Power Plant
`Health Monitoring (cid:173)
`The Human Factor
`
`M J Ward
`Advanced Technologist
`Product Support Performance
`Rolls-Royce pic, Derby, England
`
`Presented to:
`
`Royal Aeronautical Society
`New Zealand Division
`
`Tenth Annual Symposium - Wellington
`
`Projections for the year 2010
`The Human Factor
`13-14 February 1992
`
`BOEING
`Ex. 1015
`
`

`
`Power Plant Health Monitoring· The Human Factor
`
`M JWard
`Advanced Technologist· Product Support Performance
`Rolls-Royce pic, Derby, England
`
`ABSTRACT/OVERVIEW
`
`There are numerous publications of
`technical papers describing how the
`authors set up and ran condition
`_monitoring systems for just about
`,nything from single pieces of
`machinery to vast chemical plants or
`whole fleets of aircraft. Even a quick
`glance at all this work reveals a
`bewildering array of techniques,
`ranging from a man with a pen~il, note
`pad and graph paper reading existing
`dials on a control panel to specially
`designed and built computer systems
`that automatically gather, store and
`process data from dedicated
`instrumentation, before presenting
`reports on equipment health to the
`operators. Each worker in the field of
`condition monitoring claims his
`approach to be the most effective and
`;rumpets the successes he appears to
`have achieved while, one suspects,
`playing down the difficulties and
`failures.
`
`The bemused observer of all this
`frenzied activity is bound to wonder
`what's in it for him and the equipment
`he is operating: he feels, subjectively
`at least,
`that monitoring should be
`able to help him because it will
`increase his knowledge about the way
`his equipment is working, but is unsure
`whether the benefits that may be
`achieved justify the costs and time put
`into setting up and running a condition
`monitoring system. What level of
`
`system should he go for? A man with
`pencil and paper where all the
`interpretation of results is a human
`process, or a sophisticated computer
`system where the interpretation is
`handled by the machine? with the cost
`of human resource continually rising
`and the cost of computer systems
`falling, where do you draw the line?
`Technology versus human intervention?
`What are the cost/benefit trade-offs?
`and what are the human factors
`involved? This paper looks at the
`various types of Engine Condition
`Monitoring (ECM) systems that have been
`employed in gas turbine aero engines
`from the perspective of one engine
`manufacturer and comments on the
`efficiency and dependability of these
`systems as well as looking at proposals
`for the future. Also examined are the
`particular facets of the human
`interface with such systems.
`
`REASONS FOR/BENEFITS OF CONDITIQN
`MONITORING
`
`One of the major reasons why so much
`interest has been shown in condition
`monitoring in recent years is due to
`the continually rising costs involved
`in operation of an airline. As the
`costs of buying and running plant and
`equipment have risen, so requirements
`have emerged for reducing the costs of
`operations and maintenance. Figure 1
`shows the cost breakdown of an
`"average" airline, based on ICAO
`statistics for 1989. While, obviously
`
`1
`
`BOEING
`Ex. 1015
`
`

`
`all of the cost categories could
`benefit from some form of monitoring,
`ECM systems have been designed to help
`reduce costs in the first two
`(separated) categories. aircraft fuel
`
`Generel, edmln,
`Insulllnce & ,"p'naes
`
`'4~
`
`Right cr"" ..11'''11 7$
`& e"pens"
`
`O,pl1tClatlon &
`amortl..Uon
`
`l1chllng, sal•• & promoUon
`17~
`
`Fig. 1 Average airline cost breakdown
`
`and oil costs, and maintenance and
`overhaul. These cost categories depend
`more than any others on equipment
`health and are therefore the
`traditional areas where condition
`monitoring techniques are applied.
`
`Sharper
`maintenance
`
`Greater
`operational
`efficiency
`
`Fig. 2 The benefits of engine condition monitoring
`
`they add up to over 25%
`Between them,
`of the total running costs of a typical
`airline.
`
`2
`
`Figure 2 summarises the potential
`benefits of ECM which can help to
`reduce the above costs in many ways:
`
`a)By reducing the amount of fuel and
`other consumables used,' ensuring that
`the equipment is running as efficiently
`as possible.
`This can be achieved by.
