JP2021140297A - Abnormality diagnosis device - Google Patents

Abstract

To provide a technology that creates a new diagnosis model which can be used for diagnosis even if there is no diagnosis model that matches the condition of the running data on a vehicle to be diagnosed.SOLUTION: When unknown conditions (an environment, a vehicle, and a driver) are contained in the running data on a vehicle to be diagnosed, among the plurality of probability distributions that approximate the already-learnt and already-existing distributions of the pieces of running data on the plurality of vehicles and of drivers, and the mixed distributions of such probability distributions, the one that is similar to the diagnosis condition is utilized to obtain a new mixed distribution as a predicted distribution. A diagnosis model is created on the basis of the predicted distribution, thereby determining an abnormality.SELECTED DRAWING: Figure 10

Description

The present invention relates to an abnormality diagnostic device used in a vehicle or the like.

In the diagnosis of abnormalities in vehicle parts or vehicle functions, the behavior of the vehicle differs depending on the environment in which the vehicle travels (for example, general roads, highways, road surface conditions, time zones, weather, etc.). Even if an attempt is made to diagnose an abnormality in driving data on an expressway using a vehicle component diagnostic model learned based only on the basis, the accuracy of the diagnosis will be low.

In order to improve the accuracy of this diagnosis, for example, there is a technique described in Patent Document 1. In Patent Document 1, the diagnostic model required for the diagnosis of each automobile part is stored in the diagnostic model storage unit in association with the target part name and the environmental information. In the brake diagnosis model, "deceleration" and "deceleration distance" are set as feature amounts, in the engine diagnosis model, "engine speed" and "engine coolant temperature" are set as feature amounts, and in the battery diagnosis model, "battery voltage". "And" acceleration "are set as feature quantities.

JP-A-2019-123351

In Patent Document 1, since it is assumed that a diagnostic model for each environment according to a part is learned in advance, there is a problem that an abnormality diagnosis cannot be performed if there is no corresponding diagnostic model. ..

An object of the present invention is to provide an abnormality diagnostic apparatus capable of generating a new diagnostic model that can be diagnosed even when there is no diagnostic model that meets the conditions of data relating to the movement of a moving body to be diagnosed. be.

In the present invention, when unknown conditions (environment, moving body, operator) are included in the data related to the movement of the moving body to be diagnosed, the moving data of a plurality of existing moving bodies and the operator learned in advance From a plurality of probability distributions that approximate the distribution and their mixture ratios, a new mixture distribution is obtained as a predicted distribution using the one that is close to the condition to be diagnosed, and a diagnostic model is generated based on the predicted distribution, resulting in an abnormality. Is to be determined.

In order to realize the above, the present invention presents a diagnosis obtained from learning movement data in which the moving body and an operator are combined in an abnormality diagnosis device that determines the possibility of abnormality of the moving body from the movement data of the moving body. A data distribution that generates a probability distribution of the feature amount of the target part and calculates the mixture ratio of the combination of the moving body and the operator to be diagnosed in order to obtain a mixed distribution that approximates the probability distribution of the combination of the moving body and the operator to be diagnosed. Diagnosis is performed with an approximation unit, a probability distribution storage unit that stores the parameters of the probability distribution generated by the data distribution approximation unit, and a mixture ratio storage unit that stores the mixture ratio calculated by the data distribution approximation unit. Based on the combination of the moving body and the operator, the mixed distribution of the combination of the moving body and the operator diagnosed from the probability distribution stored in the probability distribution storage unit and the mixing ratio stored in the mixing ratio storage unit. Is provided, and a diagnostic unit for determining an abnormality by comparing the generated mixture distribution with the distribution of moving data to be diagnosed is provided.

According to the present invention, it is possible to provide an abnormality diagnostic device capable of generating a new diagnostic model that can be diagnosed even when there is no diagnostic model that meets the conditions of data relating to the movement of the moving body to be diagnosed. can.

Issues, configurations and effects other than those described above will be clarified by the description of the following embodiments.

