway equipment pieces to be used to predict deterioration of the railway equipment, using the maintenance accuracy as an explanatory variable, and stores the model in the storage unit. The processor outputs the overall maintenance cost for maintenance based on the deterioration prediction of the deterioration prediction model for each of the railway equipment pieces. When estimating the deterioration prediction model for each of the multiple railway equipment pieces, the processor performs processing to estimate a deterioration prediction model for a second railway equipment piece that includes, as an explanatory variable, the deterioration prediction of a deterioration prediction model for a first railway equipment piece that has already been estimated and stored in the storage unit. [Effects of the Invention]

[0010] According to the present invention, it is possible to derive a lower-cost maintenance method for a railway, which is an example of a system of systems, while taking into consideration the relationship between the maintenance of each facility and device. Note that problems, configurations, and effects other than those described above will become clear from the following description of the embodiment of the invention. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a configuration diagram showing an example of a railway maintenance decision support system according to a first embodiment. [Figure 2] 10 is a flowchart showing an example of a process for estimating a deterioration prediction model and determining the maintenance accuracy when the maintenance cost is minimized. [Figure 3A] FIG. 10 is a diagram showing an example of input data regarding a method for determining maintenance accuracy at each time point. [Figure 3B] FIG. 10 is a diagram showing an example of output data regarding a method for determining maintenance accuracy at each time point. [Figure 4] FIG. 10 is a diagram showing an example of a data structure in which measurement values ​​of railway facilities are registered. [Figure 5] FIG. 4 is a diagram showing an example of environmental information. [Figure 6] FIG. 4 is a diagram showing an example of transportation information. [Figure 7] 10 is a flowchart showing an example of a method for calculating maintenance costs. [Figure 8] FIG. 10 is a diagram showing an example of a data structure representing maintenance costs. [Figure 9] FIG. 10 is a diagram showing an example of output when the total maintenance cost is lowest. [Figure 10] FIG. 10 is a configuration diagram showing an example of a railway maintenance decision support system according to a second embodiment. [Figure 11] 10 is a flowchart showing an example of a process for estimating a failure prediction model and determining the maintenance accuracy when the maintenance cost is minimized. [Figure 12] FIG. 10 is a diagram showing a specific example of failure information used when calculating each coefficient. [Figure 13]FIG. 11 is a diagram showing an example of a data structure of a counting result according to the third embodiment. [Figure 14] FIG. 10 is a diagram showing an example of a comparative display of maintenance costs. [Figure 15] FIG. 10 is a configuration diagram showing an example of a railway maintenance decision support system according to a fourth embodiment. [Figure 16] 10 is a flowchart showing an example of a process for determining model estimation in the case of maintenance that handles both deterioration and failure, and maintenance accuracy for each piece of equipment when maintenance costs are minimized. [Figure 17] FIG. 1 is a diagram showing an example of a configuration made up of multiple facilities and devices. [Figure 18] 10 is a flowchart showing an example of model estimation for each railway facility configured based on the configuration example. [Figure 19] This is an activity diagram showing an example of on-site railway maintenance. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Parts with the same reference numerals represent the same objects, and the basic configurations and operations are similar. The embodiment is an example for explaining the present invention, and for clarity of explanation, some details have been omitted or simplified as appropriate. The present invention can also be implemented in various other forms. Examples of various types of information may be described using expressions such as "table," "list," and "queue," but the various types of information may also be expressed using data structures other than these. For example, various types of information such as "XX table," "XX list," and "XX queue" may also be expressed as "XX information." When describing identification information, expressions such as "identification information," "identifier," "name," "ID," and "number" are used, but these are interchangeable. In the embodiment, the target is facilities and equipment that have a maintenance relationship, such as rails and wheels, or air brakes and electric brakes, and a support system is described that presents more appropriate maintenance methods based on the maintenance relationships between multiple facilities and equipment.

[0013] First Embodiment The first embodiment relates to a railway maintenance decision support system that estimates a model that enables deterioration prediction based on sensor information sent from railway facilities and equipment, and further estimates the maintenance relationships between facilities, etc., thereby presenting the maintenance accuracy of each facility and device that can minimize overall maintenance costs. Figure 1 is a configuration diagram showing an example of a railway maintenance decision support system.

[0014] A railway maintenance decision support system (railway maintenance support system) is composed of a computer device equipped with information processing resources such as an input interface, memory devices (main memory and auxiliary memory), a processor (e.g., CPU), a display device (e.g., LCD display), a communication unit, and a bus that connects these together.

[0015] The input interface constitutes the input unit 101. When a maintenance manager issues a processing execution instruction, a deterioration prediction model is estimated and a search calculation for the maintenance accuracy that minimizes the overall cost is performed. The storage device (storage unit) uses a main storage device (e.g., memory) and an auxiliary storage device (e.g., HDD), and the main storage device stores programs that realize each function (deterioration prediction model estimation unit 102, deterioration prediction unit 103, optimum search unit 104, result creation unit 105). The auxiliary storage device stores various information to be used (sensing data 106, environmental information 107, transportation information 108, deterioration information 109), as well as the processing results of the programs, such as a deterioration prediction model 110, a maintenance cost prediction 111, and maintenance accuracy 112. The main storage device may also store a program that determines maintenance accuracy, a program that calculates maintenance costs, etc. The auxiliary storage device stores programs executed by the processor.

[0016] The processor is stored in the processing unit 113 and functions as an arithmetic execution unit that executes the processing of the program, thereby enabling various functions to be performed. The display device corresponds to the output unit 114, and allows the item setting status and processing results to be confirmed.

[0017] The communication unit 115 is configured to include an interface used for communication. The processing unit 113 sends and receives sensor information sent from each railway facility and device to be managed that is external to the railway maintenance decision support system via the communication unit 115, and stores the information in an auxiliary storage device. The bus connects the processor, input interface, storage device, display device, and interface, and contributes to the realization of functions by passing information between them.

[0018] The sensor information is sent to the communication unit 115 via the network. For example, the results of sensing by a sensor device 117 mounted on a railway vehicle 116 are sent to the communication unit 115 via the network. Similarly, sensing data sensed by a sensor device 119 that observes track-related equipment 118 is sent to the communication unit 115. Furthermore, sensing is also performed on electrical circuit-related equipment 120 by a sensor device 121, and the sensing data is sent to the communication unit 115. Note that the number of facilities and devices to be managed is arbitrary and is not limited to the numbers shown in this figure. At a minimum, two or more related facilities or devices are required.

