TW201908163A - Abnormality detecting device, abnormality detecting method, and computer readable recording medium - Google Patents

Abstract

Embodiments of the present invention achieve highly accurate anomaly detection. According to one embodiment, an anomaly detection device includes: a condition generator, a threshold setter and an anomaly detector. The generator generates a plurality of conditions for classifying a difference between a predicted value of a state of a vehicle and a measured value of the state of the vehicle based on travel information of the vehicle, the predicted value being based on a control command value and a prediction model. The threshold setter sets a plurality of thresholds for the conditions. The anomaly detector performs anomaly detection on the vehicle based on the prediction model, the thresholds, and the conditions.

Description

Abnormality detection device, abnormality detection method and computer readable recording medium

An embodiment of the present invention relates to an abnormality detection device, abnormality detection method, and computer-readable recording medium.

In order to maintain the safe and stable operation of the railway, the maintenance management or inspection of the railway vehicles must be carried out on a daily basis. Previously, maintenance and management centered on periodic inspections of railway vehicles were carried out, but in recent years, in order to ensure higher safety of railways, the following technology is being developed: the use of sensor values obtained from railway vehicles Or vehicle information such as control values, for diagnosis or status monitoring, thereby early detection of abnormalities.

The following technologies exist: a technology for performing abnormality detection by comparing the value measured by a sensor on a railway vehicle with a threshold value; or a model that reproduces the normal system operation of a railway vehicle and uses the model Techniques for detecting anomalies or anomalies. However, it is difficult to accurately detect anomalies based on the same threshold value in railway vehicles whose driving conditions dynamically change in time series due to changes in route gradient or weather, passengers getting on and off, driver's operation, and the like.

[Problems to be Solved by the Invention] Embodiments of the present invention provide an abnormality detection device, an abnormality detection method, and a computer program that realize high-accuracy abnormality detection. [Means to solve the problem]

The abnormality detection device as an embodiment of the present invention includes a condition generation unit, a threshold setting unit, and an abnormality detection unit. The condition generating unit generates a plurality of conditions that distinguish the difference between the predicted value of the state of the vehicle based on the control command value and the prediction model and the measured value of the state of the vehicle based on the travel information of the vehicle. The threshold setting unit sets a plurality of thresholds for the plurality of conditions. The abnormality detection unit performs abnormality detection of the vehicle based on the prediction model and the plurality of thresholds.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In addition, in the drawings, the same constituent element is denoted by the same number, and the description is appropriately omitted.

FIG. 1 is a block diagram showing an example of an abnormality detection system according to the first embodiment of the present invention.

The abnormality detection system of FIG. 1 includes an abnormality detection device 100, a vehicle system 500, an environmental information system 600, a terminal 700, an input device 800, and a screen display device 900. Hereinafter, the outline of the abnormality detection system of FIG. 1 will be described.

The abnormality detection device 100 has a learning mode and an operation mode. The abnormality detection device 100 is provided in the learning mode based on at least one of the measurement information of the railway vehicle (hereinafter, vehicle) obtained from the vehicle system 500 and the environmental information of the vehicle obtained from the environmental information system 600. Function of an abnormality detection model related to the braking system of the vehicle (abnormality detection model generation unit 200). The driving information of the vehicle of the present embodiment includes at least one of the measurement information of the vehicle obtained from the vehicle system 500 and the environmental information of the vehicle obtained from the environmental information system 600. The abnormality detection model includes a prediction model that predicts the state of the vehicle, and a threshold value related to the deviation (deviation) of the predicted value of the self-prediction model. As an example, the state of the vehicle is the deceleration of the vehicle.

The abnormality detection device 100 is provided in the operation mode, and uses a prediction model and a threshold value to detect abnormality of the vehicle (abnormality detection unit 110 ). The so-called anomaly detection refers to determining whether there is an anomaly. Anomaly detection is also called anomaly determination. The abnormality detection is performed by comparing the deviation and the threshold value of the difference between the predicted value as a prediction model and the actual measured value of the state of the vehicle obtained from the vehicle.

In addition, the abnormality detection device 100 has a function (threshold value setting unit 220) for setting a threshold value for a plurality of conditions (hereinafter, driving conditions) based on the travel information of the vehicle. In this case, the abnormality detection device 100 generates an abnormality detection model including one prediction model and multiple thresholds corresponding to multiple driving conditions. In the anomaly detection using the anomaly detection model, from the plurality of thresholds contained in the anomaly detection model, the driving information corresponding to the driving condition where the vehicle is placed, that is, the object of anomaly detection (current driving information ) The threshold value of the satisfied driving conditions, and use the selected threshold value and prediction model.

The creation of multiple driving conditions and the setting of multiple thresholds are performed in the learning mode. In the creation of driving conditions, the vehicle's driving information and the results of abnormality detection (deviations between the predicted value of the prediction model and the actual measured value of the state value, etc.) are used. The learning mode and operation mode can be switched automatically or by instructions from maintenance personnel, etc., or they can be executed simultaneously.

When an abnormality is detected, the abnormality detection device 100 displays, as an example, a location where the abnormality is detected, an abnormality detection model for abnormality detection, driving information for abnormality detection, a predicted value obtained from a prediction model, etc., on the screen In the display device 900. In this way, the monitoring by the railway operator is supported.

Here, the vehicle brake system of this embodiment will be briefly described. FIG. 2 shows an example of a brake level and a configuration of a brake and an air cushion for a specific wheel of a vehicle. Furthermore, the braking level is actually located in the cab of the tram. Hereinafter, the braking system of the vehicle that is the target of abnormality detection by the abnormality detection device 100 will be described with reference to FIG. 2.

A brake lever 10 as an example of a controller provides a means for the driver to perform a brake operation. The driver moves the brake lever from below to above, whereby the vehicle can be braked. The values of 1 to 8 displayed on the brake lever 10 are the braking level (braking phase). The larger the value, the stronger the braking force acting on the vehicle. The number of levels here is an example and does not exclude the use of vehicles with more or fewer levels. The braking level corresponds to an example of the control command value for the vehicle or brake.

Furthermore, the braking operation of the vehicle is not limited to the braking operation by the driver. For example, in vehicles equipped with automatic train stop devices (Automatic Train Stop: ATS), automatic train control devices (Automatic Train Control: ATC), and automatic train driving devices (Automatic Train Operation: ATO), these devices exist instead of driving Personnel to perform the braking operation. In this case, as an example, commands related to brake braking output from these devices correspond to control command values.

In FIG. 2, the wheels 30 of the vehicle traveling on the rail 20 are shown. As one of the means for braking the vehicle with a brake, there is a tread brake 42. Here, for the sake of explanation, only one wheel is shown, but actually, a plurality of pairs of left and right wheels are provided.

The tread brake 42 uses a cylinder as power. By increasing the brake cylinder pressure, which is the pressure in the cylinder 43, the brake shoe 41 is pressed against the tread surface of the wheel 30 that is the surface in contact with the rail. The frictional force between the wheel 30 and the brake pad 41 becomes the braking force of the tread brake 42.

In this way, since the tread brake utilizes the friction force of the brake pad, there is a possibility that the brake pad will wear out due to continuous use, and the braking force may decrease. The tread brake is an example of a mechanical brake used for a vehicle. In addition, there are disk brakes that use a pad or the like to press a disk fixed to a wheel shaft against a wheel to obtain a braking force. The braking force of the brake varies according to the wear state of the brake pad or pad. When an abnormality is detected in the brake system by this abnormality detection device, the operator can also confirm the brake pads or discs of the brake system, etc. to confirm whether there is actually no abnormality.

