AU775127B2

AU775127B2 - Method and system for analyzing fault log data for diagnostics and repairs of locomotives

Info

Publication number: AU775127B2
Application number: AU40603/00A
Authority: AU
Inventors: David Richard Gibson; Nicholas Edward Roddy; Anil Varma
Original assignee: General Electric Co
Current assignee: General Electric Co
Priority date: 1999-04-02
Filing date: 2000-03-31
Publication date: 2004-07-15
Anticipated expiration: 2020-03-31
Also published as: CA2388489A1; CA2388489C; MXPA01009848A; AU4060300A; WO2000060465A8; CA2331217A1; WO2000060465A1

Description

WO 00/60465 PCT/US00/08662 METHOD AND SYSTEM FOR ANALYZING FAULT LOG DATA FOR DIAGNOSTICS AND REPAIRS OF LOCOMOTIVES BACKGROUND OF THE INVENTION The present invention relates generally to diagnostics of locomotive or other large land-based, self-powered transport equipment, and, more specifically, to a system and method for processing historical repair data and fault log data to facilitate analysis of malfunctioning machine equipment.

A machine, such as a locomotive or an off-road haul vehicle, includes elaborate controls and sensors that generate faults when anomalous operating conditions of the machine are encountered. Typically, a field engineer will look at a fault log and determine whether a repair is necessary.

Approaches like neural networks, decision trees, etc., have been employed to learn over input data to provide prediction, classification, and function approximation capabilities in the context of diagnostics.

Often, such approaches have required structured and relatively static and complete input data sets for learning, and have produced models that resist real-world interpretation.

Another approach, Case Based Reasoning (CBR), is based on the observation that experiential knowledge (memory of past experiences or cases) is applicable to problem solving as learning rules or behaviors. CBR relies on relatively little pre-processing of raw knowledge, focusing instead on indexing, retrieval, reuse, and archival of cases. In the diagnostic context, a case refers to a problem/solution description pair that represents a diagnosis of a problem and an appropriate repair.

CBR assumes cases described by a fixed, known number of des.crintive attrihbtes. Conventional CBR systems assume a corpus of fully

I:

P \WPDOCS\LMM\wy Sma pSpwi. ir-mionsV doc-12/05/04 -2valid or "gold standard" cases that new incoming cases can be matched against.

U.S. Patent No. 5,463,768 discloses an approach which uses error log data and assumes predefined cases with each case associating an input error log to a verified, unique diagnosis of a problem. In particular, a plurality of historical error logs are grouped into case sets of common malfunctions. From the group of case sets, common patterns, i.e., consecutive rows or strings of data, are labelled as a block. Blocks are used to characterise fault contribution for new error logs that are received in a diagnostic unit.

For a continuous fault code stream where any or all possible fault codes may occur from zero to any finite number of times and where the fault codes may occur in any order, predefining the structure of a case is nearly impossible.

Therefore, there is a need for a system and method for processing historical repair data and fault log data, which is not restricted to sequential occurrences of fault log entries and which provides weighted repair and distinct fault cluster combinations, to facilitate analysis of new fault log data from a malfunctioning machine.

According to one aspect of the present invention there is provided a method for analysing fault log data from a machine undergoing diagnostics, comprising: receiving new fault log data (232) comprising a plurality of faults from the machine; selecting a plurality of distinct faults (233) from the new fault log data; generating at least one distinct fault cluster (236) from the plurality of distinct faults; generating a plurality of weighted repair and distinct fault cluster combinations; and identifying at least one repair (238) for the at least one distinct fault cluster using the plurality of weighted repair and distinct fault cluster combinations.

According to another aspect of the present invention there is provided a system. for analysing fault log data from a machine undergoing diagnostics, comprising: a directed weight data storage unit (26) adapted to store a plurality of weighted repair and distinct fault cluster combinations; a processor (12) adapted to receive new fault log data (232) comprising a plurality of faults from the malfunctioning equipment; a processor (12) for selecting a plurality of distinct faults (233) from the new fault log data; a processor (12) for 9**9 generating at least one distinct fault cluster (236) from the selected plurality of distinct faults; a processor (12) for generating a plurality of weighted repair and distinct fault cluster combinations; and a processor (12) for identifying at least one repair (238) for the at least one distinct fault cluster using the plurality of predetermined weighted repair and o distinct fault cluster combinations.

According to yet another aspect of the present invention there is provided a method 9 35 for processing repair data comprising a plurality of repairs and fault log data comprising a plurality of faults from one or more machines to facilitate analysis of a malfunctioning machine, comprising: generating a plurality of cases (52, 54, 56, 60) from the repair data and the fault log data, each case comprising a repair and a plurality of distinct faults; ?.%WPDOCS\LM BTwy Smewo SpmifiW.6=s743990 gwp~. .21/04 -2Agenerating, for each of the plurality of cases, at least one repair and distinct fault cluster combination and assigning, to each of the repair and distinct fault cluster 0 0 0*00 0 0 0000 0 0 0 0000 0000 0 0 0090 0000 0000 0 0 0 0 0 000 0 0 0 0000 0* 00 P.%WPDOCSc~b'q-54399G d.-30 Oo-b 2001 -3combinations, a weight (100), whereby weighted repair and distinct fault cluster combinations facilitate prediction of at least one repair for the malfunctioning machine.

According to yet another aspect of the present invention there is provided a system for processing repair data comprising a plurality of repairs and fault log data comprising a plurality of faults from one or more machines to facilitate analysis of a malfunctioning machine, comprising: a repair log data storage unit (20) adapted to store a plurality of repairs; a fault log data storage unit (22) adapted to store a plurality of faults; a processor adapted to generate a plurality of cases (52, 54, 56, 60) from the repair data and the fault log data, each case comprising a repair and a plurality of distinct faults; the processor adapted to generate, for each of the plurality of cases, at least one repair and distinct fault cluster combination and the processor adapted to assign, to each of the repair and distinct fault cluster combinations, a weight (100) determined by dividing (108) i: il a number of times the combination occurs in cases comprising related repairs by a number of times the combination occurs in the plurality of cases, whereby weighted repair and distinct fault cluster combinations facilitate prediction of at least one repair for the malfunctioning machine.