It is a block diagram which shows the structure about Example 1 of this invention. It is a block diagram which shows the detailed structure of the data extraction part 112 in FIG. It is a figure which shows an example of the data which concerns on Example 1 of this invention. It is a figure which shows an example of the divided data which concerns on Example 1 of this invention. It is a figure which shows the extraction period of the data which concerns on Example 1 of this invention. It is a figure which shows an example of the combination of the data in the driver and the vehicle which concerns on Example 1 of this invention. It is a flowchart which shows the operation of the data distribution approximation part which concerns on Example 1 of this invention. It is a figure which shows an example of the approximation to the probability distribution of the feature amount data by the data distribution approximation part 113 concerning the Example of the invention. It is a figure which shows an example of the approximation to the probability distribution of the feature amount data by the data distribution approximation part 113 concerning the Example of the invention. It is a figure which shows an example of the approximation to the probability distribution of the feature amount data by the data distribution approximation part 113 concerning the Example of the invention. It is a figure which shows an example of the approximation to the probability distribution of the feature amount data by the data distribution approximation part 113 concerning the Example of the invention. It is a figure which shows an example of the approximation to the mixture distribution of the feature amount data by the data distribution approximation part 113 which concerns on Example 1 of the invention. It is a figure which shows an example of the approximation to the mixture distribution of the feature amount data by the data distribution approximation part 113 which concerns on Example 1 of the invention. It is a figure which shows an example of the mixture distribution according to the driving environment of the vehicle which concerns on Example 1 of this invention. It is a figure which shows an example of the mixture distribution according to the driving environment of the vehicle which concerns on Example 1 of this invention. It is a figure which shows an example of the mixture distribution according to the driving environment of the vehicle which concerns on Example 1 of this invention. It is a flowchart which shows the process in the data distribution approximation part 113 when the learning driving data of the combination of a new vehicle and a driver, or the combination of a vehicle and a new driver which concerns on Example 1 of this invention is obtained. It is a figure which shows an example of the probability distribution which concerns on Example 1 of this invention. It is a figure which shows an example of the probability distribution which concerns on Example 1 of this invention. It is a figure which shows an example of the probability distribution which concerns on Example 1 of this invention. It is a block diagram of the diagnostic unit 12 including the detailed block of the diagnostic model generation unit 122 according to the first embodiment of the present invention. It is a flowchart which shows the operation in the mixture ratio search unit 1222 and the mixture distribution generation unit 1223 which concerns on Example 1 of this invention. It is a block diagram which shows the detailed structure of the abnormality degree calculation unit 123 which concerns on Example 1 of this invention. It is a figure which shows the predicted distribution 1301 when the feature quantity about Example 1 of this invention is made one-dimensional. It is a figure which shows an example of the data point which spreads in the feature quantity space about Example 1 of this invention. It is a figure which shows an example which extracted a plurality of probability distribution parts concerning Example 1 of this invention. It is a figure which shows an example of the mixture distribution which concerns on Example 1 of this invention. It is a figure which shows an example of the mixture distribution which concerns on Example 1 of this invention. It is a block diagram which shows the whole structure for performing the online diagnosis of the vehicle which concerns on Example 2 of this invention.

Hereinafter, examples of the present invention will be described with reference to the accompanying drawings. Similar components are designated by the same reference numerals, and the same description will not be repeated.

The various components of the present invention do not necessarily have to be independent of each other, and one component is composed of a plurality of members, a plurality of components are composed of one member, and a certain component is different. It is allowed that a part of one component overlaps with a part of another component.

FIG. 1 is a block diagram showing a configuration according to a first embodiment of the present invention. FIG. 1 assumes a configuration in which an abnormality diagnosis of a vehicle is performed in an offline state.

In FIG. 1, the abnormality diagnosis device 1 includes a learning unit 11, a diagnosis unit 12, a probability distribution storage unit 13, and a mixing ratio storage unit 14. These functions are realized by loading a program stored in an auxiliary storage device such as a hard disk included in the abnormality diagnosis device 1 into a main storage device such as a semiconductor memory and executing this by an arithmetic unit such as a CPU. In the following description, such a well-known operation will be omitted as appropriate.

The abnormality diagnosis device 1 is a device that determines the possibility of an abnormality related to the vehicle from the data related to the running of the vehicle. Connect with a network or the like.

The learning unit 11 includes a learning data input unit 111, a data extraction unit 112, and a data distribution approximation unit 113. The learning travel data 2 in which a plurality of vehicles and a plurality of drivers are combined, which has been accumulated in advance, is input to the learning data input unit 111.

The data extraction unit 112 extracts a feature amount necessary for diagnosing the part to be diagnosed from the learning travel data 2 input to the learning data input unit 111. The extracted feature amount approximates the feature amount space with a plurality of probability distributions and a mixed distribution using the probability distribution in the data distribution approximation unit 113, and mixes the probability distribution with the probability distribution storage unit 13 and the mixture ratio of the probability distribution. It is stored in the ratio storage unit 14.

The diagnosis unit 12 includes a diagnosis data input unit 121, a diagnosis model generation unit 122, an abnormality degree calculation unit 123, and a diagnosis result output unit 124. The acquired diagnostic travel data 3 is input to the diagnostic data input unit 121. The diagnostic model generation unit 122 generates a prediction distribution from the plurality of probability distributions stored in the probability distribution storage unit 13 and the mixing ratio of the plurality of probability distributions stored in the mixing ratio storage unit 14, and is based on the prediction distribution. Output the diagnostic model. The abnormality degree calculation unit 123 calculates the abnormality degree of the vehicle from the diagnostic model and the diagnostic data. The diagnosis result output unit 124 outputs the determination result of the abnormality of the vehicle from the calculated abnormality degree and the trend of the abnormality degree in the past.

The learning travel data 2 and the diagnostic travel data 3 are travel data collected from the vehicle in advance. Here, it is desirable that the learning travel data 2 is data on which the vehicle is likely to be normal, for example, data in which no component failure or emergency maintenance is performed immediately after the data acquisition. Further, the diagnostic travel data 3 is not included in the learning travel data 2. The diagnostic travel data 3 and the learning travel data 2 are stored in a storage device installed in the vehicle, and the data is taken out from this storage device after the travel is completed.

