Method and apparatus for diagnosing plant anomaly

• US 5,914,875 A
• Filed: 01/10/1997
• Issued: 06/22/1999
• Est. Priority Date: 01/11/1996
• Status: Expired due to Fees

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Takizawa, Yoji, Monta, Kazuo
• Primary Examiner(s): Peeso, Thomas R.
• Assistant Examiner(s): NGUYEN, CUONG H
• Application Number:
  US08/782,888
• Time in Patent Office: 893 Days
• Field of Search: 364/186, 364/184, 376/245, 395/180, 395/183.02, 395/183.01, 395/183.22
• US Class Current: 700/79
• CPC Class Codes: G05B 23/0272 Presentation of monitored results, e.g. selection of status reports to be displayed; Filtering information to the user
• CPC Class Codes: G05B 23/0281 Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis
• CPC Class Codes: Y04S 10/52 Outage or fault management, e.g. fault detection or location

Abstract

Determined results of which measured data of a plant is normal/abnormal are mapped to a plant model in an abstract function level hierarchically represented with flow structures of mass/energy and goals to be accomplished by the respective flow structures so as to detect a failure propagation network. In the failure propagation network, a flow structure at the lowest hierarchical level of a network with a top goal of the highest priority is selected as an object to be diagnosed first. In other words, flow structures that are functionally important and close to the origin of anomaly are preferentially diagnosed. The diagnosis for flow structures detects abnormal elements through hypothesis and test. The abnormal element are evaluated with respective mass/energy balances or the like. With respect to the evaluated result, the origin of failure is inferred based on the rule of experience. When the abnormal element has a detailed flow structure, it is diagnosed. Thus, an abnormal position can be detected in more detail.

Claims

1. A method for diagnosing a plant anomaly corresponding to measured data supplied from the plant, the method comprising the steps of:

sampling the measured data;

deciding from the sampled measured data whether a state of each function of the plant is normal; and

inferring an origin of the plant anomaly from a plant model in an abstract function level, the plant model being composed of model elements, by determining whether a state of each of the model elements is normal starting downward from the top of a hierarchy of the plant model corresponding to information obtained in the deciding step, the plant model having a goal and a flow structure, the goal representing a purpose to be accomplished by a lower flow structure and a condition for supporting a higher flow structure, and the flow structure being represented by flow of either of mass or energy through a flow function; and

detecting a failure propagation network having each model element which has not been determined as a normal element in the model element determining step.

2. The method as set forth in claim 1, wherein the inferring step includes the steps of:

determining a diagnosing order of each flow structure of the failure propagation network; and

diagnosing the flow structure of the failure propagation network to detect a failed flow function corresponding to the determined diagnosing order.

3. The method as set forth in claim 2, wherein the diagnosing order determining step includes the step of assigning a higher priority of diagnosis to the lower flow structure of a network unit linked to the goal having a higher priority in the failure propagation network, the network unit being separated from other units in the failure propagation network.

4. The method as set forth in claim 2, wherein the diagnosing step includes a first hypothesis step of assuming that each flow function on each path linked between flow functions determined as abnormal with measured data not through a flow function determined as normal with the measured data is abnormal in the flow structure to be diagnosed.

5. The method as set forth in claim 2, wherein the inferring step includes the step of diagnosing a more detailed flow structure into which the detected failed flow function is decomposed to detect the origin of the plant anomaly in detail.

6. The method as set forth in claim 4, wherein the diagnosing step includes the steps of:

estimating a state amount related to each flow function assumed as abnormal in the first hypothesis step;

evaluating whether or not each flow function determined as abnormal by the first hypothesis step is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

detecting an origin of failure of the flow structure based on a rule of experience corresponding to an evaluated result.

7. The method as set forth in claim 6, wherein the diagnosing step includes a second hypothesis steps of:

assuming that a flow function adjacent to the flow function determined as abnormal with the measured data is abnormal, when any failed flow function cannot be identified in the origin detecting step; and

assuming that each flow function on each path linked between the abnormal flow functions not through the flow function determined as normal with the measured data is abnormal.

8. The method as set forth in claim 7, wherein the diagnosing step includes the steps of:

estimating a state amount related to each flow function assumed as abnormal in the second hypothesis steps; and

evaluating whether or not each flow function determined as abnormal by the second hypothesis steps is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function.

9. The method as set forth in claim 6, wherein the diagnosing step includes the steps of:

estimating a state amount related to each flow function assumed as normal in the first hypothesis step, when any failed flow function cannot be identified in the origin detecting step; and

evaluating whether or not each flow function assumed as normal in the first hypothesis step is normal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function.

10. The method as set forth in claim 6, wherein the diagnosing step includes a third hypothesis steps of:

assuming that a flow function determined as normal with the measured data is abnormal, said flow function being most close to a flow function determined as abnormal with the measured data, when any failed flow function cannot be identified in the origin detecting step; and

assuming that each flow function on each path linked between the abnormal flow functions not through the other flow function determined as normal with the measured data is abnormal.

