`Engineers, Part F: Journal of Rail and Rapid
`Transit
`
`http://pif.sagepub.com/
`
`
`Diesel Locomotive Reliability Improvement by System Monitoring
`K N Fry
`Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit
`DOI: 10.1243/PIME_PROC_1995_209_248_02
`
` 1995 209: 1
`
`The online version of this article can be found at:
`
` http://pif.sagepub.com/content/209/1/1
`
`Published by:
`
`http://www.sagepublications.com
`
`
`
`On behalf of:
`
`Institution of Mechanical Engineers
`
`
`
`
`
`
`Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid TransitAdditional services and information for
`
`
`can be found at:
`
`Email Alerts:
`
`
`http://pif.sagepub.com/cgi/alerts
`
`Subscriptions:
`
`http://pif.sagepub.com/subscriptions
`
`
`
`Reprints:
`
`
`http://www.sagepub.com/journalsReprints.nav
`
`
`
`http://www.sagepub.com/journalsPermissions.navPermissions:
`
`
`
`>>
`
`Version of Record
`
`- Jan 1, 1995
`
`What is This?
`
`
`
`Downloaded from
`
`pif.sagepub.com
`
` by guest on June 3, 2013
`
`IPR2013-00412 - Ex. 1002
`Toyota Motor Corp., Petitioner
`1
`
`
`
`
`
`
`
`
`
`
`Proceedings of the
`Institution of
`Mechanical Engineers
`
`I MECHE
`
`Part F
`Journal of
`Rail and Rapid Transit
`
`1995 Vol209 No Fl
`
`Contents
`
`Editorial
`
`ISSN 0954-4097
`
`Diesel locomotive reliability improvement by system monitoring
`KNFry
`
`Experimentation and modelling of climbing phenomena of guided transport
`vehicles
`P Drazetic, R Tassin, Y Ravalard, B Marguet and H Lagneau
`
`Phenolic composites--novel developments in products and processing
`M R Orpin
`
`Design for the alleviation of transportation fatality risk by the implementation
`of new light rail transit
`R Huston, P Cardimen and K Halperin
`
`Fail-safe on-board databus for automatic train protection
`i-I Kirrmann, B Eschermann and D Forsgran
`
`Vertical interaction between train and track with soft and stiff railpads---full-scale
`experiments and theory
`M Fermer and J C 0 Nielsen
`
`Analysis of modified railway passenger truck designs to improve lateral
`stability /curving behaviour compatibility
`S Narayana Swamy, R V Dukkipati and M 0 MOsman
`
`11
`
`19
`
`25
`
`33
`
`39
`
`49
`
`Communications are invited on these papers (see inside back cover)
`
`Published for the Institution of Mechanical Engineers by Mechanical Engineering Publications Limited,
`London and Birmingham, Alabama
`
`The Proceedings of the Institution of Mechanical Engineers is a ten-part journal. Part F is published twice a year. The two
`issues are numbered separately within the volume, i.e. Fl, F2.
` by guest on June 3, 2013
`pif.sagepub.com
`Downloaded from
`
`2
`
`
`
`Diesel locomotive reliability improvement by system
`monitoring
`
`KN Fry, BSc
`British Rail Research. Railway Technical Centre, Derby
`
`System monitoring for reliability (SMR) involves monitoring critical parts of a vehicle and informing the owning business of an
`impending fault. Diesel locomotives offer the largest opportunity for such systems and British Rail Research has developed a system
`designed to improve Class 47locomotive reliability.
`The vehicle-mounted equipment comprises a computer that continuously monitors the condition of the vehicle through sensors at key
`points. The computer is connected to a radio telephone and modem and a GPS satellite navigator. The key elements in the success of
`the system are the automated analysis of data on-board the vehicle and its ability to call for help ahead of the occurrence of service
`failures. The business interface is through a Windows based information display which runs on a personal computer connected to the
`public telephone network. This controls the display of messages from monitored vehicles and allows vehicles to be interrogated to check
`on current condition.
`When fully implemented, a reduction in technical casualties of 40 per cent is anticipated. There are additional financial benefits from
`efficiency improvements and vehicle maintenance cost savings.
`
`Key words: system monitoring for reliability, diesel locomotives
`
`1 INTRODUCI'ION
`The Vehicle Systems Unit of British Rail Research has
`undertaken a series of projects over many years con(cid:173)
`cerned with the development of condition monitoring
`for railway rolling stock and diesel locomotives in par(cid:173)
`ticular. This work has recently concentrated on the
`development of systems to improve the reliability of
`vehicles. System monitoring for reliability (SMR) is the
`name given to monitoring critical parts of a vehicle in
`order to improve its reliability by informing the owing
`business of an impending fault.