Next, the configuration of the data extraction unit 112 will be described. FIG. 2 is a block diagram showing a detailed configuration of the data extraction unit 112 in FIG.

In FIG. 2, the data extraction unit 112 includes a driver / vehicle-specific data collection unit 1121, a specific travel mode extraction unit 1122, a feature amount calculation unit 1123, and a standardization processing unit 1124.

The driver / vehicle-specific data collection unit 1121 receives the input data input to the learning data input unit 111, and divides the data for each combination of the driver and the vehicle. An example of the input data is shown in FIG.

FIG. 3A is a diagram showing an example of data relating to the first embodiment of the present invention. FIG. 3B is a diagram showing an example of divided data according to the first embodiment of the present invention. FIG. 3C is a diagram showing a data extraction period according to the first embodiment of the present invention. FIG. 3D is a diagram showing an example of a combination of data in a driver and a vehicle according to a first embodiment of the present invention.

In FIG. 3A, the input data can distinguish a unique driver (driver name, driver ID, etc.), a unique vehicle or a unique vehicle type, and a unique vehicle specification (vehicle name or vehicle). ID, vehicle model name, vehicle model ID, ID for each vehicle spec, etc.) and sensor data are included. The sensor data is provided from the vehicle or a data collection terminal mounted on the vehicle, and includes time-series acceleration, speed, and position information, as well as internal data of the vehicle (engine speed) according to the vehicle parts to be diagnosed. Etc.) and vehicle operation data (brake operation, etc.) are preferably included.

Further, as shown in FIG. 3A, the input data includes the driving data of a plurality of vehicles even if the same driver is used, or the driving data of a plurality of drivers even if the same vehicle is used. There is. In the logistics industry and the like, each of a plurality of bases has a plurality of vehicles and a plurality of drivers. The driver has a driving habit peculiar to the driver, and there is a difference in driving skill between a skilled driver and a new driver, so that the operation for the vehicle is different. In this embodiment, in order to create a diagnostic model in consideration of such a driver difference and distinguish the combination of the vehicle and the driver, the driver / vehicle-specific data collection unit 1121 uses the driver and the driver as shown in FIG. 3B. The data is divided for each combination of vehicles.

The specific travel mode extraction unit 1122 extracts travel mode data necessary for diagnosing the vehicle parts to be diagnosed. The driving mode is based on idling, acceleration, constant speed, and deceleration. In the traveling section from the departure of the vehicle to the destination, idling, acceleration, constant speed, and deceleration are repeatedly performed, and a plurality of traveling modes are generated. In diagnosing vehicle parts, necessary driving data is obtained from this driving mode. For example, in the diagnosis of the braking device, traveling data relating to deceleration, that is, braking operation is required from the traveling mode, and other than that is unnecessary. Therefore, as shown in FIG. 3C, the specific running mode extraction unit 1122 analyzes the running mode in which running and stopping are repeated, and sets one running mode from the start to the stop of running to a plurality of running modes such as acceleration, constant speed, and deceleration. It is divided into running data (divided running data 1122a), and only the deceleration running data, that is, the data of the braking period 1122b is extracted from each of the divided running data 1122a.

The feature amount calculation unit 1123 calculates a feature amount suitable for the vehicle part to be diagnosed from the extracted time-series data of the traveling mode. Feature quantities include statistical values of time-series data (average value, variance value, maximum value, minimum value, etc.), operation time of vehicle parts to be diagnosed (brake operation time, etc.), and operation results (mileage during brake operation). Etc.) is desirable.

As shown in FIG. 3D, the standardization processing unit 1124 standardizes each feature amount for each run calculated by the feature amount calculation unit, and aggregates each feature amount for each combination of the driver and the vehicle.

In the diagnostic model according to this embodiment, a plurality of probability distribution parts 42, 43, 44 as shown in FIG. 14B are extracted from the distribution of each data point 41 spreading in the feature space as shown in FIG. 14A, and these parts are extracted. By determining the mixing ratio with respect to the above, a mixing distribution 45 for each combination of the driver and the vehicle as shown in FIG. 14C or FIG. 14D is created and used as a predicted distribution. By using this prediction distribution, when the driving data to be diagnosed is obtained, the probability based on each obtained data point can be obtained. When there is an abnormality in the vehicle part to be diagnosed, the distribution of each data point in the predetermined feature space is different, so that the probability at each data point is low. Taking advantage of this property, the diagnostic model calculates the degree of anomaly based on the probability of data points based on the predicted distribution. The data extraction unit 112 identifies the driving mode required for the vehicle part to be diagnosed, extracts data for each combination of the vehicle and the driver, calculates the feature amount required for the vehicle part to be diagnosed from the extracted data, and distributes the data. It is transmitted to the approximation unit 113.

The data distribution approximation unit 113 generates a mixture distribution that is the above-mentioned predicted distribution. As a method for generating a mixture distribution, there is a method for estimating variational Bayes, and by applying this algorithm, it is possible to generate a mixture distribution for each combination of a driver and a vehicle.