In addition to the wear of parts, the braking force of the brake also varies according to the load on the vehicle. A load compensating device 50 is mounted on the vehicle of FIG. 2. The load compensation device 50 includes an air cushion 51. By detecting the air cushion pressure of the air cushion 51, the load that the vehicle bears can be measured. As the braking control of the vehicle, in addition to the operation of the brake lever 10, the braking force of the brake may be adjusted according to the air cushion pressure detected by the load compensation device 50. In this way, regardless of the increase or decrease in the load on the vehicle, the desired deceleration can be achieved.

In the braking system of the vehicle, in order to supplement the braking force of the mechanical brake, an electric brake may also be used in combination. The electric brake will be described using FIG. 3. FIG. 3 is a configuration example of a power generating brake and a regenerative brake of a vehicle.

The vehicle of FIG. 3 is equipped with a main motor 60a and a main motor 60b. When a power generating brake is used, the main motor 60a, the main motor 60b and the resistor 70 constitute a closed circuit, which further converts the electric power of the main motor into heat energy.

On the other hand, when the regenerative brake is used, electric power generated by the main motor 60 a and the main motor 60 b is transmitted from the pantograph 80 toward the overhead line 90. Alternatively, when a battery is mounted in the vehicle, the battery may be charged using the generated electric power. In this way, in the regenerative brake, the main motor 60a and the main motor 60b are used as generators, and the kinetic energy is converted into electric power, thereby ensuring the braking force.

The mechanical brake or the generator brake is an example, and even when another type of brake is used in the brake system, it can be used as the object of abnormality detection by the abnormality detection device 100.

Here, since the configuration of the brake system is relatively complicated, and the characteristics of the brake or the brake system may vary due to multiple factors or conditions, it is difficult to accurately detect the abnormality of the vehicle's brake system.

For example, multiple brakes with different characteristics are used together in a vehicle's braking system. In addition, as described above, the braking force of the brake system of the vehicle changes according to the load. For example, with regard to passenger vehicles, the number of passengers varies greatly depending on the time period or operating interval. Therefore, the braking force of the braking system varies greatly within a short period of time. Regarding cargo vehicles, the load also varies greatly depending on the cargo load. In addition, the tendency of the slope or cant varies depending on the route or section where the vehicle travels, and there is a possibility that the deceleration when the brake operation is performed fluctuates. Furthermore, the difference in weather conditions, such as the presence or absence of precipitation and the temperature, may change the physical properties of the components that constitute the brake system, and affect the characteristics of the brake system. In addition, the characteristics of the brake operation vary depending on the driver, and sometimes there are differences in the characteristics of the brake due to the individual differences of each vehicle.

Therefore, in the present embodiment, the threshold value used is switched according to the driving conditions of the vehicle in which the appropriate threshold value corresponding to the conditions inside and outside the vehicle is used, thereby easily performing high-precision abnormal detection. In this way, the risk of misdetection is reduced for the presence or absence of abnormalities, and the early detection of abnormalities and the safe and stable operation of the railway are maintained.

Hereinafter, the abnormality detection device 100 of FIG. 1 will be described in more detail. In the following description, the braking system that is subject to abnormality detection by the abnormality detection device 100 may be a brake device for a specific wheel of an arbitrary vehicle of the railway, or may be a plurality of entire brake devices of the entire vehicle, or may be The entirety of the braking devices of multiple vehicles or groups. The object of abnormality detection is not only the braking system, but also the power system, air conditioning system, power system, etc. In addition, the object of the abnormality detection is not limited to railway vehicles, and may be any vehicle equipped with wheels such as automobiles, construction machines, and airplanes. It does not exclude devices or systems other than vehicles.

The abnormality detection device 100 includes a vehicle information collection unit 101, an environmental information collection unit 102, a data processing unit 103, an abnormality detection model generation unit 200, a condition generation unit 230, an abnormality detection unit 110, a report unit 120, and a screen generation unit 130.

The abnormality detection model generation unit 200 includes a model generation unit 210 and a threshold setting unit 220.

The vehicle information collection unit 101 acquires measurement information (also referred to as measurement data) related to the vehicle from various sensors of the vehicle system 500 in the vehicle. As examples of sensors, there are sensors that detect the braking operation of the vehicle as a control command value, sensors that detect the deceleration of the vehicle, sensors that detect the driving speed, and sensors that measure the load on the vehicle Sensors, etc. In addition, various sensors can also be considered. The measurement information includes the detected value of the sensor (control command value, etc.), the measured value of the sensor (driving speed, load, deceleration, etc.), etc. In addition, when the deceleration is calculated based on the value of the speed sensor in the vehicle system 500, the calculated deceleration can also be obtained as part of the measurement information.

The type of measurement information to be acquired (the type of sensor or the type of control command value) can be set arbitrarily. The period for obtaining measurement information can be set arbitrarily. For example, regarding the measurement information related to the driving speed of the vehicle, the value is obtained within a short sampling period in milliseconds. Regarding the sensor that measures the load on the vehicle, the value is obtained within the sampling period in minutes.

The environmental information collection unit 102 obtains the environmental information of the vehicle (also referred to as driving environment data) from the environmental information system 600. As examples of environmental information, there are information related to operation routes or information related to weather. As an example of the information related to the operation route, there is information related to the operation route such as the slope or slope of each section (difference between the height of the track inside and outside the railway). As examples of weather-related information, there are weather-related information such as weather, temperature, precipitation, wind speed, and barometric pressure. The environmental information can be obtained by acquiring the information accumulated in the database in the above-ground system, or by acquiring the information transmitted from an external server. The type of environmental information to be acquired and the frequency of acquisition can be set arbitrarily.

The abnormality detection device 100 may be installed outside the vehicle such as a facility of a railway operation management company or an operation direction center as an above-ground device, or may be installed on the vehicle as an on-board device. The installation form of the abnormality detection device 100 is not particularly limited.

When the abnormality detection device 100 is installed outside the vehicle as an above-ground device, as an example, the measurement information of the vehicle system 500 in the vehicle is received via the on-board coil, the above-ground coil of the answering machine, and the above-ground information network. That is, the vehicle system 500 transmits data to the ground-based information network via the ground coil, etc., and the abnormality detection system receives the data via the ground-based information network. Metal cables, coaxial cables, optical cables, telephone lines, wireless, Ethernet (registered trademark), etc. can be used in the information network on the ground, but the method is not particularly limited. The anomaly detection device 100 can also receive data from the environmental information system 600 via an above-ground information network.

When the abnormality detection device 100 is an on-board device, the abnormality detection device 100 obtains data from the vehicle system 500 via an information network in the vehicle. The information network in the vehicle includes an Ethernet or a wireless area network (Local Area Network, LAN), etc., but it can also be a network using other methods. The abnormality detection device 100 may also use the coil on the vehicle or the coil on the ground of the answering machine to obtain the data of the environmental information system 600 connected to the information network on the ground.

The input device 800 provides an interface for maintenance personnel to operate. The input device 800 includes a mouse, a keyboard, a voice recognition system, an image recognition system, a touch panel, or a combination of these. Maintenance personnel can input various instructions or data from the input device 800 into the abnormality detection device 100 and perform operations.

