`
`A quoi peut Atre attribui le bruit de fond ?
`
`Tout dabord, aux conditions dans lesquelles sont faites les lectures par les pilotes
`
`-
`
`cadrans petits et peu prcis ;
`;
`I'instrumentation commune
`
`la lecture peut Ftre faite indifftremment par le pilote ou le co-piJote sur
`
`- conditions de vol difficiles i stabiliser pendant la piriode de 5 minutes recommandies par le constructeur
`
`-
`
`-
`
`-
`
`; en thorie les pilotes doivent faire
`charges ilectriques, hydrauliques et prilivements d'air mal identifiis
`les relevis darns des conditions toujours identiques, mais les impiratifs de vol ne le permettent pas toujours
`et parfois mime il ny a aucune consigne donnie au pilote de ce cit;
`
`les r6sultats sont tris sensibles A la tempirature extrieure de lair
`
`ensuite, les erreurs de transcription et de saisie des opirateurs au sol.
`
`11 est A noter qu'un bruit de fond important correspond, pour les PRATT & WHITNEY, A une dfectuosit6 du
`moteur, du type encrassement du compresseur.
`
`II semble qu'une bonne coopiration pilotes - responsables techniques aide a am~liorer le bruit de fond.
`
`Autres difficltis :
`
`tous Jes 6v~nements de maintenance
`Le constructeur recommande de marquer en paralile avec ]a courbe
`susceptibles d'intiresser la vie du moteur (lavages compresseur / turbine, changement dlquipements, riglages,
`jugement Ai sa
`etc...), pour que lopirateur chargd de linterpretation des courbes ait tous les 61iments de
`; on constate frequemment en pratique des oublis ou des retards importants.
`disposition
`
`L'exploitation au sol des relev6s du pilote ou de lenregistreur est faite priodiquement, giniralement une
`ou deux fois par semaine, et
`le constructeur est fr6quernment consulti pour lever les doutes
`; Iensemble
`100 heures de voJ entre )a
`jusqu'
`des dilais peut demander
`lecture et lexploitation dfinitive ; ceci
`suppose un materiel suffisamment robuste pour supporter une digradation rapide. C'est le cas pour le PT 6
`qui est un moteur s~r et bien connu
`; c'est moins ivident pour le CT 7 et surtout le PW 120.
`
`Pour que les seuils d'alerte prconiss par le constructeur aient une signification, if faut que ]a surveillance
`ait 6ti mise en place dis la sortie d'une rivision gnirale ou d'une visite de section chaude
`; aujourd'hui,
`ce r~flexe de
`faire syst6matiquement au moins une visite section chaude avant mise en place du Trend
`]a facilit6 avec laquelle les PT 6 sont vendus ou
`Monitoring nest pas entr6 dans les moeurs
`; de plus,
`transf&rs entre compagnies ne facilite pas la rigueur d'applcation des consignes du constructeur.
`
`5 - RESULTATS
`
`Ie Trend Monitoring fait ressortir tris rapidement les probJmes affectant Jes 6qui-
`D'une maniiire g~n~rale,
`pements piriph6riques du moteur (vannes, harnais, transmetteurs de vitesse et de couple, les thermocouples)
`laborieuse
`mais la recherche de panne qui permet d'identifier
`le composant responsable est souvent assez
`les sp~cialistes qui d~pouillent les courbes identifient plus facilement les pilotes de Iavion que Ie composant.
`
`introduction du Trend Monitoring dans une petite compagnie est lachat d'un
`La premire consquence de
`tr~s complet de bancs d'essais, appareillages, boroscopes, etc... permettant
`lot
`la vrification prcise des
`instruments de base, des injecteurs et du moteur. Ceci arnmliore consid6rablement Ie suivi g~n6ral des machines
`et c'est en soi un premier bon risultat.
`
`L'utilisation des cahiers de signatures de pannes donnies par le constructeur est tris difficile
`: linterprtation
`des courbes par la plupart des exploitants leur permet tout au plus de dire qu'il y a "quelque chose". La lecture
`des courbes 4tablies A la main (figure 2) est particuliirement d~licate
`les pentes sont faibles.