[0019] Next, with reference to FIG. 2, an example of a process for estimating a deterioration prediction model in a railway maintenance decision support system, estimating total maintenance costs, and determining the maintenance accuracy of each piece of equipment that minimizes the maintenance cost will be described. FIG. 2 is a flowchart showing an example of a process for estimating a deterioration prediction model and determining maintenance accuracy. In this description, there are two railway equipment A and railway equipment B as maintenance management targets, and the deterioration of railway equipment B is affected by the maintenance accuracy of railway equipment A. In other words, railway equipment A and railway equipment B have a maintenance relationship. Note that in the present invention, it is sufficient for there to be two or more pieces of target equipment, but additional processing for the case where there are three or more pieces of target equipment will be described later.

[0020] First, the processing unit 113 determines the maintenance accuracy of target facility A (railway facility A) and target facility B (railway facility B) (201). When determining the maintenance accuracy, the processing unit 113 executes a program to calculate the maintenance accuracy. There are two methods for this determination: a period-based determination method and a point-in-time determination method, but both can be processed in the same way. Here, a specific example of the point-in-time determination method will be shown and explained. Here, FIG. 3A shows an example of input data for the method for determining the maintenance accuracy at each point in time.

[0021] In Figure 3A, equipment ID 301 is an ID indicating which equipment it refers to, and measurement ID 302 indicates the measurement item for that equipment. Maintenance date and time 303 indicates the maintenance date and time, i.e., the date and time of inspection or repair. Measurement value 304 indicates the value measured for measurement ID 301. Inspection standard 305 indicates the inspection standard at the time of measurement. The 160-170 in the first line of this figure means that if the measurement value is between 160 and 170, the inspection standard is met. Next, Figure 3B shows an example of the processing result from a diagram illustrating a method for determining maintenance accuracy. Note that the same numbers are used for the same items as in Figure 3A, so their explanation will be omitted. Required accuracy median 306 indicates the median value of the required accuracy. For example, if inspection standard 305 is 160-170, the required accuracy median is determined to be (160 + 170) / 2 = 165. Required maintenance accuracy 307 indicates the deviation based on inspection standard 305 being 1 (1 if it is the same value as the required maintenance accuracy). There are several possible calculation methods, but for example, it can be calculated as required maintenance accuracy = (required accuracy median - (upper limit of inspection standard - lower limit of inspection standard) / 2) / required accuracy median. Maintenance accuracy 308 indicates the actual maintenance accuracy. For example, the calculation method is maintenance accuracy = 1 - (required accuracy median - measured value) / required accuracy median. Thus, if the required maintenance accuracy is, for example, 0.95, this means that a maintenance accuracy between 1 and 0.95 obtained from the measurement results using the measurement ID is considered normal. This maintenance accuracy is used to perform the processing from 202 onwards in Figure 2.

[0022] Returning to the explanation of Fig. 2, next, the processing unit 113 estimates a deterioration prediction model for railway facility A (202). In Fig. 1, this process corresponds to the deterioration prediction model estimation unit 102. Examples of methods for estimating a deterioration prediction model include a regression estimation method using a linear model and an estimation method using machine learning. In this explanation, regression estimation using a linear model will be used as an example, but the estimation method for a deterioration prediction model is not limited to this.

[0023] When estimating a linear model, the explanatory variables (x1, x2, x n ) candidates and a response variable y. The response variable y is a measurement value that indicates the deterioration to be predicted. Possible explanatory variable candidates include other measurement values ​​of railway equipment A and environmental information around the facility (temperature, humidity, weather, rainfall, amount of solar radiation, time of day, etc.), but one of the features of the present invention is that the maintenance accuracy of facility A is included in the explanatory variables. Here, an example of a data structure in which measurement values ​​of railway equipment are registered will be described using FIG. 4. Also, an example of environmental information will be described using FIG. 5.

[0024] 4, equipment ID 401 indicates the ID of the equipment, measurement ID 402 indicates the ID of the measurement item for that equipment, measurement date and time 403 is the date and time of measurement, and measurement value 404 indicates the measurement value.

[0025] 5 shows an example of environmental information, where date / time 501 indicates the date and time, and temperature 502 indicates the temperature at that date and time. Similarly, humidity 503 indicates humidity, precipitation 504 indicates the amount of rainfall, and solar radiation 505 indicates the amount of solar radiation. Of course, when traveling a long distance by train, this environmental information may be prepared for each section where the environmental conditions may change. In addition, in this embodiment, regression estimation is performed by adding the maintenance accuracy shown in FIG. 3B.

[0026] Multiple regression estimation is performed from the combination of these explanatory variable candidates and the target variable. During multiple regression estimation, unnecessary explanatory variables are deleted, but processing is performed in such a way that explanatory variables for maintenance accuracy are not deleted when deleting variables. This results in a degradation prediction model being estimated as a result of regression estimation. The estimated regression equation is the following formula.

[0027]

number

[0028] where y A is the deterioration prediction for equipment A, v A is the maintenance accuracy of equipment A, w1, w2,...w k are the explanatory variables (k≦n) remaining from the candidate explanatory variables, and t is the elapsed time. This is the deterioration prediction model for railway equipment A. Note that these explanatory variables w1, w2,...w k The variables in the table may be explanatory variables that take into account transportation information.

[0029] The transportation information will be explained with reference to Figure 6. Figure 6 shows an example of transportation information. Train number 601 indicates the train number, and is an ID that identifies which train it is. Date and time 602 indicates the date and time. Running line 603 indicates the running line, and indicates which line the train is running on. Running position 604 indicates the running position, and is expressed, for example, as the cumulative distance from the starting station. Speed ​​605 indicates the speed, and acceleration 606 indicates the acceleration. Running weight 607 indicates the running weight. When transportation information is taken into account, for example, if the deterioration effect on railway equipment A only acts when a train is passing, a certain explanatory variable w i may be treated as taking the values ​​shown below.

[0030]

number

[0031] Here, when using the data structure shown in Figure 6, for example, if data exists for the relevant train number, date and time, running route and running position, it will be 1, and if not, it will be 0. Alternatively, if it is due to the number of train passes, it can be done as follows.