`i)
`enabling on-wing adjustments of
`variable stator vanes, bleed valves,
`fan trim balance, etc. to be made
`without dedicated ground runs.
`this
`clearly has operational advantages,
`as well as saving fuel and
`mechanics time.
`enabling a better standard of
`engine rework to be achieved, by
`correlating rework activity (such
`as blade polishing) with fuel burn
`improvements.
`iii) indicating when engine washing may
`be required.
`
`ii)
`
`b)By reducing the possibility of
`failures of the equipment and hence
`reducing the associated repair and
`dislocation costs, and such intangible
`effects as loss of customer confidence,
`loss of quality and loss of operator
`reputation (often underestimated).
`
`C)By optimising the maintenance of the
`equipment, to achieve the correct
`balance between too little maintenance
`(leading to breakdowns) and too much
`(leading to unnecessary labour and
`parts costs).
`
`Clearly, reducing fuel and consumable
`usage leads to easily quantifiable
`savings: however, it is becoming widely
`recognised that equipment down-time due
`to repair and maintenan~e represents
`lost opportunities for making profits,
`Which, while less easy to quantify, can
`be very significant indeed.
`
`BOEING
`Ex. 1015
`
`

`
`The current levels of gas turbine
`reliability have changed the rework
`
`-228
`
`Shop vlalt
`pet rate
`1000 hours
`
`1.4
`
`1.2
`
`1.0
`
`0.8
`
`0.6
`
`0.4
`
`1st generation
`
`;"'_""'"",,---,
`
`1/'-...•OAnA ~ -,'5~2:4
`0.2 ~524G/~~ '"
`o r.:i
`535E4 3rd generation
`o
`8
`10
`12
`14
`Yellnll In servlc.
`
`2
`
`4
`
`6
`
`16
`
`18
`
`20
`
`Fig. 3 RB211 derivative reliability improvements
`
`criteria more towards performance
`related issues rather than those of
`mechanical failure.
`
`Figure 3 gives a diagrammatic
`representation of the shop visit rate
`for the RB211 family. This shows how
`the derivative approach has improved
`the basic reliability from generation
`to generation to a level where third
`generation engines have a shop visit
`rate of less than 0.1 per thousand
`hours, ie.
`the engines are consistently
`taying on-wing for more than 10,000
`hours.
`
`The twin requirements of health
`monitoring therefore become:
`
`a) the ability to schedule maintenance
`to allow the most cost effective rework.
`
`b) the need to protect against
`disruptive unscheduled removals.
`
`The cost involved in scheduled
`maintenance and overhaul can be reduced
`by:
`
`a) focusing attention on the items of
`equipment that actually require
`maintenance: airframe, APU, engine and
`(with the correct instrumentation
`fitted) module deterioration can be
`diagnosed, enabling maintenance and
`overhaul activity to be concentrated
`on the Areas that require it.
`
`b) allowing the operator to plan shop
`workload ahead of time, based on the
`health of all the engines in the fleet.
`Maintenance and overhaul "peaks"
`could be evened out,
`leading to better
`shop utilisation.
`
`cj allowing the operator to simulate the
`effects of maintenance action, allowing
`trade-offs to be carried out between
`the costs of component efficiency
`improvements due to rework and the
`resulting operational benefits.
`
`ECM can provide a great deal of
`information on which to base decisions
`on scheduled maintenance and overhaul
`action. Depending on the
`information
`instrumentation available,
`about problems requiring maintenance
`action can be produced down to module
`level (for engines) or individual
`control surface level (for airframes),
`with attendant cost savings.
`such a
`level of understanding can also help to
`reduce spare parts holdings.
`Indeed,
`the implementation of such concepts as
`Maintenance Programme Design,
`Resource/Activity Matching
`("Rightsizing") and Just-in-Time
`Inventory systems can be highly
`dependant on having access to the
`information that ECM can provide.
`
`The cost involved in unscheduled
`maintenance and overhaul can be reduced
`by:
`
`a) providing warnings of impending
`failures,
`thus reducing the
`
`3
`
`BOEING
`Ex. 1015
`
`

`
`possibilities of operational disruption
`(remote engine removals,
`inflight
`shutdowns, aborts, delays, diversions
`and cancellations): these events can
`be very costly indeed, both financially
`and in terms of loss of passenger
`confidence, as illustrated in the
`actual example shown in Figure 4.
