Bayesian approach to identifying sub-module failure

• US 8,458,525 B2
• Filed: 03/19/2010
• Issued: 06/04/2013
• Est. Priority Date: 03/19/2010
• Status: Expired due to Fees

• Current Assignee: Hamilton Sundstrand Space Systems International Inc (Raytheon Technologies Corporation d/b/a RTX)
• Original Assignee: Hamilton Sundstrand Space Systems International Inc (Raytheon Technologies Corporation d/b/a RTX)
• Inventors: Pargellis, Andrew, Sutin, Brian M.
• Primary Examiner(s): MASKULINSKI, MICHAEL C
• Application Number:
  US12/727,437
• Publication Number:
  US 20110231703A1
• Time in Patent Office: 1,173 Days
• Field of Search: None
• US Class Current: 714/37
• CPC Class Codes: G05B 23/0281 Quantitative, e.g. mathematical distance; Clustering; Neural networks; Statistical analysis

Abstract

A diagnostic device identifies failed sub-modules within a larger system based on error codes received from the system. The device stores a likelihood matrix that correlates each sub-module with each possible error code and maintains a likelihood value corresponding to the probability of a failed sub-module generating a corresponding error code and stores a prior probability of failure associated with each sub-module based on prior observational data. In response to received error codes, the device calculates a posterior probability of failure for each of the plurality of sub-modules based on a product of the likelihood values corresponding to the received error codes and the prior probability of failure associated with each sub-module. Based on the calculated posterior probability, the device identifies the sub-module with the highest posterior probability of failure as the failed sub-module.

Claims

1. A method of identifying sub-module failure within a system having a plurality of sub-modules, the method comprising:

calculating a prior probability of failure associated with each sub-module based on a lambda-smoothing equation

2. The method of claim 1, wherein calculating the posterior probability of failure includes multiplying the likelihood values selected based on the received error codes with the prior probability of failure to generate the posterior probability of failure with respect to each sub-module.

3. The method of claim 2, wherein calculating the posterior probability of failure is based on the equation

4. A diagnostic device for identifying failed sub-modules within a larger system, the diagnostic device comprising:

an input for receiving error codes from the system;

a memory for storing a likelihood matrix that correlates each sub-module with each possible error code and maintains a likelihood value corresponding to the probability of a failed sub-module generating a corresponding error code and stores a prior probability of failure associated with each sub-module based on prior observational data, wherein the prior probabilities included in the likelihood matrix are calculated with a lambda-smoothing algorithm that is based on a number of observed failures associated with a particular sub-module, a number of sub-modules included within the system, and a number of total observed failures associated with the system; and

a processor for executing an algorithm that calculates a posterior probability of failure for each of the plurality of sub-modules based on a product of the likelihood values corresponding to the received error codes and the prior probability of failure associated with each sub-module, wherein the processor generates an output identifying the sub-module with the highest posterior probability of failure as the failed sub-module.

5. The diagnostic device of claim 4, wherein the output generated by the processor includes an ordered list of possible sub-module failures based on the posterior probabilities of failure calculated for each of the plurality of sub-modules.

6. The diagnostic device of claim 4, wherein the diagnostic device is internal to the larger system.

7. The diagnostic device of claim 4, wherein the diagnostic device is external to the larger system.

8. The diagnostic device of claim 4, wherein the lambda-smoothing algorithm is defined by the equation

9. The diagnostic device of claim 4, wherein the posterior probability of failure for each sub-module is calculated based on a Bayesian method defined by the equation

10. A diagnostic device for identifying failed sub-modules within a larger system, the diagnostic device comprising:

memory means for storing a likelihood matrix that includes likelihood values corresponding to a probability of a failed sub-module generating error codes and a prior probability of failure associated with each sub-module, wherein the likelihood values are calculated with a lambda-smoothing algorithm that is based on a number of observed failures associated with a particular sub-module, a number of sub-modules included within the system, and a number of total observed failures associated with the system;

processor means for calculating a posterior probability of failure for each of the plurality of sub-modules based on likelihood values selected that correspond to the error codes and the prior probability of failure associated with each sub-module; and

output means for identifying based on the calculated posterior probability of failure associated with each sub-module the sub-module most likely to have failed.

11. The diagnostic device of claim 10 wherein the lambda-smoothing algorithm is defined by the equation

12. The diagnostic device of claim 10, wherein the posterior probability of failure for each sub-module is calculated based on a Bayesian method defined by the equation

13. The diagnostic device of claim 10 further comprising means for receiving one or more of the error codes from the system.