Next, the outline of the operation of the data distribution approximation part when the variational Bayesian estimation method is applied with reference to FIG. 4 will be described. FIG. 4 is a flowchart showing the operation of the data distribution approximation unit according to the first embodiment of the present invention.

In step S101, the D-dimensional feature data extracted by the data extraction unit 112 is divided into M clusters, and the average vector and covariance matrix of each cluster are calculated. In FIG. 3D, the feature amounts for each run of each combination of the driver and the vehicle are summarized, but in step S101, the combination of the driver and the vehicle is not particularly concerned and is divided into M pieces. Here, D is the number of features used for diagnosis, and M is the maximum number of probability distributions that are components of the mixture distribution. For each data divided into M pieces, the average vector and the covariance matrix are obtained. Here, the data is divided into M pieces without considering the distribution of the data, but it may be divided into M pieces by applying a k-means algorithm or the like.

In step S102, the distribution of data is assumed to be the superposition of the Gaussian distribution of each cluster, and the burden rate for each cluster divided in step S101 is calculated for each feature data point. Here, the burden rate indicates the probability that each data point for each combination of the driver and the vehicle belongs to each cluster.

In step S103, the mixing ratio is calculated based on the total burden rate on each cluster for each combination of the driver and the vehicle.

In step S104, the mean vector and the covariance matrix are updated based on the burden factor. As a result, the center of each cluster moves, and the data spread of each cluster changes.

In step S105, the likelihood is calculated.

In step S106, it is determined whether or not the likelihood has converged, that is, whether or not the increase in the likelihood has subsided. Under the covariance matrix, the processes of steps S102 to S105 are repeated until the likelihoods converge.

In FIG. 4, the mixing ratio of a specific cluster (probability distribution) may be close to zero. In this case, the mixing ratio may be regarded as 0. Further, when the mixing ratio of a specific cluster is 0 in any combination of driver and vehicle, the number of clusters is M', and the average vector and covariance matrix of the corresponding cluster need not be stored.

The parameters of the probability distribution of each cluster obtained in FIG. 4 are stored in the probability distribution storage unit 13. For example, in the case of Gaussian distribution as described above, the mean vector and the covariance matrix are stored as parameters. Further, the mixing ratio calculated for each combination of the vehicle and the driver is stored in the mixing ratio storage unit 14.

Next, an example of approximation of the feature data to the probability distribution will be described with reference to FIG. 5A to 5D are diagrams showing an example of approximation of the feature amount data to the probability distribution by the data distribution approximation unit 113 according to the embodiment of the present invention.

FIG. 5A is a plot of D-dimensional feature data points 1131 for feature m and feature n. On the other hand, the data distribution approximation unit 113 generates the probability distributions 1132, 1133, 1134 as shown in FIGS. 5B, 5C, and 5D, respectively. The center of each probability distribution 1132, 1133, 1134 is determined by the average vector obtained in FIG. 4, and the spread and direction of the probability distribution are determined by the covariance matrix obtained in FIG.

6A and 6B are diagrams showing an example of approximation of the feature amount data to the mixture distribution by the data distribution approximation unit 113 according to the first embodiment of the present invention. In FIGS. 6A and 6B, the data distribution approximation unit 113 sets a mixing ratio for each combination of the driver and the vehicle with respect to the probability distributions 1132, 1133, 1134 shown in FIGS. 5B, 5C, and 5D. Mixture distributions 1135, 1136 are generated based on the mixing ratio, such as FIG. 6A for a driver-vehicle combination and FIG. 6B for another driver-vehicle combination.

That is, the data distribution approximation unit 113 generates a probability distribution of the feature amount of the part to be diagnosed obtained from the learning driving data 2 in which the vehicle and the driver are combined, and approximates the probability distribution of the combination of the vehicle and the driver to be diagnosed. Calculate the mixing ratio of the vehicle and driver combination to be diagnosed in order to obtain the mixing distribution.

Further, the data distribution approximation unit 113 generates a mixture distribution according to the traveling environment of the vehicle. 7A to 7C are diagrams showing an example of a mixture distribution according to the traveling environment of the vehicle according to the first embodiment of the present invention. It is assumed that the prediction distribution as shown in FIG. 7A is generated when a lot of driving environment data has been acquired for a certain combination of a driver and a vehicle. The state of the vehicle changes depending on the environment in which the vehicle travels, even if the combination of the same vehicle and the driver is used. For example, if the environment in which a vehicle travels, such as a general road and an expressway, is different, the state of the vehicle changes even if the combination of the same vehicle and the driver is used. Compared to general roads, on highways, idling time tends to be short, acceleration tends to be large, constant speed time tends to be long, and deceleration tends to be large. In addition, the amount of steering wheel operation tends to be smaller on expressways than on ordinary roads. In this embodiment, a mixture distribution according to the environment in which the vehicle travels is also created. For example, a mixture distribution according to the driving environment is generated as shown in FIG. 7B on a general road and FIG. 7C on an expressway. The driving environment can be distinguished by the average speed value, position information, road matching information based on position information, area information, and area-based weather information. The data distribution approximation unit 113 generates a probability distribution and a mixing ratio corresponding to the environment by using the learning driving data in which the vehicle and the driver are combined. Then, the data distribution approximation unit 113 stores the probability distribution generated for each environment and the mixture ratio for generating the mixture distribution for each environment in the probability distribution storage unit 13 and the mixture ratio storage unit 14, respectively. The traveling environment of the vehicle can be defined by the data extraction ratio such as the traveling data at the time of braking operation, the traveling data at the time of accelerator operation, and the traveling data at the time of steering wheel operation in the traveling section. For example, if the traveling data in a certain section includes almost no braking operation, it is highly possible that the vehicle is traveling on an expressway. On the other hand, if the braking operation is frequently included, it is highly possible that the vehicle is traveling in a congested section. In this embodiment, the environment is a range in which the extraction ratio of at least one of the driving data at the time of braking operation, the driving data at the time of accelerator operation, and the driving data at the time of steering wheel operation in the traveling section is predetermined. It is assumed that the vehicle is running on the condition that it is in.