According to yet another aspect of the present invention there is provided an article of manufacture comprising: a computer program product comprising computer usable medium having computer readable program code means embodied therein for 20 causing the processing of repair data comprising a plurality of repairs and fault log data comprising a plurality of faults from one or more machines to facilitate analysis of a malfunctioning machine, the computer readable program code means in said article of manufacture comprising: computer readable program code means for causing a computer to generate a plurality of cases (52, 54, 56, 60) from the repair data and the fault log data, 25 each case comprising a repair and a plurality of distinct faults; computer readable program code means for causing a computer to generate, for each of the plurality of cases, at least S one repair and distinct fault cluster combination and computer readable program code means for causing a computer to assign, to each of the repair and distinct fault cluster combinations, a weight (100) determined by dividing (108) a number of times the combination occurs in cases comprising related repairs by a number of times the combination occurs in the plurality of cases, whereby weighted repair and distinct fault cluster combinations facilitate prediction of at least one repair for the malfunctioning machine.

According to yet another aspect of the present invention there is provided a method for analysing fault log data from a malfunctioning machine, comprising: receiving new fault log data comprising a plurality of faults from the malfunctioning machine; selecting a plurality of distinct faults from the new fault log data; generating at least one distinct fault cluster from the plurality of distinct faults; and predicting at least one repair for the at least one distinct fault cluster using a plurality of predetermined weighted repair and distinct fault cluster combinations.

P:.WPDOCSc:i pmA5 4 39J1 doct OObar 2001 -3A- According to yet another aspect of the present invention there is provided a system for analysing fault log data from a malfunctioning machine, comprising: a directed weight data storage unit adapted to store a plurality of weighted repair and distinct fault cluster combinations; a processor adapted to receive new fault log data comprising a plurality of faults from the malfunctioning machine; the processor adapted to select a plurality of distinct faults from the new fault log data; the processor adapted to generate at least one distinct fault cluster from the selected plurality of distinct faults; and the processor adapted to predict at least one repair for the at least one distinct fault cluster using the plurality of predetermined weighted repair and distinct fault cluster combinations.

According to yet another aspect of the present invention there is provided an article of manufacture comprising: a computer program product comprising computer usable medium having computer readable program code means embodied therein for causing the analysis of fault log data from a malfunctioning machine, the computer readable program code means in said article of manufacture comprising: computer readable program code means for causing a computer to receive new fault log data Scomprising a plurality of faults from the malfunctioning machine; computer readable program code means for causing a computer to identify a plurality of distinct faults from the new fault log data; computer readable program code means for causing a computer to 20 generate at least one distinct fault cluster from the identified plurality of distinct faults; and computer readable program code means for causing a computer to predict at least one repair for the at least one distinct fault cluster using a plurality of predetermined weighted repair and distinct fault cluster combinations.

Preferred embodiments of the invention will be hereinbefore described with 25 reference to the accompanying drawings, and in those drawings: FIG. 1 is one embodiment of a block diagram of a system of the present invention for automatically processing repair data and fault log data from a locomotive or other large land-based, self powered transport equipment and diagnosing malfunctioning equipment; FIG. 2 is an illustration of a portion of repair log data; FIG. 3 is an illustration of a portion of fault log data; FIG. 4 is a flowchart describing the steps for generating a plurality of cases, and repair and fault cluster combustions for each case; FIG. 5 is an illustration of a case generated according to the flowchart of FIG. 4; Wn flnlrft~tP 10 OZZ11I~\ LL WA fI-ftA~ I LUUIUOU FIGS. 6A-6C are illustrations of repair and fault cluster combinations for the case shown in FIG. FIG. 7 is a flowchart describing the steps for determining a weight for each of the repair and fault cluster combinations; FIGS. 8A-8C are illustrations of partial lists of single, double, and triple fault clusters for related repairs; FIGS. 9A-9C are illustrations of partial lists of single, double, and triple fault clusters for related and unrelated repairs; FIGS. 10A-10C are illustrations of partial lists of weighted repair and distinct fault cluster combinations; FIG. 11 is a flowchart describing the steps for adding a new case to the case database and updating the weighted repair and distinct fault cluster combinations; FIG. 12 is an illustration of a portion of new fault log data from a malfunctioning machine; FIG. 13 is a flowchart describing the steps for analyzing the new fault log data from a malfunctioning machine and predicting one or more possible repair actions; FIG. 14 is an illustration of the distinct faults identified in the new fault log, shown in FIG. 12, and the number of occurrences thereof; FIGS. 15A-15D are illustrations of distinct fault clusters for the distinct faults identified in FIG. 14; FIG. 16 is a printout of weighted repair and fault cluster combinations by the system shown in FIG. 1 for the fault log shown in FIG.

12, and a listing of recommended repairs; FIG. 17 is a flowchart further describing the step of predicting repairs from the weighted repair and fault cluster combinations shown in FIG. 16; and

U.

WO /61 0465 Dl1"T/TQAn/AnQPL' d 5 I J |II J;-5- FIG. 18 is one embodiment of a flowchart describing the steps for automatically analyzing new fault log data from a malfunctioning machine and predicting one or more possible repair actions.

DETAILED DESCRIPTION OF THE INVENTION FIG. 1 diagrammatically illustrates one embodiment of a diagnostic system 10 of the present invention. In one aspect, system provides a process for automatically harvesting or mining repair data comprising a plurality of related and unrelated repairs and fault log data comprising a plurality of faults from a locomotive or other large landbased, self-powered transport equipment, and generating weighted repair and distinct fault cluster combinations which are diagnostically significant predictors to facilitate analysis of new fault log data from malfunctioning machine equipment.

Although the present invention is described with reference to a locomotive, system 10 can be used in conjunction with any machine in which operation of the machine is monitored, such as a chemical, an electronic, a mechanical, or a microprocessor machine.

