METHODS SYSTEMS AND APPARATUS FOR DETERMINING WHETHER BUILT-IN-TEST FAULT CODES ARE INDICATIVE OF AN ACTUAL FAULT CONDITION OR A FALSE ALARM

• US 20120011403A1
• Filed: 07/08/2010
• Published: 01/12/2012
• Est. Priority Date: 07/08/2010
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Bharadwaj, Raj Mohan, Bock, Joel, Scandura, Phil
• Granted Patent: US 8,621,305 B2
• Time in Patent Office: Days
• US Class Current: 714/30
• CPC Class Codes: G06F 11/2215 to test error correction or detection circuits
• CPC Class Codes: G06F 11/263 Generation of test inputs, e.g. test vectors, patterns or sequences ; with adaptation of the tested hardware for testability with external testers

Abstract

Methods and apparatus are provided for determining whether a built-in-test fault code (BITFC) data sequence generated by a built-in-test (BIT) of a particular module of a complex system is indicative of an actual fault condition. A regression function is generated for the particular module based on stored BITFC data sequences generated by the BIT and stored repair data for that module from a fault history database. Later, during operation of the particular module, the BIT generates a new BITFC data sequence. A processor can then load the new BITFC data sequence and execute the regression function with respect to the new BITFC data sequence to determine whether the new BITFC data sequence is indicative of an actual fault condition at the particular module or is indicative of a false fault condition at the particular module.

Claims

1. A method for determining whether a built-in-test fault code (BITFC) data sequence generated by a built-in-test (BIT) of a particular module of a complex system is indicative of an actual fault condition, the method comprising:

generating a regression function for the particular module based on stored built-in-test fault code (BITFC) data sequences generated by the BIT for the particular module and stored repair data for that module, wherein the regression function correlates some particular BITFC data sequences with a finding of an actual fault condition found (FF) result at the particular module, and correlates other particular BITFC data sequences with a finding of no fault condition found (NFF) result at the particular module;

executing the BIT at the particular module during operation of the particular module to generate a new BITFC data sequence; and

applying, at a processor, the regression function to the new BITFC data sequence to determine whether the new BITFC data sequence is indicative of an actual fault condition at the particular module or is indicative of a false fault condition at the particular module.

2. A method according to claim 1, wherein the step of generating a regression function, comprises:

loading, from a fault history database, the stored BIT fault code data sequences generated by the BIT for the particular module and the repair data, wherein the repair data comprises actual fault condition found (FF) result data and no fault condition found (NFF) result data corresponding to the stored BIT fault code data sequences for the particular module, wherein the actual fault condition found (FF) result data indicates that a particular BIT fault code data sequence generated by the BIT corresponded to an actual fault condition found (FF) result, and wherein the no fault condition found (NFF) result data indicates that the particular BIT fault code data sequence generated by the BIT corresponded to a false alarm (NFF) result;

processing the stored BIT fault code data sequences and the repair data to generate sorted words that are sorted by length and frequency, wherein each word corresponds to a plurality of symbols that are part of a symbol sequence, wherein each symbol is an identifier that corresponds to a particular BIT fault code, and wherein each word ranges between two and n symbols in length;

using the sorted words to generate feature vectors comprising a first feature vector for a first data class that corresponds to the no fault condition found (NFF) result data, and a second feature vector for a second data class that corresponds to actual fault condition found (FF) result data; and

generating, based on the feature vectors, the regression function.

3. A method according to claim 2, further comprising:

storing the regression function in memory for subsequent use in analyzing the new BITFC data sequence.