`
`Event: IFSD and dlveralon wlltl
`re~ Slta engine ...movDt
`
`u_........._r..r._.
`Io\Or«ooM1!RY, "r... De< , tAP)
`_"'Nll"oI,n...,_.o_
`rondr,. _
`"''''''' ""_ ........
`one, - . PlIed .... _. ".d tho
`piIro1-""*"""'_............
`WIlA FlIahl no 0 _ I.... "~I1·
`110I= ... 'on IoIJm, FI
`1Iod "'ode •
`_IftOrlonck>olId_
`l!Io...,lD
`11, Te•. Erlb'r!""
`o.IlM-'<><I ""
`,,,,,'I'''' ....... I',,t
`_k.''''lllD
`" ...ledr"T.... """'lIupl_.
`
`_or"'"'"np~fC,
`
`likely hidden coslS incurred
`(not Including shop v1llt coslS):
`- p8ftS8I1OlIl!IlrBlI8rBmld
`
`10 o\het ._nllS
`
`$IOOOll
`
`""'-~
`
`,,.,
`
`, - _....- .-
`....~, ...._....-
`-- -
`"'..
`....
`_....- -
`
`-~- -
`
`• 1nmIft au! 01 posllIon
`CW88!I~ down
`
`removllllll '8n'1018l1llll,
`lruddngllld!llbol.-
`
`$15000
`
`~~_ !u\lnflMlnllflloM
`_ Crew costs. IlInding t8e3.
`~1rkllII,<!Ilc
`
`Olhere
`
`$12000
`
`$2000
`
`$70000
`
`performance and consequently use less
`fuel.
`
`condition monitoring may be defined as
`a system (or series of systems) that
`gather, store and process data from an
`operator's equipment in order to assist
`in making decisions about the operation
`and maintenance of such equipment.
`It
`is the key element that enables the
`operator to change from a strategy of
`reaction to unplanned events (such as
`excessive fuel usage, breakdowns, etc.)
`to one of pre-emptive action to
`foreseen events. All in a timescale
`which aids management,
`improves
`efficiency,
`reduces disruption and
`increases the profitability of the
`equipment being used.
`
`. Worth 3% of aircraft annual fuel burn
`
`TECHNIQUES USED IN CONpITION MONITORING
`
`Fig. 4 An example of hidden costs
`
`b) reducing secondary damage, by
`indicating the need for maintenance
`and overhaul in time.
`
`As well as making a contribution to
`cost savings in the above areas, ECM
`can also help in fleet management, by:
`
`.) allowing the operator to match
`aircraft to particular route
`requirements (such as hot engines on
`cool routes, high performance aircraft
`on longer routes).
`
`b) providing statistics to aid fleet
`planning and utilisation studies.
`
`Beyond enhancing the operational
`reliability, utilisation and safety of
`operation, Engine Condition Monitoring
`helps to reduce the burden that today's
`aircraft place on the environment.
`Simply because engines which get the
`best possible maintenance have the best
`
`Clearly, man has been monitoring
`machinery for centuries, if only by
`listening to it!
`"Funny noises" in
`anything from ancient waterrnills to
`modern motor cars have always been a
`signal for the alert operator to get
`things checked over:
`"It sounds
`different, so something must be wrong".
`
`(up
`With aircraft engine maintenance,
`until the late 1960's), it was cornmon
`practice to overhaul an engine after a
`predefined number of flight hours or
`cycles whether it needed it or not, but
`with the increasing maturity of jet
`engines it has been recognised that it
`is the "condition" that counts.
`
`In the early days, condition monitoring
`mainly consisted of monitoring
`mechanical parameters such as oil
`system filter condition, oil pressure,
`temperature, engine vibration and
`magnetic chip detector (MCD) debris.
`This type of data was manually recorded
`and analysis of MCD debris required the
`assistance of a qualified inspector
`
`4
`
`BOEING
`Ex. 1015
`
`

`
`with a turn around time of up to two
`hours. A schematic diagram of the MCD
`debris analysis process is shown in
`Figure 5. This illustrates how a track
`of the accumulated debris can be used
`to detect incipient failures. A
`database of debris samples was usually
`kept using a basic card index system.