`;
`
`Deux exemples caractiristiques :
`
`Le moteur, objet de la figure 2, surveillA A titre expirimental par Trend Monitoring, a
`tL d6posi normalement
`i lissue de son potentiel diclari ;
`il prisentait tin
`tat micanique tr~s ben avec le compresseur encrasse.
`Le constructeur, consulti pour avis sur la mani~re de mettre en oeuvre le Trend Monitoring, a estimi que
`le moteur dtait en mauvais 6tat.
`
`Dans )a figure 7 ci-dessous, qui repr~sente la totaliti de la vie dun PW 120 (environ 2 200 heures depuis neuf),
`les paramitres de base semblent typiques dune ditirioration de section chaude et ont atteint les seuils d'alerte ;
`cependant, pour des raisons non connues, le moteur a 6tA
`laiss6 en service et a eu du pompage avec surchauffe
`violente au dicollage et destructions importantes. Une autre compagnie a eu exactement la mime allure de
`le moteur A temps
`courbe mais a su d6poser
`celui-ci prisentait des briures importantes en bout de pales
`de turbines.
`
`''A
`
`.
`
`I
`
`-
`
`iAR
`
`BOEING
`Ex. 1034, p. 243
`
`
`
`24-8
`
`III ~
`
`~
`
`~
`
`-
`
`--
`
`v
`
`tA.
`
`---------
`
`45-
`
`1
`
`Comparer avec la figure 6 qui reprisente on exemple recent de dc~tnrioration importante de chambres de combustion.
`
`Figure 7
`
`PT 6
`
`turbine
`ie Ceu en bout de pales de
`les derives semblent provoqu~es essentiellement par
`Sur cette machine,
`du g6n~rateur de gaz
`les
`dans 1'exp~rience que nous en avons et avec
`limites actuelles, ce
`leo provoque
`Papplication des consignes de depose avant qu'il y ait eu r~ellement des degradations importantes. Nous navons
`pas vu de contre exemple caract~ristique.
`PWt 120
`
`franiaises, 3 ont etil d~tect~s par Trend Monitoring
`Sur 12 cas dincidents survenus a des compagnies
`;mais
`les 4 inciden.s analysiis comme susceptibles d'6re d~tectes au Trend Monitoring, 3 'ont 6t effectivemnent
`sur
`le liime cas est celni cit6 ci-dessus.
`
`Dans 2 cas de moteurs depos~s suite an Trend Monitoring. un pr~sentait des dommages
`I pale distributeur HP),
`lantre des br~ilures
`(injecteur d~fectueux et l~g~re br~lure sur
`de pales sur les 2 premiers 6tages de turbine.
`
`reels mais mineurs
`importantes en bout
`
`faible en regard de l'importance des dtgats constates sur plusieurs moteurs
`;il
`La sensibilit6 semble tres
`l'6solution du rapport des vitesses NH
`l'inter pre tat ion des courbes one notion qui est
`ya, d'autre part, dans
`par rapport any sitesses NL. Cette notion nest pas apprr~hendee de
`la m~me mani~re par
`tons Ins utilisateurs
`et nest pas facile ii exploiter.
`
`CT 7
`
`I scume pale a et6 d6tect6 au Trend
`impact de corps 6trangers ayant d~form6
`Nous axons vu un cas on un
`figure 5) et un cas de depose suite a
`indications Trend Monitoting ayant effectis-ement une
`Monitoring (voir
`d~tirioration importante de la section chaude.
`
`La sensibilit6 semble suffixante.
`
`6
`
`-CONCLUSIONS
`
`if nkcessite one attention tr~s soutenne et nne connaissance
`Le "Trend Monitoring" existe, il ext utilisable mais
`thermodynamnique,
`rmccanique du moteur mais des principes de son fonictionnement
`fine, non seulement de
`Ia
`cxs courbes en I'absence de consignes type "toot on rien" sor
`'ensemrble
`xi
`'on veut
`interpreter correctement
`des paramitres ;en fait, if faut ur petit bureau d'6tudes. Dans one petite compagnie, cc n'est pas 6vident
`et le changement d'une seule personne 1,ent amener Ia perte quasi complete d'une longue experience.