The definition of the environment can also be defined by the position information of the vehicle, the map information based on the position information, and the weather information. For example, determining the road type (expressway, general road) based on the position information of the vehicle, estimating the slope of the road on which the vehicle is traveling from the map information, and weather information (rainfall and snowfall) based on the position information. It is possible to define the environment by estimating the road surface condition by acquisition.

Next, in a situation where the probability distribution and the mixing ratio for a combination of a plurality of drivers and vehicles have already been learned, learning driving data obtained by combining a new vehicle and a driver or a vehicle and a new driver that are not learned is obtained. The processing of the data distribution approximation unit 113 at the time of occurrence will be described.

FIG. 8 is a flowchart showing processing by the data distribution approximation unit 113 when learning running data of a new vehicle and driver or a combination of vehicle and new driver according to the first embodiment of the present invention is obtained.

In FIG. 8, in step S201, the D-dimensional feature amount data obtained by the new vehicle and driver or the combination of the vehicle and the new driver extracted by the data extraction unit 112 is divided into MM'clusters, and each of them is divided into MM'clusters. Calculate the mean vector and covariance matrix of the cluster. Here, M'is the final number of clusters in FIG. If M'is a value equal to or greater than M, the value of M may be increased. Further, as in the case of FIG. 4, here, the distribution of the feature amount data due to the combination of the new vehicle and the driver or the combination of the vehicle and the new driver is not considered, and the data is divided into MM's. It may be divided into M pieces by applying a means algorithm or the like.

In step S202, the burden rate for each cluster divided in step S201 is calculated for each feature data point. Here, the burden rate indicates the probability that each data point for each combination of the driver and the vehicle belongs to each cluster.

In step S203, the mixing ratio is calculated based on the total burden ratio for each cluster.

In step S204, the mean vector and covariance matrix of the new cluster are updated based on the burden factor. This moves the center of the new cluster and changes the data spread of each cluster.

In step S205, the likelihood is calculated.

In step S206, it is determined whether or not the likelihood has converged, that is, whether or not the increase in likelihood has subsided. Under the vector and covariance matrix, the processes of steps S202 to S205 are repeated until the likelihoods converge.

By the above processing, the mixing ratio can be generated using the learned probability distribution, so that even when the amount of driving data due to the combination of the new vehicle and the driver or the combination of the vehicle and the new driver is small, the accuracy is high. A diagnostic model that can be diagnosed can be generated. That is, the data approximation unit in the first embodiment uses a new vehicle and a driver, or a new vehicle and a new driver when learning driving data obtained by combining a new vehicle and a driver or a vehicle and a new driver is obtained. Learning by combining a new vehicle and a driver, or a vehicle and a new driver from a mixture ratio generated by using the probability distribution learned from the combined learning driving data and the probability distribution stored in the probability distribution storage unit 13. Generate a mixture distribution of driving data.

Specific examples are shown in FIGS. 9A to 9C. 9A-9C are diagrams showing an example of a probability distribution according to the first embodiment of the present invention. In FIG. 9A, the data points 1137 in the combination of the new vehicle and the driver or the vehicle and the new driver and the learned probability distribution 1132, 1133, 1134 are simultaneously shown on the feature space. When the processing according to FIG. 8 is completed, a new cluster corresponding to the data point 1137 is created, and the corresponding probability distribution 1138 as shown in FIG. 9B is generated. Eventually, a mixture distribution of 1139 is generated for the new vehicle and driver, or the combination of vehicle and new driver, as shown in FIG. 9C.

Next, the operation of the diagnostic unit 12 will be described with reference to FIG. FIG. 10 is a block diagram of the diagnostic unit 12 including the detailed block of the diagnostic model generation unit 122 according to the first embodiment of the present invention.

The diagnostic model generation unit 122 includes a new driver / vehicle combination detection unit 1221, a mixture ratio search unit 1222, a mixture distribution generation unit 1223, and a mixture distribution acquisition unit 1225.

