Knowledge data base processing system and expert system

• US 5,493,729 A
• Filed: 07/20/1994
• Issued: 02/20/1996
• Est. Priority Date: 03/14/1990
• Status: Expired due to Term

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Namba, Shigeaki, Nigawara, Seiitsu, Kohmoto, Hiroshi
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/277,366
• Time in Patent Office: 580 Days
• Field of Search: 395/51, 395/61, 395/76
• US Class Current: 706/52
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

Described is a processing system for knowledge data base in which indices representing the degrees of certainty of causal relations between an event and plural events relevant to the first-mentioned event are stored. The system has an input unit for inputting information on those actually experienced among events inferred based on the knowledge data base, an updating unit for updating the indices so that, among the plural causal relations, the certainty of the causal relation corresponding to the actually experienced event inputted by the input means is made higher relative to the certainties of the other causal relations, and a storing unit for storing back the thus-updated indices again in the knowledge data base. This allows an expert system with the processing system to perform inference with higher certainty.

Claims

1. A processing system for processing knowledge data of a knowledge data base for an electric power plant included in an inference system in which indices representing degrees of certainty of causal relations between a given event occurring at a device of said electric power plant and plural events relevant to said given event are stored, said processing system comprising:

means for receiving information regarding which of said plural events has actually occurred with respect to said given event, said plural events being inferred by said inference system based on said knowledge data base by using said indices;

means for updating said indices by increasing in a predetermined mapping relation, among said causal relations, a certainty of a causal relation corresponding to said one of said plural events having actually occurred with respect to said given event relative to certainties of said causal relations of said other plural events; and

means for storing back said updated indices again in said knowledge data base,

said indices being degrees of certainty for inferences of causes/effects of said plural events with respect to said given event.

2. The processing system of claim 1, wherein said updating means performs updating processing each time said causes/effects are identified after inference of said causes/effects.

3. The processing system of claim 1, wherein said updating means performs updating processing based on statistical data of a predetermined number of actual occurrences.

4. The processing system of claim 1, further comprising means for providing a user with information regarding results of inferences based on said knowledge data base and said updated indices.

5. A processing system for processing knowledge data of a knowledge data base included in an inference system in which certainty factors representing degrees of certainty of causal relations between a given event and plural events relevant to the said given event are stored, said processing system comprising:

means for calculating a degree of frequency of actual occurrences of a causal event (or a subsequent event) inferred by said inference system with respect to said given event based on said knowledge data base by using said certainty factors; and

means for updating said certainty factors of plural causal relations, which are relevant to said given event, in accordance with said degree of frequency of actual occurrences of said causal event (or subsequent event), so calculated, the thus updated certainty factors being stored back to said knowledge data base.

6. A processing system for processing knowledge data of a knowledge data base included in an inference system in which certainty factors representing degrees of certainty of causal relations between a given event and plural events relevant to said given event are stored, said processing system comprising:

means for calculating a degree of frequency of actual occurrences of a causal event (or a subsequent event) inferred by said inference system with respect to said given event based on said knowledge data base by using said certainty factors; and

means for updating Said certainty factors of plural causal relations, which are relevant to said given event, in accordance with said degree of frequency of actual occurrences of said causal event (or the subsequent event), so calculated, the thus updated certainty factors being stored back to said knowledge data base,

wherein said degree of frequency of actual occurrences of said causal event is used as a significant degree when an incrementing number of inferences of said causal event (or the subsequent event) for said given event has reached at least a predetermined number, said predetermined number being variably presentable, and

wherein the difference between said certainty factor of said causal event (or the subsequent event) and said degree of frequency of actual occurrences of said causal event (or the subsequent event) is used for updating said certainty factor.

7. The processing system of claim 6, wherein said updating means performs said updating such that said difference becomes closer to 0.

8. The processing system of claim 6, further comprising means for estimating a limit value of said frequency of actual causal event occurrence.

9. The processing system of claim 8, wherein said updating, itself, is modified for either suppressing or promoting updating of said certainty factor in accordance with said difference between said frequency and said limit value of said frequency.

