Method and apparatus for generation of multi-branched diagnostic trees

• US 5,272,704 A
• Filed: 08/18/1989
• Issued: 12/21/1993
• Est. Priority Date: 08/18/1989
• Status: Expired due to Fees

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: General Electric Company
• Inventors: Jolly, Christopher H., Zalondek, Kevin C., Tong, David W.
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/395,826
• Time in Patent Office: 1,586 Days
• Field of Search: 364/513, 395/55, 371/15.1, 371/23
• US Class Current: 714/26
• CPC Class Codes: G06F 11/2257 using expert systems
• CPC Class Codes: H04L 41/0659 by isolating or reconfiguring faulty entities

Abstract

A method and apparatus for automated generation of multi-branched diagnostic trees is described. The present invention generally is practiced on model-based diagnostic systems and utilizes a candidate generator, constraint propagator and best measurement generator, along with a model of a system to be diagnosed, to generate nodes and branches of a diagnostic tree.

Claims

1. A method for automated generation of a multi-branched diagnostic tree for isolation of faults in a system comprising the steps of storing a model of the system to be diagnosed in a memory, said model including terminals indicating measurement points in the system to be diagnosed, inputting into the system measurements of taken at the measurement points in the system,

propagating the variables through the model for automated generation of a diagnostic tree;

determining an operating value for each of said terminals;

selecting a granularity for each node of said diagnostic tree;

establishing at least one terminal as a root node and establishing the root node as a selected node;

generating a plurality of branches extending from said selected node, the number of branches extending from said selected node corresponding to each selected granularity, and selecting a branch value to correspond to each of said generated branches;

selecting one of said branches of said selected node not terminating in a leaf node for analysis;

propagating the branch value of said selected branch through said model thereby generating a propagated value for said terminals;

identifying conflicts between propagated values at each terminal and generating conflict candidate sets;

terminating the selected branch with a leaf node denoting no further fault isolation is possible if no conflict set exists;

terminating the selected branch with a node denoting multiple faults if multiple, nonintersecting conflict sets exits;

terminating the selected branch with a leaf node denoting a single fault if one conflict set having a single component exists;

attempting to identify a next measurement to be made if at least one conflict set exists,

terminating the selected branch with a leaf node denoting no further fault isolation is possible if no next measurement can be identified,

terminating the selected branch with a measurement node corresponding to an identified next measurement and establishing said measurement node as the selected node and returning to the step of generating a plurality of branches if a next measurement can be identified;

finding a branch not terminated with a node and establishing the found branch as the selected branch and returning to the step of propagating the branch value of said selected branch;

finding a measurement node without established branches, establishing said measurement node as the selected node, and returning to the step of generating a plurality of branches if a measurement node is established; and

stopping tree generation if all branches are terminated with leaf nodes.

2. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein selecting the granularity of each node of said tree comprises the step of:

setting the granularity of each node to equal at least two if at least one component of said system to be diagnosed has an analog input or output.

3. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein the granularity of each node is equal to the granularity of the root node.

4. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein the granularity of each node may vary.

5. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein the granularity of each node is selected by a system user.

6. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein the granularity of each node is selected automatically.

7. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein establishing a terminal of said model as a root node comprises the steps of:

identifying an outout terminal of said model; and

establishing said identified output terminal as said root node.

8. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein establishing a terminal of said circuit model as a root node comprises the steps of:

identifying a component most likely to be faulty based on prior experience;

identifying an output terminal of said identified component; and

establishing said identified output terminal as said root node.

9. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein measurement at a plurality of terminals is established as a root node.

10. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein said selected value is a single numerical value.

11. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein said selected value is a range of numerical values.

12. A method for automated generation of a diagnostic tree in accordance with claim 1 further comprising the step of:

finding a node denoting multiple faults, establishing said multiple fault node as a selected node and returning to step (d) before stopping tree generation.

13. A method for automated generation of a diagnostic tree in accordance with claim 1 wherein a node denoting multiple faults is established as a leaf node.

14. An automated diagnostic computer system, comprising:

input means for entering and editing information in the system and generating a model of a system to be diagnosed, said model including terminals indicating measuring points in the system to be diagnosed,

a memory in the system for storing said model;

means for making predictions about terminal values using said model and assigning predicted values to terminals of said model;

means for generating conflicts between said assigned values and actual measured values and between different sets of assigned values;

means for identifying a next measurement to be made for isolating a fault;

means for determining an operating value for each of said terminals;

means for selecting a granularity for each node of said diagnostic tree;

means for establishing a first terminal as a root node and establishing the root node as a selected node;

means for generating a plurality of branches extending from said selected node, the number of branches extending from said selected node corresponding to said selected granularity, and selecting a branch value to correspond to each of said generated branches;

means for selecting one of said branches of said selected node not terminating in a leaf node for analysis;

means for propagating the branch value of said selected branch through said model thereby generating a propagated value for said terminals;

means for identifying conflicts between propagated values at each terminal and generating conflict candidate sets;

means for terminating the selected branch with a leaf node denoting no further fault isolation is possible if no conflict set exists;

means for terminating the selected branch with a node denoting multiple faults if multiple, nonintersecting conflict sets exists;