The screen display device 900 displays the data or information output by the abnormality detection device 100 in the form of a static image or a moving image. As an example, the screen display device 900 is a liquid crystal display (Liquid Crystal Display, LCD), an organic electroluminescence display, a fluorescent display tube (Vacuum Fluorescent Display (VFD)), etc., but other methods may also be used Display device.

Multiple input devices 800 and screen display devices 900 can be provided. For example, the input device 800 and the screen display device 900 may be installed in the operation command center and the cab of the vehicle, respectively.

In addition, the input device 800 and the screen display device 900 may be one integrated device. For example, when there is a display with a touch panel function, the same device can serve as both the input device 800 and the screen display device 900.

The abnormality detection device 100 includes an information database 310, a model database 320, and a detection result database 330 as databases.

In FIG. 1, the database 310, the database 320, and the database 330 are all disposed inside the abnormality detection device 100. However, the configuration method of the database is not particularly limited. For example, a part of the database may also be arranged in an external server or storage device. The database can be installed by a relational database management system or various Unstructured Query Language (Not Only Structured Query Language, NoSQL) systems, but other methods can also be used. As the storage format of the database, it can be Extensible Markup Language (XML), JavaScript Object Notation (JSON), Comma-Separated Values (CSV), etc., or binary Forms and other forms. All databases in the anomaly detection device 100 need not be implemented in the same database system and storage format, and may be mixed with multiple methods.

The information database 310 internally stores the measurement information obtained by the vehicle information collection unit 101 and the environmental information obtained by the environmental information collection unit 102. It is also possible to insert a memory medium such as a memory element storing measurement information and environmental information into the abnormality detection device 100 and use the memory medium as the information database 310.

Examples of the information database 310 are shown in FIGS. 4 and 5. The driving information (measurement information and environmental information) is stored in the form of the table 310a shown in FIG. 4 and the table 310b shown in FIG.

The "data identifier (ID)" column of the table 310a of FIG. 4 stores the identification number of the entry stored in the table 310a. The material ID becomes the primary key. Each data ID is associated with a table 310b as shown in FIG. The form 310b is also stored in the information database 310. The "Time" column stores the creation time of the entry. In this example, entries are generated at fixed sampling times. However, the entries can also be generated on other criteria, such as in units of sections set in advance on the rails.

The name of the driver who performed the brake operation is stored in the “driver” column of the table 310a. In the case where the main body performing the braking operation is not the driver, but an ATS, ATC, ATO, or other device, the name of the operated device or an identifier indicating the device may be stored instead of the driver's name.

The weather-related information obtained from the environmental information system 600 is stored in the "weather" column of the table 310a.

The temperature-related information obtained from the environmental information system 600 is stored in the “temperature” column of the table 310a. The information related to the temperature may be the actual measured value or a label that grades the actual measured value. In the example shown in the figure, the conversion table 310c of FIG. 6 is used to store the label of any level of T1, T2, T3, T4, T5, T6, and T7 converted from the real-value temperature. For example, when the temperature is -11°C, it is converted to grade T1, when the temperature is 15°C, it is converted to grade T4, and when the temperature is 33°C, it is converted to grade T6. In addition, when creating a prediction model to be described later, data such as converting the level T1 to 1, T2 to 2, and T3 to 3, and converting the level name to an arbitrary integer can also be used as an explanatory variable (explanatory variable).

As in the example of the "temperature" column, in the information database 310, it is also possible to store processed information that has been calculated or converted from measurement information or environmental information.

In the column 310a, the "ride rate" column stores the percentage of the ride rate as an indicator of the load on the vehicle. In order to express the load, other indicators can also be used. As an example, the occupancy rate is defined by the ratio of the capacity of passenger vehicles to the number of passengers of passenger vehicles. The ride rate is mostly inferred from the air cushion pressure of the load compensation device, so the air cushion pressure can also be used directly as an indicator.

The air cushion pressure is the actual measured value of the sensor, and is not a value that is indirectly estimated using a conversion table or formula like the ride rate. Therefore, there is a possibility of reducing errors in model generation. However, the value of the air cushion pressure depends on the manufacturer or model of the load compensation device installed in the vehicle, and there is also a lack of versatility. Therefore, it may be considered that the difference in the load compensation device of each vehicle can be absorbed by using the commonly used index such as the occupancy rate.

In the "Slope" column of Table 310a, a value indicating the slope of the route is stored in permil units. The so-called micro-percentage refers to the value representing the height difference of every 1000 m of the horizontal distance in meters. Thousandths are examples, and values in other units can also be stored.

The slope of the millimeter unit is stored in the "bevel" column of the table 310a. Here, values in other units can also be stored in the same way.

In addition, the "wind speed" column and the "air pressure" column are displayed in the table 310a. There may also be a row storing other information such as the current position, the current section in the railroad track, and so on.

The table 310b as shown in FIG. 5 stores the time information and the actual value of the braking level, deceleration and other information for the corresponding entry in the table 310a of FIG. 4. The example of FIG. 5 is a table 310b corresponding to the document ID: 2560 of FIG. 4. The entries of the table 310b are generated at shorter time intervals than the table 310a. The generation interval (sampling interval) of the entries of the table 310b may be the same as the table 310a. In addition, although the table 310b and the table 310a are different tables, a table obtained by integrating these tables may be used.

The data stored in the information database 310 can also be processed. For example, the data processing unit 103 causes the content of each table stored in the information database 310 to be displayed on the screen display device 900. The maintenance personnel or the driver uses the input device 800 to perform processing operations on the data. The data processing section 103 performs data processing according to the processing operation.

In addition, the interval at which information or data is acquired by the vehicle information collection unit 101 or the environmental information collection unit 102 can also be adjusted. For example, the data processing unit 103 accepts the designated operation of the acquisition interval from the maintenance personnel or the driver via the input device 800, and adjusts the acquisition interval according to the content of the operation.

The abnormality detection model generation unit 200 uses the data stored in the information database 310 to create an abnormality detection model of the vehicle's braking system. The anomaly detection model includes a prediction model and one or more thresholds. The model generation unit 210 generates a prediction model, and the threshold setting unit 220 generates a threshold. The generated anomaly detection model is stored in the model database 320.

7 is an example of the model database 320. One or more anomaly detection models can be stored in the model database 320. Each anomaly detection model is identified by the model ID. In the row of the prediction model, data indicating a matching prediction model is stored, or the address (pointer) of the memory in which the prediction model is stored is stored. The data representing the prediction model includes, for example, a deceleration model. One or more threshold values are stored in the threshold value column. When storing multiple thresholds, the driving conditions corresponding to each threshold are also stored (details will be described later).

As an example, when the abnormality detection device 100 is started or when a system is newly added as an abnormality detection target, the abnormality detection model is generated by the learning mode. When there are a plurality of systems that are subject to abnormality detection, an abnormality detection model is generated for each system.

The anomaly detection model is created using data samples (feature vectors) extracted from the information database 310.

The data sample (feature vector) contains more than one explanatory variable. As an example of explaining the variable, the value of the braking level (control command value) of the table 310b is used. In addition to this, other types of values (speed, etc.) in the driving information or the specifications of the vehicle (as an example, the size, weight, etc. of the vehicle) can be used as explanatory variables. You can also calculate multiple variables contained in the driving information to generate explanatory variables. In addition, here, the target variable of the prediction model is set as the deceleration. The data sample can be generated in the unit of entries in the brake information table 310b, and it can also reduce the granularity of time, aggregate multiple consecutive entries into one, and generate a data sample based on these.