`Records of oil consumption may also
`have been taken to help monitor the
`engine. This type of system relied and
`still relies on subjective human
`experience built up over many years of
`exposure with the airline and its
`operation.
`
`Fig. 5 Magnetic chip detector debris analysis
`
`~owever, it was not until the early
`aeventies that the use of additional
`maintenance methods, required to
`monitor the engine performance level
`became established, particularly after
`the fuel crisis of 1973. Maintenance
`became modular, and the fixed time
`between overhauls gave way to the idea
`of maintenance depending on condition.
`
`What makes today's condition monitoring
`different to that used previously is
`that:
`
`a) Data is gathered systematically from
`the equipment, even when nothing
`appears to be wrong.
`
`b) Data is recorded for subsequent
`analysis and interpretation.
`
`some processing of the
`C) ~n most cases,
`data is done to convert measurements
`to meaningful
`information.
`
`The evolution of condition monitoring
`(as defi:ned above) shows a clear
`progression, see Figure 6. As the
`
`Basic mechanical
`parameters
`-Mea
`.Ollfllters
`_Vibmtlon
`• Borascope
`• Oil consumption
`
`Q...."'II1II
`.~~ ~
`'9;i
`
`Requirement for additional
`"condition" monitoring of
`• Engine performance
`• Fleet condition
`
`Integrated approach means that:
`- Data gathered systematl<::slly
`- Proc8998d to convert meansurements
`to meenlngfullnfonnation
`- Combined database for entire operation
`
`Fig. 6 Condition monitoring evolution
`
`complexity of operations has increased,
`from isolated recordings of instrument
`readings and other data sources fitted
`mainly for safety purposes on
`to what
`individual pieces of equipment,
`is now,
`integrated recording of all
`instruments and data sources (some of
`which have been fitted specifically for
`monitoring) on all pieces of equipment
`in use. Conversion of measurements
`into meaningful
`information/maintenance
`actions has become vital since the
`volume of available data has, and is,
`continuing to increase almost
`exponentially in a climate where the
`cost of the human resource required to
`interpret any results is also
`increasing rapidly.
`The degree to
`which data is processed to convert to
`meaningful information is fundamental
`to the type of human involvement
`subsequently required,
`the operator
`
`5
`
`BOEING
`Ex. 1015
`
`

`
`cannot afford to waste human resource
`as a basic data interpreter, this has to
`be delegated to the machine as far as
`possible.
`
`As the amount of information being
`recorded and processed has increased,
`the advantages of an integrated,
`equipment wide, computerised approach
`to condition monitoring have become
`apparent, namely:
`
`ie.
`was in proper working order,
`consistent with some form of prediction
`for the partiCUlar engine type.
`For
`this type of ·'single-shot" analysis the
`database of knowledge consisted of a
`graphical history of the particular
`parameter being monitored spanning the
`last few flights of the aircraft and
`the comparative results were available
`to the crew in a matter of minutes of
`recording the data.
`
`a) correlation of information from
`different data sources on the same
`piece of equipment enables a better
`picture of the status of that
`particular item to be built up.
`
`The limitations with this approach were
`that the prediction method was
`relatively coarse and it was difficult
`to build up a picture of the overall
`fleet condition.
`
`b) correlation of data from different
`pieces of equipment enables a picture
`of the health of the entire operation
`to be obtained.
`
`An analogy for this is the recording of
`data on a patients health at a Doctor's
`Surgery: separate records of blood
`pressure, pulse rate, age, height and
`weight, etc. are obviously less useful
`than a combined record of all the
`information (since many medical
`conditions require information from
`~everal such data sources before a
`~nfident diagnosis of a complaint can
`be given). Also, easy access to all
`information from all individuals gives
`important clues about the health of the
`whole population and can lead to
`effective diagnosis of health problems
`affecting everybody (such as
`contaminated food and water supplies).