[0032]

number

[0033] Here, num is the number of passes up to the elapsed time t. Similarly, when using the data structure of Figure 6, for example, the number of passes is the number of data items that exist for the relevant train number, date and time, running route and running position. Alternatively, the value of the transport information data can be used. For example, w i (t) may be the value of the running weight for the relevant train number, date and time, running route, and running position. This allows the influence of running weight to be taken into account in the deterioration. Note that the explanatory variables in Equation 4, Equation 6, Equation 7, etc., which will be explained below, may also be explanatory variables that take transportation information into account.

[0034] Next, the processing unit 113 executes the deterioration prediction model estimation unit 102 to estimate a deterioration prediction model for railway facility B while adding the deterioration information for railway facility A (203). Here again, regression estimation of a linear model is used as an example, but the estimation method is not limited to this. The deterioration prediction model for railway facility B is also estimated in the same way, but estimation is performed by adding the deterioration information for railway facility A as the first explanatory variable candidate. The regression equation estimated in this way is the following formula.

[0035]

number

[0036] where y B is the deterioration prediction for equipment B, y A (t) is the deterioration prediction for equipment A at elapsed time t, v B is the maintenance accuracy of equipment B, w1, w2,...w jis the explanatory variable remaining from the candidate explanatory variables (j≦n), and t is the elapsed time. This is the deterioration prediction model for railway equipment B.

[0037] Next, using the estimated deterioration prediction models for railway facility A and railway facility B, the maintenance costs are predicted when the maintenance accuracy of railway facility A and railway facility B is changed. Loop 204 indicates the start of the loop when changing the maintenance accuracy of railway facility A, and loop 205 indicates the start of the loop when changing the maintenance accuracy of railway facility B. Loop 206 indicates the start of the loop of time change in the simulation. In this embodiment, the maintenance accuracy of railway facility A is changed to 1, 2...k, and the maintenance accuracy of railway facility B is changed to 1, 2...j, and the maintenance cost is calculated at the set maintenance accuracy.

[0038] First, the processing unit 113 performs a deterioration prediction for railway facility A (207). In Fig. 1, this process corresponds to the deterioration prediction unit 103. When predicting the deterioration of railway facility A, the deterioration prediction model of Equation 1 is used, and the values ​​of each explanatory variable at time t are acquired and set from the sensing data and environmental information in the auxiliary storage device. The maintenance accuracy is set using the maintenance accuracy set at the start of loop 204. Once the values ​​of each explanatory variable have been set, Equation 1 is used to calculate y A This prediction result is y A Let (t).

[0039] Next, the processing unit 113 performs deterioration prediction for the railway facility B (208). In FIG. 1, this process corresponds to the deterioration prediction unit 103. As in the case of the deterioration prediction for the railway facility A, the deterioration prediction model of Equation 4 is used, and the values ​​of each explanatory variable at time t are obtained from the sensing data and environmental information in the auxiliary storage device, and in addition, the y obtained in 207 is used. A (t) is used to set the maintenance accuracy. The maintenance accuracy set at the start of loop 204 and the start of loop 205 is used to set the maintenance accuracy. After the values ​​of each explanatory variable are set, use Equation 4 to set y B This prediction result is y B Let (t).

[0040] Next, the processing unit 113 predicts the maintenance cost of railway facility A (209). The processing unit 113 executes a program for calculating the maintenance cost. There are various ways to calculate the maintenance cost of railway facility A depending on the facility and the maintenance method, but here, the maintenance cost will be explained as three costs: inspection cost, repair cost when an abnormality is found during inspection, and restoration cost when deterioration leads to a breakdown during transportation operations.

[0041] FIG. 7 shows an example of a method for calculating maintenance costs. First, the processing unit 113 determines whether deterioration has led to a failure between time (t-1) and time t (701). One possible method for determining this is to determine whether the amount of deterioration of the target railway equipment exceeds a preset threshold. If no failure has occurred, the process proceeds to 703. If a failure has occurred, the processing unit 113 adds up the restoration cost (702). The restoration cost is calculated based on preset values, including the cost of removing the target equipment and the loss of opportunity to use the equipment until it is removed.

[0042] Next, the processing unit 113 determines whether an inspection period is included between time (t-1) and time t (703). The inspection period indicates when the railway equipment is inspected, for example, once a month, and if there is an inspection period between time (t-1) and time t, it is assumed that the inspection has been carried out. If the inspection period is not included, the processing proceeds to 708. If the inspection period is included, the processing unit 113 adds the inspection cost (704). The inspection cost is also a preset value, and the addition is carried out according to that value.

[0043] Next, the processing unit 113 determines whether any abnormalities were found during the inspection (705). If no abnormalities are found, the process proceeds to 708. If any abnormalities are found, it is assumed that repairs have been made, and the degradation is changed to no degradation (706). When handling no degradation using a degradation prediction model, it is possible to achieve a no degradation state by, for example, resetting the elapsed time t to zero. Next, the processing unit 113 adds the repair cost (707). The repair cost is also a preset value, and additions are made according to that value. Finally, the restoration cost, inspection cost, and repair cost are added together to arrive at the maintenance cost (708).

[0044] Returning to the explanation of Figure 2, the processing unit 113 similarly predicts the maintenance cost of railway facility B (210). The processing unit 113 executes a program for calculating the maintenance cost. The specific processing method is the same as in 209, and the cost can be calculated using, for example, the maintenance cost calculation method in Figure 7.

[0045] Next, the processing unit 113 determines whether the time loop has ended (211), and if the maintenance cost has not been calculated by the end time, the time is updated and the process returns to 206. If the end time has been reached, the process proceeds to 212. In this way, the maintenance costs of railway facility A and railway facility B when the maintenance accuracy of railway facility A is k and the maintenance accuracy of railway facility B is j are calculated (212). The calculation method is to add up the maintenance costs of railway facility A and railway facility B calculated at each time in 209 and 210 for each railway facility.

[0046] Next, processing unit 113 determines whether the loop for changing the maintenance accuracy of railway equipment B has ended (213), and if the end of the maintenance accuracy has not been reached, the maintenance accuracy is updated and the process returns to 205. Otherwise, the process proceeds to 214. Next, the maintenance cost of railway equipment A and the maintenance cost of railway equipment B for each maintenance accuracy (1, 2...j) of railway equipment B when the maintenance accuracy of railway equipment A is k are output (214). Next, processing unit 113 determines whether the loop for changing the maintenance accuracy of railway equipment A has ended (215), and if the end of the maintenance accuracy has not been reached, the maintenance accuracy is updated and the process returns to 204. Otherwise, the process proceeds to 216.