`This paper describes the development of a system for
`monitoring Class 47 locomotives. It begins with the
`guiding philosophy for SMR. It then goes on to
`describe the component parts; the on-board equipment
`and analysis of data; the communication of information
`to and from the vehicle; and the information display
`system. Finally, there is a review of the current position.
`
`The MS was reuived on 25 November 1993 and was accepted for publication on
`22 December 1994.
`
`l PHILOSOPHY OF SYSTEM MONITORING FOR
`RELIABILITY
`1.1 The economic backgrouad
`Studies into maintenance and maintenance-related costs
`of typical types of rolling stock have been undertaken to
`determine the most cost-effective areas for the applica(cid:173)
`tion of condition monitoring and diagnostic systems.
`The four main vehicle types used by British Rail have
`been examined; diesel locomotives, electric locomotives,
`diesel and electric multiple units.
`The costs have been broken down by vehicle system
`and subsystem into five areas, three directly associated
`with maintenance; exams, repairs and overhauls and
`two indirectly related; the cost of unreliability and
`unavailability. The results, given in Fig. 1, show that the
`largest total cost is for diesel locomotives, and the
`largest element of this total is the cost of unreliability.
`Examination of the reasons for unreliability showed
`that the total is made up of a small number of causes
`(see Fig. 2), many of which are easily monitored and
`
`.!!
`.!:!
`.<::
`~ ,_
`8.
`~ g
`
`1t ., .. ~
`
`0 Unavailability
`~
`Reliability
`0
`Overhauls
`
`Repairs
`
`Exams
`
`fl1 •
`
`Diesel locomotive Electric locomotive
`
`Diesel
`multiple unit
`
`Electric
`multiple unit
`
`Vehicle type
`Fig. 1 Rolling stock maintenance and maintenance-affected costs
`
`F01693 C IMccbE 1995
`
`Proc lnstn Medl Engrs Vol 209
`
`Downloaded from
`
`pif.sagepub.com
`
` by guest on June 3, 2013
`
`3
`
`
`
`2
`
`K N FRY
`
`8
`
`6
`
`4
`
`I
`8. .. > .,
`
`~
`~ .a
`~
`Traction control
`interlocks
`
`D Not preventable
`•
`Preventable
`
`Battery
`condition
`
`Loss of coolant
`and low oil
`pressure
`
`Brake condition
`Fig. 2 Causes of Class 47 failures in service
`
`allow warning to be given. About 40 per cent are con(cid:173)
`sidered preventable through appropriate monitoring
`and notification of impending failure.
`
`2.2 Reliability emphasis
`The factors described in Subsection 2.1
`led to the
`concept of system monitoring for reliability which is
`specifically aimed at reducing
`in-service failure of
`equipment rather than reducing maintenance costs or
`increasing availability. This very specific approach gives
`a number of advantages:
`
`(a) the number of measurands is significantly reduced;
`(b) the maintenance philosophy of the vehicles does not
`change, so it is quicker and easier to implement;
`(c) data analysis is generally easier since faults severe
`enough to cause a vehicle to fail are more easily
`identified than the comparatively smaller changes
`associated with a need for maintenance.
`
`2.3 Importance of on-board analysis
`Most approaches to vehicle monitoring have involved
`fitting data logging equipment and analysing the data
`after they have been downloaded. This requires the
`routine download of a large amount of data, the major(cid:173)
`ity of which will indicate that the vehicle is healthy. It
`also introduces a delay in fault identification and will
`miss the majority of faults likely to affect service reli(cid:173)
`ability on a daily basis.
`An emphasis on reliability improvement requires that
`in order to achieve a fast response to developing faults,
`the analysis of data must be automated and done on(cid:173)
`board the vehicle; the vehicle must also be able to call
`for help. It is this requirement for a high degree of
`vehicle system 'intelligence', in conjunction with a com(cid:173)
`munications system where the vehicle can call for help
`or be called at any time that is the key to successful
`SMR.
`The move towards an 'intelligent vehicle' gives two
`other major benefits. Firstly, it considerably reduces the
`
`Part F: Journal of Rail and Rapid Transit
`
`amount of data requiring transmission, which is partic(cid:173)
`ularly advantageous where communication is by radio.
`Secondly, it opens up the possibility of providing infor(cid:173)
`mation to the driver or train crew in those situations
`where it can be usefully acted upon.
`The provision of a continuous communications link
`between the owning business and an 'intelligent vehicle'
`allows information on condition to be provided on
`demand. Such vehicle interrogation may be useful as a
`check on condition just prior to assignment or for mon(cid:173)
`itoring the development of a fault already reported.