`
`DATA COT.I.ECTTON METHOpS
`
`Originally, Flight Engineers were used
`to record the inflight values from
`cockpit instrumentation at regular
`intervals during stable cruise. A
`slide rule or "Dunn Wheel" nomogram was
`used to ascertain whether the engine
`
`For a more centralised approach to ECM,
`there was a progression towards using a
`ground based computer to analyse the
`recordings. This involved transmitting
`the data back to base where
`mathematical algorithms could provide
`consistent analysis of the whole fleet.
`
`This approach had, and continues to
`have, one major drawback,
`in that there
`are inherent time delays introduced by
`having to transmit the data back to
`base (either by telex,
`fax, voice,
`etc.). This,
`together with manual
`transcription of the data into a form
`suitable for input to the computer
`means that the results of the analysis
`are not immediately available for
`interpretation. This process is also
`susceptible to instrumentation reading
`accuracy and transcription errors.
`
`There was a major step forward with the
`introduction of Digital Flight Data
`Recorders (DFDR'S) or Quick Access
`Recorders (QAR's). These are able to
`record the main aircraft and engine
`parameters, during the whole or
`selected segments of a flight, onto
`magnetic cartridge. These cartridges
`can be downloaded via ground replay
`
`6
`
`BOEING
`Ex. 1015
`
`

`
`Addressing and Reporting system).
`These datalinks allOW a multitude of
`messages/data to be sent between an
`aircraft and the airline ground base
`using VHF communication satellites or
`ground network systems.
`
`We have now therefore arrived at a
`system, illustrated in Figure 7, which
`can acquire and transmit selected data
`to a central ground station and provide
`analysed output within minutes of an
`event.
`The datalink also allows ground
`personnel to request additional data at
`anytime during a flight to aid
`troubleshooting •
`
`'"
`II
`""'. 'Teet Cell ..
`
`..',.
`'0-........
`
`REPORTS
`via
`Data Unk
`
`MAIN BASE
`Ground based Software
`~ Performance Analysis
`- Aircraft
`- Engine
`Maintenance Englneerlng'-_..:-"'M'=U'-__--'
`
`"An;'al~Y$~a~d~O~u~Pu~t-;-: -,
`
`- Trends
`- Alert Messages
`
`•
`
`Fig. 7 Condition monitoring - System overview
`
`The human element has been removed from
`the day to day decision making in
`acquisition and analysis of data, but
`conversely it puts emphasis on the
`human interpreter who now needs to be
`available 24 hours a day to support a
`round the clock operation. This may
`not be practical in todays environment
`and may necessitate changes to working
`practices to ensure that the
`appropriate personnel are on-hand and
`available to make decisions.
`
`systems into a central computer which
`can then dissect the time history of
`the flight in order to select
`appropriate cruise,
`take-off and other
`types of data as necessary. This
`technique, currently in use by a number
`of airlines eliminates transcription
`errors and provides some improvement in
`response time.
`
`In today's latest generation aircraft,
`there are onboard Aircraft Condition
`Monitoring systems (ACMS) which
`continually monitor the aircraft ARINC
`databuses.
`The ACMS is programmed to
`-elect/record data in the form of
`.ceports" when various trigger logic
`and stable frame criteria are
`satisfied. These reports can again be
`stored on cartridge or "floppy disk" as
`the medium for transfer to the ground
`based analysis programs. This form of
`data sampling at source vastly reduces
`the quantity of data recorded/stored.
`
`The use of ACMS systems has also meant
`that there have been tremendous
`advances in the overall quality of the
`pata being recorded in terms of both
`accuracy and repeatability. This,
`coupled with a corresponding
`~mprovement in today's sensor
`dliability has meant that the
`integrated approach to condition
`monitoring has become a reality.
`
`there is still the problem
`However,
`especially for aircraft on long haul
`routes, where the aircraft may be away
`from home base for anything up to 3-4
`days.
`consequently the data on the
`cartridge or "floppy disk" may be 3-4
`days old and a fault or trend may well
`have been established unbeknown to the
`operator. This problem is being
`addressed by various operators who are
`now employing near "real time"
`datal inks within their operation, such
`as ACARS (Aircraft Communications
`
`7
`
`BOEING
`Ex. 1015
`
`

`
`GRmJND l\Np,X.ySTS SOFTWARE
`
`The objective of the ground analysis
`software is to process sufficient
`information in a manageable form to
`allow the operator to make effective
`decisions. Gone is the mass of output
`relying on the experience/intuition of
`the human to find the problem,
`this has
`been replaced by elegant software
`systems that can reduce this avalanche
`of data to a manageable form that can
`be readily interpreted.