4. A method according to claim 2, wherein the step of processing the stored BIT fault code data sequences and the repair data to generate sorted words that are sorted by length and frequency, comprises:

determining unique BIT fault codes from the BIT fault code data sequences;

mapping each unique BIT fault code to a corresponding symbol, wherein each symbol corresponds to a particular BIT fault code;

converting the stored BIT fault code data sequences into a symbol sequence;

segmenting the symbol sequence into subsets of words, wherein each subset of words comprises subsequences of identical symbol length, wherein the symbol length of each subsequence varies from two symbols to a maximum word length N, wherein the subsets collectively make up a collection of subsets;

counting a number of occurrences of words of each unique symbol length in the collection of subsets;

generating word counts for each unique word, for each unique length from the collection of subsets;

for each subset of words, sorting word counts by the number of occurrences of each word, in descending order, from the most frequently occurring words to the least frequently occurring words;

converting the sorted word counts to sorted word frequencies by dividing the sorted word counts by total number of words, wherein each word frequency is the frequency of occurrence in each word of similar length in the current collection of sequences;

converting the sorted word counts to sorted word frequencies by dividing the sorted word counts by the total number of words, wherein each word frequency is the frequency of occurrence in each word of similar length in the historical database.

5. A method according to claim 4, wherein the step of segmenting, comprises:

segmenting the symbol sequence into sets of words, wherein each set of words includes words that comprise a particular number of consecutive symbols that are part of the symbol sequence, wherein the particular number is different in each set of words, and wherein the particular number of consecutive symbols in any word ranges between two for the second set of words and N, where N is the maximum word length specified as a number of symbols.

6. A method according to claim 4, wherein the step of sorting comprises:

for each set of words;

determining the number of occurrences of each word in the particular set of words to generate word counts for that set of words; and

arranging the word counts for that set of words in descending order from the most frequently occurring words to least frequently occurring words.

7. A method according to claim 4, further comprising the steps of:

comparing the word frequencies for each word to a minimum word frequency threshold; and

discarding any words having a corresponding word frequency that is less than the minimum word frequency threshold to filter low frequency words from the symbol sequence.

8. A method according to claim 4, further comprising the steps of:

comparing the word complexity of each word to a word complexity threshold; and

discarding any words having a corresponding word complexity that is less than the word complexity threshold to filter low complexity words from the symbol sequence.

9. A method according to claim 2, wherein the step of using the sorted words to generate feature vectors, comprises:

concatenating sorted words that correspond to the first data class to generate the first feature vector that corresponds to no-fault condition found (NFF) result data; and

concatenating other sorted words that correspond to the second data class to generate the second feature vector that corresponds to actual fault condition found (FF) result data.

10. A method according to claim 2, wherein the step of generating, based on the feature vectors, the regression function, comprises:

computing, based on the first feature vector, parameters for the regression function that predict likelihood of a no fault found (NFF) at the particular module when words of the first data class that correspond to particular BITFC data sequences indicative of no-fault condition found (NFF) result data are input to the regression function; and

computing, based on the second feature vector, other parameters for the regression function that predict likelihood of an actual fault condition at the particular module when words of the second data class that correspond to particular BITFC data sequences indicative of actual fault condition found (FF) result data are input to the regression function.

11. A method according to claim 1, wherein the step of applying the regression function to the new BITFC data sequence, comprises:

loading, at the processor, the new BITFC data sequence generated by the BIT at the particular module; and

executing the regression function at the processor with respect to the new BITFC data sequence to generate either;

a no-fault found (NFF) result that indicates that the new BITFC data sequence is not indicative of an actual fault condition at the particular module, or

a fault found (FF) result that indicates that the new BITFC data sequence is indicative of an actual fault condition at the particular module.

12. A method according to claim 1, wherein the complex system is an aircraft and wherein the particular module is an avionics module of the aircraft.

13. A method according to claim 1, wherein the built-in-test (BIT) is executed at the particular module to perform a particular diagnostic test with respect to that particular module to determine whether that particular module is operating correctly, and

wherein the BIT generates a BIT fault code when the particular module is determined not to be is operating correctly, and wherein the BITFC fault code data sequence comprises BIT fault codes generated by the BIT.