[0047] In this way, the maintenance costs of railway facility A and railway facility B for each maintenance accuracy of railway facility A and each maintenance accuracy of railway facility B are output (216). Figure 8 shows an example of a data structure that represents these maintenance costs.

[0048] Railway facility A maintenance accuracy 801 indicates the maintenance accuracy of railway facility A, and railway facility B maintenance accuracy 802 indicates the maintenance accuracy of railway facility B. Note that each maintenance accuracy is expressed as a summary of the error from a reference maintenance accuracy (for example, the maintenance accuracy set at the start of the loop), but maintenance accuracy may be expressed using other methods. Railway facility A inspection cost 803 indicates the inspection cost of railway facility A at the corresponding maintenance accuracy, and railway facility B inspection cost 804 indicates the inspection cost of railway facility B at the corresponding maintenance accuracy. Railway facility A repair cost 805 indicates the repair cost of railway facility A at the corresponding maintenance accuracy, and railway facility B repair cost 806 indicates the repair cost of railway facility B at the corresponding maintenance accuracy. Railway facility A restoration cost 807 indicates the restoration cost of railway facility A at the corresponding maintenance accuracy, and railway facility B restoration cost 808 indicates the restoration cost of railway facility B at the corresponding maintenance accuracy. Total maintenance cost 809 indicates the total maintenance cost. In this way, it is possible to grasp the total maintenance cost for each level of maintenance accuracy of railway equipment. If the error range includes multiple levels of maintenance accuracy, a representative maintenance accuracy value may be found from these levels of maintenance accuracy, and the values ​​of each cost (803 to 808) may be found based on this maintenance accuracy.

[0049] Next, a search is made for the maintenance accuracy that results in the lowest total maintenance cost (217). The search is performed by processing unit 113, and in Fig. 1, this process corresponds to the optimum search unit 104. When the maintenance cost data structure is as shown in Fig. 8, the search extracts the record that results in the lowest total maintenance cost 809, and by referencing the values ​​of maintenance accuracy 801 for railway facility A and maintenance accuracy 802 for railway facility B at that time, the desired combination of maintenance accuracy can be identified.

[0050] Finally, the processing unit 113 outputs (218) the maintenance accuracy of railway facility A and railway facility B when the total maintenance cost is lowest, and the processing is completed. In Figure 1, this process corresponds to the result creation unit 105. Figure 9 shows an example of output when the total maintenance cost is lowest. The maintenance accuracy output unit 901 for railway facility A shows the change in the maintenance accuracy of railway facility A. In this example, it shows the change when the current maintenance accuracy is below 10% error, and the maintenance accuracy is changed to below 3% error to below 17% error. Furthermore, the change in maintenance accuracy compared to the current state may be displayed on a scale to make it easier to understand. The numerical values ​​displayed in the maintenance accuracy output unit 901 for railway facility A, ranging from -200% to +200%, serve as the scale. Note that there are various ways to determine the scale, but one example is to calculate and output {current maintenance accuracy} / {proposed maintenance accuracy} × 100% (note that a negative value indicates a worsening of maintenance accuracy, and a positive value indicates an improvement of maintenance accuracy). The maintenance accuracy output section 902 for railway equipment B shows the change in the maintenance accuracy of railway equipment B, and like the maintenance accuracy output section 901 for railway equipment A, it shows the value of the maintenance accuracy itself and a scale showing the percentage change from the current state.

[0051] Pointer 903 indicates the current maintenance accuracy of railway facility A. Pointer 904 indicates the current maintenance accuracy of railway facility B. Pointer 905 indicates the maintenance cost after the change, and in this example indicates the maintenance accuracy of railway facility A that will minimize the overall maintenance cost. Pointer 906 indicates the maintenance cost after the change, and in this example indicates the maintenance cost of railway facility B that will minimize the overall maintenance cost. The pointers (903 to 906) can be in a form that appropriately uses symbols, letters, numbers, etc.

[0052] Each square in the table shows the overall maintenance cost for each maintenance accuracy of railway equipment A and railway equipment B, with the lowest cost areas shown in a heat map with a dark background and the highest cost areas in a white background. By presenting changes in maintenance costs in this way, a display that is easy for maintenance managers to understand can be made, and they can determine whether other combinations of maintenance accuracy are likely to be effective, making it possible to encourage changes to the current maintenance management method.

[0053] Maintenance accuracy change display 907 is a display example that specifically shows a change in maintenance accuracy that minimizes the overall maintenance cost. In this display example, the current maintenance accuracy specified by pointer 903 and pointer 904 and information on the maintenance accuracy at the maintenance cost that minimizes the overall cost specified by pointer 905 and pointer 906 are displayed. In this display example, the change result is output based on the scale of the percentage change from the current state in the maintenance accuracy output section 901 for railway facility A and the maintenance accuracy output section 902 for railway facility B, and the maintenance cost value.

[0054] Furthermore, the display example of the maintenance accuracy change display 907 may also present cases other than the maintenance accuracy that results in the minimum maintenance cost. The maintenance accuracy change display 907 may be any display that shows the overall maintenance cost change and the results of the maintenance accuracy change, and for example, when the maintenance manager uses the selection unit 908 to select an appropriate maintenance cost and clicks on a location that is not the minimum maintenance cost, information on the change in maintenance accuracy and maintenance cost from the current state to the location specified by the click may be displayed.

[0055] Furthermore, the present invention can be applied not only to the maintenance related to deterioration described above, but also to maintenance related to failures. Deterioration is a model that predicts the progression of deterioration for a specific type of railway equipment, but because failures often occur probabilistically, predictions are made by calculating the failure probability distribution function for a specific type of railway equipment. In the second embodiment below, the configuration and differences in the case of maintenance related to failures will be explained, and content that is the same as that already explained will be omitted.

[0056] Second Embodiment Fig. 10 shows an example of the system configuration of a railway maintenance decision support system for maintenance related to failures. The differences from deterioration include a failure prediction model estimation unit 1001, a failure prediction unit 1002, failure information 1003, and a failure prediction model 1004. The differences between each unit will be explained using Fig. 11, which shows an example of processing content.

[0057] Figure 11 is a flowchart showing an example of the process of estimating a failure prediction model in the railway maintenance decision support system in the case of maintenance related to a failure, estimating the total maintenance cost, and determining the maintenance accuracy of each piece of equipment that minimizes the maintenance cost. Note that the same processes are assigned the same numbers as in Figure 2. In the case of maintenance related to a failure, processing unit 113 estimates a failure prediction model for railway equipment A (1101). In Figure 10, this process corresponds to the failure prediction model estimation unit 1001. In the case of a failure, unlike deterioration prediction, changes do not occur gradually, but often occur suddenly.

