Abnormality diagnostic device and method therefor

1. An abnormality diagnostic device comprising:

a vehicle database configured to store a plurality of pieces of sensor data of a diagnostic object in a railroad vehicle; and

processing circuitry configured to:

divide the plurality of pieces of sensor data according to values of a learning item to obtain a plurality of sets of sensor data,

generate a plurality of models each associating a first variable assigned first sensor data with a second variable assigned second sensor data, based on the sets of sensor data,

calculate feature amounts of the models based on first sensor data of each set and the models,

select at least one value from the values of the learning item based on the feature amounts to obtain a data selection condition based on the selected at least one value,

extract first sensor data satisfying the data selection condition from the vehicle database, and

diagnose abnormality of the diagnostic object based on the extracted first sensor data and a diagnostic model representing a relation between first sensor data and abnormality of the diagnostic object.

2. The abnormality diagnostic device according to claim 1 , wherein:

the learning item is at least one of a time range, sensor data, an environmental factor, a vehicle classification number in the railroad vehicle, and an operation pattern of the railroad vehicle, and

the data selection condition is a condition based on the at least one of a time range, sensor data, an environmental factor, a vehicle classification number in the railroad vehicle, and an operation pattern of the railroad vehicle.

3. The abnormality diagnostic device according to claim 2 , wherein the time range includes at least one of a time period, a day of the week, a month and a season, the sensor data includes at least one of a vehicle occupancy, latitude and longitude, and a section on a route, and the environmental factor includes at least one of weather, an air temperature and humidity.

4. The abnormality diagnostic device according to claim 1 , wherein the model is a regression model relating first sensor data to second sensor data.

5. The abnormality diagnostic device according to claim 1 , wherein the feature amount is at least one of a determination coefficient, a correlation coefficient, and a residual of the model.

6. The abnormality diagnostic device according to claim 1 , wherein the processing circuitry generates a plurality of the data selection conditions, and diagnoses abnormality of the diagnostic object for each of data selection conditions.

7. The abnormality diagnostic device according to claim 6 , wherein the processing circuitry outputs a comprehensive diagnostic result based on diagnostic results for the data selection conditions.

8. The abnormality diagnostic device according to claim 1 , wherein the processing circuitry learns the data selection condition so that an accuracy rate for history data relating to abnormality of the railroad vehicle becomes high.

9. The abnormality diagnostic device according to claim 1 , wherein the processing circuitry selects a value of the learning item for which the feature amount satisfies a learning condition.

10. An abnormality diagnostic method comprising:

providing a vehicle database configured to store a plurality of pieces of sensor data of a diagnostic object in a railroad vehicle;

dividing the plurality of pieces of sensor data according to values of a learning item to obtain a plurality of sets of sensor data;

generating a plurality of models each associating a first variable assigned first sensor data with a second variable assigned second sensor data, based on the sets of sensor data;

calculating feature amounts of the models based on first sensor data of each set and the models;

selecting at least one value from the values of the learning item based on the feature amounts to obtain a data selection condition based on the selected at least one value;

extracting first sensor data satisfying the data selection condition from the vehicle database; and

diagnosing abnormality of the diagnostic object based on the extracted first sensor data and a diagnostic model representing a relation between first sensor data and abnormality of the diagnostic object.