Hereinafter, a method of generating a prediction model will be described. Imagine a regression model used as a prediction model. The model generating unit 210 uses the information database 310 to obtain a feature vector X=(x ₁ , x ₂ , x ₃ , ..., x _n ) having the explanatory variable as an element.

Then, the model generation unit 210 performs multiple regression analysis (multiple regression analysis) to obtain the equation (1) that predicts the deceleration as the target variable. [Mathematical formula 1] (1) Here, y is the target variable, x _n is the explanatory variable, and b _n is the partial regression coefficient. In addition, in order to absorb the difference between the measurement units of each explanatory variable, the target variable and all explanatory variables are normalized to an average value of 0 and a dispersion of 1, whereby a standard partial regression coefficient can be used as the partial regression coefficient b _n . There can be one or more explanatory variables.

Model generation using complex regression analysis is an example. In addition, support vector regression (Support Vector Regression), self-regression, and other methods can also be used to make a prediction model of the target variable.

Cross-validation can also be used when making prediction models. For example, the data sample can be divided into multiple sets, at least one of the sets can be used as the test data for verification, and the other sets can be used to make the model. With this, the performance of the generated model can be confirmed.

In the embodiment of the present invention, it is assumed that the information database stores the information acquired after the brake system that becomes the abnormality detection is in a normal state. Therefore, the generated prediction model can be said to model the normal state of the brake system of the vehicle. However, it is also acceptable to store a part of the information when the brake fails in the information database.

In order to suppress the number of explanatory variables used in the production of the prediction model, for example, variable selection method or principal component analysis (Principle Components Analysis) may be used to adjust the number of explanatory variables. It is effective when different interpretation variables are related to each other, or when calculation time and processing load reduction are required.

The so-called variable selection method refers to a method of generating a model by selecting a partial set of explanatory variables that are effective for prediction from a set of all explanatory variables. A model that uses one or a few explanatory variables can be initially generated, and then a model with additional explanatory variables can be generated one by one, and a useful explanatory variable can be selected. Conversely, a model with many explanatory variables can be generated initially, and then one that will explain the variable The model is removed one by one, by which specific explanatory variables are useful. In addition, a genetic algorithm can also be used to select explanatory variables.

The so-called principal component analysis refers to the generation of new explanatory variables by solving the inherent value problem of the correlation matrix or variance-covariance matrix used for the data generated by the model, and summarizes The number of dimensions. The use of the new explanatory variable obtained by principal component analysis as the variable x _{n of} equation (1), and performing regression analysis is called principal component regression.

It shows an example of using the variable selection method. When the object of abnormality detection is the vehicle's braking system, and the target variable of the prediction model is set to the deceleration of the brake, for example, initially, only the brake level that is presumed to have the highest correlation with the deceleration is used to explain the variable Forecasting model. Thereafter, other explanatory variables that are considered to be related to the deceleration, such as the driving speed of the vehicle, are sequentially added to the prediction model, and the prediction accuracy is confirmed. Select the explanatory variables when the desired prediction accuracy can be obtained.

Next, the threshold value set in the prediction model by the threshold value setting unit 220 will be described. As a method of using the threshold value, when the difference between the predicted value of the target variable calculated by the prediction model (here, the predicted value of the deceleration) and the measured value of the deceleration (actually measured value) exceeds the threshold value, There is an abnormal decision. Making an abnormal decision is also called detecting abnormality. The difference between the measured value and the predicted value is called divergence. There may be both a case where the actual measured value is greater than the predicted value and a case where the actual measured value is less than the predicted value, so the value of the deviation may adopt the sign of either positive or negative. When focusing on the absolute value of the distance from the self-predicted value, and the sign is not a problem, the absolute value of the difference can also be defined as a deviation.

FIG. 8 shows an example of the determination method using the threshold value of the normal distribution. The graph of FIG. 8 shows the normal distribution 400 of divergence, the horizontal axis is the divergence, and the vertical axis is the probability density. A deviation between the predicted value and the measured value of a plurality of prediction models is obtained, and it is assumed that the plurality of deviations are distributed according to a normal state, and a normal distribution 400 is produced. The data used to obtain multiple deviations may be data samples used to generate the prediction model, test data, other driving information not related to the generation of the prediction model, or any combination of these. When the deviation from the deviation is larger, the normal distribution 401 and the normal distribution 402 indicated by the broken lines become distributions whose edges further expand.

Using the normal distribution 400, a threshold value for the prediction model is set. As an example, if the standard deviation is set to σ, the value of a constant multiple of the standard deviation such as 2σ or 3σ is set as the threshold. When 2σ is set as the threshold value, if the deviation exceeds 2σ, an abnormality is detected in the abnormality detection. If such a threshold value is set, about 95% of the measured value is judged as no abnormality (normal). As another setting example of the threshold value, a deviation value corresponding to a predetermined probability (for example, upper X percentage points or lower X percentage points) or its absolute value may be set as the threshold value. The method of determining the threshold value described here is an example, and does not exclude the use of other methods. For example, it may be assumed that a distribution other than the normal distribution determines the threshold, and maintenance personnel, drivers, and others may also set the threshold based on experience.

The abnormality detection unit 110 uses the abnormality detection model (prediction model and threshold value) stored in the model database 320 to perform abnormality detection of the system that is the target of abnormality detection. A feature vector is generated based on the driving information used for anomaly detection, and the generated feature vector and prediction model are used to predict the deceleration. Compare the deviation and threshold between the predicted deceleration and the measured deceleration. If it is below the threshold, it is determined to be normal, and if it is greater than the threshold, it is determined to be abnormal. The abnormality detection unit 110 stores information in the detection result database 330 according to the result of abnormality detection. The abnormality detection unit 110 notifies the screen generation unit 130 and the report unit 120 of information related to the result of abnormality detection.

FIG. 9 shows an example of the detection result database. According to the time series, the braking level, the actual measured value of the deceleration, the predicted value of the deceleration obtained by the prediction model (ID0001), and the presence or absence of abnormality detection are stored. In the example of FIG. 9, no abnormality is detected at any time. It is also possible to add the driving information at the corresponding time (such as driver, weather, temperature, ride rate, slope, slope, wind speed, air pressure, etc.) to the detection result database.

The above shows an example of setting one threshold value to the prediction model, but in this embodiment, the threshold value setting unit 220 may set a plurality of threshold values corresponding to the driving conditions to the prediction model. In this case, the abnormality detection unit 110 specifies the driving conditions satisfied by the current driving information when performing the abnormality detection. Moreover, abnormality detection is performed using the threshold value corresponding to the specified driving condition. The driving condition required for the threshold setting unit 220 to set a plurality of thresholds is generated by the condition generating unit 230.

The condition generating unit 230 uses the detection result database 330 and the driving information for anomaly detection to generate a plurality of driving conditions (a plurality of conditions) that distinguish the difference between the predicted value of the prediction model and the actual measured value corresponding to the value of the difference . Hereinafter, the operation of the condition generating unit 230 will be described in detail.

The condition generation unit 230 uses the detection result database 330 and the driving information to create a classifier (for example, a decision tree) that predicts the level corresponding to the deviation.