[0058] Fig. 12 shows a specific example of failure information 1003 used when determining the coefficients of Equation 6, which will be explained later. Equipment type ID 1201 is an equipment type ID, and indicates, for example, the model of railway equipment A (assuming there is multiple data for equipment identical to railway equipment A). Measurement period 1202 indicates the measurement period, and number of failures 1203 indicates the number of failures for the equipment type ID that occurred during that measurement period. In this way, in the case of failures, in order to handle the occurrence of events probabilistically, model prediction is performed using data on the number of occurrences for each type of railway equipment, rather than for specific railway equipment.

[0059] Weibull distribution estimation is often used in reliability engineering as one of the models for dealing with the occurrence of events probabilistically. This embodiment will also be explained using failure probability estimation based on the Weibull distribution. Generally, if the failure probability distribution function of a certain piece of equipment is assumed to be a Weibull distribution, it is given by Equation 5.

[0060]

number

[0061] Here, m is the Weibull coefficient and η is the scale. In the case of railway equipment, there is often little operation of ground equipment during times when trains are not running, and there are also many times when railway vehicles are not running (when they are stored in a garage, undergoing inspection, etc.), so in reality, the time of day when the equipment is actually operating and environmental factors often have an effect, rather than simply the time t. Therefore, the failure prediction model for railway equipment A is given by Equation 6, which is a partially modified Weibull distribution.

[0062]

number

[0063] where z A is the failure prediction model for railway equipment A, m is the Weibull coefficient, η is the scale, s A is the correction time, о1, о2, о k is A The explanatory variables are о1, о2, о. k For g, similar to the case of deterioration prediction, candidate explanatory variables are extracted from the sensing data 106 and environmental information 107. In addition, by using information on passing through a position where a failure may occur, transportation information may be taken into account as described above. There are various ways to give the formula g, but for example, it can be assumed to be a linear regression formula and then first calculated as s A Then, we use the cumulative hazard method or maximum likelihood method used in Weibull distribution estimation to calculate z A For example, based on the measurement period 1202 and number of failures 1203 of the failure information 1003, the actual measured value z A ^(t),t=t0,t1,t k At the same time, the predicted value z is calculated based on Eq. A (t), t=t0,t1, t k , and calculate the deviation from the difference. Then, move s in the direction that reduces the deviation. A By correcting the equation above, we can finally determine the above equation 6.

[0064] Next, processing unit 113 adds the failure information of railway facility A to estimate a failure prediction model for railway facility B (1102). In Fig. 10, this process corresponds to failure prediction model estimation unit 1001. The failure prediction model for railway facility B at this time is given by Equation 7.

[0065]

number

[0066] where z B is the failure prediction model for railway equipment B, z A (t) is z at time t A value, m is the Weibull coefficient, η is the scale, s B is the correction time, о1, о2, о k is B As with the deterioration prediction model, one of the features of this invention is that the explanatory variables include the failure probability of type A of railway equipment in the estimation. B , s B We will estimate the following.

[0067] The next difference from deterioration is the failure prediction (1103) of railway equipment A. Failure prediction is performed by the processing unit 113, which corresponds to the failure prediction unit 1002 in FIG. 10. In the case of failure prediction, since the probabilistic formula 6 is given as the failure probability distribution function, in this example, the failure probability z at this time t is once calculated. A After calculating (t), a random number between 0 and 1 is generated separately, and that random number is z A If the time t exceeds the threshold, it is predicted that a failure has occurred. Similarly, the processing unit 113 executes the failure prediction unit 1002 to predict a failure of the railway equipment B (1104). At this time, as in the failure prediction of the railway equipment A, in this example, the failure probability z B After calculating (t), a random number between 0 and 1 is generated separately, and that random number is z B If (t) is exceeded, it is predicted that a failure has occurred. After that, by performing the same process as for deterioration prediction, it is possible to calculate the maintenance cost that minimizes the overall cost for the failure.

[0068] Next, a third embodiment will be described. In the third embodiment, an example will be described in which the change in the maintenance cost of each railway facility over time is presented in an easy-to-understand manner. Note that a description of the same content as that already explained will be omitted.

[0069] <Third embodiment> FIG. 13 is a diagram showing an example of the data structure of another aggregation result in 216 of FIG. 2. It is basically an extension of FIG. 8, and the numbers in FIG. 8 are used for the same parts. Only the changes will be explained below. Elapsed period 1301 indicates the elapsed time, and corresponds to the time change t in the simulation described in 206 of FIG. 2. Accumulated total maintenance cost 1302 indicates the accumulated total maintenance cost, and is the cumulative total of total maintenance cost 709 up to the elapsed time t. In this way, data is organized for each elapsed period 1301.

[0070] Figure 14 shows an example of a comparative display of maintenance costs using the aggregation results of Figure 13. Current maintenance plan 1401 shows the change in cost for the current maintenance plan. The graph on the far left is a graph of the change in maintenance cost over time for railway facility A, with the horizontal axis representing elapsed time and the vertical axis representing maintenance cost. The horizontal and vertical axes of all graphs in Figure 14 below are the same. For the data, the elapsed period 1301 in Figure 13 is used as the value on the horizontal axis, and the total value of the codes (803, 805, 807) for records that match the current maintenance accuracy, accumulated up to the elapsed time, is used as the value on the vertical axis. The graph in the middle is a graph of the change in maintenance cost over time for railway facility B. For the data, similarly, the elapsed period 1301 in Figure 13 is used as the value on the horizontal axis, and the total value of the codes (804, 806, 808) for records that match the current maintenance accuracy, accumulated up to the elapsed time, is used as the value on the vertical axis. The graph on the far right shows the total maintenance costs for railway equipment A and railway equipment B.

[0071] Maintenance change proposal 1402 shows the cost changes of the maintenance change proposal. As with current maintenance proposal 1401, the graph on the far left shows the change in maintenance cost over time for railway facility A, the graph in the middle shows the change in maintenance cost over time for railway facility B, and the graph on the far right shows the total maintenance cost for railway facility A and railway facility B. The method of displaying the graphs using the information in Figure 13 is the same as for current maintenance proposal 1401.

