Apparatus abnormality diagnosis method and system
A technique relating to an apparatus abnormality diagnosis system, capable of easily creating and adding/updating an diagnosis model with respect to an initial and new failure case, and appropriately and efficiently achieving diagnosis of abnormality and instruction of operation using the model. In the abnormality diagnosis system, an diagnosis model creating process unit creates a structured abnormality model expressing a structured abnormality of maintenance operation type to an alarm and apparatus event relating to the maintenance operation type by a graph network structure based on acquisition of maintenance operation data. And, by synthesizing the structured abnormality model with an existing structured abnormality model, the diagnosis model is updated.
1. An apparatus abnormality diagnosis method of performing an apparatus abnormality diagnosis process for estimating a cause for the sign of abnormality or the abnormality by using information processing with a computer when a sign of abnormality of a target apparatus is detected or an abnormality occurs, and making an instruction for operation of the cause or maintenance operation details, wherein
in a process of creating or updating an diagnosis model of the apparatus for use in the apparatus abnormality diagnosis process,
at least one of apparatus event data and alarm data output from at least one of the apparatus and an external abnormality detection system is used, and the creating or updating process includes:
a first process step of obtaining, in a predetermined unit of time, maintenance operation data regarding the apparatus and at least one of the alarm data and the apparatus event data;
a second process step of, by using the data obtained in the first process step, creating a structured abnormality between a relevant maintenance operation and at least one of an apparatus event and alarm associated therewith as a structured abnormality model represented by a graph network structure; and
a third process step of, by synthesizing the new structured abnormality model created in the second process step with a structured abnormality model created so far and recalculating a probability of occurrence among nodes updated in the graph network structure, updating the diagnosis model formed so as to include the resultant structured abnormality model;
wherein
in the process of creating or updating the diagnosis model of the apparatus,
three pieces of data output from the apparatus, comprising sensor data, the apparatus event data, and the alarm data, and alarm data output from the external detection system detecting a sign of abnormality or an abnormality are used,
in the first process step, the maintenance operation data includes information of a date and time and type of the maintenance operation and components, the apparatus event data includes information of a type of an operation state of the apparatus and a date and time of occurrence, the alarm data includes information of an alarm type and a date and time of occurrence, and data of the apparatus event and the alarm occurring within a predetermined time period is obtained based on the date and time of the maintenance operation, and
in the second process step, a structured abnormality between the type of the maintenance operation and an apparatus event type and alarm type associated therewith is created as a structured abnormality model represented by the graph network structure,
wherein the first process step, the second process step and the third process step are effected by a hardware processor.
2. The apparatus abnormality diagnosis method according to claim 1 , including a fourth process step of performing the apparatus abnormality diagnosis process by using at least one of the new apparatus event data and alarm data and the diagnosis model updated in the third process step.
3. The apparatus abnormality diagnosis method according to claim 2 , wherein
in the fourth process step, when the new alarm data is obtained, the maintenance operation data and the apparatus event data and alarm data are obtained in a previous predetermined unit of time to create structured abnormality data representing an occurrence pattern regarding a relevant alarm, a graph network part matching the structured abnormality data this time is extracted from the diagnosis model updated in the third process step and, based on the result, maintenance operation candidate information is output.
4. The apparatus abnormality diagnosis method according to claim 2 , wherein
the fourth process step includes a process step of displaying information of making an instruction for operation of the cause or maintenance operation details on a screen of an information processing apparatus to be used by an administrator or a maintenance operator of the system.
5. The apparatus abnormality diagnosis method according to claim 1 , including a process step of displaying information of the diagnosis model or a structured abnormality model therein on a screen of an information processing apparatus to be used by an administrator or a maintenance operator of the system.
6. An apparatus abnormality diagnosis method of performing an apparatus abnormality diagnosis process for estimating a cause for the sign of abnormality or the abnormality by using information processing with a computer when a sign of abnormality apparatus is detected or an abnormality occurs, and making an instruction for operation of the cause or maintenance operation details, wherein
in a process of creating or updating an diagnosis model of the apparatus for use in the apparatus abnormality diagnosis process,
at least one of apparatus event data and alarm data output from a least one of the apparatus and an external abnormality detection system is used, and the creating or updating process includes:
a first process step of obtaining, in a predetermined unit of time, maintenance operation data regarding the apparatus and at least one of the alarm data and the apparatus event data;
a second process step of, by using the data obtained in the first process step, creating a structured abnormality between a relevant maintenance operation and at least one of an apparatus event and alarm associated therewith as a structured abnormality model represented by a graph network structure; and
a third process step of, by synthesizing the new structured abnormality model created in the second process step with a structured abnormality model created so far and recalculating a probability of occurrence among nodes updated in the graph network structure, updating the diagnosis formed so as to include the resultant structured abnormality model;
wherein the structured abnormality model includes, for each node in the graph network, data of the probability of occurrence corresponding to all combinations of occurrence patterns with one or more nodes having an input relation to a relevant node and occurrence patterns with one or more nodes having an output relation from the relevant node, and
wherein the first process step, the second process step and the third process step are effected by a hardware processor.