`
`imp~rativement complti et recoupk par
`le Trend Monitoring doit 6tre
`fait de sex possibilit~s limit~es,
`On
`d'autres methodes de contr~les non basees sur
`les performances en vol, telles que mexures de performances
`fr~quentes, vixites diverses, etc... Ceci
`an sol par micaniciens,
`inspections endoscopiques systc~matiques et
`existe diji pour le PW 120 Ct le CT 7 et
`les inspections endoscopiques, en particulier, sont assez bien ma~tris~es
`par Ia plupart des compagnies. La p~riodicit6 de ces inspections ext moms bien maltris~e en raison de Ia faible
`experience gen~rale de cex machines et Ia tendance ext au resserrement.
`
`la section chande ext admissible, saul
`Ia butde priodique de vixite de
`Dans ces conditions, 1'61imination de
`pour
`le PT6. En effet, de par
`Ia conception et linstallation du moteur qni presente tres pen de points d'accis
`et qui sont masquis par de nombreux 6quipements,
`'inspection
`Ia section chande de ce moteur se pr~te mal ii
`endoscopique ;de ce fait, nous avons ete amenex a maintenir pour le PT6 une butee en heures pour [a revision
`de Ia section chaude, a titre de precaution, axec
`'espoir que I'exp~rience future de chaque exploitant permetrra
`de la rendre inutile.
`
`L'emploi d'un ordinateur et d'une
`cas.
`
`imprimante pour cxs traces ext
`
`indispensable, A notre avis, dans
`
`tons
`
`lex
`
`vY
`
`, i
`
`BOEING
`Ex. 1034, p. 244
`
`
`
`DISCUSSION
`
`24-9
`
`C. SPRUNG
`Quelles sont lea diff6rences, du point de vue performances, puissances
`et cofsoastions, entre lea tolerances do materiel neuf et celles
`acceptees en maintenance? Vous avez pr~cisL& que cette tol~rance eat
`
`de ± 2% per rapport A is valeur nominale pour du materiel neuf.
`Author's Reply:
`Les tol~rances indiqu6es Boat ceiles appliqoees par rappoit aux
`relev~s fait en vol par Ia o~thode du "trend monitoring". Elles sont
`completement indipendantes de celles utilis~es en maintenance, take
`si elles s'en rapprochent.
`
`BOEING
`Ex. 1034, p. 245
`
`
`
`25-1
`
`GAS PATH ANALYSIS AD NGINE PERFOREACK
`NONITOINO IN A CHINOOK HELICOPTER
`by
`D.E. Glenny
`Aeronautical Iemearch Laboratory
`Defence Science and Technology Organisation
`OPO BOX 4331
`Melbourne Victoria 3001
`Australia
`
`Periodic and consistent assessment of engine performance in military helicopters is
`essential if in-service operating margins are not be be eroded by harsh environmental
`conditions. Manually initiated GO-NO-GO pre-flight checks (HIT etc) or ad-hoc in-flight
`performance checks rarely provide sufficiently reliable data for maintenance or
`In contrast, performance assessment methods based on gas path
`diagnostic purposes.
`analysis principles and engine/aircraft data, automatically recorded during flight,
`ARL has investigated a number of these
`offer a potentially attractive alternative.
`alternatives, and has carried out an in-service trial on a Boeing CH47C Chinook
`helicopter operated by the RAAF. In the trial existing aircraft/engine Instrumentation
`was compiemented by specially designed probes located at module interfaces whilst the
`The performance-fault
`data were recorded on an ARL designed acquisition system.
`algorithms used in the analyses were configured for a range of engine operating
`Results for the trial are presented in terms of deviations, from pre-
`speeds.
`established base-line conditions. Statistical analyses using linear regression fits and
`Kalman Filtering techniques have been investigated to minimise the effects of data
`uncertainty. The applicability of the procedures, including thermodynamic analyses and
`equipment, are discussed in terms of fleetwide adoption for the Chinook.