Divide the deviation between the predicted value and the measured value in the test result database into multiple levels (deviation levels). For example, if the deviation is below the threshold A, the deviation is set to a small deviation A, if the deviation is greater than the threshold A and the threshold B is not met, then the deviation is set to a moderate degree of deviation B, if the deviation If it is more than the threshold value B, it is set as the deviation level C with a large deviation. The threshold value A may be the threshold value initially set for the prediction model, but it is not limited to this. As an example of other rank classification, it can also be classified as follows. Sort the deviations in ascending or descending order. If the size of the divergence is arranged in ascending order, the set of the top 20% of the data with the smallest deviation is set to the deviation level A, the set of the data with the second smallest deviation of 60% is set to the deviation level B, and the remaining 20% of the data set is set to deviate from level C. The ratio of each level can be arbitrarily determined, and is not limited to the values shown here.

The condition generating unit 230 selects any one of the deviation levels A to C for each entry of the detection result database 330 as shown in FIG. 9 according to the deviation of the entry, and assigns the selected deviation level. Furthermore, the driving information corresponding to each item is specified based on the table 310a of FIG. 4 and associated with each item. However, when driving information is already included in the detection result database 330, this action is not required. As a result, as shown in FIG. 10, a data group that associates the deviation level with the driving information is generated in each entry of the detection result database. The data set may be stored in the internal buffer of the condition generating unit 230, or may be stored in a memory device or database (not shown). In the data set of Fig. 10, the deviation levels all become A. Use this data set as a learning material, set the deviation level as the target variable, and set the other items as the explanatory variables, thereby generating a decision tree. Furthermore, when creating a data set, when there are unnecessary items in the database (FIG. 9, FIG. 4, etc.) on which the production is based, there is no need to use the items. For example, if there is no need for anomaly detection items, it may not be included in the data set in FIG. 10.

Here, it is assumed that there is no data in which an abnormality is detected in the data group, but when there is data in which an abnormality is detected, the data may also be excluded from the data group. In addition, when the abnormality detection unit 110 detects an abnormality, the maintenance personnel may also confirm the correctness of the detection result. For example, when an abnormality is detected, a confirmation screen (refer to FIG. 16 described later) that confirms the detection result is displayed on the screen display device 900. When it is judged that the detection result is a false detection, the maintenance personnel inputs an instruction of this meaning. Based on this instruction, the condition generation unit 230 or other processing unit corrects the detection result of the detection result database 330. In the data group, the data detected abnormally can also be excluded when making the decision tree.

FIG. 11 shows an example of the generated decision tree. The decision tree predicts the deviation level corresponding to the target variable based on two explanatory variables related to precipitation and ride rate. Node 1001a, node 1001b, and node 1001c are end nodes corresponding to the target variable. The top node is called the root node. Nodes other than the end node and the root node are called intermediate nodes. The root node and the intermediate node are the nodes for explaining variables. The end node is a node that deviates from the grade (target variable node). The node 1001a, the node 1001b, and the node 1001c correspond to the deviation level A, the deviation level B, and the deviation level C, respectively. The decision tree classifies "when sunny (no precipitation) and the ride rate is less than 90%" as a departure class A, and "when sunny (no precipitation) and the ride rate is greater than 90%" as a departure class B, classify "when rainy (no precipitation)" as a deviation from level C.

The condition generating unit 230 acquires the conditions included in the path from each departure level node (end node) to the root node as the travel conditions corresponding to each departure level. As a result, the conditions corresponding to the path from the departure level A to the root node "when clear (no precipitation), and the ride rate is 90% or less" corresponding to the path from the departure level B to the root node are obtained respectively The condition "When it is sunny (without precipitation) and the ride rate is greater than 90%", and the condition "When it is rainy (without precipitation)" corresponding to the path from the departure level C to the root node is used as the driving condition A, the driving condition B, Driving condition C.

The threshold value setting unit 220 sets a threshold value for each of the driving conditions A to C. Specifically, the driving information used for the generation of the prediction model (or the driving information not used for the generation of the prediction model) is classified into groups A to C that satisfy the driving conditions A to C. Anomaly detection is performed on the driving information classified into group A, and the deviation is calculated based on the detection result. The distribution of deviation (normal distribution, etc.) is obtained, and the threshold value A is set using this distribution (refer to the description in FIG. 8 ). With this, the threshold value A for the driving condition A is set. Similarly, the threshold B corresponding to the driving condition B is also set for the group B, and the threshold C corresponding to the driving condition C is also set for the group C.

The distribution corresponding to the deviation of the group A becomes smaller because the standard deviation becomes smaller, so the threshold value A becomes a smaller value. The distribution corresponding to the deviation of the group C becomes larger because the standard deviation becomes larger, so the threshold value C becomes a larger value. The threshold value B corresponding to the group B becomes a value between the threshold value A and the threshold value C.

Threshold A is the threshold used when the occupancy rate is less than 90% on sunny days, Threshold B is the threshold used when the occupancy rate is greater than 90% on sunny days, and C is the precipitation limit Threshold when used when the quantity is greater than 0. When abnormality detection is performed corresponding to such driving conditions, by switching the threshold value, abnormality detection that accurately reproduces the brake characteristics can be realized. Here, only precipitation and ride rate are used for driving conditions, but other items such as air temperature and humidity can be used as driving conditions.

There are multiple ID3, C4.5, etc. in the algorithm for learning the decision tree, but any algorithm can also be used. In order to prevent noise or excessive learning, pruning of the decision tree can also be carried out. Decision tree learning is an example, and other classifiers can also be used. Which of the multiple items of driving information is used in the decision tree depends on the algorithm or the learning materials used.

Here, the decision tree is used to set the threshold of each driving condition, but the maintenance personnel can also refer to the expert's opinion to set the threshold for each driving condition.

Each group of driving conditions and threshold values generated and set by the condition generating unit 230 and the threshold value setting unit 220 is stored in the corresponding cell of the threshold value column of the model database 320. An example of storing multiple threshold values and the driving conditions corresponding to the multiple threshold values for model 0001 is shown in FIG. 12.

In this way, when an abnormality detection model in which a plurality of thresholds are set in the prediction model is used, the abnormality detection unit 110 specifies the traveling condition that satisfies the current traveling information among the plurality of traveling conditions when performing abnormality detection. Furthermore, abnormality detection is performed using the threshold value and the prediction model corresponding to the specified driving conditions. Hereinafter, a specific example of the operation of the abnormality detection unit 110 in this case will be described.

FIG. 13 is a diagram illustrating an operation example of the abnormality detection unit 110. The upper part of Fig. 13 shows the braking level. The middle section shows the deceleration of the brake. The lower part shows the deviation between the measured value and the predicted value obtained by the prediction model. In the prediction model, the braking level corresponds to the explanatory variable, and the deceleration corresponds to the target variable.

At time t1, an operation is performed to bring the braking level to the fourth level. The braking system accepts this operation and applies a braking force to the vehicle. Therefore, the deceleration of the vehicle increases and thereafter stabilizes at a fixed value. In the interval between time t1 and time t2, the predicted value of the deceleration deviates slightly from the measured value (measured value), but the deviation between the predicted value and the measured value becomes approximately the same as the deviation between the predicted value and the measured value. range. In this section, the driving environment of the vehicle satisfies the driving condition A. The threshold α corresponds to the driving condition A.

At time t2, the driving conditions satisfied by the vehicle's driving environment change from A to B. The abnormality detection unit 110 detects a change in driving conditions and changes the threshold used to β.

The driving environment and its changes can be detected based on the measured value or control command value contained in the measurement information, the route data or the weather data contained in the environment information. In addition, the driving environment and its changes can be detected based on clear instructions from the driver or command center, and wireless signals received from the ground coil.

