State estimation apparatus, method, and non-transitory computer readable medium

1. A system for estimating a state of a monitoring target, the system comprising:

a sensor that measures an operation of the monitoring target; and

an information processing apparatus including:

acquisition circuitry configured to acquire first data from the sensor; and

processing circuitry configured to:

calculate a distribution representing a magnitude of a value and a frequency of the magnitude of the value based on the first data acquired by the acquisition circuitry;

divide a width of the distribution into a plurality of division intervals;

detect a first division interval in which a deviation of the distribution becomes a maximum value among the plurality of division intervals and a second division interval in which the deviation of the distribution is less than a predetermined value among the plurality of division intervals;

determine one or more division intervals existing between the first division interval and the second division interval as a specific range used for estimating the state of the monitoring target;

estimate the state of the monitoring target based on a number or a ratio of the number of the first data whose values are within the specific range among values of the first data acquired by the acquisition circuitry, or input the first data whose values are within the specific range into a state estimation model based on a learned neural network and obtain an output of the state estimation model as an estimation result of the state of the monitoring target; and

with respect to the state:

when the number or the ratio is a first threshold value or smaller, determine a completely normal state,

when the number or the ratio is greater than the first threshold value and equivalent to or smaller than a second threshold value, determine a normal state with a possibility of trouble, and

when the number or the ratio is greater than the second threshold value, determine a normal state with a strong possibility of trouble.

2. The system according to claim 1 , wherein the processing circuitry is configured to determine, as the specific range, a third division interval including a steepest slope of the distribution or a fourth division interval which is adjacent to the third division interval and which is farther from a maximum value of the magnitude of the distribution among the plurality of division intervals.

3. The system according to claim 1 , wherein

the acquisition circuitry is configured to acquire second data indicating an occurrence time of a variation factor of the value, and

the processing circuitry is configured to:

calculate a first distribution representing a relationship between:

magnitudes of values included in the first data acquired during a unit time period including the occurrence time, and

frequencies of the magnitudes of the values; and

determine the specific range based on the first distribution.

4. The system according to claim 1 , wherein

the acquisition circuitry is configured to acquire second data indicating an occurrence time of a variation factor of the value, and

the processing circuitry is configured to:

calculate a first distribution representing a relationship between:

a magnitude of a value included in the first data acquired during a first unit time period including the occurrence time, and

a frequency of the magnitude of the value;

calculate a second distribution representing a relationship between:

a magnitude of a value included in the first data acquired during a second unit time period neighboring the first unit time period, and

a frequency of the magnitude of the value; and

determine the specific range based on a difference between the first distribution and the second distribution.

5. The system according to claim 4 , wherein the processing circuitry is configured to determine a division interval including a portion in which an amount of variation between the first distribution and the second distribution is greatest as the specific range.

6. The system according to claim 4 , wherein, in a case where a plurality of occurrence times of the variation factor are present, the specific range with a largest corresponding variation amount is selected among specific ranges determined for respective occurrence times.

7. The system according to claim 1 , wherein the processing circuitry is configured to:

assess whether the value included in the first data is an outlier; and

remove the value assessed as the outlier from the first data.

8. The system according to claim 1 , wherein the processing circuitry is configured to perform smoothing of at least data of the first data, the data having a value included in the specific range.

9. The system according to claim 1 , further comprising output circuitry configured to output at least one of the distribution and the specific range.

10. The system according to claim 1 , wherein

the monitoring target is an automatic ticket gate,

the monitoring target is a mail sorter, or

the monitoring target is a bill validator.

11. The system according to claim 1 , wherein the first division interval is a division interval including a protrusion of the distribution and the second division interval is a division interval including a tail of the distribution.

12. An information processing method for estimating a state of a monitoring target, the method comprising:

acquiring first data from a sensor measuring an operation of the monitoring target;

calculating a distribution representing a magnitude of a value and a frequency of the magnitude of the value based on the first data acquired;

dividing a width of the distribution into a plurality of division intervals;

detecting a first division interval in which a deviation of the distribution becomes a maximum value among the plurality of division intervals and a second division interval in which the deviation of the distribution is less than a predetermined value among the plurality of division intervals;

determining one or more division intervals existing between the first division interval and the second division interval as a specific range used for estimating the state of the monitoring target;

estimating the state of the monitoring target based on a number or a ratio the number of the first data whose values are within the specific range among values of the first data acquired, or inputting the first data whose values are within the specific range into a state estimation model based on a learned neural network and obtaining an output of the state estimation model as an estimation result of the state of the monitoring target; and

with respect to the state:

when the number or the ratio is a first threshold value or smaller, determining a completely normal state,

when the number or the ratio is greater than the first threshold value and equivalent to or smaller than a second threshold value, determining a normal state with a possibility of trouble, and

when the number or the ratio is greater than the second threshold value, determining a normal state with a strong possibility of trouble.

13. A non-transitory computer readable medium storing a program, which when executed by a computer causes the computer to perform a method for estimating a state of a monitoring target, the method comprising:

acquiring first data from a sensor measuring an operation of the monitoring target;

calculating a distribution representing a magnitude of a value and a frequency of the magnitude of the value based on the first data acquired;

dividing a width of the distribution into a plurality of division intervals;

detecting a first division interval in which a deviation of the distribution becomes a maximum value among the plurality of division intervals and a second division interval in the deviation of the distribution is less than a predetermined value among the plurality of division intervals;

determining one or more division intervals existing between the first division interval and the second division interval as a specific range used for estimating the state of the monitoring target;

estimating the state of the monitoring target based on a number or a ratio of the number of the first data whose values are within the specific range among values of the first data acquired, or inputting the first data whose values are within the specific range into a state estimation model based on a learned neural network and obtaining an output of the state estimation model as an estimation result of the state of the monitoring target and

with respect to the state:

when the number or the ratio is a first threshold value or smaller, determining a completely normal state,

when the number or the ratio is greater than the first threshold value and equivalent to or smaller than a second threshold value, determining a normal state with a possibility of trouble, and

when the number or the ratio is greater than the second threshold value, determining a normal state with a strong possibility of trouble.