[0072] Change effect 1403 shows a comparison graph of the change effect. The processing involves overlaying and displaying the total maintenance cost created in the current maintenance plan 1401 and the total maintenance cost created in the maintenance change plan 1402. By presenting the effect of changes in maintenance costs due to changes in maintenance accuracy in this way, it becomes possible for the maintenance manager to confirm whether the changes are meaningful (for example, whether they have led to significant cost reductions).

[0073] The present invention can also be applied to both deterioration and failure. By handling both deterioration and failure, it becomes possible to deal with cases where, for example, the deterioration state of one piece of railway equipment changes the probability of damage to another piece of railway equipment. In the following fourth embodiment, differences in the configuration and processing when handling both deterioration and damage will be described, and explanations of content that is the same as those already described will be omitted.

[0074] <Fourth embodiment> Fig. 15 shows an example of the system configuration of a railway maintenance decision support system when handling both deterioration and failure. Differences from Fig. 1 or Fig. 10 include a deterioration prediction model estimation unit 1501 and a failure prediction model estimation unit 1502. The differences will be explained using Fig. 16 showing the processing contents.

[0075] Figure 16 is a flowchart showing an example of the process of estimating a deterioration or failure prediction model in a railway maintenance decision support system in the case of maintenance that handles both deterioration and failure, estimating the total maintenance cost, and determining the maintenance accuracy of each piece of equipment that minimizes the maintenance cost. Note that the same processes are assigned the same numbers as in Figure 2.

[0076] Processing unit 113 estimates the deterioration prediction model and failure prediction model for railway facility A (1601). In FIG. 15, this processing corresponds to deterioration prediction model estimation unit 1501 and failure prediction model estimation unit 1502. For actual model estimation, the same processing as 202 and 1101 already explained is performed. Next, processing unit 113 estimates the deterioration prediction model and failure prediction model for railway facility B by adding the deterioration and failure information for railway facility A (1602). In FIG. 15, this processing corresponds to deterioration prediction model estimation unit 1501 and failure prediction model estimation unit 1502. At this time, the deterioration prediction model is given by Equation 8.

[0077]

number

[0078] where y B is the predicted deterioration value of equipment B, y A (t) is the predicted deterioration value of equipment A at elapsed time t, z A (t) is the failure prediction value of equipment A at elapsed time t, v B is the maintenance accuracy of equipment B, w1, w2,...w j is the explanatory variable (j≦n) remaining from the candidate explanatory variables, and t is the elapsed time. This is an equation that includes the predicted deterioration and failure values ​​of equipment A as explanatory variables. This is the deterioration prediction model for railway equipment B.

[0079] Similarly, the failure prediction model is given by Equation 9.

[0080]

number

[0081] where z B is the failure prediction model for railway equipment B, z A (t) is the failure prediction value of equipment A at elapsed time t, y A (t) is the predicted deterioration value of equipment A at elapsed time t, m is the Weibull coefficient, η is the scale, s B is the correction time, о1, о2, о k isB The explanatory variables are the predicted deterioration and failure values ​​of equipment A, which are included in the explanatory variables. This is the deterioration and failure prediction model for railway equipment B.

[0082] The next change is the deterioration prediction and failure prediction of railway equipment A by the processing unit 113 (1603). Since the deterioration prediction is a determination of whether the deterioration has exceeded a reference value, and the failure prediction is a determination of whether a value exceeding the failure probability of the failure prediction model has been obtained in a random trial, if either the deterioration prediction model or the failure prediction model determines that an abnormality has occurred, it is determined that an abnormality has occurred. Similarly, for the deterioration prediction and failure prediction of railway equipment B, if either one becomes abnormal, it can be determined that an abnormality has occurred (1604). By performing the following process in the same manner as the flow in Figure 2, it becomes possible to output the maintenance accuracy required to reduce overall maintenance costs when dealing with both deterioration and failure.

[0083] The present invention can also be applied to an object consisting of three or more facilities or devices. In the fifth embodiment, processing for handling three or more railway facilities or devices will be described. Note that a description of the same content as that already described will be omitted.

[0084] Fifth Embodiment Figure 17 shows an example of a configuration made up of multiple facilities and devices. In this example of a railway system of systems, railway facility A is connected to railway facility B and railway facility C, and similarly, railway facility B is connected to railway facility D and railway facility E, and railway facility D is connected to railway facility F and railway facility G. It also shows that railway facility E is connected to railway facility G and railway facility H. These connection relationships indicate the maintenance relationships of each piece of railway facility. In the case of such equipment, it is advisable to perform model estimation from downstream facilities while referencing the configuration information.

[0085] FIG. 18 is a flowchart showing an example of model estimation for each piece of railway equipment based on the configuration example. First, the processing unit 113 extracts a group of equipment included in the equipment's configuration information and sets it as set M. Also, the equipment that has already been modeled is set as set N (1801). Note that N is initially an empty set. Here, the configuration information is information that expresses the railway equipment configuration example in FIG. 17 as data. In other words, the configuration information indicates the connection relationships of each piece of railway equipment as a system of systems, and includes information indicating which piece of railway equipment's maintenance accuracy will be affected by a change in the maintenance accuracy of a certain piece of railway equipment. The configuration information is, for example, like a connection matrix in graph theory, where each piece of equipment is i and the number of pieces of railway equipment is n, and the element (i, j) represents whether railway equipment i is connected to railway equipment j (for example, 1 is given if there is a connection from i to j, -1 is given if there is a connection in the opposite direction, and 0 is given if there is no connection).

[0086] Next, the processing unit 113 extracts a group of equipment from the set M that has no equipment under it, and sets it as P (1802). Specifically, by referring to the configuration information described above, for all j, (i, j), any i for which j does not have a value of 1 becomes an element of P. Next, the processing unit 113 performs model prediction for each piece of equipment i in P (1803). This model prediction is performed by the deterioration prediction model estimation unit 102 in FIG. 1 or the failure prediction model estimation unit 1001 in FIG. 10 (note that detailed processing of these has already been explained, and will be omitted here).

[0087] Next, the processing unit 113 adds each element of model-predicted P to the model-predicted set N, and removes P from M to create a new set M (1804). Finally, the processing unit 113 determines whether M has become an empty set (1805), and if it is not an empty set, the processing returns to 1802. If it is an empty set, the processing ends. In this way, by using the connection information of railway equipment and performing model estimation starting from the lower-level railway equipment, it is possible to perform model estimation for all railway equipment.

