Fault diagnosis expert system

• US 5,127,005 A
• Filed: 09/19/1990
• Issued: 06/30/1992
• Est. Priority Date: 09/22/1989
• Status: Expired due to Term

• Current Assignee: Ricoh Company Limited
• Original Assignee: Ricoh Company Limited
• Inventors: Katoh, Eiichi, Oda, Toshihiko
• Primary Examiner(s): ATKINSON, CHARLES
• Application Number:
  US07/584,801
• Time in Patent Office: 650 Days
• Field of Search: 371/15.1, 395/575, 395/50, 395/51, 395/52, 395/916, 364/275.7, 364/276.3
• US Class Current: 714/26
• CPC Class Codes: G05B 23/0278 Qualitative, e.g. if-then rules; Fuzzy logic; Lookup tables; Symptomatic search; FMEA
• CPC Class Codes: G06F 11/2257 using expert systems
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: Y10S 706/916 Electronic or computer, internal or network, circuit

Abstract

A fault diagnosis expert system for locating and eliminating sources of a machine trouble. The system comprises a storage part for storing an expert knowledge constructed in a hierarchical search tree structure in which cause-to-effect links connecting high-level effect events to low-level cause events are pre-defined and all possible low-level cause events are pre-enumerated for each high-level effect event, a user interface part for providing a user with questions and responses regarding the state of the machine trouble, an inference part for inferring a cause of the machine trouble with the expert knowledge and the user information, and an outputting part for allowing inference results from the inference part to be displayed for the user. The expert knowledge includes a cause candidate criterion pre-defined as an event used for selection of searching paths when the inference part makes a decision as to which path of the search tree should be traced back. The cause-to-effect links include an identifier indicative of a difference between self-dependency causes and other-dependency causes.

Claims

1. A fault diagnosis expert system for locating and eliminating sources of a machine trouble comprising:

storage means for storing a fault diagnosis expert knowledge constructed in a hierarchical structure of a search tree in which cause-to-effect links connecting high-level effect events to low-level cause events are pre-defined and all possible low-level cause events are pre-enumerated for each high-level effect event, said low-level cause events being candidates for a cause producing said high-level effect events;

user interface means for providing a user with questions and responses concerning state of the machine trouble, allowing the user to supply user information in reply to the questions from the user interface means and allowing the user interface means to provide the user with the responses in reply to the user information;

inference means for inferring a cause of the machine trouble with the fault diagnosis expert knowledge stored in said storage means and the user information supplied from the user; and

outputting means for allowing inference results supplied from said inference means to be displayed for the user,

wherein the fault diagnosis expert knowledge stored in the storage means includes a cause candidate criterion which is pre-defined as an event used for the inference means to trace back appropriate nodes within the search tree when the inference means infers the cause of the machine trouble to locate a true cause thereof.

2. A fault diagnosis expert system as claimed in claim 1, wherein said inference means checks, during inference of selecting a low-level cause event from a high-level effect event by tracing back nodes through the search tree, whether an event given as the cause candidate criterion with respect to one of the cause-to-effect links being connected between the high-level effect event and the low-level cause event already takes place before the low-level cause event is selected from the high-level effect event, the inference means then making a decision that the high-level cause event is a cause candidate of the low-level effect event only when said event already takes place.

3. A fault diagnosis expert system as claimed in claim 2, wherein a certainty factor is pre-defined for each of the cause-to-effect links within the storage means, said certainty factor indicating the degree of likelihood of the low-level cause event being a cause producing the high-level effect event.

4. A fault diagnosis expert system as claimed in claim 3, wherein said cause-to-effect links pre-defined by said fault diagnosis expert knowledge stored in said storage means have an identifier for indicating a difference between self-dependency causes and other-dependency causes, said self-dependency causes having specific troubleshooting measures for eliminating sources of the machine trouble or preventing occurrence of the machine trouble, said other-dependency causes having no specific troubleshooting measures and being followed by low-level self-dependency causes downstream of said other-dependency cause through the search tree, said low-level self-dependency causes having specific troubleshooting measures for eliminating sources of the machine trouble or preventing occurrence of the machine trouble.

