Cause inferring device

Details Cause inferring device Cause inferring device

1997-02-04

1999-09-07

Downs, Robert W.

Data processing: artificial intelligence

Knowledge processing system

Knowledge representation and reasoning technique

39518301, 39518302, G06F 944

Patent

active

059501839

DESCRIPTION:

BRIEF SUMMARY
TECHNICAL FIELD

The present invention relates to a cause inferring device whereby a true cause is inferred using a knowledge base in which knowledge expressing the relationship between various phenomenon items and various cause items is represented in the form of a matrix or in the form of a decision tree, and more particularly to an inferring device that is suited to application to fault diagnosis of machines.


BACKGROUND ART

Conventional methods of fault diagnosis of machines such as construction machinery include the following two well known methods.
One is the method of diagnosis in which the various phenomenon items (inspection items) are treated as nodes where branching is effected in accordance with the answer (e.g. YES, NO) constituting the result of the associated inspection, so as to lead to the cause which represents the final conclusion.
Since the knowledge indicating the relationship between the various inspection items and the various cause items is expressed in a decision tree structure, this is called FTA diagnosis (Fault Tree Analysis).
In another method of diagnosis, degree of relationship data indicating the degree of relationship of the various inspection items and the various cause items are arranged in the fashion of a matrix in which either the inspection items or cause items are rows while the other one of these are the columns; frequency of occurrence data are input indicating the degree to which the prescribed inspection item of the various inspection items occurs. The likelihood of a cause item can then be calculated from these frequency of occurrence data that have been input and degree of association data, arranged in matrix fashion; the cause is inferred from these calculated certainties. Since the knowledge regarding the causal relationship between the inspection items and the fault cause items is represented in the form of a matrix, this is called matrix fuzzy diagnosis.
Furthermore, a technique whereby knowledge of a decision tree structure for FTA diagnosis is converted into knowledge arranged in matrix fashion for matrix fuzzy diagnosis is disclosed in Japanese Patent Publication H. 3-116330 and so is already public knowledge.
Also, methods of inference called incident base inference (ID3 etc.) are widely known, in which an efficient method of categorization resulting from extraction of data characteristics from incident data is represented in the form of a decision tree.
This incident base inference is a technique whereby general rules are compiled from a collection of past incidents (problem and solution set) and when a new incident is presented the solution is found by using these rules; this is utilized as one method of knowledge acquisition.
In incident base inference, a decision tree is compiled whereby classes (categorized item: is a melon, is an apple, etc.) are categorized using for example the properties of the collection of past incidents (question item: what color, what size, etc.) and property values (values that the reply to the question may take: green, red, or large, small, etc.).
A characteristic advantage of matrix fuzzy diagnosis is that the candidate fault causes can be narrowed down even if the inspection item frequency of occurrence is unanswerable or even if inspection results are input in which the frequency of occurrence is expressed in terms of uncertainty with a numerical value in the range 0 to 1.
However, there was the drawback that if the information provided by the inspection items is insufficient the precision with which the causes are narrowed down was poor.
Also, there was the problem that, although the cause of the fault is output represented by a certainty, it was not possible to present effective inspection items that would further narrow down the cause from among a plurality of candidate fault causes for which the same certainty is expressed.
Furthermore, when the inspections are to be carried out, the inspection items are simply displayed as a list, so it was not possible to ascertain which inspection, of the plurality of inspecti

REFERENCES:
patent: 5214653 (1993-05-01), Elliott, Jr. et al.
patent: 5253333 (1993-10-01), Abe
patent: 5293323 (1994-03-01), Doskocil et al.
patent: 5301258 (1994-04-01), Haysahi
patent: 5528516 (1996-06-01), Yemini et al.
patent: 5729452 (1998-03-01), Smith et al.
F. Pipitone, "The FIS Electronics Troubleshooting System," IEEE Computer, pp. 68-76, Jul. 1986.
J. Poshtan and R. Doraiswami, "Influence Matrix Approach to Fault Diagnosis and Controller Tuning," Proc. Canadian Conf. on Electrical and Computer Engineering, pp. 810-813, Sep. 1994.

Affiliated with

Also associated with

Rating

Cause inferring device does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Cause inferring device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Cause inferring device will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-1815378