`
`Nomenclature
`
`A
`BB
`C
`CVA
`FF,WF
`HIT
`IAS
`IGV
`IFM
`Ka b
`KAB
`N 1
`N2
`p
`P
`PPI
`T
`TEAC
`TOR
`WA
`Ir
`
`(o/i%)
`
`Area
`Bleed Band
`Corrected
`Coefficient of Variation
`Fuel Flow
`Health Indicatr Test
`Indicated Air Speed
`Inlet Guide Vanes
`In Flight Monitoring
`Influence Coefficient
`Fault Matrix Coefficient
`Gas Generator Speed
`Power Turbine Speed
`Static Pressure
`Total Pressure
`Power Performance Indicator
`Temperature
`Turbine Engine Analysis Check
`Torque
`Mass Flow
`Mean
`
`1. INTRODUCTION
`
`6 Pressure Ratio
`8 Temperature Ratio
`n Efficiency
`c Standard Deviation (STD)
`1--5 Station Numbers
`Subscripts
`Compressor
`Compressor Turbine
`Power Turbine
`
`c
`ct
`pt
`
`The arduous operating environment of the military helicopter requires that effentive
`power assurance checks are carried out routinely prior to, or during flight. Procedures
`currently used are configured around well established methods such as HIT, TEAC
`(topping) and PPI : the procedure used by an operator has usually been a function of the
`country of origin, or major user of the helicopter. The results of these checks are
`invariably treated as GO-NG-GO indicators and are rarely retained from one flight to the
`next for trending or prognostic purposes. In contrast to these methods the RAAF has, on
`its Iroquois helicopters, used an in flight performance assessment/monitoring method
`(IFM), Reference 1, which compares actual engine performance with prespecified baseline
`values. Deviations in torque and exhaust gas temperature are monitored routinely to
`give an indication of engine condition or performance deterioration. This procedure was
`implemented because of a long history of severe blade erosion and stability -
`surge
`problems experienced by the T53 engine, and the fact that the HIT method was carried out
`at relatively low power levels whilst daily use of topping checks was counter productive
`in terms of engine life. Notwithstanding the potential gains available by the adoption
`of an IFM procedure the viability of the method is limited by scatter due to visual
`observations of aircraft instrumentation and the difficulty in maintaining a steady
`state power setting during the monitoring period. Because only a limited number of gas
`path parameters were recorded during the IFM test it was also difficult to Isolate
`performance decrements to given engine components or modules.
`
`i
`
`BOEING
`Ex. 1034, p. 246
`
`
`
`25-2
`
`Recognising these problems a combined RAAF/ARL programme was initiated to investigate
`the effectiveness of using automatically recorded engine/aircraft data, combined with
`gas path analyses techniques to assess performance deterioration and component
`degradation in a military helicopter. At the request of the RAAF the investigation was
`centred on the Lycoming T55-Lll engine in the Chinook helicopter. This paper describes
`the Chinook engine performance monitoring programme, it covers development of fault
`algorithms, specification of instrumentation and data recorder, and finally data
`analysis of an in service trial.
`
`2. PERFOMAiNCE AIALYSIS
`
`Currently fault diagnosis and module assessment methods available to helicopter
`maintenance personnel are simple fault tree logics, ie:
`
`If
`
`A + B then C,
`
`or if
`
`A - B then D.
`
`More complex star charts have been proposed but have rarely been incorporated into
`Comprehensive fault tree libraries can be derived from engine
`maintenance manuals.
`This capability was
`thermodynamic models in which component faults can be implanted.
`not fully developed at ARL at the time this investigation was commenced, and therefore
`Gas path analyses such as those proposed by Staples and
`could not be used.
`Saravanamutto, Reference 2, were investigated but were rejected in favour of the more
`"apparently" definitive/analytic differential gas path analyses suggested by Urban,
`This latter method offered a potential to trend data measurands and
`Reference 3.
`predict component efficiencies, and hence to diagnose faults to a model or component
`level using analytical methods. The Urban technique of differential gas path analysis
`is based on relating changes in independent component performance parameters ( WA, n I
`Areas etc) to changes in dependent engine measurands (P,T) via a series of analytically
`derived influence coefficients. The technique has been used by both Shapiro, Reference
`4, and Cockshutt, Reference 5, in the analysis of compressible flow and gas turbine
`Urban has further developed the method by formulating diagnostic
`cycle studies.
`equations or fault matrices for given operating conditions. In its most simplistic form
`it Just relates changes in gas path components to changes in measured engine parameters
`whilst in its more developed algorithms attempts have been made to compensate for
`instrumentation errors and multiple faults by use of modern estimation theory
`Similar diagnostic procedure are now commerically available from
`(References 6 and 7).
`These
`most major engine manufacturers TEMPER/GEM(GE), COMPASS (RR) and GTEVA (P & W).