Exemplary system 10 includes a processor 12 such as a computer UNIX workstation) having a hard drive, input devices such as a keyboard, a mouse, magnetic storage media tape cartridges or disks), optical storage media CD-ROMs), and output devices such as a display and a printer. Processor 12 is operably connected to and processes data contained in a repair data storage unit 20 and a fault log data storage unit 22.

Repair data storage unit 20 includes repair data or records regarding a plurality of related and unrelated repairs for one or more locomotives. FIG. 2 shows an exemplary portion 30 of the repair data contained in repair data storage unit 20. The repair data may include a customer identification number 32, a locomotive identification or unit pu& WO 00/60465 PCT/USOO/08662 -bnumber 33, the date 34 of the repair, the repair code 35, a repair code description 36, a description of the actual repair 37 performed, etc.

Fault log data storage unit 22 includes fault log data or records regarding a plurality of faults occurring prior to the repairs for the one or more locomotives. FIG. 3 shows an exemplary portion 40 of the fault log data contained in fault log data storage unit 22. The fault log data may include a customer identification number 42, a locomotive identification number or unit 44, the date 45 that the fault occurred, a fault code 46, a fault code description 48, etc. Additional information may include various sensor readings, temperature sensor readings, pressure sensor readings, electrical sensor readings, engine power readings, etc.

From the present invention, it will be appreciated by those skilled in the art that a repair data storage unit and a fault log data storage unit may contain repair data and fault log data for a plurality of different locomotives.

FIG. 4 is a flowchart of an exemplary process 50 of the present invention for selecting or extracting repair data from repair data storage unit 20 and fault log data from fault log data storage unit 22, and generating a plurality of diagnostic cases, which are stored in a case storage unit 24. As used herein, the term "case" comprises a repair and one or more distinct faults or fault codes. As also used herein, the term "distinct fault" is a fault or a fault code which differs from other faults or fault codes so that, as described in greater detail below, if the fault log data includes more than one occurrence of the same fault or fault code, the similarly occurring fault or fault code is identified only once.

With reference still to FIG. 4, process 50 comprises, at 52, selecting or extracting a repair from repair data storage unit 20 (FIG. 1).

Given the identification of a repair, the present invention searches fault log data storage unit 22 (FIG. 1) to select or extract, at 54, distinct faults -occurring ov a preuetrn e period of tc por to the repair.

UV.AU11111r UVG-. a jJIlU%'L t1tliUU JeIIUU U1 tU11C p11,1 LU Uk Ulk 1.

WO 00/60465 PCT/USOO/08662 The predetermined period of time may extend from a predetermined date prior to the repair to the date of the repair. Desirably, the period of time extends from prior to the repair, 14 days, to the date of the repair. It will be appreciated that other suitable time periods may be chosen. Desirably, the same period of time is chosen for generating all of the cases. As will be appreciated further by the discussion below, it is the selection of distinct faults which is important and not the order or sequence of their arrangement in the fault log data.

At 56, the number of times each distinct fault occurred during the predetermined period of time is determined. A repair and the one or more distinct faults are generated and stored as a case, at FIG. 5 illustrates a case 70 generated by process 50 (FIG. 4).

Exemplary case 70 comprises a file name 72 which lists, for example, a repair or repair code 2322 which corresponds to replacement of an electronic fuel injection pump, a customer identification number 74, a locomotive identification number or unit 76, a start date 78 and an end date over which faults are selected, a listing of the distinct fault or fault codes 82 which occurred between start date 78 and end date 80, and the number of times each distinct fault or fault code occurred 84.

In this exemplary case 70, fault code 7A5D indicates that the locomotive diesel engine failed to reach full operating power, fault code 7A4A indicates an air conditioner compressor failed to start, and fault code 76D5 indicates a fault reset. Case 70 may also list additional information 86 such as various average sensor readings, average temperature sensor readings, average pressure sensor readings, average electrical sensor readings, average engine power readings, etc., for distinct faults 82.

With reference again to FIG. 4, at 62, repair and distinct fault cluster combinations are generated. For exemplary case 70 (FIG. three repair code 2322 and single fault cluster repair code 2322 and fault code 7A5D, repair code 2322 and fault code 7A4A, and repair code WO 00 0465~ nP'Fn rPnn I~OLLI

V

-8- 2322 and fault code 76D5) combinations are illustrated in FIG. 6A. Three repair code 2322 and double fault clusters repair code 2322 and fault codes 7A5D and 7A4A, repair code 2322 and fault codes 7A5D and 76D5, and repair code 2322 and fault codes 7A4A and 76D5) combinations are illustrated in FIG. 6B, and one repair code 2322 and triple fault cluster repair code 2322 and fault codes 7A5D, 7A4A, and 76D5) combination is illustrated in FIG. 6C. From the present description, it will be appreciated by those skilled in the art that a case having a greater number of distinct faults would result in a greater number of repair and fault cluster combinations.

Process 50 is repeated by selecting another repair entry from the repair data to generate another case, and to generate a plurality of repair and fault cluster combinations. Case data storage unit 24 desirably stores a plurality of cases comprising related and unrelated repairs and a plurality of repair and distinct fault cluster combinations.

FIG. 7 is a flowchart of an exemplary process 100 of the present invention for generating weighted repair and fault cluster combinations based on the plurality of cases generated in process Process 100 comprises, at 101, selecting a repair and fault cluster combination, and determining, at 102, the number of times the combination occurs in cases comprising related repairs. The number of times the combination occurs in the plurality of cases of related and unrelated repairs, all repairs is determined at 104. A weight is determined at 108 for the repair and distinct fault cluster combination by dividing the number of times the combination occurs in cases comprising related repairs by the number of times the distinct fault cluster occurs in the plurality of cases comprising related and unrelated repairs.

o1 Wn In/fnAKC ~rmrrrrrrrn mnrr r lr• YL IUUUIU5a Process 100 is further described in greater detail with reference to FIGS. 8-10 and exemplary case data storage unit 24 (FIG. 1) which contains, for example, 500 cases covering 60 different repairs in which 38 cases relate to repair code 2322.