The new driver / vehicle combination detection unit 1221 estimates the driving environment of the corresponding driving data from the diagnostic driving data that combines the vehicle and the driver input to the diagnostic data input unit 121, and also supports the combination of the driver and the vehicle. It is determined whether or not the mixing ratio of the predicted distribution is stored in the mixing ratio storage unit 14. If it is not stored, it is determined that the combination is a new driver and a vehicle, or a combination of a driver and a new vehicle, and the process proceeds to the process of the mixing ratio search unit 1222. On the other hand, if it is stored, the process proceeds to the process of the mixture distribution acquisition unit 1225.

In the mixing ratio search unit 1222, a new driver and vehicle detected by the new driver / vehicle combination detection unit 1221 or a new driver and vehicle from the mixing ratio for each environment of the driver / vehicle combination stored in the mixing ratio storage unit 14 or If there is something that matches the driving environment of the data regarding the combination of the driver and the new vehicle, the mixing ratio is acquired in the order of vehicle> driver priority. If there are multiple items with the same priority, it is possible to take the average of the applicable mixing ratios or classify the environmental conditions in more detail and select the ones that match. For example, if there are environmental conditions, the mixing ratio is selected according to the priority of environment, vehicle, and driver.

The mixture distribution generation unit 1223 generates a mixture distribution (predicted distribution) based on the mixture ratio obtained from the mixture ratio search unit 1222 and the corresponding probability distribution stored in the probability distribution storage unit 13.

The mixture distribution acquisition unit 1225 acquires the mixture ratio corresponding to the existing combination of the driver and the vehicle stored in the mixture ratio storage unit 14 and the probability distribution corresponding to the mixture ratio stored in the probability distribution storage unit 13. , Generate a mixture distribution (predicted distribution).

FIG. 11 is a flowchart showing the operation of the mixture ratio search unit 1222 and the mixture distribution generation unit 1223 according to the first embodiment of the present invention.

In step S301, in the mixing ratio for each combination of the driver and the vehicle stored in the mixing ratio storage unit 14, the driving environment condition of the data regarding the combination of the new driver and the vehicle or the combination of the driver and the new vehicle is set. It is determined whether or not there is a match, and if so, in step S302, a mixture distribution is generated based on the mixture ratio of the corresponding environment. If not, the process proceeds to step S303.

In step S303, among the mixing ratios for each combination of the driver and the vehicle stored in the mixing ratio storage unit 14, are there any vehicle conditions that match the new driver and the vehicle or the combination of the driver and the new vehicle? Whether or not it is determined, and if there is, in step S304, a mixture distribution is generated based on the mixture ratio of the corresponding vehicle. If not, the process proceeds to step S305.

In step S305, is there any mixture ratio for each combination of the driver and the vehicle stored in the mixture ratio storage unit 14 that matches the driver conditions for the new driver and the vehicle or the combination of the driver and the new vehicle? Whether or not it is determined, and if there is, in step S306, a mixture distribution is generated based on the mixture ratio of the corresponding driver. If not, a mixture distribution is generated in step S307 with the probability distributions having an equal mixture ratio.

In the above, when there are a plurality of applicable mixing ratios, a mixed distribution is generated by taking the average of each mixing ratio.

The abnormality degree calculation unit 123 calculates the abnormality degree from the mixture distribution generated by the mixture distribution generation unit 1223 or the mixture distribution acquisition unit 1225.

FIG. 12 is a block diagram showing a detailed configuration of the abnormality degree calculation unit 123 according to the first embodiment of the present invention. The abnormality degree calculation unit 123 includes a specific traveling mode extraction unit 1231, a feature amount calculation unit 1232, a standardization processing unit 1233, and an abnormality degree calculation unit 1234.

The specific travel mode extraction unit 1231 and the feature amount calculation unit 1232 have the same functions as the specific travel mode extraction unit 1122 and the feature amount calculation unit 1123 in the data extraction unit 112 (see FIG. 2) of the learning unit 11.

The mixture distribution generated by the mixture distribution generation unit 1223 or the mixture distribution acquisition unit 1225 and the diagnostic running data 3 are input to the abnormality degree calculation unit 123.

The standardization processing unit 1233 and the feature amount calculation unit 1232 hold the average value vector and the standard deviation vector when standardized by the standardization processing unit 1124 in the data extraction unit 112 of the learning unit 11. Use this value for standardization.

The anomaly degree calculation unit 1234 calculates the anomaly degree a by the following equation, using the mixture distribution generated / acquired by the mixture distribution generation unit 1223 or the mixture distribution acquisition unit 1225 as the predicted distribution p (x).

a = -1 / (number of diagnostic data points) x (total of ln p (x) at each diagnostic data point x)
By taking the logarithmic equation of p (x) as described above, the probability of data deviating from the predicted distribution becomes low, so that it becomes a large negative value. Further, since the sum of ln p (x) at each point x is taken, the larger the amount of data deviating from the predicted distribution, the larger the negative value. Furthermore, since it is divided by the number of diagnostic data points, the value is the average value of the degree of abnormality per data point, and by adding a minus sign, the degree of abnormality increases as it deviates from the predicted distribution. It has become.

That is, the abnormality degree calculation unit 123 in the diagnosis unit 12 diagnoses from the probability distribution stored in the probability distribution storage unit 13 and the mixture ratio stored in the mixture ratio storage unit 14 based on the combination of the vehicle and the driver to be diagnosed. A mixture distribution of the combination of the vehicle and the driver is generated, and the generated mixture distribution is compared with the distribution of the diagnostic driving data to determine the abnormality.