11. The method as set forth in claim 10, wherein the diagnosing step includes the steps of:

estimating a state amount related to each flow function assumed as abnormal in the third hypothesis steps; and

evaluating whether or not each flow function determined as abnormal by the third hypothesis steps is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function.

12. A method for diagnosing a plant anomaly corresponding to measured data supplied from the plant, the method comprising the steps of:

sampling the measured data;

deciding from the sampled measured data whether a state of each function of the plant is normal; and

inferring an origin of the plant anomaly with a plant model in an abstract function level corresponding to information obtained in the measured data deciding step, the plant model having a flow structure represented by flow of mass or energy through a flow function,

wherein the inferring step includes the steps of

a first hypothesis step of assuming that each flow function on each path linked between flow functions determined as abnormal with measured data not through a flow function determined as normal with the measured data is abnormal in the flow structure;

estimating a state amount related to each flow function assumed as abnormal in the first hypothesis step;

evaluating whether each flow function determined as abnormal by the first hypothesis step is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

detecting an origin of failure of the flow structure based on a rule of experience corresponding to an evaluated result.

13. The method as set forth in claim 12, wherein the inferring step includes the steps of

second hypothesis steps of assuming that a flow function adjacent to the flow function determined as abnormal with the measured data is abnormal, when any failed flow function cannot be identified in the detecting step, and assuming that each flow function on each path linked between the abnormal flow functions not through the flow function determined as normal with the measured data is abnormal;

estimating a state amount related to each flow function assumed as abnormal in the presuming step; and

evaluating whether or not each flow function determined as abnormal by the second hypothesis steps is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function.

14. The method as set forth in claim 12, wherein the inferring step includes the steps of:

estimating a state amount related to each flow function assumed as normal in the first hypothesis step, when any failed flow function cannot be identified in the detecting step; and

evaluating whether or not each flow function assumed as normal in the first hypothesis step is normal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function.

15. The method as set forth in claim 12, wherein the inferring step includes the steps of:

third hypothesis steps of assuming that a flow function determined as normal with the measured data is abnormal, said flow function being most close to a flow function determined as abnormal with the measured value, when any failed flow function cannot be identified in the detecting step, and assuming that each flow function on each path linked between the abnormal flow functions not through the other flow function determined as normal with the measured data is abnormal;

estimating a state amount related to each flow function assumed as abnormal in the third hypothesis steps; and

evaluating whether or not each flow function determined as abnormal by the third hypothesis steps is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function.

16. An apparatus for diagnosing an anomaly of a plant corresponding to measured data supplied from the plant, comprising:

means for storing a plant model hierarchically representing function of the plant in an abstract function level, the plant model having a flow structure and a goal as a model element, the flow structure being represented by flow of either of mass or energy through a flow function, and the goal representing a purpose to be accomplished by a lower flow structure and a condition for supporting a higher flow structure;

measured amount deciding means for obtaining a state amount related to the plant model corresponding to the measured data and comparing the state amount with a threshold thereof so as to decide whether or not the state amount is normal;

means for inferring an origin of the anomaly of the plant using the plant model corresponding to information supplied from said measured amount determining means; and

a display unit for displaying an inference process and result supplied from said inferring means,

wherein said inferring means includes

means for determining whether a functional state of each model element is normal downward from the top in a hierarchy of the plant model based on the information supplied from the measured amount determining means; and

means for detecting a failure propagation network having each model element which has not been determined as normal by said model element determining means.

17. The apparatus as set forth in claim 16, wherein said model element determining means has means for determining that a model element including a state amount determined as abnormal by said measured amount determining means is abnormal.

18. The apparatus as set forth in claim 16, wherein said inferring means includes

means for assigning a higher priority of diagnosis to the lower flow structure of a network unit linked to the goal having a higher priority in the failure propagation network, the network unit being separated from other units in the failure propagation network;

and

means for diagnosing the flow structure corresponding to the priorities of diagnosis of the flow structures in the failure propagation network so as to detect a failed flow function.

19. The apparatus as set forth in claim 16, wherein said display unit displays the inference process and result using the failure propagation network.

20. The apparatus as set forth in claim 18, wherein said diagnosing means has means for diagnosing a more detailed flow structure into which the detected failed flow function is decomposed so as to detect the origin of the anomaly in detail.

21. The apparatus as set forth in claim 18, wherein said diagnosing means includes:

first hypothesis means for assuming that each flow function on each path linked between flow functions determined as abnormal by said measured amount determining means not through a flow function determined as normal by said measured amount determining means is abnormal in the flow structure to be diagnosed;

means for estimating a state amount related to each flow function assumed as abnormal by the first hypothesis means;

means for evaluating whether or not each flow function determined as abnormal by said first hypothesis means is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

first detecting means for detecting an origin of failure of the flow function based on rule of experience corresponding to an evaluated result.

