Method of testing and predicting failures of electronic mechanical systems

• US 5,581,694 A
• Filed: 10/17/1994
• Issued: 12/03/1996
• Est. Priority Date: 10/17/1994
• Status: Expired due to Fees

• Current Assignee: The United States of America As Represented By The Secretary of Agriculture
• Original Assignee: United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration
• Inventors: Patterson-Hine, Frances A., Iverson, David L.
• Primary Examiner(s): Beausoliel, Jr., Robert W.
• Assistant Examiner(s): TU, CHRISTINE TRINH LE
• Application Number:
  US08/332,198
• Time in Patent Office: 778 Days
• Field of Search: 395/183.02, 395/183.13, 395/183.15, 395/62, 395/63, 395/65
• US Class Current: 714/26
• CPC Class Codes: G06F 11/2257 using expert systems

Abstract

A method employing a knowledge base of human expertise comprising a reliability model analysis implemented for diagnostic routines is disclosed. The reliability analysis comprises digraph models that determine target events created by hardware failures human actions, and other factors affecting the system operation. The reliability analysis contains a wealth of human expertise information that is used to build automatic diagnostic routines and which provides a knowledge base that can be used to solve other artificial intelligence problems.

Claims

1. A method of testing and predicting failures of electronic and mechanical systems each comprising a plurality of units, said method comprising the steps of:

(a) receiving data samples indicative of the operating status of each of said plurality of units of said system;

(b) analyzing said data samples to determine any failures of any of said plurality of units and generating respective failure signals thereof;

(c) directing said failure signals to a digital computer having a memory for storing a knowledge base comprising objects separately stored in memory and each object having a plurality of slots, said knowledge base having encapsulation, inheritance and polymorphism characteristics, said knowledge base further comprising a digraph model that determines predetermined target events and comprises nodes and AND gates that are connected by directed edges, each of said nodes corresponding to a particular failure of said plurality of units and each of said AND gates having a plurality of inputs, said nodes and AND gates being connected by the directed edges so that propagation therebetween and exit therefrom determines said target events, said propagation being determined by executable singletons and doubletons, each of said nodes, said AND gates and said doubletons being assigned as an individual object in memory;

(d) classifying said failure signal as one of said singletons and doubletons and generating singleton and doubleton failure signals therefor;

(e) identifying said singleton generated failure signal with its respective node, then delivering said node identification to its object and monitoring for the occurrence of its respective predetermined target event; and

(f) identifying said doubleton generated failure signal with its respective node as well as its own identity and delivering said node identification to its object and said doubleton identification to its object, and monitoring for the occurrence of any predetermined target event.

2. A method of testing and predicting failures of electronic and mechanical systems each comprising a plurality of units, said method comprising the steps of:

(a) receiving data samples indicative of the operating status of each of said plurality of units of said system;

(b) analyzing said data samples to determine any failures of any of said plurality of units and generating respective failure signals thereof;

(c) directing said failure signals to a digital computer having a memory for storing a knowledge base comprising objects separately stored in memory and each object having a plurality of slots, said knowledge base having encapsulation, inheritance and polymorphism characteristics, said knowledge base further comprising a digraph model that determines predetermined target events and comprises nodes and AND gates that are connected by directed edges, each of said nodes corresponding to a particular failure of said plurality of units and each of said AND dates having a plurality of inputs, said nodes and AND gates being connected by the directed edges so that propagation therebetween and exit therefrom determines said target events, said propagation being determined by executable singletons and doubletons, each of said nodes, said AND gates and said doubletons being assigned as an individual object in memory;

(d) classifying said failure signal as one of said singletons and doubletons and generating singleton and doubleton failure signals therefor, said step of classifying said failure signals and generating said singleton signal further comprises the steps of;

(i) marking one of said plurality of slots of said node of said object of said respective failed node as failed; and

(ii) identifying other nodes that said marked node propagates thereto by way of said directed edges and marking said other propagated nodes as failed in one of said plurality of slots of said objects of each of said other propagated nodes; and

(iii) monitoring for the occurrence of any target events;

(e) identifying said singletons generated failure signal with its respective node, then delivering said node identification to its object and monitoring for the occurrence of its respective predetermined target event; and

(f) identifying said doubleton generated failure signal with its respective node as well as its own identity and delivering said node identification to its object and said doubleton identification to its object, and monitoring for the occurrence of any predetermined target event.

3. The method of testing and predicting failures of electronic and mechanical systems according to claim 2 further comprising the step of identifying said propagated nodes that are reached by said directed and propagating edges as being inputs to one of a group consisting of an AND gate object and a multiple output node object and then marking one of said slots of one or more respective node objects that are reached by said directed and propagating edges as an object which is equivalent to said objects of said group.

4. The method of testing and predicting failures of electronic and mechanical systems according to claim 3, wherein said step of classifying and generating said doubleton failure signal further comprises the steps of:

(a) analyzing each failure node to determine if its directed edge is connected to an input of an AND gate object and also determining if a failure node object has been marked as an equivalent node;

(b) constructing doubletons for each AND gate object by determining a cross product singleton node objects each of which cross product being an input of said AND gate object analyzed in step 4(a);

(c) creating a unique said doubleton object for at least two cross product inputs to said constructed AND gate object;

(d) sorting said AND gate objects into upstream and downstream AND gate objects and giving priority to said sorted upstream AND gate objects over said downstream AND gate objects in the processing involved with testing and predicting of failures;

(e) propagating said AND gate objects having an input which has its object marked as equivalent to form a downstream_-- andins list;

(f) checking said downstream_-- andins list to determine if any associated AND gate objects have been qualified and if any target events has been generated;

(g) checking said AND gate objects which have their object marked as being an equivalent object to determine if the objects marked as equivalent propagate through said qualified AND gates and classifying such AND gate objects as hidden doubletons; and

(h) repeating steps (b) through (g) for said hidden doubletons.