A specific example will be described with reference to FIG. FIG. 13 is a diagram showing a predicted distribution 12231 when the feature amount according to the first embodiment of the present invention is one-dimensional.

In FIG. 13, when the value of the feature amount x is b, the probability p (b) is a large value. On the other hand, when the value of the feature amount x is a, the probability p (a) is a small value. Therefore, according to the above formula for calculating the degree of abnormality, when the diagnostic data contains many points such as the point a that deviates from the predicted distribution, the degree of abnormality becomes high.

The diagnosis result output unit 124 determines and outputs a diagnosis result, that is, whether or not there is a possibility of an abnormality in the vehicle, based on the abnormality degree calculated by the abnormality degree calculation unit 1234. In this determination, a threshold value is set for the degree of abnormality and the trend of the degree of abnormality, and when the threshold value is exceeded, it is determined to be abnormal.

In addition, the diagnosis result output unit 124 not only determines whether the diagnosis result is normal or abnormal, but also informs the driver, the owner of the vehicle, and the manager of the possibility of abnormality when the determination result of abnormality is obtained. You may want to output additional information to get credit. For example, it is conceivable to show the latest trend of the degree of abnormality or the comparison between the average value of the degree of abnormality at the time of normal and the degree of abnormality at the time of determining the abnormality.

Furthermore, it is also possible to show the features that had a large effect on the deterioration of the degree of abnormality. For example, when a braking distance is selected as a feature amount when detecting an abnormality in a braking device, if this feature amount has a great influence on the deterioration of the degree of abnormality, in addition to the abnormality determination, "braking distance is It is also possible to indicate what kind of abnormality is concrete, such as "It is growing."

According to the abnormality detection device according to the present embodiment, even if the driving data is based on a new vehicle and driver that does not have a trained diagnostic model or a combination of a vehicle and a new driver, the trained diagnostic model can be used. Data distribution characteristics that are considered to be similar can be extracted and used for diagnosis.

Further, according to the abnormality detection device according to the present embodiment, even if the amount of driving data due to the combination of the new vehicle and the driver or the combination of the vehicle and the new driver is small, the learned probability distribution and the probability of newly learning are obtained. Since a diagnostic model can be generated using the distribution, highly accurate diagnosis can be performed.

Next, Example 2 of the present invention will be described. The second embodiment is to diagnose the vehicle online.

FIG. 15 is a block diagram showing an overall configuration for performing an online diagnosis of a vehicle according to a second embodiment of the present invention. Those having the same functions as those shown in FIG. 1 (Example 1) are designated by the same reference numerals, and detailed description thereof will be omitted.

The vehicle 4 is connected to the network 6 via the communication base station 5. The traveling data of the vehicle 4 is uploaded from the communication base station 5 to the data collection server 7 via the network 6 at a specific timing. Here, the specific timing is a periodic timing such as every 1 second, every 10 seconds, every 1 minute, every one run (from start to stop), every move from the departure point to the destination, etc. It is the timing of.

The data collection server 7 includes a data acquisition unit 71, a data transmission unit 72, a diagnosis result acquisition unit 73, and a diagnosis result transmission unit 74. The data acquisition unit 71 uses the travel data uploaded from the vehicle 4 for learning travel data. It is stored as 2 and output as online data to the data transmission unit 72.

The data transmission unit 72 transmits the online data received from the data acquisition unit 71 to the abnormality diagnosis device 1.

The diagnosis result acquisition unit 73 acquires the diagnosis result output from the abnormality diagnosis device 1 and outputs the diagnosis result to the diagnosis result transmission unit 74.

The diagnosis result transmission unit 74 transmits the diagnosis result to the vehicle 4 corresponding to the received diagnosis result, or to the base of the vehicle 4 (not shown) or the management source of the vehicle 4.

The abnormality diagnosis device 1 is changed from the diagnosis unit 12 in FIG. 1 to the online diagnosis unit 15.

The operation of the diagnostic data input unit and the diagnosis result output unit of the online diagnosis unit 15 is changed as compared with the diagnosis unit 12.

The diagnostic data input unit 151 inputs diagnostic data in accordance with the transmission timing from the data transmission unit 72 of the data collection server 7.

The diagnosis result output unit 152 transmits the determined diagnosis result (presence or absence of possibility of abnormality) to the diagnosis result acquisition unit 73 of the data collection server 7.

With the above configuration, according to the second embodiment, even when the diagnostic data is acquired online, the diagnosis can be performed with high accuracy as in the case of offline.

In the present invention, attention is paid to the point that the distribution of the driving data is different for each vehicle, the distribution of the driving data is different depending on the driver of the vehicle, and the distribution of the driving data is different depending on the driving environment of the vehicle. And Example 2 is described, but the present invention is not limited to vehicles, and is applicable to equipment and operators such as construction machines and aircraft that operate in a complicated environment. That is, it may be a combination of a moving body and an operator who operates the moving body. In the case of a combination of a moving body and an operator, the learning travel data 2 is the learning travel data, and the diagnostic travel data 3 is the diagnostic travel data (diagnosis travel data). In the case of the above embodiment, the moving body is a vehicle, and the operator is a driver for operating the vehicle.