7. The apparatus abnormality diagnosis method according to claim 6 , including a fourth process step of performing the apparatus abnormality diagnosis process by using at least one of the new apparatus event data and alarm data and the diagnosis model updated in the third process step.
8. The apparatus abnormality diagnosis method according to claim 7 , wherein
in the fourth process step, when the new alarm data is obtained, the maintenance operation data and the apparatus event data and alarm data are obtained in a previous predetermined unit of time to create structured abnormality data representing an occurrence pattern regarding a relevant alarm, a graph network part matching the structured abnormality data this time is extracted from the diagnosis model updated in the third process step and, based on the result, maintenance operation candidate information is output.
9. The apparatus abnormality diagnosis method according to claim 7 , wherein
the fourth process step includes a process step of displaying information of making an instruction for operation of the cause or maintenance operation details on a screen of an information processing apparatus to be used by an administrator or a maintenance operator of the system.
10. The apparatus abnormality diagnosis method according to claim 6 , including a process step of displaying information of the diagnosis model or a structured abnormality model therein on a screen of an information processing apparatus to be used by an administrator or a maintenance operator of the system.
11. An apparatus abnormality diagnosis method of performing an apparatus abnormality diagnosis process for estimating a cause for the sign of abnormality or the abnormality by using information processing with a computer when a sign of abnormality of a target apparatus is detected or an abnormality occurs, and making an instruction for operation of the cause or maintenance operation details, wherein
in a process of creating or updating an diagnosis model of the apparatus for use in the apparatus abnormality diagnosis process,
at least one of apparatus event data and alarm data output from at least one of the apparatus and an external abnormality detection system is used, and the creating or updating process includes:
a first process step of obtaining, in a predetermined unit of time, maintenance operation data regarding the apparatus and at least one of the alarm data and the apparatus event data;
a second process step of, by using the data obtained in the first process step, creating a structured abnormality between a relevant maintenance operation and at least one of an apparatus event and alarm associated therewith as a structured abnormality model represented by a graph network structure; and
a third process of, by synthesizing the new structured abnormality model created in the second process step with a structured abnormality model created so far and recalculating a probability of occurrence among nodes updated in the graph network structure, updating the diagnosis model formed so as to include the resultant structured abnormality model;
wherein
in the process of creating or updating the diagnosis model of the apparatus,
the first process step includes:
(A) a process step of obtaining a history of previous maintenance operation data regarding the relevant apparatus; and
(B) a process step of obtaining a history of previous alarm data and apparatus event data regarding the relevant apparatus, and
the second and third process steps include:
(C) a process step of calculating transition data of the number of times of occurrence in a predetermined unit of time for each maintenance operation type by using the history obtained in the process step (A);
(D) a process step of calculating transition data of the number of times of occurrence in a predetermined unit of time for each alarm type and each apparatus event type by using the history obtained in the process step (B);
(E) a process step of calculating a correlation coefficient between transition data of the number of times of occurrence for said each maintenance operation type and transition data of the number of times of occurrence for said each apparatus event type and alarm type;
(F) a process step of extracting, from the correlation coefficients calculated in the process step (E), a maintenance operation type and an apparatus event type and an alarm type having a correlation coefficient satisfying a threshold, and creating a graph network among the extracted data;
(G) a process step of creating one or more groups of a maintenance operation type and an apparatus event type and an alarm type having a closed-circuit relation with a structured computation for the graph network created in the process step (F); and
(H) a process step of setting a directivity of the graph network based on an order of a date and time of occurrence of each data in the groups created in the process step (G),
wherein the first process step, the second process step and the third process step are effected by a hardware processor.
12. The apparatus abnormality diagnosis method according to claim 11 , including a fourth process step of performing the apparatus abnormality diagnosis process by using at least one of the new apparatus event data and alarm data and the diagnosis model updated in the third process step.
13. The apparatus abnormality diagnosis method according to claim 12 , wherein
in the fourth process step, when the new alarm data is obtained, the maintenance operation data and the apparatus event data and alarm data are obtained in a previous predetermined unit of time to create structured abnormality data representing an occurrence pattern regarding a relevant alarm, a graph network part matching the structured abnormality data this time is extracted from the diagnosis model updated in the third process step and, based on the result, maintenance operation candidate information is output.
14. The apparatus abnormality diagnosis method according to claim 12 , wherein
the fourth process step includes a process step of displaying information of making an instruction for operation of the cause or maintenance operation details on a screen of an information processing apparatus to be used by an administrator or a maintenance operator of the system.
15. The apparatus abnormality diagnosis method according to claim 11 , including a process step of displaying information of the diagnosis model or a structured abnormality model therein on a screen of an information processing apparatus to be used by an administrator or a maintenance operator of the system.