`recent additions to the engine diagnostic library utilize similar basic principles but
`incorporate varying enhancement techniques such as weighted least squares, Kalman
`Filtering or proprietary routines to derive the most likely faulty component.
`
`This particular investigation only used a simple version of differential gas path
`analysis so limiting its complexity and compulational requirements. The procedure is
`justifiable in the case of a simple single spool gas turbine and has the added advantage
`of allowing faults or deficiencies in the analysis itself to be more readily
`identified. The use of the simplified procedures does not preclude, at a later stage,
`the adoption of a more complex analysis. The task was basically aimed at investigating
`methods for acquiring, analysing and interpreting data obtained in flight from an
`operational military helicopter and so establishing a technology base in engine
`monitoring procedtires.
`
`2.1 Nodule Assessment
`
`RAAF operational experience with the Lycoming T53 engine in the Iroquois helicopter
`indicated that major deteriorations in performance occurred due to erosion in the
`As it was
`compressor rotor blades and erosion or deposition in the turbine nozzles.
`expected that similar problems would occur in the Lycoming T55-Lll engine the
`The T55-L11 engine has a single
`investigation was based on the Chinook helicopter.
`spool gas generator driving an essentially constant speed power turbine, it consists of
`4 gas path modules, Figure I shows its configuration and the engine station numbering.
`The condition of gas path modules may be deduced from changes in their respective flow
`areas and efficiencies, however to assess the level of degradation of an individual
`component or discriminate between modules then many more measurands may be required for
`analysis than are practically available in a particular installation. A study of the
`T55 thermodynamic cycle shows that at least 13 aircraft/engine measurands are required
`to define a "corrected or normalized" operating point from which basic component
`performances can be assessed. The particular types of measurands determine the level
`and accuracy of the fault isolation capability.
`
`In the Chinook/T55- 7 parameters (N1 , N2 , TOR, T5, P1 , T1 , and IAS) are readily
`For this investigation an additional 6
`available from the existing wiring harness.
`parameters (P2 , P 3, T 3, FF, IGV and BB position) were incokporated into the aircraft-
`engines.
`
`Using these measurands the following basic performance parameters could be derived.
`
`BOEING
`Ex. 1034, p. 247
`
`
`
`p3 /P1
`
`T3 /T1
`
`T5 /T1
`
`Corrected Engine Speed
`
`Compressor Pressure Ratio
`
`Compressor Temperature Ratio
`
`Power Turbine Temperature Ratio
`
`FF/ 64
`
`Corrected Fuel Flow
`
`TORC
`
`Where
`
`8
`
`a
`
`Corrected Torque
`
`T1/ 2 88 .1 5
`1
`
`114.7
`
`P
`
`Temperature Ratio, and
`
`Pressure Ratio
`
`25-3
`
`NIC
`
`CPR
`
`CTR
`
`TTR
`
`WFC
`
`TORC
`
`Of these parameters at least one, either Nl- or p /P1 , is required to specify a
`reference point. Using basic performance paraeters anA the gas turbine cycle equations
`a series of influence coefficients relating changes in efficiencies and areas, etc to
`changes in pressures and temperatures can be determined. Typically
`
`AT 3
`
`= Kal AN 1 + Ka2 ATM + Ka3 AWA + Ka4
`
`Anc + Ka5 Anct + Ka6 AA4
`
`Ap 3
`
`= Kb
`
`1 AN, + Kb2 AT ..
`
`The infuence coefficients Ks1--n' Kb--- n, etc. are evaluated from engine design
`parameters and are presented n matrix form for each operating point.
`
`a T 3
`
`A P 3
`
`Kbl
`
`Kb2
`
`Kal
`
`Ka2
`
`Ka3
`
`Ka4
`
`A N
`
`A T4
`
`A WA
`
`A n o
`
`A n ct
`
`A A4
`A npt
`
`A TORC
`
`Kc
`
`A FFC
`
`a T5
`
`A A5
`
`where for this paper
`
`T
`T
`-T - -I) baseline
`
`AT 3
`
`ATORC
`
`= TOR C - TORC) baseline etc,
`
`can be evaluated at a reference operating point of either P3 /P 1 or NlC
`Inversion of
`the matrix generates a fault matrix for the engine at the given reference operating
`point.
`The fault matrix represents a set of linear equations relating faults in the
`component or module to changes in corrected variables at given prespecified operating
`points. For example using NIC as a reference.
