Diagnostic expert system for hierarchically decomposed knowledge domains

• US 5,619,621 A
• Filed: 07/15/1994
• Issued: 04/08/1997
• Est. Priority Date: 07/15/1994
• Status: Expired due to Fees

• Current Assignee: Storage Technologies Corporation (Oracle Corporation)
• Original Assignee: Storage Technologies Corporation (Oracle Corporation)
• Inventors: Puckett, Timothy L.
• Primary Examiner(s): Hafiz, Tariq R.
• Application Number:
  US08/275,499
• Time in Patent Office: 998 Days
• Field of Search: 395/50-54, 395/60-61, 395/65, 395/75, 395/76, 395/77
• US Class Current: 706/47
• CPC Class Codes: G06F 11/2257 using expert systems

Abstract

A forward chaining expert system is disclosed for use in analyzing categorization problems such as in diagnosing device malfunctions. The expert system is particularly useful in diagnosing malfunctions in a knowledge domain having a hierarchical functional decomposition. In such a knowledge domain, the hierarchical decomposition is used in constructing a collection hierarchically related rule sets which constitute the rule base for the expert system. The novel inferencing engine of the expert system iteratively examines rules within a selected rule set and: (a) determines the veracity of rule premises when compared with data within a fact base and (b) examines any statuses returned from the performance of rule consequents. Thus, upon encountering a rule consequent returning success, the inferencing engine selects a new rule set at a lower level in the rule set hierarchy and applies the new rule set to the data retained in the fact base. By allowing the inferencing engine to iteratively select one or more rule sets from a partially ordered collection of rule sets whereby the partial order is related to the knowledge domain hierarchical decomposition, the inferencing engine is able to provide problem solutions in increasingly greater detail when selecting increasingly more specific rule sets. Thus, the inferencing engine need not backtrack to examine alternative solutions.

Claims

1. A method for analyzing data comprising:

establishing a hierarchical decomposition of a device into a plurality of components, each of the components outputting diagnostic data items, said hierarchical decomposition including a plurality of levels of said components of the device with each lower level component having a related higher level component such that a lower level component provides a functional capability of its related higher level component;

storing in storage means a rule set hierarchy having a plurality of rule sets, said rule set hierarchy having a hierarchical relationship between rule sets of said plurality of rule sets wherein;

(A1) each rule set of said plurality of rule sets includes rules for inferring a performance of a related component of said plurality of components; and

(A2) for each pair, R₁ and R₂, of said rule sets, R₁ is a lower level rule set related to R₂ as a higher level rule set, when said related component for R₁ is a lower level component to said related component for R₂ ;

selecting, as a designated rule, an eligible rule from a current rule set of said plurality of the rule sets, when such an eligible rule exists, wherein said designated rule has a premise with a maximal number of predicates for evaluation using informational data items stored in a fact base;

evaluating said premise of said designated rule for determining a premise value;

performing, when said premise value is a predetermined value, a consequent of said designated rule for determining additional informational data items using said diagnostic data items, said performance of said consequent at least providing a result;

using said result for determining whether to add said additional informational data items to said fact base and for determining a new instantiation of said current rule set, wherein;

(B1) when Said result includes a first predetermined value, said new instantiation of said current rule set must be a predetermined eligible rule set that is at a lower level to said current rule set;

(B2) when said result includes a second predetermined value, said new instantiation of said current rule set must be an eligible rule set that is;

(a) unrelated to said current rule set, and

(b) is at a lower level than each previous instantiation of said current rule set;

setting said current rule set to ineligible for reconsideration as said current rule set;

repeating said steps of selecting, evaluating, performing, using, and setting until a rule cannot be selected for said designated rule;

stopping the analysis of the diagnostic data items output by components of the device; and

outputting an analysis related to a performance of the device using information from performing a consequent of at least one rule selected as said designated rule.

2. A method, as claimed in claim 1, wherein:

said step of establishing a hierarchical decomposition of a device includes establishing a functional decomposition of the device wherein a component of a lower level of said components performs a subfunction of a related component of a higher level of said components.

3. A method, as claimed in claim 1, wherein:

said step of storing includes;

(C1) providing one or more rules in a higher level rule set for use in diagnosing a fault in a related higher level component; and

(C2) storing in said storage means one or more rules for each lower level rule set, each said lower level rule set for use in diagnosing a fault in a related lower level component.

4. A method, as claimed in claim 1, wherein:

the data items output by a higher component of a higher level of said components includes data used in detecting a fault in the higher component and for each lower component of a lower level of said components, the data items output by the lower component includes data used in detecting a fault in the lower component.

5. A method, as claimed in claim 1, wherein:

said step of storing a rule set hierarchy includes incorporating in said storage means a predetermined linkage for linking a first rule in a higher rule set of a higher level of said rule set hierarchy with a predetermined lower rule set of a lower level of said rule set hierarchy so that when said first rule is said designated rule and (B1) of claim 1 occurs, then said predetermined lower rule set becomes said current rule set.

6. A method, as claimed in claim 1, wherein:

said step of performing said consequent of said designated rule includes inputting to said fact base fact data having informational content provided by the informational data items output by a component related to said current rule set.

7. A method, as claimed in claim 1, wherein:

said step of performing includes examining said result obtained from performing said consequent so that said predetermined eligible rule set at a lower level to said current rule set is selected as a new instantiation of said current rule set when said result indicates successful performance of said consequent.

8. A method, as claimed in claim 1, wherein:

said consequent includes a query of data items output by the device.