`
`considerable resource has been invested
`into the development of ground based
`~oftware by all engine manufacturers in
`order to obtain the maximum amount of
`knowledge from the measurements
`available and to present the results to
`the operator in a form that is clear,
`concise, easy to interpret an~, above
`all, correct.
`The operator needs to be
`presented with the optimum amount of
`information to enable him to make his
`maintenance decisions -
`too little and
`the troubleshooter does not have enough
`to understand the problem fully,
`too
`much and the relevant part is difficult
`to find,
`thus destroying his
`effectiveness.
`
`it was recognised by Rolls-Royce that
`as condition monitoring systems have
`become more and more sophisticated the
`ground based software required to
`support such systems should consist of
`two basic parts:
`
`a) Completely general features for
`storing displaying and manipulating
`data,
`that can be applied to any piece
`of equipment being monitored.
`
`b) Application-dependant features,
`tailored to the requirements of the
`equipment being monitored.
`
`the human skills required
`In many ways,
`to develop the two parts are very
`different: the general features which,
`typically, absorb 75% to 85% of the
`effort needed to create a condition
`monitoring system require software
`expertise,
`to ensure that the latest
`techniques of data storage and
`retrieval, statis~ical analysis, data
`input and output, and software
`structuring are employed. Compared to
`this, are the application-dependant
`features requiring engineering
`expertise, to ensure that the correct
`data is being gathered and appropriate
`analysis methods are being used.
`Original Equipment Manufacturers
`(OEM's) can therefore concentrate on
`the engineering aspects of monitoring
`rather than having to re-invent the
`wheel to create a full blown condition
`monitoring system.
`
`(Condition Monitoring and
`COMPASS,
`Performance Analysis software system),
`the latest generation condition
`monitoring software from Rolls-Royce,
`has been developed with the above
`points in mind. Figure 8 illustrates
`how the system interfaces with the
`incoming data and also defines the
`areas of responsibility for data
`acquisition/handling.
`
`__ E~.m ""I_~D11y
`
`- - - M
`"'\11
`
`"',U __IbIllly ..llh
`"I..,w.- lnvoIY.....nl
`
`N~m··...lJU
`
`T.~nd pial.,
`
`......
`
`hnrchllOtB,
`~.Yplol.
`
`r.....'
`
`~""'pr
`
`""'!l.
`lon.
`
`",tot","
`
`m"'"I""""".
`
`Fig. 8 Schematic of Aircraft/COMPASS interface
`
`8
`
`BOEING
`Ex. 1015
`
`

`
`~he COMPASS computer program consists
`of two basic parts:
`
`Rolls-Royce see the following unique
`advantages over users of other systems:
`
`alA Neutral Host, within which the
`application dependent routines reside,
`providing all the general facilities
`required to turn a set of analysis
`routines into a fully-fledged condition
`Monitoring system (ie.
`the ~rend, Plot
`and utility Modules - see Figure 8).
`~hese facilities include:
`
`i)
`ii)
`iii)
`iv)
`
`V)
`vi)
`
`data management and storage
`data plotting and display
`statistical analysis
`data smoothing and trend
`recognition
`alert generation and processing
`system "housekeeping" functions
`
`b) ~he application-dependant routines
`(Analysis Module - see Figure 8),
`appropriate to the range of equipment
`being monitored.
`
`• Supplied fi1ln~ system
`and Interlac.& may be
`replec.ed by ope...,lors
`'yllem c..d
`0'"" IlIln..
`Interface.
`
`7",/7"-i7"-/,L-/~,L~_~Neutrol COI,IPASS
`HOlt SoltwlIrc
`
`Englne/Alroraft Mgnulocturers
`Engino/A1rcroft Dependent
`Proprietary Monitoring
`Diogno,tic Routine,
`
`Fig. 9 COMPASS neutral host concept
`
`~he Neutral Host concept is illustrated
`in Figure 9 and shows how analysis
`modules from various OEM's can be
`integrated within the general ~rend,
`plot and utility Modules.