The present invention is not limited to the above-described examples, and includes various modifications. For example, the above-described embodiment has been described in detail in order to explain the present invention in an easy-to-understand manner, and is not necessarily limited to those having all the described configurations. In addition, it is possible to replace a part of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the fairness of one embodiment. Further, it is possible to add / delete / replace a part of the configuration of each embodiment with another configuration. Further, each of the above configurations, functions, processing units, processing means and the like may be realized by hardware by designing a part or all of them by, for example, an integrated circuit. Further, each of the above configurations and functions may be realized by software by the processor interpreting and executing a program that realizes each function. Information such as programs, tables, and files that realize each function can be stored in a memory, a hard disk, a recording device such as an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

1 ... Abnormality diagnostic device, 2 ... Learning driving data, 3 ... Diagnostic driving data, 4 ... Vehicle, 5 ... Communication base station, 6 ... Network, 7 ... Data collection server, 11 ... Learning unit, 12 ... Diagnostic unit, 13 ... Probability distribution storage unit, 14 ... Mixed ratio storage unit, 15 ... Online diagnostic unit, 41 ... Data points, 42, 43, 44 ... Probability distribution parts, 45 ... Mixed distribution, 71 ... Data acquisition unit, 72 ... Data transmission Unit, 73 ... Diagnosis result acquisition unit, 74 ... Diagnosis result transmission unit, 102 ... Data extraction unit, 111 ... Learning data input unit, 112 ... Data extraction unit, 113 ... Data distribution approximation unit, 121 ... Diagnosis data input unit, 122 ... Diagnosis model generation unit, 123 ... Abnormality calculation unit, 124 ... Diagnosis result output unit, 151 ... Diagnosis data input unit, 152 ... Diagnosis result output unit 1121 ... Driver / vehicle-specific data collection unit 1122 ... Specific driving mode extraction Units 1122a ... Driving data, 1122b ... Brake period 1123 ... Feature amount calculation unit 1124 ... Standardization processing unit 1131 ... Feature amount data points 1132,1133,1134 ... Probability distribution, 1135, 1136 ... Mixed distribution, 1137 ... Data points, 1138 ... probability distribution, 1139 ... mixed distribution, 1221 ... new driver / vehicle combination detection unit, 1222 ... mixed ratio search unit, 1223 ... mixed distribution generation unit, 1225 ... mixed distribution acquisition unit, 1231 ... specific driving mode extraction Unit, 1232 ... Feature amount calculation unit, 1233 ... Standardization processing unit, 1234 ... Abnormality calculation unit, 12231 ... Prediction distribution

Claims (5)

In an abnormality diagnostic device that determines the possibility of an abnormality in the moving body from the moving data of the moving body.
To generate a probability distribution of the feature amount of the part to be diagnosed obtained from the learning movement data obtained by combining the moving body and the operator, and to obtain a mixture distribution that approximates the probability distribution of the combination of the moving body and the operator to be diagnosed. The data distribution approximation part that calculates the mixture ratio of the combination of the moving body and the operator to be diagnosed in
A probability distribution storage unit that stores the parameters of the probability distribution generated by the data distribution approximation unit, and a probability distribution storage unit.
A mixing ratio storage unit that stores the mixing ratio calculated by the data distribution approximation unit, and a mixing ratio storage unit.
Based on the combination of the moving body and the operator to be diagnosed, the combination of the moving body and the operator to be diagnosed from the probability distribution stored in the probability distribution storage unit and the mixing ratio stored in the mixing ratio storage unit. A diagnostic unit that generates a mixed distribution and compares the generated mixed distribution with the distribution of moving data to be diagnosed to determine anomalies.
An abnormality diagnostic device characterized by being equipped with. In claim 1,
An abnormality diagnosis device characterized in that the moving body is a vehicle and the operator is a driver of the vehicle. In claim 2,
The data distribution approximation unit generates the probability distribution and the mixing ratio corresponding to the environment by using the learning movement data in which the vehicle and the driver are combined.
When the mixing ratio of the combination of the vehicle and the driver is not stored in the mixing ratio storage unit in the movement data to be diagnosed, the diagnosis unit selects the environment and the vehicle from the mixing ratio stored in the mixing ratio storage unit. An abnormality diagnostic device characterized in that a mixture ratio is selected according to a driver's priority and a mixed distribution of a combination of a vehicle and a driver corresponding to the movement data to be diagnosed is generated. In claim 3,
In the environment, the extraction ratio of at least one of the driving data at the time of braking operation, the driving data at the time of accelerator operation, and the driving data at the time of bundle operation in the traveling section is within a predetermined range. An abnormality diagnosis device characterized in that the vehicle is in a running state as a condition. In claim 3,
The environment is an abnormality diagnosis device characterized in that at least one of road type information, road slope estimation result, and road surface condition estimation result based on the position information of the vehicle is a condition.

Images