22. The apparatus as set forth in claim 21, wherein said diagnosing means includes:

second hypothesis means for assuming that a flow function adjacent to the flow function determined as abnormal by said measured amount determining means is abnormal when any failed flow function cannot be identified by said first detecting means and assuming that each flow function on each path linked between the abnormal flow functions not through the flow function determined as normal by said measured amount determining means is abnormal;

means for estimating a state amount related to each flow function assumed as abnormal by said second hypothesis means;

means for evaluating whether or not each flow function determined as abnormal by said second hypothesis means is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

means for detecting an origin of failure of the flow function based on the rule of experience corresponding to an evaluated result.

23. The apparatus as set forth in claim 24, wherein said diagnosing means includes:

means for estimating a state amount related to each flow function assumed as normal by said first hypothesis means when any failed flow function cannot be identified by said first detecting means;

means for evaluating whether or not each flow function assumed as normal by said first hypothesis means is normal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

means for detecting an origin of failure of the flow function based on the rule of experience corresponding to an evaluated result.

24. The apparatus as set forth in claim 21, wherein said diagnosing means includes:

third hypothesis means for assuming that a flow function determined as normal by said measured amount determining means and most close to a flow function determined as abnormal by said measured amount determining means is abnormal when any failed flow function cannot be detected by said first detecting means and assuming that each flow function on each path linked between the abnormal flow functions not through the other flow function determined as normal by said measured amount determining means is abnormal;

means for estimating a state amount related to each flow function assumed as abnormal by said third hypothesis means;

means for evaluating whether or not each flow function determined as abnormal by said third hypothesis means is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

means for detecting an origin of failure of the flow function based on the rule of experience corresponding to an evaluated result.

25. An apparatus for diagnosing an anomaly of a plant corresponding to measured data supplied from the plant, comprising:

means for storing a plant model in an abstract function level, the plant model having a flow structure represented by flow of mass or energy through a flow function;

measured amount deciding means for obtaining a state amount related to the plant model corresponding to the measured data and comparing the state amount with a threshold thereof so as to decide whether or not the state amount is normal;

means for inferring an origin of the anomaly of the plant using the plant model corresponding to information supplied from said measured amount determining means; and

a display unit for displaying an inference process and result supplied from said inferring means,

wherein said inferring means includes;

first hypothesis means for assuming that each flow function on each path linked between flow functions determined as abnormal by said measured amount determining means not through a flow function determined as normal by said measured amount determining means is abnormal in the plant model;

means for estimating a state amount related to each flow function assumed as abnormal by said first hypothesis means;

means for evaluating whether or not each flow function determined as abnormal by said first hypothesis means is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

first detecting means for detecting an origin of failure of the flow function based on a rule of experience corresponding to an evaluated result.

26. The apparatus as set forth in claim 25, wherein said inferring means includes:

second hypothesis means for assuming that a flow function adjacent to a flow function determined as abnormal by said measured amount determining means is abnormal when any failed flow function cannot be identified by said first detecting means and assuming that each flow function on each path linked between the abnormal flow functions not through the flow function determined as normal by said measured amount determining means is abnormal;

means for estimating a state amount related to each flow function assumed as abnormal by said second hypothesis means;

means for evaluating whether or not each flow function determined as abnormal by said second hypothesis means is abnormal with the estimated amount state based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

means for detecting an origin of failure of the flow function based on the rule of experience corresponding to an evaluated result.

27. The apparatus as set forth in claim 25, wherein said inferring means includes

means for estimating a state amount related to each flow function assumed as normal by said first hypothesis means when any failed flow function cannot be identified by said first detecting means;

means for evaluating whether or not each flow function assumed as normal by said first hypothesis means is normal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

means for detecting an origin of failure of the flow function based on the rule of experience corresponding to an evaluated result.

28. The apparatus as set forth in claim 25, wherein said inferring means includes:

third hypothesis means for assuming that a flow function determined as normal by said measured amount determining means and most close to a flow function determined as abnormal by said measured amount determining means is abnormal when any failed flow function cannot be detected by said first detecting means and assuming that each flow function on each path linked between the abnormal flow functions not through the other flow function determined as normal by said measured amount determining means is abnormal;

means for estimating a state amount related to each flow function assumed as abnormal by said third hypothesis means;

means for evaluating whether or not each flow function determined as abnormal by said third hypothesis means is abnormal with the estimated state amount based on a deviation from a normal value, a balance of input and output of the flow function, and a restriction corresponding to a type of the flow function; and

means for detecting an origin of failure of the flow function based on the rule of experience corresponding to an evaluated result.