FIG. 8A is a portion 110 of the plurality single fault clusters generated from the 38 cases in case data storage unit 24 related to repair code 2322. As shown in FIG. 8A, repair code 2322 and single fault cluster or fault code 76D5 combination occurred 24 times, repair code 2322 and single fault cluster or fault code 7A5D combination occurred 23 times, and repair code 2322 and single fault cluster or fault code 7A4A combination occurred once. Also observed are other fault codes which occurred in other cases involving repair code 2322.

FIG. 8B illustrates a portion 112 of the plurality of double fault clusters generated from cases in case storage unit 24 related to repair code 2322. As shown FIG. 8B repair code 2322 and double fault cluster or fault codes 7A5D and 76D5 combination occurred 20 times, repair code 2322 and double fault cluster or fault codes 7A5D and 7A4A combination occurred once, and repair code 2322 and double fault cluster or fault codes 7A4A and 76D5 occurred once. FIG. 8C illustrates a portion 114 of the plurality of triple fault clusters generated from cases in case storage unit 24 related to repair code 2322 in which repair code 2322 and triple fault cluster or fault 7A5D, 7A4A, and 76D5 combination occurred once.

For cases involving repair code 2322 and having distinct faults numbering greater than three additional fault clusters of four, n, (not shown) are generated.

FIG. 9A-9C are portions 120, 122, and 124, of the single, double, and triple fault clusters, respectively, generated from all of the plurality of cases related and unrelated repairs) in case storage unit 24.

As shown in FIG. 9A, single fault cluster or fault code 76D5 occurred in 24 out of all the cases, single fault cluster or fault code 7A5D occurred in O1.

Wn nn/Afl;4 1K)4" Rti TP~ AOALL V V rL, u)uuu3VV 84 out of all the plurality of cases, and single fault cluster or fault code 7A4A occurred in 4 out of all of the plurality of cases.

As shown in FIG. 9B double fault cluster or fault codes and 76D5 occurred in 73 out of all of the plurality of cases, double fault cluster or fault codes 7A5D and 7A4A occurred once out of all of the plurality of cases, and double fault cluster or fault codes 7A4A and 76D5 occurred once out of all of the plurality of cases.

As shown in FIG. 9C, triple fault cluster or fault codes 7A4A, and 76D5 occurred once out of all of the plurality of cases.

For cases having distinct faults numbering greater than three additional fault clusters of four, five, n, (not shown) are generated.

Weighted repair and fault cluster combinations are determined and stored in a directed weight data storage unit 26. Partial listings of the weighted repair and fault cluster combinations are best shown in FIGS. 10A-10C.

For example, FIG. 10A illustrates a portion 130 of the weighted repair and single distinct fault cluster combinations. As shown in FIG. 10A, repair code 2322 and single fault cluster or fault code combination has a weight of 0.27 or 27% 23/84), repair code 2322 and single fault cluster or fault code 76D5 and has a weight of 0.09 or 9% 24/268), repair code 2322 and single fault cluster or fault code 7A4A has a weight of 0.25 or 25% 1/4).

FIG. 10B illustrates a portion 140 of the weighted repair and double distinct fault cluster combinations. As shown in FIG. 10B, repair code 2322 and double fault cluster or fault codes 7A5D and 76D5 combination has a weight of 0.27 or 27% 20/73), repair code 2322 and double fault cluster or fault codes 7A5D and 7A4A has a weight of UI UV/O I UDu 1ai I.,u kiu UU.UI! IdIL codIe;I Us IdUL or codes 7A4Ae.g., and 76D5rep has a weight 2322 a0.25d double fa(e.g.,ult clus1/4)ter or fault codes 7A4A and 76D5 has a weight of 0.25 or 25% 1/4).

U.L

Wf l lA5 nl"'TT'A nL~ I -11- Uuuo FIG. 10C illustrates a portion 145 of the weighted repair and triple distinct fault cluster combination. As shown in FIG. 10 OC, repair code 2322 and triple fault cluster or fault codes 7A5D, 7A4A, and 76D5 has a weight of 1.0 or 100% 1/1).

For cases having distinct faults numbering greater than three additional directed weights for fault clusters of four, five, n, (not shown) are generated.

Once the weighted repair and distinct fault cluster combinations are determined, they can be used to analyze a malfunctioning machine in a number of ways. For example, distinct fault clusters can be generated from new fault log data from a malfunctioning locomotive and readily compared, manually or automatically, to weighted repair and distinct fault cluster combinations for prediction of one or more repairs as described in greater detail below.

As shown in FIG. 11, a process 150 of the present invention provides updating directed weight data storage unit 26 to include one or more new cases. For example, a new repair and fault log data 25 (FIG. 1) from a malfunctioning locomotive is received at 152. At 154, a plurality of repair and distinct fault cluster combinations for the plurality of the distinct fault is generated.

The number of times each fault cluster occurred for related repairs is updated at 155; and the number of times each fault cluster occurred for all repairs are updated at 156. Thereafter, the weighted repair and distinct fault cluster combinations are redetermined at 158.

As noted above, the system provides prediction of repairs from fault log data from a malfunctioning machine. Desirably, after verification of the repair(s) for correcting a malfunction the new case can be inputted and updated into the system.

O1 WO 00/60465 -12- PCT/US00/08662 From the present invention, it will be appreciated by those skilled in the art that the repair and fault cluster combinations may be generated and stored in memory when generating the weights therefor, or alternatively, be stored in either the case data storage unit, directed weight storage unit, or a separate data storage unit.

Processor 12 is also operable to receive new fault log data 200 for analysis thereof. FIG. 2 shows in greater detail an exemplary portion 220 of the new fault log data 200 which may include a customer identification number 222, a locomotive identification number or unit 224, the dates 225 the faults occurred, the fault codes 226, and a fault code description 228. Additional information may also include various sensor readings, temperature sensor readings, pressure sensor readings, electrical sensor readings, engine power readings, etc. Desirably, the new fault log data includes faults occurring over a predetermined period of time prior such as a predetermined number of days 14 days). It will be appreciated that other suitable time periods may be chosen.