`
`2.4 Information required
`System monitoring is really only half the story. In order
`for the service reliability to be improved not only must
`information about a fault be provided, but the informa(cid:173)
`tion must be suitably acted upon in order to remedy the
`situation. If appropriate action is not carried out, the
`vehicle will fail just as it would without monitoring. In
`this respect the presentation of information to the end
`user is of paramount importance.
`An important part is the information content. The
`majority of rolling stock is maintained by means of
`component replacement to facilitate rapid return to
`service and so the information provided should support
`this philosophy. In other words, faults in equipment
`need only be diagnosed down to the level of 'replaceable
`unit' or the level of action required to allow the vehicle
`to continue running, such as 'top up with coolant'.
`Diagnosis to this depth is particularly important in the
`case of a vehicle reporting a developing fault but a long
`way from a repair depot. Should the vehicle be brought
`back to the depot, repaired at an outstation, repaired by
`a mobile maintenance team or left for a while? If the
`vehicle does need to return to depot, diagnosis to
`replaceable unit level allows an indication beforehand
`of what spares and depot resources are required (for
`example, under cover, crane, pit, manpower etc.) This
`speeds up repair time considerably. Similarly, if a
`mobile maintenance team needs to be sent to the vehicle
`they will know what equipment to take with them.
`
`IC)IMechE 1995
`
`Downloaded from
`
`pif.sagepub.com
`
` by guest on June 3, 2013
`
`4
`
`
`
`DIESEL LOCOMOTIVE RELIABILITY IMPROVEMENT BY SYSTEM MONITORING
`
`3
`
`Information from the system must be sent to the
`business maintenance controllers. They respond to calls
`from outstations reporting problems with vehicles and
`arrange repair or a replacement vehicle.
`For fault diagnosis, messages should be sent from the
`vehicle immediately, but this is not necessarily the case
`when prognosis is involved. There are situations where,
`if the existing set of circumstances were to continue, the
`vehicle would fail, but the fault may be naturally reme(cid:173)
`died before it occurs. For example, a battery draining
`with the engine stopped may be just about to be
`charged following an engine start, or a vehicle with a
`coolant leak may be just running on to a depot to have
`its coolant topped up. In these circumstances the
`approach has been to define a failure limit and make
`predictions of the remaining time to failure. Messages
`can then be generated a set time before failure is esti(cid:173)
`mated. The maintenance controller can then decide on
`whether the fault will require action based upon the
`duty of the vehicle.
`
`3 ON-BOARD SYSTEM
`
`3.1 Equipment fitted
`The vehicle-mounted equipment comprises a computer
`that continuously monitors the condition of the vehicle
`through sensors at key points. The computer is con(cid:173)
`nected to a radio telephone and modem allowing the
`system to ring out with fault messages or be inter(cid:173)
`rogated by the owning business at any time. The com-
`
`puter knows the vehicle's position through connection
`to a GPS satellite navigator.
`The equipment is housed in three rugged steel enclo(cid:173)
`sures sealed to IP66 and protected against the electro(cid:173)
`magnetic environment of the locomotive. Also fitted are
`a number of transducers mounted directly on to existing
`components, some small enclosures containing trans(cid:173)
`ducers and appropriate interconnection via high specifi(cid:173)
`cation cable sealed into flexible conduit. Two aerials are
`also used, a short whip aerial mounted on the end of the
`vehicle, and a small flat antenna mounted on the roof
`for the GPS navigator. This equipment is designed to
`retrofit into the vehicle without interfering with its
`normal operation and maintenance. A general sche(cid:173)
`matic is shown in Fig. 3.
`
`3.1.1 Computer
`Hardware. The on-board computer is an industry stan(cid:173)
`dard VME bus-based system made up of single height
`eurocards. The use of the VME bus standard allows a
`system to be made up in a very modular and flexible
`way using equipment from one or a number of sup(cid:173)
`pliers. The system can be easily expanded to include
`additional processing power, memory, communications
`or monitoring channels.
`Software. Most computers use an operating system as
`the master supervisor of their resources; memory, pro(cid:173)
`cessing time and input/output devices such as sensors,
`modems and disk drives. The operating system also
`provides an interface between the computer and the
`
`Tube to bis i
`~-····
`
`6 TC
`
`!