`
`AT = KAI AT 3 + KA 2 AP 3 + KA3 ATOR + KA4 AFF + KA5 AT 5
`A typical numeric fault matrix Is given in Figure 2 for NIC = 95%.
`Calculation of
`trends in the independent variables WA, TM, etc, require fault matrices at each
`operating point for NIC or p /F1 if a fixed analysis point, as in the HIT method, is not
`n thfs-investigation it was found to be sufficient to generate
`to be prespecified.
`fault matrices at 5 engine speeds 80, 85, 90, 95 and 100% NIC and to interpolate
`linearly, for Intermediate points.
`
`ie : KAl/9 5 .6 = M N9 5 .6 + C
`
`Figure 3 shows typical fault coefficient for KA2 , KAI etc, for the
`A TM and delta
`compressor efficiency, algorithms where M and C are the slope and constant of the linear
`equation.
`
`___ a
`
`• '
`
`' "
`
`I;
`
`BOEING
`Ex. 1034, p. 248
`
`
`
`25-4
`
`For this analysis X C was taken as the reference baseline as It was an existing
`paranter, displayed In the cockpit and used by operators for driving and setting
`purposes.
`More importantly compressor erosion is directly related to changes in
`compressor pressure ratio p2 /p 1 , and this parameter could be readily trended, and
`treated independently of any tault matrix.
`
`To determine the respective deltas for each measurand, the individual variables were
`compared with engine data from both computer mod6l predictions and actual engine
`It was found that, as the model predictions were derived from engine
`tests.
`specification data, large deltas or deviations in measured and calculated variables were
`generated, consequently to give more representative trend lines only actual engine test
`data were used in the subsequent analysis. Furthermore, to avoid the effects of bleed
`band on the trend lines and the diagnostics, all data for N1C < 83.5% were ignored.
`
`3.
`
`MONITORING UNIT
`
`A schematic of the Data Monitoring Unit is given in Figure 4. It consists of 3 separate
`parts :
`
`Data Acquisition System
`Instrumentation and Wiring
`Data Transcription and Reduction.
`
`3.1
`
`Data Acquisition System
`
`A 16 channel analogue and digital data recorder (16 CAD) was designed and manufactured
`at ARL in the late 70's for Engine Performance' Monitoring work in the RAAF Macchi Jet
`Trainer, Reference 8 gives details of the basic system. The original 16 CAD systems
`used an 8 bit, tri-tone, recording mode, however the resolution was too small for engine
`performance trending. For this investigation two 16 CAD recorders were combined and
`converted to allow the recording of 16 channels at 10 bits and 14 channels at 8 bits.
`The system was based on small, rugged, low cost consumer available cassette tape
`recorders, and had performed more than adequately in the low vibration environment of
`the Macchi jet trainer.
`However problems arose with the low frequencies emanating
`from the helicopter rotor blades, the vibration predominantly affecting the data tape
`and its transport mechanism. Data were recorded on commercial C-90 compact cassettes in
`30 second blocks at scan rates of 15 channels/sec : each block was therefore made up of
`15 sets of measurands. The data recording was initiated by the pilot and activated by a
`crew man using a control box located in the galley way leading to the cockpit.
`Recording was carried out at least once per flight, once steady state operating
`conditions had been achieved. The data recorder control box imprinted a date-time-group
`and event number on the tape, however engine numbers had to be recorded by hand. Data
`tapes were removed once per week and forwarded to ARL for analysis.
`
`3.2
`
`Instrumentation and Wiring
`
`17 analogue and 5 digital channels, comprising pressures, temperatures, engine speeds,
`fuel flows and discretes were recorded for both engines and the aircraft : a list of the
`instrumentation is given in Figure 4.
`Four of the instrumentation channels were non
`aircraft standard, they were :
`Compressor inlet pressure - Total
`Compressor outlet pressure - Static
`Compressor outlet temperature, and
`Fuel Flow
`
`Special probes for P1 , p2 and T3 , Figure 5, were manufactured by Hawker de Havilland
`Australia, whilst the full flow metering system, based on Faure Herman Turbine Flow
`meters, was detailed by the RAAF. Figure 6 shows a fuel flow meter installed in the
`Chinook No. 1 engine fuel line, whilst a typical pressure transducer Installation Is
`given in Figure 7. The installation of probes, wiring looms, data recorder and break in
`points to existing instrumentation lines were carried out by RAAF personnel at No. 3
`Aircraft Depot Amberley.
