Unit space generating device, plant diagnosing system, unit space generating method, plant diagnosing method, and program
A unit space generating device ( 141 ) which generates a unit space for use when diagnosing an operating state of a plant on the basis of a Mahalanobis distance is provided with: a sampling data acquiring unit ( 141 A) which acquires a sampling data group comprising a plurality of state quantities of the plant, measured with a fixed period; an adoption determining unit ( 141 C) which, on the basis of an adoption probability calculated each time the sampling data group is acquired, determines whether the sampling data group is to be adopted as a unit space generation data group serving as the basis for a unit space; a unit space generating unit ( 141 D) which generates unit spaces on the basis of a plurality of the adopted unit space generation data groups; and an output value acquiring unit ( 141 B) which acquires an output value of the plant corresponding to the sampling data group.
1. A unit space generating device that generates a unit space used in a case of diagnosing an operation state of a plant on the basis of a Mahalanobis distance, the device comprising:
a sampling data acquisition unit that acquires a sampling data group including a plurality of state quantities of the plant, which are measured for each fixed cycle;
an adoption determination unit that determines to adopt the sampling data group as a unit space generation data group, which is based on the unit space, on the basis of an adoption probability calculated each time the sampling data group is acquired;
a unit space generation unit that generates the unit space on the basis of a plurality of the adopted unit space generation data groups; and
an output value acquisition unit that acquires an output value of the plant corresponding to the sampling data group,
wherein when a first sampling data group is acquired at a period where a load change is not occurred and a second sampling data group is acquired at a period where a load change is occurred and has a larger absolute value of at least one of a change rate and a change amount of the acquired output value than the first sampling data group, the adoption determination unit calculates the adoption probability such that an adoption probability of the second sampling data group is higher than an adoption probability of the first sampling data group, and calculates the adoption probability by adding a probability ratio with respect to the absolute value of the change rate or the change amount of the acquired output value to a probability in which a value of the probability is set such that the sampling data group is adopted once in a predetermined period.
2. The unit space generating device according to claim 1 , wherein the change rate is represented by an absolute value of a second derivative of the output value.
3. The unit space generating device according to claim 1 , wherein the change amount is represented by an absolute value of a first derivative of the output value.
4. A plant diagnosing system that diagnoses an operation state of a plant, the system comprising:
a Mahalanobis distance calculation unit that calculates a Mahalanobis distance on the basis of a state quantity of the plant acquired in a case of diagnosing the operation state of the plant; and
a plant state determination unit that determines whether or not the operation state of the plant is normal on the basis of a unit space and the Mahalanobis distance,
wherein the unit space is generated using the unit space generating device according to claim 1 .
5. A unit space generating method of generating a unit space used in a case of diagnosing an operation state of a plant on the basis of a Mahalanobis distance, the method comprising:
a sampling data acquisition step of acquiring a sampling data group of a state quantity of the plant, which is measured for each fixed cycle;
an adoption determination step of determining to adopt the sampling data group as a unit space generation data group, which is based on the unit space, on the basis of an adoption probability calculated each time the sampling data group is acquired;
a unit space generation step of generating the unit space on the basis of a plurality of the adopted unit space generation data groups; and
a step of acquiring an output value of the plant corresponding to the sampling data group,
wherein in the adoption determination step,
when a first sampling data group is acquired at a period where a load change is not occurred and a second sampling data group is acquired at a period where a load change is occurred and has a larger absolute value of at least one of a change rate and a change amount of the acquired output value than the first sampling data group, the adoption probability is calculated such that an adoption probability of the second sampling data group is higher than an adoption probability of the first sampling data group, and the adoption probability is calculated by adding an adoption probability ratio with respect to the absolute value of the change rate or the change amount of the acquired output value to a probability in which a value of the probability is set such that the sampling data group is adopted once in a predetermined period.
6. A plant diagnosing method of diagnosing an operation state of a plant, the method comprising:
a Mahalanobis distance calculation step of calculating a Mahalanobis distance on the basis of a state quantity of the plant acquired in a case of diagnosing the operation state of the plant; and
a plant state determination step of determining whether or not the operation state of the plant is normal on the basis of a unit space and the Mahalanobis distance,
wherein the unit space is generated using the unit space generating method according to claim 5 .
7. A program that causes a computer of a unit space generating device, which generates a unit space used in a case of diagnosing an operation state of a plant on the basis of a Mahalanobis distance, to function, wherein the program is stored in a non-transitory storage device, the program causing the computer to execute:
a sampling data acquisition step of acquiring a sampling data group of a state quantity of the plant, which is measured for each fixed cycle;
an adoption determination step of determining to adopt the sampling data group as a unit space generation data group, which is based on the unit space, on the basis of an adoption probability calculated each time the sampling data group is acquired;
a unit space generation step of generating the unit space on the basis of a plurality of the adopted unit space generation data groups; and
a step of acquiring an output value of the plant corresponding to the sampling data group,
wherein in the adoption determination step,
when a first sampling data group is acquired at a period where a load change is not occurred and a second sampling data group is acquired at a period where a load change is occurred and has a larger absolute value of at least one of a change rate and a change amount of the acquired output value than the first sampling data group, the adoption probability is calculated such that an adoption probability of the second sampling data group is higher than an adoption probability of the first sampling data group, and the adoption probability is calculated by adding an adoption probability ratio with respect to the absolute value of the change rate or the change amount of the acquired output value to a probability in which a value of the probability is set such that the sampling data group is adopted once in a predetermined period.
8. A program that causes a computer of a plant diagnosing system which diagnoses an operation state of a plant to function, the program causing the computer to execute:
a Mahalanobis distance calculation step of calculating a Mahalanobis distance on the basis of a state quantity of the plant acquired in a case of diagnosing the operation state of the plant; and
a plant state determination step of determining whether or not the operation state of the plant is normal on the basis of a unit space and the Mahalanobis distance,
wherein the unit space is generated using the program according to claim 7 .