Fault diagnosis and prognosis using diagnostic trouble code markov chains

• US 8,498,776 B2
• Filed: 11/17/2009
• Issued: 07/30/2013
• Est. Priority Date: 11/17/2009
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations LLC (General Motors Company)
• Inventors: Singh, Satnam, Khare, Vineet R., Chougule, Rahul
• Primary Examiner(s): ARTHUR JEANGLAUDE, GERTRUDE
• Application Number:
  US12/620,466
• Publication Number:
  US 20110118932A1
• Time in Patent Office: 1,351 Days
• Field of Search: 701/1, 701/2, 701/29.2, 701/29.9, 701/30.2, 701/31.8, 701/31.4, 701/34.3, 701/36, 701/37, 701/33.4, 714/26, 714/E11.029, 706/12, 706/52, 340/438
• US Class Current: 701/33.4
• CPC Class Codes: G05B 19/0428 Safety, monitoring G05B19/0423 takes precedence
• CPC Class Codes: G05B 2219/24076 Markov model for safety analysis
• CPC Class Codes: G05B 2219/2637 Vehicle, car, auto, wheelchair
• CPC Class Codes: G05B 23/0232 based on qualitative trend analysis, e.g. system evolution

Abstract

A system and method for fault diagnosis includes receiving information defining a relationship between failure modes and diagnostic trouble codes and extracting diagnostic trouble code data, including set times, frequency data and diagnostic trouble code sequence information for a plurality of diagnostic trouble codes relating to a plurality of failure modes. The system and method further include constructing a Markov chain using the diagnostic trouble code data for each of the plurality of failure modes, training the Markov chain to learn a set of state parameters using the diagnostic trouble code data, and computing a likelihood of a diagnostic trouble code sequence for each of the plurality of failure modes using the trained Markov chains.

Claims

1. A fault diagnosis and prognosis system, comprising:

a plurality of electronic control units, wherein at least one of the plurality of electronic control units is configured to;

receive information defining a relationship between failure modes and diagnostic trouble codes;

extract diagnostic trouble code data for a plurality of diagnostic trouble codes relating to a particular failure mode;

construct a Markov chain using the diagnostic trouble code data for the particular failure mode;

train the Markov chain to learn a set of state parameters using the diagnostic trouble code data;

compute a likelihood of a diagnostic trouble code sequence for the particular failure mode using the trained Markov chain; and

predict a trend using the trained Markov chain.

2. The system of claim 1, wherein the diagnostic trouble code data includes set times, frequency data and diagnostic trouble code sequence information.

3. The system of claim 1, wherein the diagnostic trouble code data is extracted from field failure data and historical data.

4. The system of claim 1, wherein the particular failure mode is selected from an ambiguity group containing a plurality of failure modes.

5. The system of claim 4, wherein a Markov chain is constructed for each of the plurality of failure modes in the ambiguity group.

6. The system of claim 5, wherein a likelihood of a diagnostic trouble code sequence is computed for each of the plurality of failure modes in the ambiguity group.

7. The system of claim 6, wherein the likelihood for each of the plurality of failure modes is ranked in order of most likely to fail.

8. The system of claim 1, further including predicting a remaining time to next failure state using the trained Markov chain.

9. The system of claim 1, wherein the set of state parameters includes initial and transition probabilities for each state of the Markov chain.

10. The system of claim 1, wherein the at least one of the plurality of electronic control units is a central diagnostic electronic control unit.

11. The system of claim 1, wherein the information defining a relationship between failure modes and diagnostic trouble codes is in the form of a diagnostic matrix.

12. The system of claim 1, wherein each of the diagnostic trouble codes represent a state of the Markov chain.

13. The system of claim 12, wherein the Markov chain is constructed to include a no diagnostic trouble code state to diagnosis intermittent failures.

14. A method for fault diagnosis, comprising:

receiving information defining a relationship between failure modes and diagnostic trouble codes;

extracting diagnostic trouble code data, including set times, frequency data and diagnostic trouble code sequence information for a plurality of diagnostic trouble codes relating to a plurality of failure modes;

constructing a set of Markov chains using the diagnostic trouble code data for each of the plurality of failure modes;

training the set of Markov chains to learn a set of state parameters using the diagnostic trouble code data; and

computing a likelihood of a diagnostic trouble code sequence for each of the plurality of failure modes using the trained set of Markov chains.

15. The method of claim 14, further including predicting a trend over a given horizon using a trained Markov chain from the set of Markov chains.

16. The method of claim 14, further including determining a fault diagnosis by ranking a likelihood for each of the plurality of failure modes in order of most likely to fail.

17. The method of claim 16, further including communicating the fault diagnosis to an on-line communication system.

18. The method of claim 16, further including communicating the fault diagnosis to an off-line communication system.

19. A computer-readable non-transitory medium tangibly embodying computer-executable instructions for:

receiving a diagnostic matrix defining a relationship between failure modes and diagnostic trouble codes;

extracting diagnostic trouble code data, including set times, frequency data and diagnostic trouble code sequence information for a plurality of diagnostic trouble codes relating to a plurality of failure modes;

constructing a set of Markov chains using the diagnostic trouble code data for each of the plurality of failure modes, wherein each of the diagnostic trouble codes represent a state of a Markov chain from the set of Markov chains;

training the set of Markov chains to learn a set of state parameters using the diagnostic trouble code data, wherein the set of state parameters includes initial and transition probabilities for each state of the Markov chain from the set of Markov chains;

computing a likelihood of a diagnostic trouble code sequence for each of the plurality of failure modes using the trained set of Markov chains;

determining a fault diagnosis by ranking a likelihood for each of the plurality of failure modes in order of most likely to fail; and

predicting a remaining time to next failure state using the trained Markov chain.

20. The system of claim 19, wherein a Markov chain from the set of Markov chains is constructed to include a no diagnostic trouble code state to diagnosis intermittent failures.