5. A fault diagnosis expert system as claimed in claim 4, wherein said identifier of each of the cause-to-effect events is expressed by a special value of the certainty factor, said inference means checking that the certainty factor of a cause-to-effect link is equal to said special value, and identifying the low-level cause event of the cause-to-effect link as an other-dependency cause.

6. A fault diagnosis expert system for locating and eliminating sources of a machine trouble comprising:

storage means for storing a fault diagnosis expert knowledge constructed in a hierarchical structure of a search tree in which cause-to-effect links connecting high-level effect events to low-level cause events are pre-defined and all possible low-level cause events are pre-enumerated for each high-level effect event, said low-level cause events being candidates for a cause producing said high-level effect events;

user interface means for providing a user with questions and responses regarding state of the machine trouble, allowing the user to supply user information in reply to the questions from the user interface means and allowing the user interface means to provide the user with the responses in reply to the user information;

inference means for inferring a cause of the machine trouble with the fault diagnosis expert knowledge stored in said storage means and the user information supplied from the user; and

outputting means for allowing inference results supplied from said inference means to be displayed for the user,

wherein the cause-to-effect links pre-defined within the fault diagnosis expert knowledge stored in the storage means have an identifier for indicating a difference between self-dependency causes and other-dependency causes, said self-dependency causes having specific troubleshooting measures for eliminating sources of the machine trouble or preventing occurrence of the machine trouble, said other-dependency causes having no said specific troubleshooting measures and being followed by low-level self-dependency causes downstream of said other-dependency cause through the search tree, said low-level self-dependency causes having said specific troubleshooting measures.

7. A fault diagnosis expert system as claimed in claim 6, wherein a certainty factor is pre-defined for each of the cause-to-effect links within the storage means, said certainty factor indicating degree of likelihood of the low-level cause event being a cause producing the high-level effect event.

8. A fault diagnosis expert system as claimed in claim 7, wherein said identifier of each of the cause-to-effect events is expressed by a special value of the certainty factor, said inference means checking that the certainty factor of a cause-to-effect link is equal to said special value, to identify the low-level cause event of the cause-to-effect link as an other-dependency cause.

9. A fault diagnosis expert system as claimed in claim 8, wherein said inference means, after checking that an inferred cause event is an other-dependency cause, enumerates low-level self-dependency causes downstream of said inferred cause event through the search tree, said inference means making a decision that said low-level self-dependency causes are cause candidates of the machine trouble.

10. A fault diagnosis expert system as claimed in claim 9, wherein some of said cause-to-effect links have supporting conditions for facilitating an inference process of locating a true cause of the machine trouble, said supporting condition being represented by a conditional statement, a positive weight and a negative weight.

11. A fault diagnosis expert system as claimed in claim 10, wherein said positive weight of the supporting condition is added to the certainty factor to increase the value of the certainty factor when the conditional statement is affirmative with respect to the state of the machine observed, while said negative weight is added to the certainty factor to decrease the value of the certainty factor when the conditional statement is negative.

12. A fault diagnosis expert system as claimed in claim 11, wherein said fault diagnosis expert knowledge stored in the storage means includes a cause candidate criterion which is pre-defined as an event used for the inference means to trace back appropriate nodes within the search tree when the inference means infers the cause of the machine trouble to locate a true cause thereof.

13. A fault diagnosis expert system as claimed in claim 12, wherein said inference means checks, during inference of selecting a low-level cause event from a high-level effect event by tracing back nodes through the search tree, whether an event given as the cause candidate criterion with respect to one of the cause-to-effect links being connected between the high-level effect event and the low-level cause event already takes place before the low-level cause event is selected from the high-level effect event, the inference means then making a decision that the high-level cause event is a cause candidate of the low-level effect event only when said event already takes place.