`~he Neutral Host has been written by
`sD-scicon (UK) Ltd, a software house
`that is independant of OEM's.
`
`alA fully integrated approach to
`condition monitoring which reduces the
`need for separate condition monitoring
`programs for each engine type/fleet,
`and "pulls together" all the
`information bei~g gathered to give
`maximum visibility of the health of
`the entire operation (together with
`reduced implementation, support,
`training and running costs).
`
`improvements available for all
`b) Host
`equipment being monitored.
`
`c) Better support (creation,
`installation, documentation etc,
`a dedicated software house).
`
`from
`
`In line with the development philosophy
`that has been outlined previously,
`COMPASS Trend Module has been designed
`to report by exception.
`~he operator
`is only alerted to problems as and when
`they appear and is subsequently guided
`towards the data that will help solve
`them.
`~his is opposed to being
`presented with large quantities of data
`output listings for manual
`interpretation.
`
`Current operators of the COMPASS system
`are:
`
`British Airways
`cathay Pacific
`Lufthansa
`cyprus Airways
`IAE (International Aero Engines)
`
`Commensurate with COMPASS philosophy
`both British Airways and Cathay Pacific
`are in the process of extending their
`range of analytical routines for
`COMPASS in order to have one condition
`monitoring system for their entire
`operation.
`
`9
`
`BOEING
`Ex. 1015
`
`

`
`COMPASS is currently available on the
`IBM mainframe and DEC Microvax
`environments. However, recent trends
`in the rapid development of computer
`hardware have established a clear
`requirement to make software systems
`such as COMPASS available in the PC
`environment.
`xt is becoming
`increasingly apparent that the human
`tends to feel more "comfortable" with a
`dedicated PC system (or networked PC
`system) rather than the more complex
`interface with a central mainframe
`system.
`
`"E'lJTtJRE ENHANCEMENTS
`
`Figure 10 gives a schematic
`representation of how the evolution of
`condition monitoring systems has seen
`the progression from the use of human
`expertise to interpret data through to
`todays situation where sophisticated
`computer systems both acquire the data
`and convert it into useful information
`(maintenanace actions) using
`pre-defined rules.
`The next step would
`be to remove the need for a human to
`define the rules for interpretation,
`but provide a system which is capable
`of generating/learning it's own rules.
`This situation is represented by the
`extreme right hand side of the graph in
`Figure 10, and it is being suggested
`
`100% Human
`
`EcM
`output
`interpretation
`
`100% Machine
`
`+ Alerts and maintenance
`recommendations
`+ Artificial intelegence
`
`111llO 1870 11180 1VllO 2000
`
`Time
`
`Fig. 10 The changing emphasis in data interpretation
`
`10
`
`that this could be achieved with the
`emerging computing techniques that
`collectively fall under the term
`"Artificial J:ntelligence".
`
`Artificial xntelligence techniques are
`claimed to be able to encapsulate and
`replay on demand the sorts of
`procedures and thought processes that
`are normally associated with human
`experts in a particular field,
`thus
`making such expertise more widely
`available,
`freeing the human expert
`from the more routine applications of
`his knowledge and allowing him to
`concentrate on areas less amenable to
`automation.
`Some airlines have been
`investigating the application of such
`techniques to their operations for some
`yeare now, and they and others have
`turned their attention to using these
`programming methods for aircraft and
`Engine Condition Monitoring and
`maintenance.
`
`A constituent part of the proposed
`Artificial Intelligence is "neural net"
`software. Essentially, a neural net is
`a software "structure" that can be
`IItrained" (using examples of ordinary
`and out-of-the-ordinary occurrences) to
`recognise significant changes in a
`stream of input data by differences in
`the way the network responds: it is
`intended to replicate certain aspects
`of the human nerve system, hence the
`name.
`
`In general terms, the proposed system
`would consist of the following
`constituent parta (see Figure 11):
`
`alThe current COMPASS system, using
`physical laws encoded as analytical
`functions,
`reducing the large amount
`of raw data input to the system to a
`smaller quantity of corrected and
`"filtered tl data.