`All instrumentation wiring looms were terminated in the
`aircraft cabin heater bay where the data recorder was located.
`The complete
`installation was covered by a RAAF Draft Modification Order - DM0 178.
`
`Instrumentation calibration was carried out initially at ARL and then following
`installation In the aircraft.
`Calibrations were updated at approximately six month
`intervals and coincided with pressure transducer changes and investigations into torque
`meter system irregularities.
`The only problems experienced with the instrumentation
`were with incorrect positioning of the P1 probes (1800 out of phase), fatigue failures
`In the T3 thermocouple, and repeatability problems in the torque meter system.
`Throughout the trial most of the equipment was found to be rzliable : the only component
`affected by the helicopter vibration was the tape recording unit. The instrumentation
`transducers used in this trial were selected primarily for their repeatability as
`against overall accuracy; Table 1 gives a summary of the instrumentation and their basic
`specifications.
`
`1
`
`BOEING
`Ex. 1034, p. 249
`
`
`
`TABLR
`
`1
`
`M3ASUUIMEAT
`
`TRANUTTRR
`
`ACCURACY
`
`REPEATABILITT
`
`No. of BITS
`
`25-5
`
`IAS
`
`SETRA 239
`
`PRESSURES
`
`ROSEMOUNT 1332A
`
`TEMPERATURES
`
`ANALOG DEVICES
`2B 52 TYPE K
`
`TORQUE
`
`MAGNETIC RELUCTANCE
`
`FUEL FLOW
`
`FAURE HERMAN
`
`N 1 - N2
`
`BLEED BAND
`
`PRESSURE SWITCH IN
`IN FCU LINE
`
`± .1% FS
`
`a .1% PS
`
`a .1% FS
`
`a 2.5%
`
`a .25%
`
`a .1%
`
`* .02 FS
`
`* .1% PS
`
`a .11 FS
`
`NOT KNOWN
`
`a .25%
`
`.1%
`
`10
`
`10
`
`10
`
`10
`
`10
`
`10
`
`8
`
`IGV
`POTENTIOMETER
`8
`----------------------------------------------------------------------------------------
`
`3.3 Data Transcription and Reduction
`
`The engine/aircraft data were transcribed from the trn-tone format on the cassette tape
`to octal using an ARL designed transcription unit, Reference 8. Transcription of data
`proved most difficult due to apparent variations in tape speed, caused (it is thought)
`by helicopter vibration. On many occasions data strings appeared to be corrupted and
`were automatically deleted by the error code algorithms. A manual scan of data during
`conversion using a data "break out box" indicated that most of the data had in tact been
`correctly recorded. Because of the simple design of the system it was not possible to
`syncronize helicopter data record speed with transcription play back speed even though a
`time pulse had been imprinted on the tape.
`Repeated transcriptions at modified play
`back speeds, a 10% of nominal, would generate different data sets : a collation of these
`transcriptions could on occasions be used to form a complete data block. Throughout the
`trial the transcripticn unit and the data tape play speed proved to be the most
`unsatisfactory component of the monitoring investigation.
`
`Data reduction was undertaken on a DEC LSI-11/23 computer using instrumentation
`calibrations to convert to engineering units. Data correction, analysis, trending and
`plotting were carried out using standard routines written in FORTRAN.
`
`4. RESULTS AND DATA ANALYSIS
`
`The data acquisition system was installed In Chinook helicopter A15-009 for 2 years when
`it was removed prior to the helicopter under going a major servicing : the
`instrumentation, wiring looms were however left in the aircraft. In the course of the
`trial 18 engines changes occurred, 7 on the starboard and 11 on the port side. A post
`analysis of the engine records showed that only 7 removals were indicative of faults
`which could have been diagnosed by performance trending, ie FCU, air leaks from
`compressor gallery etc. Examinations of all monitored data revealed that only 7 engines
`had data trends with more than 60 points, and of these only two were associated with the
`7 engines removed for gas path related problems. One of these was due to an FCU change
`whilst the other was a high HIT reading which had occurred at the begining of a trend
`record and was corrected by a change in the T5 harness without a permanent engine
`removal.