10. A processing system for processing knowledge data of a knowledge data base for an electric power plant included in an inference system in which certainty factors representing degrees of certainty of causal relations between a given event occurring at a device of the electric power plant and plural events occurring at other devices of the electric power plant relevant to said given event are stored, each of said certainty factors being a number between zero and one, said inference system inferring a cause/effect with respect to said given event based upon said knowledge data base by using said certainty factors, said processing system comprising:

certainty-factor updating means for, based upon information regarding which one of said plural events has actually occurred, increasing, in a predetermined mapping relation, a certainty factor of one of plural causal relations between said given event and said one of said plural events, said one of said plural events having actually occurred, the thus updated certainty factors being stored back to said knowledge data base; and

means for normalizing said certainty factors of said plural causal relations between said given event and said plural events such that a sum of said certainty factors is equal to one.

11. A processing system for processing knowledge data of a knowledge data base for an electric power plant included in an inference system in which certainty factors representing degrees of certainty of causal relations between a given event occurring at a device of the electric power plant and plural events occurring at other devices of the electric power plant relevant to said given event are stored, said processing system comprising:

means for updating said certainty factors based upon information regarding which one of said plural events has actually occurred with respect to said given event by increasing a certainty factor of a causal relation between said given event and said one of said plural events relative to certainty factors of causal relations between said given event and others of said plural events, the thus updated certainty factors being stored back to said knowledge data base;

means for storing historical information of said updated certainty factors from said knowledge data base in said inference system;

means for classifying events into at least three levels such that events of higher certainty factors are classified as semidetermined events, events of lower certainty factor are classified as rare events and events of intermediate certainty factors are classified as unstable events based upon said historical information of said Updated certainty factors from said knowledge data base said inference system; and

means for providing a user with results of said classification performed by said classifying means.

12. A processing system for processing knowledge data of a knowledge data base for an electric power plant included in an inference system in which certainty factors representing degrees of certainty of causal relations between a given event occurring at a device of the electric power plant and plural events occurring at other devices of the electric power plant relevant to said given event are stored, said processing system comprising:

means for updating said certainty factors based upon events having actually occurred at a device of the electric power plant, the thus updated certainty factors being stored back to said knowledge data base;

means for storing historical information of said updated certainty factors from said knowledge data base in said inference system; and

means for discriminating a causal relation between said historical information of said updated certainty factors and historical information regarding intensities of observed physical quantities, said intensities defining individual events having actually occurred.

13. An expert system for rendering an inference, comprising:

a knowledge data base in which an inference tree diagram connects together groups of events in at least three layers extending from causal events to resultant events via intermediate events,

wherein degrees of causal relations between said events in said adjacent layers are applied., as certainty factors, to corresponding individual inter-event routes connecting events in said adjacent layers;

means for inferring events based upon said knowledge data base; and

updating means for updating said knowledge data base so that said certainty factors for said inter-event routes relevant to actually occurred events are increased,

wherein said updating means includes a certainty factor distribution curve representing distribution of certainty factors for intensities of physical quantities, said certainty factor distribution curve being provided for each event whose certainty factor varies depending on the intensity of an observed physical quantity,

wherein when the certainty factor for the intensity of a given physical quantity is updated, the certainty factor distribution curve is corrected by said updating means by conducting interpolation between the thus-updated certainty factor and other certainty factors,

wherein each inter-event route is defined by a combination of matrix elements by allocating said individual event items of the inference tree diagram as elements of a matrix consisting of N rows and M columns, N being the largest number among the numbers of the event items in the respective layers of said inference tree diagram and M being the number of said layers, and

wherein a dummy row is added by said updating means to said matrix consisting of the N rows and the M columns, a certainty factor of a given constant value is applied beforehand to an inter-event route which ends up with an element in said dummy row, and in inferential calculation with an event in said intermediate layer having been determined, a dummy element in said dummy row is used in place of the element of the event item in said intermediate layer, said dummy element being in the same column as the last-mentioned element.

14. The expert system of claim 13, further comprising a means for recording both times of occurrence of plural events in a common column of the assumed inference tree diagram and events when said plural events have occurred successively before occurrence of events in an adjacent column.

15. The expert system of claim 13, further comprising a means for notifying an operator of occurrence of plural events in a common column of said inference tree diagram as an abnormal state when said plural events have occurred successively before occurrence of events in an adjacent column.