[0088] Here, the configuration information is preferably information indicating that equipment is more susceptible to deterioration and failure as it moves downstream (i.e., the lower the level of railway equipment, the more susceptible it is to deterioration and failure), and the processing unit 113 preferably performs processing using a connection matrix based on this configuration information. This makes it possible to generate a prediction model for other equipment using predictions of equipment that is more susceptible to deterioration and failure as explanatory variables, and to output results based on equipment that is highly related to the impact of maintenance. Note that the configuration information is input into the railway maintenance decision support system using an appropriate method, and can be input by a maintenance manager, for example.

[0089] Next, a sixth embodiment will be described. In the sixth embodiment, an example of a method of utilizing the above-described railway maintenance decision support system in the field of railway maintenance will be described, with rails and wheels as maintenance targets.

[0090] Sixth Embodiment Figure 19 shows an activity diagram for a railway maintenance site. In this example, first, sensor information is sent from railway equipment A and railway equipment B, and then stored in the railway maintenance decision support system. (The railway maintenance decision support system in this example is the same as the railway maintenance decision support system in Figure 1.) For example, railway equipment A inputs sensing information about the rails (elevation, alignment, level, gauge, flatness, etc.).

[0091] Sensing information (wheel diameter, etc.) related to wheel wear is input from railway equipment B. Next, the railway maintenance decision support system outputs the maintenance accuracy that minimizes overall costs based on the above sensing information.

[0092] The calculation process is the same as the flow in Figure 2, and in the case of rails and wheels, the processing unit 113 first estimates a rail deterioration prediction model and a wheel deterioration prediction model. At this time, for the wheel deterioration prediction model, the rail deterioration state is included as an explanatory variable to estimate a wheel deterioration prediction model. Next, the processing unit 113 uses the rail and wheel deterioration prediction model to estimate the total maintenance cost when the rail maintenance accuracy and wheel maintenance accuracy are changed. Finally, the processing unit 113 outputs the combination of maintenance accuracy that will result in the lowest overall cost.

[0093] The maintenance manager refers to this calculation result (the maintenance accuracy of each piece of equipment that minimizes overall costs) and checks the amount of improved maintenance cost and the proposed changes to maintenance accuracy to decide whether to change the target maintenance accuracy for each piece of equipment. For example, the minimum overall maintenance cost is displayed as when rail distortion is made 1.2 times stricter than the current accuracy, but wheel diameter errors can be tolerated up to 1.1 times the current level, and the maintenance manager decides whether to change to that proposed target maintenance accuracy. If the target maintenance accuracy is changed, the maintenance manager presents the target maintenance accuracy to the maintenance planner. The maintenance planner then creates a specific maintenance plan (inspection plan, repair plan, replacement plan) that maintains the target maintenance accuracy. This maintenance plan is presented to each maintenance staff member, who then inspects, repairs, replaces, etc. the rails and wheels in accordance with the maintenance plan.

[0094] In this way, by clarifying the relationships between railway equipment maintenance and presenting the desired maintenance accuracy for each piece of railway equipment to the maintenance site in order to reduce overall maintenance costs, a change to a better maintenance method will be promoted, leading to a reduction in maintenance costs and ultimately to an improvement in the balance of payments for railway operators.

[0095] In other words, by applying the railway maintenance decision support system (railway maintenance support system) to railways, which are a system of systems, and estimating the maintenance accuracy of each piece of equipment and device that can reduce overall maintenance costs, it becomes easier to change the maintenance management standards for each piece of equipment and device in order to reduce overall maintenance costs. Furthermore, based on changes to the maintenance management standards made by the maintenance manager, the maintenance planner can create a maintenance plan based on the changes to the maintenance management standards, and by having the maintenance person perform maintenance based on the changed maintenance plan, overall maintenance costs can be minimized.

[0096] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to the above-described embodiments, and various design modifications can be made without departing from the spirit of the present invention as defined in the claims. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

[0097] For example, the railway maintenance decision-making support system may be deployed as a computer directly operated by a maintenance manager, or may be deployed on the cloud. In this case, components such as the input and output units of the railway maintenance decision-making support system may be omitted as appropriate, and the maintenance manager may input and output data using a communication unit. As long as appropriate processing can be performed, the railway maintenance decision-making support system may be configured, for example, with one or more computers. Furthermore, the railway maintenance decision-making support system may be configured so that the functions of the railway maintenance decision-making support system are realized by multiple devices performing processing in a distributed manner. [Explanation of symbols]

[0098] 113 Processing section 102 Deterioration prediction model estimation unit 105 Results Creation Department

Claims

1. A processor and a storage unit are included, The processor: a deterioration prediction model used to predict deterioration of railway equipment, the deterioration prediction model having, as an explanatory variable, maintenance accuracy, which is information relating to the accuracy of maintenance of the railway equipment obtained by converting sensing data acquired from sensor devices of the railway equipment, is estimated for each of a plurality of pieces of railway equipment, and stored in the storage unit; In estimating the deterioration prediction model, performing a process of estimating a second railway equipment deterioration prediction model that includes, as an explanatory variable, the deterioration prediction of the first railway equipment deterioration prediction model that has already been estimated and stored in the storage unit; Output the total maintenance cost for each railway facility based on the deterioration prediction model's deterioration prediction. A railway maintenance support system characterized by:

2. 2. The railway maintenance support system according to claim 1, The processor: calculating the overall maintenance cost by changing the maintenance accuracy and predicting deterioration using the deterioration prediction model for each railway equipment stored in the storage unit; Identifying the maintenance accuracy of each piece of railway equipment when minimizing the overall maintenance cost. A railway maintenance support system characterized by:

3. A processor and a storage unit are included, The processor: a failure prediction model used to predict failures of railway equipment is estimated for each of a plurality of railway facilities and stored in the storage unit; In estimating the failure prediction model, performing a process of estimating a second railway equipment failure prediction model that includes, as an explanatory variable, the failure prediction of the first railway equipment failure prediction model that has already been estimated and stored in the storage unit; Output the total maintenance cost for each piece of railway equipment based on the failure prediction model's failure prediction. A railway maintenance support system characterized by:

4. The railway maintenance support system according to claim 3, The processor: calculating the total maintenance cost by predicting failures while changing the maintenance accuracy, which is information regarding the accuracy of maintenance of the railway equipment, using the failure prediction model for each railway equipment stored in the storage unit; Identifying the maintenance accuracy of each piece of railway equipment when minimizing the overall maintenance cost. A railway maintenance support system characterized by:

