Method and apparatus for diagnosing difficult diagnose faults in a complex system

• US 20040073844A1
• Filed: 08/26/2003
• Published: 04/15/2004
• Est. Priority Date: 10/28/1999
• Status: Active Grant

• Current Assignee: Nicholas Edward Roddy, Unkle C. Richard
• Original Assignee: Nicholas Edward Roddy, Unkle C. Richard
• Inventors: Roddy, Nicholas Edward, Unkle, C. Richard
• Granted Patent: US 7,100,084 B2
• Time in Patent Office: Days
• US Class Current: 714/39
• CPC Class Codes: G05B 23/0229 knowledge based, e.g. expert systems; genetic algorithms

Abstract

A method and apparatus for determining the root cause of no trouble found events in a machine is disclosed. The actual faults occurring during a predetermined time interval prior to the no trouble found event are analyzed and correlated with the no trouble found events in an effort to identify those actual faults that have a high correlation with each no trouble found event. If a high correlation is not found, then the no trouble found event is analyzed off-line to determine the root cause.

Claims

1. A method for analyzing no trouble found events from one or more similar machines, wherein the no trouble found events are those for which no cause could be identified, said method comprising:

(a) receiving data representing the faults experienced by the one or more machines;

(b) receiving no trouble found event data;

(c) for the one or more machines, selecting faults occurring within a predetermined time prior to a selected no trouble found event;

(d) generating at least one distinct fault cluster from the plurality of selected faults; and

(e) determining the correlation between the selected no trouble found event and the at least one distinct fault cluster.

2. The method of claim 1 wherein the fault data includes operational parametric information occurring at or near the time when the fault occurred.

3. The method of claim 1 including a step (f) determining the root cause for the selected no trouble found event based on a high correlation with the at least one distinct fault cluster.

4. The method of claim 1 wherein the step of determining the correlation includes:

(e1) creating a plurality of cases, wherein each case comprises a specific no trouble found event and the faults selected in step (c);

(e2) creating distinct fault clusters for each of the plurality of cases, wherein the number of distinct fault clusters within each case is equivalent to the number of unique fault combinations for the faults within the case;

(e3) determining the number of occurrences of the combination of the selected no trouble found event and each cluster within all cases;

(e3) determining the number of occurrences of each fault cluster within all cases; and

(e4) wherein the correlation value is calculated by dividing the results of step (e2) by the results of step (e3).

5. An article of manufacture comprising a computer program product comprising a computer-usable medium having a computer-readable code therein for analyzing no trouble found events from one or more similar machines, said article of manufacture comprising:

a computer-readable program code module for receiving fault data representing the faults experienced by the one or more machines;

a computer-readable program code module for receiving no trouble found event data;

a computer-readable program code module for the one or more machines, for selecting faults occurring within a predetermined time prior to the selected no trouble found event;

a computer-readable program code module for generating at least one distinct fault cluster from the plurality of selected faults; and

a computer-readable program code module for determining the correlation between the selected no trouble found event and the at least one distinct fault cluster.

6. The article of manufacture of claim 5 wherein the fault data includes operational parametric information occurring at or near the time when the fault occurred.

7. The article of manufacture of claim 5 including a computer-readable program code module for analyzing the no trouble found event to determine the root cause based on a high correlation with the at least one distinct fault cluster.