Between time t2 and time t3, the predicted value of the deceleration is fixed, but the actual measured value changes significantly from time t1 to time t2. Therefore, the timing at which the deviation exceeds the threshold value β occurs three times, and the abnormality detection unit 110 detects the abnormality at each timing.

At time t3, an operation is performed to change the braking level from the fourth level to the second level. The braking system accepts this operation and reduces the braking force applied to the vehicle, so the deceleration of the vehicle is reduced. Between time t3 and time t4, the deviation is within the range of the threshold β, so no abnormality is detected.

At time t4, the driving conditions satisfied by the vehicle's driving environment return from B to A. Since the abnormality detection unit 110 has detected a change in driving conditions, the threshold value used is changed from β to α using the time t4 as a boundary. Between time t4 and time t5, the deviation is within the range of the threshold value α, so no abnormality is detected.

At time t5, the brake release operation is performed. The brake system accepts this operation and further reduces the braking force applied to the vehicle, so the deceleration of the vehicle is further reduced. After time t5, the deviation is also within the range of the threshold value α, so no abnormality is detected.

The reporting unit 120 notifies the terminal 700 used by the operator, driver, or maintenance personnel of the railway that the abnormality detection unit 110 has detected an abnormality. The notification may be performed by sending an e-mail, displaying a pop-up message on the operation screen of the terminal 700, or notifying using a predetermined device management protocol, or other means. The notification may also include detailed information about the abnormality (for example, the location (current value) on the map on which the abnormality has occurred, the identifier of the vehicle on which the abnormality has occurred, etc.). By receiving the notification, the user or maintenance personnel can know the meaning and details of the detected abnormality.

The screen generating unit 130 compares the presence or absence of abnormality detection, the current position of the vehicle detected during the abnormality detection, the abnormality detection model and threshold for abnormality detection, the data of the sensor, the predicted value obtained from the prediction model, etc. It is displayed on the screen display device 900. The screen generating unit 130 may be provided by the abnormality detection device 100, or may be provided by a terminal, a management server, or the like disposed in an information network of a vehicle information system or an above-ground system to which the abnormality detection device 100 is connected.

FIG. 14 shows an example of the main screen 901 displayed by the screen generating unit 130. The main screen 901 displays information of multiple groups. Here, assume that the screen display device 900 is located in a command room that manages and monitors a plurality of groups.

On the upper part of the main screen 901, vehicle-related information is displayed in a table format. Examples of display items include formation, presence or absence of abnormality detection, train number (identification of formation), ride rate, current position, etc., but other information can also be displayed. In the "Abnormality" column of the table, the presence or absence of abnormality detection is displayed. The exclamation mark "!" indicates that there is anomaly detection. It can be seen that an abnormality is detected in group B. The display related to the presence or absence of abnormality detection is only an example, and may also be expressed in other ways.

A map is displayed in the lower part of the main screen 901, and the current position of each group is displayed. For the group B where the abnormality is detected, the meaning of the abnormality detection and the model name used are displayed in a dialog box.

Click the matching group on the main screen 901 of FIG. 14, and then you can switch to the abnormal detail screen. The method of screen transition is not limited to this, and can also be performed by other methods such as prescribed keyboard operation.

FIG. 15 shows an example of the abnormal detail screen 902 converted by clicking the group B.

Fig. 15 shows the same chart as Fig. 8 on the right side of the screen. That is, a graph of a braking level, a graph of predicted deceleration, a graph of actual measured deceleration, and a graph of deviation are displayed. A fixed period including the time of abnormality detection is displayed separately. In order to visually confirm the time when the abnormality is detected, a vertically long bar-shaped drawing indicating that the threshold value is exceeded is displayed.

A check box (check box) is provided on the left side of the screen in FIG. 15 to select items for displaying the graph. It can also be different from this by setting the means for specifying the time range for graph display. By providing the above-mentioned interface, railway operators can grasp the details of the abnormalities and take prompt responses.

When the abnormality detection unit 110 detects an abnormality, it may also present a confirmation screen to the person in charge and allow the maintenance personnel to confirm the correctness of the detection result. FIG. 16 shows an example of the confirmation screen 903. When it is judged that the detection result is a false detection, the maintenance personnel inputs their correction instructions. The condition generation unit 230 corrects the detection result in the detection result database based on the instruction.

FIG. 17 shows the hardware configuration of the abnormality detection device of this embodiment. The abnormality detection device of this embodiment includes the computer device 100. The computer device 100 includes a central processing unit (Central Processing Unit, CPU) 151, an input interface 152, a display device 153, a communication device 154, a main memory device 155, and an external memory device 156, and these devices communicate with each other via a bus 157 connection.

The CPU (Central Processing Unit) 151 executes an abnormality detection program as a computer program on the main memory device 155. The abnormality detection program is a program that realizes the above-mentioned functional configurations of the abnormality detection device. By the CPU 151 executing the abnormality detection program, each functional configuration is realized.

The input interface 152 is a circuit for inputting operation signals from input devices such as a keyboard, a mouse, and a touch panel to the abnormality detection device.

The display device 153 displays data or information output from the abnormality detection device. The display device 153 is, for example, an LCD (liquid crystal display), a cathode ray tube (CRT), and a plasma display panel (PDP), but it is not limited thereto. The data or information output from the computer device 100 can be displayed by the display device 153.

The communication device 154 is a circuit for communicating the abnormality detection device with an external device by wireless or wire. The measurement information can be input from an external device via the communication device 154. The measurement information input from the external device can be stored in the information database 310.

The main memory device 155 memorizes the abnormality detection program, data required for execution of the abnormality detection program, data generated by execution of the abnormality detection program, and the like. The abnormality detection program is deployed on the main memory device 155 and executed. The main memory device 155 is, for example, random access memory (Random Access Memory, RAM), dynamic random access memory (Dynamic Random Access Memory, DRAM), static random access memory (Static Random Access Memory, SRAM), but It is not limited to this. The information database 310, the model database 320, and the test result database 330 may also be built on the main memory device 155.

The external memory device 156 memorizes the abnormality detection program, the data required for the execution of the abnormality detection program, the data generated by the execution of the abnormality detection program, and the like. When the abnormality detection program is executed, the programs or data are read by the main memory device 155. The external memory device 156 is, for example, a hard disk, an optical disk, a flash memory, and a magnetic tape, but it is not limited thereto. The information database 310, the model database 320, and the test result database 330 may also be built on the external memory device 156.

Furthermore, the abnormality detection program may be installed in the computer device 100 in advance, or may be stored in a compact disc-read only memory (Compact Disc-Read Only Memory, CD-ROM) or other memory medium. In addition, anomaly detection programs can also be uploaded to the Internet.

In addition, the computer 100 may be provided with one or more processors 151, an input interface 152, a display device 153, a communication device 154, and a main memory device 155, or peripheral devices such as a printer or a scanner may be connected.

In addition, the abnormality detection device may be constituted by a single computer 100, or may be constituted by a system including a plurality of computers 100 connected to each other.

18 is a flowchart of abnormality detection processing performed in the operation mode of the embodiment of the present invention. The process of the flowchart of FIG. 18 can be executed on the occasion of an action of the system to be detected as an abnormality, or it can be executed at a fixed cycle, and can also be executed at the timing when an instruction is received from a user such as a maintenance person. Execute at other times.

In step S101, the abnormality detection unit 110 obtains the travel information to be the target of abnormality detection from the information database 310.