5. 3. The railway maintenance support system according to claim 2, A display device is provided, The processor: A display using a scale showing the degree of change in maintenance accuracy between the current maintenance accuracy and the maintenance accuracy of the proposed change; and displaying on the display device a display showing the results of the overall maintenance cost change and the maintenance accuracy change for the proposed change. A railway maintenance support system characterized by:

6. 5. The railway maintenance support system according to claim 4, A display device is provided, The processor: A display using a scale showing the degree of change in maintenance accuracy between the current maintenance accuracy and the maintenance accuracy of the proposed change; and displaying on the display device a display showing the results of the overall maintenance cost change and the maintenance accuracy change for the proposed change. A railway maintenance support system characterized by:

7. 3. The railway maintenance support system according to claim 2, A display device is provided, The processor: A display showing the time change in the maintenance cost of each piece of railway equipment and the time change in the overall maintenance cost in the current maintenance plan, which is the current maintenance plan; A display showing the time change in the maintenance cost of each piece of railway equipment and the time change in the overall maintenance cost in a maintenance change plan, which is a change plan from the current situation; and displaying a comparison of the current maintenance plan and the maintenance change plan, in which the changes over time in the overall maintenance costs are superimposed on the display device. A railway maintenance support system characterized by:

8. 5. The railway maintenance support system according to claim 4, A display device is provided, The processor: A display showing the time change in the maintenance cost of each piece of railway equipment and the time change in the overall maintenance cost in the current maintenance plan, which is the current maintenance plan; A display showing the time change in the maintenance cost of each piece of railway equipment and the time change in the overall maintenance cost in a maintenance change plan, which is a change plan from the current situation; and displaying a comparison of the current maintenance plan and the maintenance change plan, in which the changes over time in the overall maintenance costs are superimposed on the display device. A railway maintenance support system characterized by:

9. 2. The railway maintenance support system according to claim 1, The processor: a failure prediction model used to predict failures of railway equipment is estimated for each of a plurality of railway facilities and stored in the storage unit; In estimating the deterioration prediction model, performing a process of estimating a deterioration prediction model for a second railway equipment by adding, to explanatory variables, the failure prediction of the first railway equipment failure prediction model that has already been estimated and stored in the storage unit; In estimating the failure prediction model, performing a process of estimating a second railway equipment failure prediction model that includes, as explanatory variables, a deterioration prediction of a first railway equipment deterioration prediction model that has already been estimated and stored in the storage unit, and a failure prediction of a first railway equipment failure prediction model that has already been estimated and stored in the storage unit; Instead of outputting the total maintenance cost, output the total maintenance cost for maintenance based on the deterioration prediction of the deterioration prediction model of each piece of railway equipment stored in the storage unit and the failure prediction of the failure prediction model of each piece of railway equipment stored in the storage unit. A railway maintenance support system characterized by:

10. The railway maintenance support system according to claim 3, The processor: a deterioration prediction model used to predict deterioration of railway equipment, the deterioration prediction model having, as an explanatory variable, maintenance accuracy, which is information relating to the accuracy of maintenance of the railway equipment obtained by converting sensing data acquired from sensor devices of the railway equipment, is estimated for each of a plurality of pieces of railway equipment, and stored in the storage unit; In estimating the failure prediction model, performing a process of estimating a failure prediction model for a second railway equipment by adding, to explanatory variables, the deterioration prediction of the deterioration prediction model for the first railway equipment that has already been estimated and stored in the storage unit; In estimating the deterioration prediction model, performing a process of estimating a deterioration prediction model for a second railway equipment, the deterioration prediction model including, as explanatory variables, a failure prediction of a failure prediction model for a first railway equipment that has already been estimated and stored in the storage unit, and a deterioration prediction of a deterioration prediction model for a first railway equipment that has already been estimated and stored in the storage unit; Instead of outputting the total maintenance cost, output the total maintenance cost for maintenance based on the deterioration prediction of the deterioration prediction model of each piece of railway equipment stored in the storage unit and the failure prediction of the failure prediction model of each piece of railway equipment stored in the storage unit. A railway maintenance support system characterized by:

11. 2. The railway maintenance support system according to claim 1, The processor: In estimating the deterioration prediction model, When three or more pieces of railway equipment are related to changes in maintenance accuracy, a deterioration prediction model is estimated in order from the railway equipment that does not affect the maintenance accuracy of other pieces of railway equipment in the configuration information indicating the connection relationships of each piece of railway equipment. A railway maintenance support system characterized by:

12. The railway maintenance support system according to claim 3, The processor: In estimating the failure prediction model, When three or more pieces of railway equipment are related to changes in maintenance accuracy in the maintenance information, which is information related to the accuracy of maintenance of railway equipment, a failure prediction model is estimated in order from the railway equipment that does not affect the maintenance accuracy of other pieces of railway equipment in the configuration information that indicates the connection relationships of each piece of railway equipment. A railway maintenance support system characterized by:

13. 2. The railway maintenance support system according to claim 1, The processor: In estimating the deterioration prediction model, By using railway transportation information and including information on trains passing through locations where railway facilities are affected by deterioration as explanatory variables, a deterioration prediction model is estimated based on explanatory variables that take into account railway transportation information. A railway maintenance support system characterized by:

14. The railway maintenance support system according to claim 3, The processor: In estimating the failure prediction model, By using railway transportation information to include information on trains passing through locations where railway equipment failures may occur as explanatory variables, a failure prediction model is estimated based on explanatory variables that take into account railway transportation information. A railway maintenance support system characterized by:

15. A railway maintenance support method performed using a railway maintenance support system having a processor and a storage unit, The processor: inputting maintenance accuracy, which is information regarding the accuracy of maintenance of railway equipment, and estimating a deterioration prediction model for predicting deterioration of the railway equipment, which uses the maintenance accuracy as an explanatory variable, for each of a plurality of pieces of railway equipment, and storing the model in the storage unit; The total maintenance cost for each railway facility based on the deterioration prediction model is output. When estimating a deterioration prediction model for each of the plurality of railway facilities, a process is performed to estimate a deterioration prediction model for a second railway facility that includes, as an explanatory variable, the deterioration prediction of the deterioration prediction model for the first railway facility that has already been estimated and stored in the storage unit. A railway maintenance support method characterized by:

Images