Advantageously, as explained below, the period of time corresponds to the period of time used for predetermining weighted repair and distinct fault cluster combinations.

FIG. 13 is a flowchart which generally describes the steps for analyzing new fault log data 200 (FIG. As shown in FIG. 13 at 232, the new fault log data comprising a plurality of faults from a malfunctioning machine is received. At 233, a plurality of distinct faults from the new fault log data is identified, and at 234, the number of times each distinct fault occurred in the new fault log data is determined. As used herein, the term "distinct fault" is a fault or a fault code which differs from other faults or fault codes so that, as described in greater detail below, if a portion of the fault log data includes more than one occurrence of the same IUlIit or IUlLt cUUo, Ui s 11111iar i ult U1 Idorfault odesU are uiuI;uL only Uonce.

As will become apparent from the discussion below, it is the selection of WO 00/60465 Pr-T/U 0cnn/a6 -13- A V V distinct faults which is important and not the order or sequence of their arrangement in the fault log data.

FIG. 14 shows the plurality of distinct faults and the number of times in which each distinct fault occurred for fault log 220 (FIG. 12). In this example, fault code 7311 represents a phase module malfunction which occurred 24 times, fault code 728F indicates an inverter propulsion malfunction which occurred twice, fault code 76D5 indicates a fault reset which occurred once, and fault code 720F indicates an inverter propulsion malfunction which occurred once.

With reference again to FIG. 13, a plurality of fault clusters is generated for the distinct faults at 236. FIGS. 15A-15D illustrate the distinct fault clusters generated from the distinct faults extracted from fault log data 200. Four single fault clusters fault code 7311, fault code 728F, fault code 76D5, and fault code 720F) are illustrated in FIG. Six double fault clusters fault codes 76D5 and 7311, fault codes 76D5 and 728F, fault codes 76D5 and 720F, fault codes 7311 and 728F, fault codes 7311 and 720F, and fault codes 728F and 720F) are illustrated in FIG. 15B. Four triple fault clusters fault codes 76D5, 7311, and 728F), fault codes 76D5, 7311, and 720F, fault codes 76D5, 728F, and 720F, and fault codes 7311, 728F, and 720F) are illustrated in FIG. and one quadruple fault cluster 76D5, 7311, 728F, and 720F) is illustrated in FIG. From the present description, it will be appreciated by those skilled in the art that a fault log having a greater number of distinct faults would result in a greater number of distinct fault clusters ones, twos, threes, fours, fives, etc.).

At 238, at least one repair is predicted for the plurality of fault clusters using a plurality of predetermined weighted repair and fault cluster ombinations. The lurality of prdeteUrmined weighI repair and fault cluster combinations may be generated as follows.

Idr wn nAnAALC ~rm nrnnn r~nr,~ r 1 4 r LI u 3uuI u ac With reference again to FIG. 1, processor 12 is desirably operable to process historical repair data contained in a repair data storage unit 20 and historical fault log data contained in a fault log data storage unit 22 regarding one or more locomotives.

For example, repair data storage unit 20 includes repair data or records regarding a plurality of related and unrelated repairs for one or more locomotives. Fault log data storage unit 22 includes fault log data or records regarding a plurality of faults occurring for one or more locomotives.

FIG. 16 illustrates a printout 250 of the results generated by system 10 (FIG. 1) for fault log 200 (FIG. in which in a top portion 252, a plurality of corresponding repairs 253, assigned weights 254, and fault clusters 255 are presented. As shown in a bottom portion 260 of printout 250, five recommendations for likely repairs actions are presented for review by a field engineer.

FIG. 17 is a flowchart of an exemplary process 300 for determining and presenting the top most likely repair candidates which may include repairs derived from predetermined weighted repair and distinct fault cluster combinations having the greatest assigned weighted values or repairs which are determined by adding together the assigned weighted values for fault clusters for related repairs.

As shown in FIG. 17, initially, a distinct fault cluster generated from the new fault log data is selected at 302. At 304, predetermined repair(s) and assigned weight(s) corresponding to the distinct fault cluster are selected from directed weight storage unit 26 (FIG.

1).

1Oz WO 00/60465 Dr-TMIicTArAn/no A U At 306, if the assigned weight for the predetermined weighted repair and fault cluster combination is determined by a plurality of cases for related and unrelated repairs which number less than a predetermined number, 5, the fault cluster is excluded and the next distinct fault cluster is selected at 302. This prevents weighted repair and fault cluster combinations which are determined from only a few cases from having the same effect in the prediction of repairs as weighted repair and fault cluster combinations determined from many cases.

If the number of cases is greater than the predetermined minimum number of cases, at 308, a determination is made as to whether the assigned value is greater than a threshold value, 0.70 or 70%. If so, the repair is displayed at 310. If the fault cluster is not the last fault cluster to be analyzed at 322, the next distinct fault cluster is selected at 302 and the process is repeated.

If the assigned weight for the predetermined weighted repair and fault cluster combination is less than the predetermined threshold value, the assigned weights for related repairs are added together at 320.

Desirably, up to a maximum number of assigned weights, 5, are used and added together. After selecting and analyzing the distinct fault clusters generated from the new fault log data, the repairs having the highest added assigned weights for fault clusters for related repairs are displayed at 324.

With reference again to FIG. 16, repairs corresponding to the weighted repair and fault cluster combinations in which the assigned weights are greater than the threshold value are presented first. As shown in FIG. 16, repair codes 1766 and 1777 and distinct fault cluster combinations 7311, 728F, and 720F, have an assigned weight of 85% and indicate a recommended replacement of the EFI.

FUd WO 00 h0465 M-r 0Cn llI -16- As also shown in FIG. 16, repairs for various fault clusters having the highest added or total weight are presented next. For example, repair code 1677 which corresponds to a traction problem has a totaled assigned weight of 1.031, repair code 1745 which corresponds to a locomotive software problem has a totaled assigned weight of 0.943, and repair code 2323 which corresponds to an overheated engine has a totaled assigned weight of 0.591.