In step S102, the abnormality detection unit 110 selects a prediction model corresponding to the system to be detected as an abnormality (here, the vehicle's braking system) from the model database 320. In addition, a threshold value corresponding to the driving conditions satisfied by the obtained driving information among the plurality of driving conditions is selected. As an example, the prediction model is a model that predicts a target variable that represents the state of the vehicle (deceleration, etc.) based on an explanatory variable that represents the control command value (braking level, etc.) for the vehicle. That is, the prediction model is a model that associates the interpretation variable representing the control command value for the vehicle with the target variable representing the state of the vehicle.

In step S103, the abnormality detection unit 110 generates a feature vector based on the acquired driving information. For example, a feature vector containing control instruction values is generated. The number of elements of the feature vector may be one or more. The abnormality detection unit 110 predicts the target variable (here, deceleration) based on the feature vector and the prediction model. That is, the abnormality detection unit 10 calculates the predicted value of the state of the vehicle based on the control command value and the predicted model.

In step S104, the abnormality detection unit 110 calculates the deviation as the difference between the predicted deceleration and the deceleration contained in the driving information, and compares the calculated deviation with the threshold value.

If the deviation is greater than the threshold value (YES), the abnormality detection unit 110 detects an abnormality and outputs information notifying abnormality detection to the screen display device 900 or the like (S105).

If the deviation is below the threshold (NO), the abnormality detection unit 110 does not detect abnormality (S106). That is, the abnormality detection unit 110 determines that the brake system of the vehicle is normal. When no abnormality is detected, information notifying that the braking system of the vehicle is normal may be output to the screen display device 900 or the like.

19 is a flowchart of a threshold value setting process related to the learning mode of the abnormality detection device. This process can be executed at a fixed cycle, and can also be executed at the timing indicated by the maintenance personnel, or at other timings. It shows an operation example when a plurality of thresholds corresponding to driving conditions are set for the prediction model. As a premise, it is assumed that the abnormality detection unit 110 performs abnormality detection based on a prediction model and a threshold value generated in advance, and stores data related to the abnormality detection in the detection result database 330.

In step S201, the condition generating unit 230 assigns a divergence level according to the divergence of the divergence value to the divergence between the predicted value and the measured value according to the detection result database 330. The condition generating unit 230 generates a data set that associates the deviation level with the driving information (see FIG. 10 ).

In step S202, the condition generating unit 230 sets each item of the data set as an explanatory variable, sets the deviation level as a target variable, and performs machine learning, etc., thereby generating a prediction based on at least one explanatory variable among a plurality of explanatory variables The classifier of the target variable. That is, a classifier is generated that associates multiple conditions related to at least one explanatory variable with multiple divergence levels. Here, a decision tree (see FIG. 11) is generated as a classifier.

In step S203, the condition generating unit 230 acquires a plurality of conditions included in the classifier as a plurality of driving conditions. In the case of a decision tree, the conditions included in the path from each deviation level node (end node) to the root node are obtained as the driving conditions corresponding to each deviation level.

In step S204, the threshold setting unit 220 sets a plurality of thresholds for a plurality of driving conditions. As an example, the threshold value setting unit determines the threshold value based on the distribution of deviations that are classified into the travel information and satisfy the deviation conditions of each driving condition. For example, the driving information used for generating the decision tree (or the driving information not used for generating the decision tree) is classified into groups satisfying each driving information. Perform anomaly detection on each group, and calculate the deviation based on the detection results. Furthermore, a probability distribution of divergence is generated (see FIG. 8). The value of the deviation corresponding to the prescribed probability (upper X percentage points, etc.) in the probability distribution, or the value based on 2 times or 3 times the standard deviation σ is determined as the threshold value.

In step S205, the threshold setting unit 220 associates the multiple thresholds and the multiple groups of the multiple driving conditions with the corresponding prediction model, and stores them in the model database 320.

In the present embodiment, the case where the target variable of the prediction model is the deceleration of the brake has been described as an example, but a prediction model that predicts other states of the vehicle, for example, the braking distance of the brake may be used instead. The braking distance can be measured, for example, by calculating the distance from the start of braking to the stop of braking, or by calculating the distance until the desired deceleration or speed is reached. In addition, a prediction model that predicts both the deceleration of the brake and the braking distance may be used. In this case, as an example, formula (1) is prepared for each of the deceleration and the braking distance. As a result, the number of target variables of the prediction model becomes two, and thus, the target variables of the prediction model may be more than one. In this case, abnormality can be detected for all target variables when the deviation exceeds the threshold, or for any target variable, if the deviation exceeds the threshold, the abnormality is detected.

According to the present embodiment, by setting the threshold value corresponding to the driving conditions, it is possible to generate many abnormality detection models suitable for the conditions related to a wide range of changes. For example, due to multiple time periods in the morning, noon, night, etc., routes in many areas such as the metropolitan area, suburbs, hills, etc., the four seasons of spring, summer, autumn, winter, rain, snow, fine weather, etc. , You can set the threshold suitable for meticulous conditions.

In the first embodiment, a plurality of thresholds are set for the same prediction model according to the driving conditions, but as a second embodiment, a plurality of abnormality detection models (prediction models and thresholds) may be generated corresponding to the driving conditions. Multiple groups of values). In this case, when anomaly detection is performed, a driving condition that satisfies the current driving information is specified, and an anomaly detection model (prediction model and threshold) corresponding to the specified driving condition is used.

The model generation unit 210 generates prediction models for a plurality of driving conditions. The threshold setting unit 220 sets a threshold corresponding to each prediction model (that is, a threshold corresponding to each driving condition).

Specifically, the model generating unit 210 generates a plurality of running conditions in the same manner as in the first embodiment. The model generating unit 210 extracts data satisfying each driving condition from the driving information, and uses the extracted data to generate a prediction model. The method for generating the prediction model is the same as the above embodiment. In addition, the threshold setting unit 220 sets the threshold corresponding to each prediction model (that is, the threshold corresponding to each driving condition) in the same manner as in the above-described embodiment. The generated prediction model, the set threshold value, and the driving conditions corresponding to both are stored in the model database 320. An example of the model database 320 of the second embodiment is shown in FIG. 20. Model 0001_A, model 0001_B, and model 0001_C are generated instead of model 0001 in FIG. 7. That is, three abnormal detection models are newly generated to replace one abnormal detection model. The generation of multiple models in place of one model as described is called model segmentation. In addition, a column storing driving conditions corresponding to the driving conditions of each model is added.

In the case where the decision tree of FIG. 11 is generated, the model 0001_A is used when the driving conditions of “when it is sunny (no precipitation) and the occupancy rate is 90% or less” are satisfied. The model 0001_B is used when the driving conditions "on a clear day (no precipitation) and the ride rate is greater than 90%" are satisfied. The model 0001_C is used when the driving conditions of "in rainy days (without precipitation)" are satisfied.

With respect to the abnormality detection model generated in the present embodiment, the model segmentation is repeated recursively, whereby many abnormality detection models suitable for conditions related to a wide range of changes can be generated. For example, due to multiple time periods in the morning, noon, night, etc., routes in many areas such as the metropolitan area, suburbs, hills, etc., the four seasons of spring, summer, autumn, winter, rain, snow, fine weather, etc. , Can generate anomaly detection models suitable for detailed conditions.

This embodiment and the first embodiment may be combined. That is, it is possible to set a plurality of thresholds corresponding to the driving conditions for each of the plurality of abnormality detection models generated by model division. With this, an abnormality detection model corresponding to more detailed conditions can be generated.