9. A method, as claimed in claim 1, wherein said step of repeating includes a second performance of said step of selecting for selecting said designated rule from a first lower level rule set is determined as said new instantiation of said current rule, wherein each rule of said first lower level rule set used for analyzing data from a related first lower level component of said plurality of components, and wherein said second performance of selecting ceases to use said first lower rule set as said current rule set when one of the following occur:

(D1) a consequent of said designated rule performs successfully; and

(D2) all rules of said current rule set have been selected as said designated rule and condition (D1) has not occurred for any designated rule from said current rule set.

10. A method, as claimed in claim 9, wherein:

said step of repeating includes iteratively choosing a previously unchosen rule set at a same level as said current rule as said new instantiation of said current rule set whenever said condition (D2) occurs, said step of iteratively choosing continuing until one of (D1) occurs and all rule sets of said same level have been chosen.

11. A method, as claimed in claim 9, wherein:

said step of repeating includes iteratively choosing a next rule set from said plurality of rule sets as said new instantiation of said current rule set when;

(E1) said next rule set is further down said rule set hierarchy than said current rule set;

(E2) condition (D1) occurs;

(E3) said next rule set is hierarchically related to said current rule set according to said hierarchical relationship between said plurality of rule sets; and

(E4) said next rule set is predetermined, via said hierarchical relationship, to become said current rule set whenever condition (D1) occurs for said current rule set as presently instantiated.

12. A method, as claimed in claim 11, wherein:

said step of stopping includes terminating analysis of the data items output by the device when condition (D1) occurs and no said next rule set exists.

13. A method, as claimed in claim 1, wherein:

said step of outputting an analysis includes outputting one of;

(F1) said additional informational data items determined in said step of performing; and

(F2) a diagnosis of a fault on the device.

14. An expert system for analyzing data comprising:

fact storage means for storing data input to and updated by said expert system during an analysis of data;

rule storage means for storing a plurality of rules, each rule in said plurality of rules having a premise with one or more conditions for evaluation using data in said fact storage means and a consequent, for determining at least one of;

(A1) new data for said fact storage means;

(A2) a conclusion inferred from data residing in said fact storage means;

rule grouping means for grouping said plurality of rules in said rule storage means into a plurality of rule sets;

inferencing means for performing each of;

(B1) selecting a rule set from said plurality of rule sets as a current rule set, said current rule set to be used in analyzing data within said fact storage means;

(B2) selecting an eligible rule from said current rule set of as a designated rule wherein said designated rule is selected whose premise has a maximal number of said conditions among eligible rules in said current set;

(B3) evaluating a premise of said designated rule, and performing a consequent of said designated rule when a premise of said designated rule is satisfied by data in said fact storage means, thereby determining at least one of (A1) and (A2) when said consequent of said current rule performs successfully;

rule set relating means for relating a first rule set in said plurality of rule sets to one or more other rule sets in said plurality of rule sets so that each second rule set of said other rule sets is associated with a predetermined rule in said first rule set, wherein when;

(C1) said first rule set is selected as said current rule set in (B1);

(C2) a predetermined second rule set of said other rule sets is associated with said designated rule selected in (B2); and

(C3) said consequent of said designated rule performs successfully in (B3), then said second rule set must be selected as said current rule set in a further performance of (B1) through (B3) by said inferencing means and said first rule set is designated as ineligible for reconsideration as said current rule set;

means for outputting resultant data when at least one of said designated rules cannot be selected in (B2) and a current rule set cannot be selected in (B1), said resultant data providing information related to one of;

(D1) data stored in said fact storage means; and

(D2) a conclusion related to said analysis of data.

15. An expert system, as claimed in claim 14, wherein:

said analysis of data includes analyzing an error log for fault data so that a fault can be detected.

16. An expert system, as claimed in claim 14, wherein:

said fact storage means includes a fact base, said fact base used for storing data used in determining when said premise of said designated rule is satisfied in (B3), said fact base retaining substantially only data not satisfying the premise of said designated rule.

17. An expert system, as claimed in claim 14, wherein:

said rule grouping means includes for each rule set of said plurality of the rule sets a different identifier identifying the rule set, said identifier used in identifying each rule within the rule set.

18. An expert system, as claimed in claim 14, wherein:

said rule grouping means partitions said plurality of rules into rule sets such that substantially each rule of said plurality of rules is contained in a single rule set.

19. An expert system, as claimed in claim 14, wherein:

said rule set relating means includes a hierarchical partial ordering of said plurality of rule sets wherein said first rule set is a parent rule set of said one or more other rule sets.

20. An expert system, as claimed in claim 14, wherein:

said rule set relating means includes a mapping means such that, for each rule, R, of said first rule set, said mapping means provides a reference to a rule set, S, of said one or more other rule sets wherein when R becomes said designated rule and said consequent of said designated rule performs successfully, then said mapping is used in determining rule set S as a next instantiation of said current rule set.

21. An expert system, as claimed in claim 14, wherein:

said inferencing means includes a breadth search means for selecting said designated rule, said breadth search means iteratively selecting as said designated rule a previously unselected rule whose premise has a maximal number of conditions or predicates for matching the data in said fact storage means.

22. An expert system, as claimed in claim 14, wherein:

said inferencing means includes pattern matching means for determining when said premise of said designated rule is satisfied by data in said fact storage means, said pattern matching means matching one or more conditions or predicates of said premise with the data in said fact storage means.