Advantageously, the top five most likely repair actions are determined and presented for review by a field engineer. For example, up to five repairs having the greatest assigned weights over the threshold value are presented. When there is less than five repairs which satisfy the threshold, the remainder of recommended repairs are presented based on a total assigned weight.

Desirably the new fault log data is initially compared to a prior fault log data from the malfunctioning locomotive. This allows determination whether there is a change in the fault log data over time. For example, if there is no change, no new faults, then it may not be necessary to process the new fault log data further.

FIG. 18 illustrates a flowchart of an exemplary automated process 500 for analyzing fault log data from a locomotive, new fault log data which is generated every day, using system 10. In particular, process 500 accommodates the situation where a prior repair is undertaken or a prior repair is recommended within the predetermined period of time over which the fault log data is analyzed. This avoids recommending the same repair which has been previously recommended and/or repaired.

At 502, new fault log data is received which includes faults occurring over a predetermined period of time, 14 days. The fault log data is analyzed, for example as described above, generating distinct fault l I iustersi u ipii p uLl gU erat 4L iault clus1te1lsi iU pltcuieineU weighted repair and fault cluster combinations.

pO.

n AnlhCi.C

T

S-17- I'PCI/USUUIUO86 At 504, the analysis process may use a thresholding process described above to determine whether any repairs are recommended having a weighted value over If no repairs are recommended, the process is ended at 506. The process is desirably repeated again with a download of new fault log data the next day.

If a repair recommendation is made, existing closed performed or completed repairs) or prior recommended repairs which have occurred within the predetermined period of time are determined at 508.

For example, existing closed or prior recommended repairs may be stored and retrieved from repair data storage unit 20. If there are no existing or recommended repairs than all the recommended repairs at 504 are listed in a repair list at 700.

If there are existing closed or prior recommended repairs, then at 600, any repairs not in the existing closed or prior recommended repairs are listed in the repair list at 700.

For repairs which are in the existing closed or prior recommended repairs, at 602, the look-back period the number of days over which the faults are chosen) is revised. Using the modified lookback or shortened period of time, the modified fault log data is analyzed at 604, as described above, using distinct fault clusters, and comparing the generated fault clusters to predetermined weighted repair and fault cluster combinations.

At 606, the analysis process may use the thresholding process described above to determine whether any repairs are recommended having a weighted value over If no repairs are recommended, the process is ended at 608 until the process is stated again with a new fault log data from the next day, or if a repair is recommended it is added to the repair list at 700.

>62 WO 0/60465 lr /T n/tI~nL 18- rk I u 1 U nU oO From the present description, it will be appreciated by those skilled in the art that other processes and methods, different thresholding values or fault log data analysis which does not use distinct fault clusters, may be employed in predicting repairs from the new fault log data according to process 500 which takes into account prior performed repairs or prior recommended repairs.

Thus, the present invention provides in one aspect a method and system for processing a new fault log which is not restricted to sequential occurrences of faults or error log entries. In another aspect, the calibration of the diagnostic significance of fault clusters is based upon cases of related repairs and cases for all the repairs.

Thus, the present invention provides in one aspect a method and system for automatically harvesting potentially valid diagnostic cases by interleaving repair and fault log data which is not restricted to sequential occurrences of faults or error log entries. In another aspect, standard diagnostic fault clusters can be generated in advance so they can be identified across all cases and their relative occurrence tracked. In still another aspect, the calibration of the significance of repair and distinct fault cluster combinations based upon cases of related repairs and cases for all the repairs is determined.

In addition, when initially setting up case data storage unit 24, a field engineer may review each of the plurality of cases to determine whether the distinct faults, and in particular, number of times the distinct faults occur, provide a good indication of the repair. If not, one or more cases can be excluded or removed from case data storage unit 24. This review by a field engineer would increase the initial accuracy of the system in assigning weights to the repair and fault cluster combinations.

oi PAWPDOCS1sRtpecciX>543M-)3oci Ooba 2001 -19- While the invention has been described with reference to preferred embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the scope of the invention. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the invention without departing from the essential scope thereof. Therefore, it is intended that the invention not be limited to the particular embodiments disclosed herein, but that the invention will include all embodiments falling within the scope of the appended claims.

Throughout this specification and the claims which follow, unless the context requires otherwise, the word "comprise", and variations such as "comprises" or S"comprising" will be understood to imply the inclusion of a stated integer or step or group ;of integers or steps but not the exclusion of any other integer or step or group of integers or steps.

The reference numerals in the following claims do not in any way limit the scope of the respective claims.

The reference to any prior art in this specification is not, and should not be taken as, an acknowledgment or any form of suggestion that that prior art forms part of the common general knowledge in Australia.

Claims

1. A method for analysing fault log data from a machine undergoing diagnostics, comprising: receiving new fault log data (232) comprising a plurality of faults from the machine; selecting a plurality of distinct faults (233) from the new fault log data; generating at least one distinct fault cluster (236) from the plurality of distinct faults; generating a plurality of weighted repair and distinct fault cluster combinations; and identifying at least one repair (238) for the at least one distinct fault cluster using the plurality of weighted repair and distinct fault cluster combinations.

2. The method of claim 1 wherein the at least one distinct fault cluster (236) comprises at least one of a single distinct fault and a plurality of distinct faults.

3. The method of claim 1 wherein each of the plurality of weighted repair and distinct fault cluster combinations are generated from a plurality of cases each case S: 20 comprising a repair and at least one distinct fault, and each of the plurality of weighted repair and distinct fault cluster combinations being assigned a weight determined by dividing (100) the number of times the combination occurs in cases comprising related repairs by the total number of times the combination occurs in said plurality of cases. 0*40 r

4. The method of claim 3 wherein identifying the at least one repair comprises *selecting at least one repair using the plurality of weighted repair and fault cluster S. combinations and adding assigned weights for distinct fault clusters for related repairs. The method of claim 1 said generating a plurality of weighted repair and distinct S 30 fault cluster combinations (50, 100) utilises a plurality of repairs and fault log data comprising a plurality of faults. comprising a plurality of faults.