14. A method according to claim 1, wherein the regression function is a logistic regression function.

15. An analysis system for a complex system that includes a plurality of modules including a particular module that regularly executes a built-in-test (BIT) to determine whether that particular module is operating correctly, wherein the BIT generates a BIT fault code when the particular module is determined not to be is operating correctly, the analysis system comprising:

memory that stores BIT fault code (BITFC) data sequences comprising a BIT fault codes generated by the BIT and repair data for that module corresponding to the BITFC data sequences, wherein the repair data comprises actual fault condition found (FF) result data and no fault condition found (NFF) result data corresponding to the stored BIT fault code data sequences for the particular module;

a first processor that generates a regression function for the particular module based on the stored built-in-test fault code (BITFC) data sequences and the stored repair data for that module, wherein the regression function correlates some particular ones of the BITFC data sequences with a finding of an actual fault condition found (FF) result at the particular module, and correlates other particular ones of the BITFC data sequences with a finding of no fault condition found (NFF) result at the particular module; and

a second processor designed to load the regression function, load a new BITFC data sequence generated by the BIT at the particular module, and execute the regression function with respect to the new BITFC data sequence to generate either;

a no-fault found (NFF) result that indicates that the new BITFC data sequence is not indicative of an actual fault condition at the particular module, or a fault found (FF) result that indicates that the new BITFC data sequence is indicative of an actual fault condition at the particular module.

16. An analysis system according to claim 15, wherein the first processor processes the stored BIT fault code data sequences and the repair data to generate sorted words that are sorted by length and frequency, wherein each word corresponds to a plurality of symbols that are part of a symbol sequence, wherein each symbol is an identifier that corresponds to a particular BIT fault code, and wherein each word ranges between two and n symbols in length.

17. An analysis system according to claim 16, wherein the first processor uses the sorted words to generate feature vectors comprising a first feature vector for a first data class that corresponds to the no fault condition found (NFF) result data, and a second feature vector for a second data class that corresponds to actual fault condition found (FF) result data, and then generates the regression function based on the feature vectors.

18. An analysis system according to claim 16, wherein the first processor:

maps each unique BIT fault code in the stored BIT fault code data sequences to a corresponding symbol to generate a symbol sequence based on the stored BIT fault code data sequences,

segments the symbol sequence into sets of words, wherein each set includes words that comprise a different particular number of consecutive symbols of the symbol sequence,

counts a number of occurrences of each word in the symbol sequence to generate word counts for each word, and then, for each set of words, sorts the word counts by the number of occurrences in descending order, and converts the sorted word counts to sorted word frequencies, wherein each word frequency is the frequency of occurrence of each word in the symbol sequence,

concatenates sorted words that correspond to the first data class to generate the first feature vector that corresponds to no-fault condition found (NFF) result data, and concatenates other sorted words that correspond to the second data class to generate the second feature vector that corresponds to actual fault condition found (FF) result data,

generates parameters for the regression function based on the first feature vector that predict likelihood of a no fault found (NFF) at the particular module when words of the first data class that correspond to particular BITFC data sequences indicative of no-fault condition found (NFF) result data are input to the regression function, wherein the no fault condition found (NFF) result data indicates that the particular BIT fault code data sequence generated by the BIT corresponded to a false alarm (NFF) result, and

generates other parameters for the regression function based on the second feature vector that predict likelihood of an actual fault condition at the particular module when words of the second data class that correspond to particular BITFC data sequences indicative of actual fault condition found (FF) result data are input to the regression function, wherein the actual fault condition found (FF) result data indicates that a particular BIT fault code data sequence generated by the BIT corresponded to an actual fault condition found (FF) result.

19. A non-transitory computer-readable storage medium with an executable program stored thereon for determining whether a built-in-test fault code (BITFC) data sequence generated by a built-in-test (BIT) of a particular module of a complex system is indicative of an actual fault condition, wherein the program instructs a processor to perform the following steps:

loading a regression function for the particular module at the processor, wherein the regression function was developed based on stored built-in-test fault code (BITFC) data sequences generated by the BIT for the particular module and stored repair data for that module, wherein the regression function correlates some particular ones of the BITFC data sequences with a finding of an actual fault condition found (FF) result at the particular module, and correlates other particular ones of the BITFC data sequences with a finding of no fault condition found (NFF) result at the particular module; and

executing the regression function to a new BITFC data sequence generated by the BIT to determine whether the new BITFC data sequence is indicative of an actual fault condition at the particular module or is indicative of a false fault condition at the particular module.