`
`4.1 Data Reduction and Screening
`
`Initial data analysis was carried out by plotting performance curves for CPR, CTR, TORC
`etc against NIC. The value of CPR etc was determined from an arithmetic mean of the
`individual values of p3 ' and P from the data block.
`The means were therefore an
`average of 15 nominally steady state points. Typical plots for CPR and CTR against an
`N IC baseline are given in Figure 8, while Figure 9 gives a cross plot of CTR versus CPR
`for the data of Figure 8. In both sets of plots the data scatter is small. The only
`measurement to exhibit large scatter bands was Torque ; these plots were much more
`inconsistent.
`Analysis of the raw data from a typical record block indicated large
`differences in the magnitude of the torque measurement.
`Variations of individual
`readings for P3, T , Torque from a data block with time (0-15 sees) are given in Figure
`10, together wits respective values of standard deviatino and coefficient of
`variation. The high value of CVA for torque appears to be a result of aperiodic "drop
`outs" in the indicating system which were not picked up by the analogue cockpit gauge.
`2
`Applications of standard statistical rejection criteria (a
`o ) for example were not
`totally succesafull in eliminating the erroneous, and obvious, torque variations and an
`alternative data rejection algorithm was developed. This algorithm used a combination
`of engineering judgement and statistical principles and incorporated a selective use of
`data means, standard deviations and coefficients of variations. Figure 11 illustrates
`
`.1,
`
`BOEING
`Ex. 1034, p. 250
`
`
`
`25-6
`
`the process on some sample torque data in rejecting an obvious data "outlier" which if
`loft in would have degraded the data mean. The rejection algorithm was used in the
`initial data reduction program and applied to all recorded data. It is interesting to
`note that it had only a significant effect on the torque values, Table 2 shows typical
`result for P3 , T5 and Torque.
`
`TABLE 2
`
`DATA
`
`MEASURAND
`
`P3
`
`T5
`TOR
`
`INITIAL
`
`FINAL
`
`a 0
`
`CVA
`
`x
`
`a
`
`CVA
`
`86.0 1.2 1.4
`
`86.4
`
`.95
`
`1.1
`
`938
`
`6
`
`.6
`
`939 3.8
`
`.4
`
`48
`
`4 8.3
`
`50.2 .65 1.3
`
`Not withstanding the above selection procedure the torque data were still marred by much
`higher scatter than the other variables, and throughout the trial it posed a significant
`obstacle to generating representative trends in many of the independent, or design
`variables. A further problem in the data analysis was the large distance between ARL
`and RAAF Amberley, and the consequent time delay in checking consistency of data. On
`many occasions at least a month elapsed between data recording and data analysis.
`Considerable data would be lost if a probe were broken or incorrectly installed. A
`further problem in the analysis was the lack of engine numbers inscribed on to the tape:
`at least twice an engine change had taken place and had not been correctly identified.
`However step changes in data trends readily indicated this fault. Notwithstanding the
`above comments the basic quality of the data was good, and considered adequate for
`diagnostic purposes provided a comprehensive check of the records was carried out.
`
`4.2 Case Studies
`
`As mentioned above little of the data recorded had immediate application to performance
`trending and gas path diagnostics, however a number of cases warranted further
`investigation, and provided good examples of problems occuring in service, three of
`these are presented as case studies
`Data Consistency - CASE I
`Deterioration in Performance - CASE II
`Component Change - CASE III
`
`It should be noted that the analysis of these cases was carried out retrospectively;
`analysis in real time may not have provided such definitive conclusions.
`
`CASE 1 - Data Consistency
`This particular case emphasses the need to monitor closely the changes in engine -
`instrumentation - calibration -
`configuration.
`In the course of the trial, the
`electrical connections for both engines N1 recording systems were reversed as part of a
`trouble shooting exercise. Failure to include this fact in the initial trend analysis
`resulted in a step change in all measured and calculated variables. It should be noted
`that NiC was the baseline parameter.
`Although the perturbations in trends were
`eliminated by application of a statistical outlier algorithm developed by Frith,
`Reference 9, it did raise doubts about the quality and consistency of the data. Cross
`correlations of both sets of engine data, in this case for corrected fuel flow, Figure
`12, immediately s