P \WPDOC b~speci?5439W0doc-0 Otoba 2001 -21-

6. The method of claim 1 wherein the receiving fault log data (232) comprises receiving a new fault log data and comparing the new fault log data to a prior fault log data.

7. A system, for analysing fault log data from a machine undergoing diagnostics, comprising: a directed weight data storage unit (26) adapted to store a plurality of weighted repair and distinct fault cluster combinations; a processor (12) adapted to receive new fault log data (232) comprising a plurality 10 of faults from the malfunctioning equipment; a processor (12) for selecting a plurality of distinct faults (233) from the new fault log data; a processor (12) for generating at least one distinct fault cluster (236) from the •selected plurality of distinct faults; 15 a processor (12) for generating a plurality of weighted repair and distinct fault cluster combinations; and a processor (12) for identifying at least one repair (238) for the at least one distinct •coS fault cluster using the plurality of predetermined weighted repair and distinct fault cluster combinations.

S8. The system of claim 7 wherein a single processor unit constitutes said processors.

9. A system of claim 7 further comprising: a processor for generating a plurality of cases (52, 54, 56, 60) from the repair data and the fault log data, each case comprising a repair and a plurality of distinct faults; a processor for generating, for each of the plurality of cases, at least one repair and distinct fault cluster combination and a processor for assigning, to each of the repair and distinct fault cluster combinations, a weight (100), whereby weighted repair and distinct fault cluster combinations facilitate identification of at least one repair for the malfunctioning equipment.

P \WPDOCSbpI\tccA54399O doc-30 Otobar 2001 -22- The system of claim 9 wherein the processor for generating the plurality of cases (52, 54, 56, 60) comprises a processor for selecting a repair from the repair data and selecting a plurality of distinct faults (54) from the fault log data over a period of time prior to the repair.

11. The system of claim 9 wherein the processor for assigning weights (100) comprises a processor for determining (102), for each repair and distinct fault cluster combination, a number of times the combination occurs in cases comprising related repairs, and a number of times the combination occurs in the plurality of cases.

12. The system of claim 11 wherein the processor for assigning a weight (100), for each repair and distinct fault cluster combination, comprises a processor for dividing (108) the number of times the combination occurs in cases comprising related repairs by the number of times the combination occurs in the plurality of cases. o

13. The system of claim 11 further comprising; a processor for generating a new case (152) from repair data and fault log data, the case comprising a repair and a plurality of distinct faults; a processor for generating, for the new case, a plurality of fault clusters (154) for the plurality of distinct faults; and a processor for predetermining a weight (155, 156, 158) for each of the plurality of repair and fault cluster combinations to include the new case.

14. The system of claim 11 further comprising: a repair log data storage unit (20) adapted to store a plurality of repairs; and a fault log data storage unit (22) adapted to store a plurality of faults.

A method for processing repair data comprising a plurality of repairs and fault log data comprising a plurality of faults from one or more machines to facilitate analysis of a malfunctioning machine, comprising: PAWPDOCS\ombc7543v9o doc-3O Oaoba 2001 23 generating a plurality of cases (52, 54, 56, 60) from the repair data and the fault log data, each case comprising a repair and a plurality of distinct faults; generating, for each of the plurality of cases, at least one repair and distinct fault cluster combination and assigning, to each of the repair and distinct fault cluster combinations, a weight (100), whereby weighted repair and distinct fault cluster combinations facilitate prediction of at least one repair for the malfunctioning machine.

16. The method of claim 15 wherein the at least one repair and distinct fault cluster combination (62) comprises a repair and at least one of a single distinct fault and a plurality of distinct faults.

17. The method of claim 15 wherein the at least one repair and distinct fault cluster combination (62) comprises a repair and a plurality of distinct faults.

18. The method of claim 15 wherein the generating the plurality of cases (52, 54, 56, 4e044* 60) comprises selecting a repair from the repair data and selecting a plurality of distinct faults (54) from the fault log data over a period of time prior to the repair.

19. The method of claim 18 wherein the generating the plurality of cases (52, 54, 56, a. *60) comprises determining (56) the number of times each of the plurality of distinct faults 4 occurs over the period of time.

The method of claim 15 wherein the assigning weights (100) comprises determining (102), for each repair and distinct fault cluster combination, a number of times the combination occurs in cases comprising related repairs, and a number of times the combination occurs in the plurality of cases.

21. The method of claim 20 wherein the assigning a weight (100), for each repair and distinct fault cluster combination, comprises dividing (108) the number of times the combination occurs in cases comprising related repairs by the number of times the combination occurs in the plurality of cases. P:\WPDOCMb\spi7743990 doc-O Octobcr 2 J 24

22. The method of claim 15 further comprising; generating a new case (152) from repair data and fault log data, the case comprising a repair and a plurality of distinct faults; generating, for the new case, a plurality of fault clusters (154) for the plurality of distinct faults; and redetermining a weight (155, 156, 158) for each of the plurality of repair and fault clusters combinations to include the new case.

23. A system for processing repair data comprising a plurality of repairs and fault log data comprising a plurality of faults from one or more machines to facilitate analysis of a malfunctioning machine, comprising: a repair log data storage unit (20) adapted to store a plurality of repairs; a fault log data storage unit (22) adapted to store a plurality of faults; S 15 a processor adapted to generate a plurality of cases (52, 54, 56, 60) from the repair data and the fault log data, each case comprising a repair and a plurality of distinct faults; the processor adapted to generate, for each of the plurality of cases, at least one .oo.oi Srepair and distinct fault cluster combination and o* the processor adapted to assign, to each of the repair and distinct fault cluster combinations, a weight (100) determined by dividing (108) a number of times the combination occurs in cases comprising related repairs by a number of times the combination occurs in the plurality of cases, whereby weighted repair and distinct fault cluster combinations facilitate prediction of at least one repair for the malfunctioning machine.