means for terminating the selected branch with a leaf node denoting a single fault if one conflict set having a single component exists;

means for attempting to identify a next measurement to be made if at least one intersecting conflict set exists, terminating the selected branch with a leaf node denoting no further fault isolation is possible if no next measurement can be identified, terminating the selected branch with a measurement node corresponding to an identified next measurement and establishing said measurement node as the selected node and if a next measurement can be identified generating a plurality of branches extending from said selected node, the number of branches extending from said selected node corresponding to said selected granularity, and selecting a branch value to correspond to each of said generated branches;

means for finding a branch not terminated with a node and establishing the found branch as the selected branch and propagating the branch value of said selected branch through said model thereby generating a propagated value for said terminals; and

means for finding a measurement node without established branches, establishing said measurement node as the selected node, and if a measurement node is established generating a plurality of branches extending from said selected node, the number of branches extending from said selected node corresponding to said selected granularity, and selecting a branch value to correspond to each of said generated branches; and

means for stopping tree generation if all branches are terminated with leaf nodes.

15. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said means for selecting the granularity of each node of said tree comprises:

means for setting the granularity of each node to equal at least two if at least one component of said system to be diagnosed has an analog input or output.

16. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said means for identifying a next measurement identify a measurement which results in a minimum number of measurements required to isolate a fault.

17. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said means for identifying a next measurement identify a measurement which results in a minimum cost to isolate a fault.

18. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein the granularity of each node is equal to the granularity of the root node.

19. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein the granularity of each node may vary.

20. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said means for selecting a granularity for each node of said diagnostic tree comprises means for providing that the granularity of each node is selected by a system user.

21. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said means for selecting a granularity for each node of said diagnostic tree comprises means for providing that the granularity of each node is selected automatically.

22. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said means for establishing a terminal of said model as a root node comprises:

means for identifying an output terminal of said model; and

means for establishing said identified output terminal as said root node.

23. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said means for establishing a terminal of said model as a root node comprises:

means for identifying a component most likely to be faulty based on prior experience;

means for identifying an output terminal of said identified component; and

means for establishing said identified output terminal as said root node.

24. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 further comprising means for establishing measurement at a plurality of terminals as a root node.

25. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said selected value is a single numerical value.

26. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 wherein said selected value is a range of numerical values.

27. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 further comprising:

means for finding a node denoting multiple faults, establishing said multiple fault node as a selected node and returning to step (d) before stopping tree generation.

28. An apparatus for automated generation of a diagnostic tree in accordance with claim 14 mean for establishing a node denoting multiple faults is as a leaf node.

29. An apparatus in accordance with claim 14 wherein said means for terminating the selected branch with a leaf node comprises means for terminating the selected branch with a leaf node which identifies multiple faulty components.

30. A method for automatically assigning a granularity to an internal node of a diagnostic tree of a model-based diagnostic reasoning system, said diagnostic tree including root nodes, integral nodes and branches, said method comprising the steps of:

a) establishing at least one terminal of the model as a root node and establishing the root node as a selected node;

b) determining a value for said one terminal;

c) selecting a granularity for said root node;

d) generating a plurality of branches extending from said root node, the number of branches extending from said root node corresponding to the selected granularity and selecting a branch value to correspond to each said generated branches;

e) selecting one of said branches of said selected node not terminating in a node for analysis;

f) propagating the branch value of said selected branch;

g) attempting to identify a next measurement to be made at a terminal and if no next measurement can be identified then terminating said selected branch in a leaf node;

h) terminating the selected branch with a node identifying the terminal to be measured if a next measurement is identified;

i) determining a number (N) of numeric ranges using the intervals propagated to the terminal to be measured;

j) generating N branches from said node identifying the terminal to be measured;

k) assigning each said branch one of said interval ranges;

l) establishing a node having at least one branch not terminating in a node as the selected node and returning to step (e); and

m) stopping tree generation when all paths of the tree terminate in leaf nodes.

31. A method for automatically assigning a granularity to each internal node in accordance with claim 30 wherein determining a number (N) of interval ranges resulting from propagating the range of values comprises the steps of:

determining if any ranges intersect; and

assigning each non-intersected portion of each said range to a respective branch and assigning each intersected portion to a respective branch.

32. A method for automatically assigning a granularity to each internal node in accordance with claim 31 further comprising the steps of:

determining a mid-point for each of said non-intersected and intersected portions where possible;

defining ranges in accordance with each of said non-intersected portions and their respective midpoints and each of said intersected portions and their respective midpoints; and

generating branches and assigning each branch one of said defined ranges.

33. A method for automatically assigning a granularity to each internal node in accordance with claim 31 further comprising the steps of:

combining intervals in a consecutive pair-wise manner;

defining ranges in accordance with said combined intervals; and

generating branches and assigning each branch one of said defined ranges.

34. A method for automatically assigning a granularity to each internal node in accordance with claim 30 wherein the granularity of said root node is user selected.

35. A method for automatically assigning a granularity to each internal node in accordance with claim 30 wherein the granularity of said root node is automatically selected.

36. A method for automatically assigning a granularity to each internal node in accordance with claim 30 wherein a maximum limit as to the number (N) of branches generated from a node is predetermined.