`
`BOEING
`Ex. 1015
`
`

`
`trained with large
`b)A neural net,
`amounts of "normal" (ie. no fault data)
`reviewing the COMPASS output to produce
`a binary "fault/no fault" assessment
`of the data and/or probabilities of
`the occurrence of specific faults of
`interest to the user.
`
`clAn expert system using the ouputs
`from the neural net together with a
`set of rules to:
`
`Raw Data
`
`Corrected
`
`Dala
`
`\\\\111
`'Filtere~ ta' Net
`~e"" t i /.
`Dala
`~ 'Patterns'
`Expert
`OJ
`
`AText +
`8~SYSlem~
`• • 0 • • '-'cEype0-- Graphlcs-----'
`
`Database
`
`/ '
`
`,
`
`,text
`
`End User
`
`Fig. 11 ECM enhancements: From data to diagnosis
`
`i)
`ii)
`
`iii)
`
`present the user with a diagnosis.
`Interact with a database of past
`diagnosis both to store the
`current diagnosis and refer to
`previous ones if necessary.
`Access a hypertext package to
`present the user with any test
`and diagrams from interpretation
`guides, maintenance manuals,
`fault isolation manuals,
`that
`complement the diagnosis.
`
`The above is seen as a viable
`enhancement to today's condition
`monitoring programs, it is clear that
`an increasing number of airlines are
`investigating the usefulness of
`Artificial Intelligence methods to
`support their operation, and an
`increasing number of software suppliers
`
`11
`
`are responding to this with products
`incorporating such techniques.
`
`THE HUMaN INTERFACE
`
`Common to all current Engine Condition
`Monitoring systems is the requirement
`for a suitable human interface. This
`may take· the form.of recognising
`numbers on a calculator display or
`shapes of debris on a magnetic chip
`detector through to specific alert
`messages pin-pointing the particular
`parameter at fault, giving recommended
`maintenance action and referring to
`supplementary data. Whichever is used,
`the human interface is vital to the
`success of any complex software package
`and must recognise the routing of the
`information to the appropriate
`recipient. An example of the potential
`community is given in Figure 12 .
`
`Potential users community
`
`• InfOrmation available
`• Alert ~ummaries {auto}
`• Engine trends
`• Module trendS
`
`• Mechanical trends
`• Test cell run analysis
`• Simulation
`
`• Fleet average data
`• Statistical data
`• Fan trim balance
`
`Fig. 12 Potential users community
`
`The importance of "people planning" in
`the operation of a Condition Monitoring
`system must be clearly recognised.
`Automated system output needs to be
`designed for rapid interpretation by
`line maintenance personnel not only by
`the computer and thermodynamic experts
`that conceived and implemented the
`system.
`
`BOEING
`Ex. 1015
`
`

`
`The following criteria must be taken
`into account:
`
`a) If following implementation, a new
`system is "dogged" by spurious alerts
`and false indications,
`the credibility
`of the operator/system is thrown into
`doubt from which it may never recover.
`conversely, if easy to read output,
`reliable alert messages,
`leaving the
`operator free from confusion are
`employed,
`then effective maintenance
`actions can be taken which spreads
`confidence throughout the whole
`operation.
`
`) In order to cater for the most
`inexperienced condition monitoring
`operators, the suppliers of ECM
`systems must provide comprehensive
`interpretation guides for the system
`output. Inherent in this is the
`knowledge of the engine itself, its
`operating characteristics and its
`failure modes.
`The most likely cause
`for shifts in a parameter or
`combination of parameters must be
`clearly identified.
`The experienced
`operator may well have a "feel" for
`these causes already, based on previous
`experience and would not require any
`further interpretation.
`
`C) with the vast amount of data that is
`available with modern aircraft
`condition monitoring systems it is
`fairly obvious that the operator can
`easily become swamped with trend
`output.
`It is here,
`that the
`that
`reporting by exception concept,
`later generation ECM systems employ,
`comes into its own.
`The operator is
`only alerted when something out of the
`ordinary occurs - vast amounts of data
`can be monitored/filtered for problems
`automatically, with no need for human
`intervention.
`
`Admittedly, resource has to be used in
`defining the "normal" engine and
`deriving suitable alert levels for the
`filtering process. This "tuning" of the
`system may actually take several
`iterations, but provided that there is
`the flexibility within the syste