In addition, the present invention is not directly limited by the above-mentioned embodiments, but may be embodied by modifying the constituent elements within the scope of the implementation stage without departing from the gist thereof. In addition, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. In addition, for example, a configuration in which several components are deleted from all the components shown in the embodiments may be considered. Furthermore, the components described in different embodiments may be combined as appropriate.

10‧‧‧brake lever

20‧‧‧ Orbit

30‧‧‧wheel

41‧‧‧brake block

42‧‧‧Tread brake

43‧‧‧Cylinder

50‧‧‧ Load compensation device

51‧‧‧Air cushion

60a, 60b‧‧‧Main motor

70‧‧‧Resistor

80‧‧‧Pantograph

90‧‧‧Wire

100‧‧‧Abnormality detection device/computer device

101‧‧‧Vehicle Information Collection Department

102‧‧‧Environmental Information Collection Department

103‧‧‧Data Processing Department

110‧‧‧ Anomaly Detection Department

111‧‧‧Model Selection Department

120‧‧‧Reporting Department

130‧‧‧Image generation unit

151‧‧‧CPU

152‧‧‧ input interface

153‧‧‧Display device

154‧‧‧Communication device

155‧‧‧Main memory device

156‧‧‧External memory device

157‧‧‧bus

200‧‧‧ Anomaly detection model generator

210‧‧‧Model generation department

220‧‧‧Provision Limit Setting Department

230‧‧‧Condition Generation Department

310‧‧‧ Information database

310a, 310b, 310c‧‧‧ Form

320‧‧‧ Model database

330‧‧‧ Test results database

400, 401, 402‧‧‧normal distribution

500‧‧‧Vehicle system

600‧‧‧Environmental Information System

700‧‧‧terminal

800‧‧‧Input device

900‧‧‧Screen display device

901‧‧‧ main screen

902‧‧‧Exception detail screen

903‧‧‧Confirmation screen

1001a, 1001b, 1001c ‧‧‧ nodes

A, B‧‧‧ Driving conditions

t, t1～t5‧‧‧‧

Alpha, beta‧‧‧limit value

σ‧‧‧ standard deviation

S101～S106、S201～S205‧‧‧Step

FIG. 1 is a block diagram of an abnormality detection system according to an embodiment of the present invention. 2 is a diagram showing a configuration example of a brake notch (brake notch), a brake, and an air spring of a railway vehicle. FIG. 3 is a diagram showing a configuration example of a power generating brake and a regenerative brake of a railway vehicle. 4 is a diagram showing an example of a table related to measurement information and environmental information. 5 is a diagram showing an example of a table related to braking information. 6 is a diagram showing an example of a conversion table. 7 is a diagram showing an example of a model database. FIG. 8 is a diagram showing an example of a method for determining a threshold value using a normal distribution. 9 is a diagram showing an example of a detection result database. FIG. 10 is a diagram showing an example of a data set for generating driving conditions. FIG. 11 is a diagram showing an example of a decision tree. 12 is a diagram showing another example of the model database. 13 is a diagram showing an example of the operation of the abnormality detection model. 14 is a diagram showing an example of a main screen output by the abnormality detection device. 15 is a diagram showing an example of an abnormality detailed screen output by the abnormality detection device. 16 is a diagram showing an example of a confirmation screen for a driver. FIG. 17 is a diagram showing the hardware configuration of the abnormality detection device of the present embodiment of the invention. 18 is a flowchart of anomaly detection processing according to an embodiment of the present invention. 19 is a flowchart of processing related to the learning mode of the abnormality detection device. 20 is a diagram showing another example of the model database.

Claims (12)

An abnormality detection device including: a condition generating unit that generates a difference between a predicted value of the state of the vehicle based on a control command value and a prediction model and a measured value of the state of the vehicle based on the driving information of the vehicle A plurality of conditions; a threshold setting unit, which sets a plurality of thresholds for the plurality of conditions; and an abnormality detection unit, based on the prediction model, the plurality of thresholds, and the plurality of conditions Carry out abnormal detection of the vehicle. The abnormality detection device according to item 1 of the patent application scope, wherein the condition generation unit uses a data set including a plurality of explanatory variables based on the driving information of the vehicle and a level corresponding to the difference to generate At least one plurality of conditions related to the interpretation variable establishes a classifier corresponding to the plurality of levels, and the plurality of conditions related to the classifier are set to the plurality of differentiating the difference condition. The abnormality detection device according to item 2 of the patent application range, wherein the threshold setting unit determines the threshold based on the distribution of the difference classified into the levels by the classifier. The abnormality detection device according to item 3 of the patent application range, wherein the threshold setting unit generates a probability distribution of the difference, and sets a value based on the standard deviation of the probability distribution in the probability distribution The value of the difference corresponding to a predetermined probability is set as the threshold value. The abnormality detection device according to any one of items 1 to 4 of the patent application range, wherein the threshold setting unit accepts the setting instructions of the plurality of thresholds via a user interface, and according to the The setting instruction to set the plurality of thresholds. The abnormality detection device according to any one of claims 1 to 5, wherein the abnormality detection unit calculates the vehicle based on the control command value corresponding to the first time point and the prediction model The predicted value of the state of the vehicle, and specifies the condition that the driving information at the first time point meets among the plurality of conditions, and the predicted value is compared with the measured value of the state of the vehicle The difference is compared with the threshold value corresponding to a specific condition, thereby detecting whether the vehicle is abnormal. The abnormality detection device according to any one of items 1 to 6 of the patent application scope, which includes a model generation unit that generates a control command value for the vehicle for the plurality of conditions A plurality of prediction models corresponding to the state of the vehicle, and the abnormality detection unit performs abnormality detection of the vehicle based on the plurality of prediction models, the plurality of thresholds, and the plurality of conditions. The abnormality detection device as described in item 7 of the patent application range, wherein the abnormality detection unit specifies the condition that the driving information corresponding to the first time point satisfies among the plurality of conditions, and according to the corresponding The prediction model of a specific condition and the control command value corresponding to the first time point, the predicted value of the state of the vehicle is calculated, and the difference between the predicted value and the measured value of the state of the vehicle is calculated by The threshold values corresponding to the specified conditions are compared, thereby detecting whether the vehicle is abnormal. The abnormality detection device according to any one of items 1 to 8 of the patent application range, wherein the control command value is a command value related to the size of the brake of the vehicle, and the state includes the vehicle's Deceleration, or air brake pressure. The abnormality detection device according to any one of claims 1 to 9, wherein the driving information includes at least one of measurement information of at least one sensor of the vehicle and environmental information of the vehicle By. An abnormality detection method that generates a plurality of conditions that distinguish the difference between the predicted value of the state of the vehicle and the measured value of the state of the vehicle based on the control command value and the prediction model based on the driving information of the vehicle, The plurality of conditions set a plurality of thresholds, and abnormality detection of the vehicle is performed based on the prediction model, the plurality of thresholds, and the plurality of conditions. A computer-readable recording medium that stores a program that causes the computer to perform the following steps: According to the driving information of the vehicle, generate a prediction value for the state of the vehicle based on the control command value and a prediction model, and the vehicle’s A plurality of conditions for distinguishing the difference between the measured values of the state; a step of setting a plurality of thresholds for the plurality of conditions; and based on the prediction model, the plurality of thresholds, and the plurality of conditions The step of detecting the abnormality of the vehicle.