24. An article of manufacture comprising: a computer program product comprising computer usable medium having computer readable program code means embodied therein for causing the processing of repair data comprising a plurality of repairs and fault log data comprising a plurality of faults from one or more machines to facilitate analysis of a malfunctioning machine, the computer readable program code means in said article of manufacture comprising: P.XWPDOMCSknbX7543*)90 d-10o- OO0br. 2001 computer readable program code means for causing a computer to generate a plurality of cases (52, 54, 56, 60) from the repair data and the fault log data, each case comprising a repair and a plurality of distinct faults; computer readable program code means for causing a computer to generate, for each of the plurality of cases, at least one repair and distinct fault cluster combination (62); and computer readable program code means for causing a computer to assign, to each of the repair and distinct fault cluster combinations, a weight (100) determined by dividing (108) a number of times the combination occurs in cases comprising related repairs by a number of times the combination occurs in the plurality of cases, whereby weighted repair and distinct fault cluster combinations facilitate prediction of at least one repair for the i. i malfunctioning machine.

A method for analysing fault log data from a malfunctioning machine, comprising: S 15 receiving new fault log data comprising a plurality of faults from the malfunctioning machine; selecting a plurality of distinct faults from the new fault log data; generating at least one distinct fault cluster from the plurality of distinct faults; and •opredicting at least one repair for the at least one distinct fault cluster using a plurality of predetermined weighted repair and distinct fault cluster combinations. .ooo.i

26. The method of claim 25 wherein the at least one distinct fault cluster comprises at least one of a single distinct fault and a plurality of distinct faults.

27. The method of claim 25 wherein the at least one distinct fault cluster comprises a plurality of distinct faults.

28. The method of claim 25 wherein each of the plurality of predetermined weighted repair and distinct fault cluster combinations are generated from a plurality of cases, each case comprising a repair and plurality of distinct faults, and each of the plurality of predetermined weighted repair and distinct fault cluster combinations being assigned a weight determined by dividing the number of times the combination occurs in cases I P.\WPDOCS IB\Thy Sn oa\SpiiCdras'7543990 spwn don.0 May. 2004 -26- comprising related repairs by the number of times the combination occurs in the plurality of cases.

29. The method of claim 28 further comprising excluding at least one predetermined weighted repair and distinct fault cluster combination in which the assigned weight is determined by related cases which number less than a predetermined number.

The method of claim 28 wherein predicting the at least one repair comprises selecting at least one repair using a predetermined weighted repair and distinct fault cluster combination having an assigned weight which is greater than a predetermined threshold value.

31. The method of claim 28 wherein predicting the at least one repair comprises selecting at least one repair using the plurality of predetermined weighted repair and fault cluster combinations and adding assigned weights for distinct fault clusters for related repairs.

32. The method of claim 25 further comprising generating a plurality of predetermined weighted repair and distinct fault cluster combinations from repair data comprising a plurality of repairs and fault log data comprising a plurality of faults.

33. The method of claim 32 wherein the generating a plurality of predetermined weighted repair and distinct fault cluster combinations comprises; generating a plurality of cases from repair data and the fault log data, each case comprising a repair and a plurality of distinct faults, generating, for each of the plurality cases, at least one repair and distinct fault S rIlctpr r, nmhiinntnn" nnd assigning, to each of the repair and distinct fault cluster combinations, a weight, wherein the assigned weight comprises dividing the number of times the combination 30 occurs in cases comprising related repairs by the number of times the combination occurs in the plurality of cases. P:\WP Xb'mb*a r" 399 -doc-0 O o. 2001 -27-

34. The method of claim 25 wherein the receiving fault log data comprises receiving a new fault log data and comparing the new fault log data to a prior fault log data.

The method of claim 25 wherein the new fault log data comprises a plurality of faults occurring over a predetermined period of time from the malfunctioning machine, and further comprising: determining at least one of a completed repair and a prior recommended repair occurring during the predetermined period of time; selecting a portion of the new fault log data which occurs after the at least one of a completed repair and a prior predicted repair; selecting a plurality of distinct faults from the selected portion of the fault log data; ii igenerating at least one distinct fault cluster from the selected plurality of distinct faults from the selected portion of the fault log data; and predicting at least one repair for the at least one distinct fault cluster from the selected portion of the fault log data using a plurality of predetermined weighted repair and distinct fault cluster combinations.

36. A system for analysing fault log data from a malfunctioning machine, comprising: a directed weight data storage unit adapted to store a plurality of weighted repair 20 and distinct fault cluster combinations; a processor adapted to receive new fault log data comprising a plurality of faults from the malfunctioning machine; the processor adapted to select a plurality of distinct faults from the new fault log data; the processor adapted to generate at least one distinct fault cluster from the selected plurality of distinct faults; and the processor adapted to predict at least one repair for the at least one distinct fault cluster using the plurality of predetermined weighted repair and distinct fault cluster combinations.

37. An article of manufacture comprising: P WPDOCS\cabkp c754399W O doc-30 Ocobr 2001 -28- a computer program product comprising computer usable medium having computer readable program code means embodied therein for causing the analysis of fault log data from a malfunctioning machine, the computer readable program code means in said article of manufacture comprising: computer readable program code means for causing a computer to receive new fault log data comprising a plurality of faults from the malfunctioning machine; computer readable program code means for causing a computer to identify a plurality of distinct faults from the new fault log data; computer readable program code means for causing a computer to generate at least one distinct fault cluster from the identified plurality of distinct faults; and computer readable program code means for causing a computer to predict at least S- one repair for the at least one distinct fault cluster using a plurality of predetermined weighted repair and distinct fault cluster combinations. 15

38. A method of analysing fault log data substantially as hereinbefore described with reference to the accompanying drawings. i

39. A system substantially as hereinbefore described with reference to the accompanying drawings. DATED this 26 h day of October 2001. GENERAL ELECTRIC COMPANY By Its Patent Attorneys DAVIES COLLISON CAVE