`
`~
`
`Cubicle end
`Battery I
`Battery V
`
`Turbine temperature
`Boost pressure
`Ambient pressure
`(inside box)
`
`:
`
`Cubicle end
`Terminal box
`+325 (loop) I
`-327
`
`1
`
`Fuel
`level
`
`3PRS Q r=
`
`L
`
`:
`
`6PRS
`6PRS
`
`Computer
`
`6PRS
`MainalternatorV ~,
`r!!!l-;;.;.;;.;;;;._-+ PC logic
`Mainalternatorl -~L---,
`~----------~~_]~~~~~~~~~., I
`74
`107
`108-V
`
`3 PRS
`6PRS
`~
`3
`P~ Ambient temperature (internal)
`
`From governor end - -- +-(
`
`Tubing 0-------i
`
`i-M,c iOOni~Rd- ···-- · ·- ---- ··· ·- ---···· ~
`; Hydraulic oil level (6W PLY FXD SKT):
`: Oil temperature (6 W PLY FXD SKT)
`; Water level (6W PLY FXD SKT)
`! Oil pressure in (manifold)
`: Oil pressure out (manifold)
`: Governor air (manifold)
`
`6WPLY
`I FXDSKT
`:
`'(}- Engine speed
`
`3 PRS
`
`Navigator aerial
`(roof cL)
`
`-----······----·--··~
`
`Navigator -a-J
`
`; Crankcase pressure
`
`~=================~:!..)
`i 5~ "!'_ -~~ ': ~-~.'?. ~~T? ..... -..... -...... F
`Fig. 3 Monitoring equipment general arrangement
`
`e IMechE 1995
`
`Proc lnstn Mech Engrs Vol 209
`
`Downloaded from
`
`pif.sagepub.com
`
` by guest on June 3, 2013
`
`5
`
`
`
`4
`
`K N FRY
`
`user, managing the basic operations of the system,
`loading and executing programs, managing input/
`output, managing a directory of files and allocating
`memory. Application programs performing data
`analysis, data storage and communications all sit on
`top of the operating system and call on it to perform
`low-level tasks.
`The operating system used in the on-board computer
`is OS-9. This is a multi-tasking operating system similar
`to UNIX but specifically developed for use in real time
`embedded systems.
`Multi-tasking means that several programs may run
`apparently simultaneously, by rapidly switching from
`one program to the next many times per second. This
`capability is used extensively for the on-board software.
`It considerably simplifies the task of managing data
`analysis, allowing for example, communications over
`the radio telephone to occur while analysing data from
`the oil system, while reading the position from the navi(cid:173)
`gator, while storing vehicle operating data to file, etc.
`Each task can be written as a separate program sim(cid:173)
`plifying development and testing, and removing any risk
`of programs interfering with each other.
`
`3.1.2 Cellular radio telephone and modem
`The computer is connected to a Racal Vodafone cellular
`radio telephone and modem. Modems for cellular radio
`require a special error correction protocol, cellular data
`link control (CDLC), which gives 100 per cent error-free
`data transmission at speeds up to 2400 bits per second.
`The modem is configured for auto-dial and auto-answer
`allowing the vehicle to call out or be called at any time.
`Development of additional hardware and a consider-
`
`able amount of software has been necessary to ensure
`reliable phone management across the two-way link.
`
`3.1.3 GPS satellite navigator
`The vehicle is equipped with a 'Navstar' XR5 GPS
`receiver which gives satellite-based positioning. Position
`information from the receiver, accurate to a mean error
`of 28 metres, is available to the on-board computer
`through a serial data link. The latitude and longitude is
`converted to eastings and northings by the computer.
`
`3.2 Data analysis
`The general structure of the data analysis performed is
`shown in Fig. 4. This structure separates the data
`analysis from the hardware allowing more generalized
`program modules to be produced.
`The program 'logger' reads all the sensors and the
`vehicle location from the satellite navigator and writes
`the data to a global buffer. The buffer holds all the mea(cid:173)
`surements made over the last minute.
`The program 'validate' takes information from the
`measurement buffer and places it in a second, similar
`buffer after :
`(a) converting it to engineering units;
`(b) adjusting it for the value of supply voltage existing
`at the time;
`(c) checking it against limits to ensure validity
`'Validate' also sets 'flags' to identify engine operating
`status and transducer and power supply faults.
`The analysis programs for individual systems then
`simply extract data from this buffer when their own cri(cid:173)
`teria for analysis are met, for example the engine has
`
`: -----+ (
`
`Logger
`
`)
`
`Data logging
`
`Data analysis
`
`( Validate
`
`)
`
`+
`~
`
`Data convened to engineering
`units, supply voltage
`compensated and checked for
`sensor faults
`
`MONBAT
`
`Current status information
`
`: Data analysis
`................. ---.- . . . . --- ..
`· Communications
`
`!
`
`44-----.llo (
`
`Status
`
`)
`
`CDLCmodem
`
`(
`
`Caller
`
`) +---
`
`Fig. 4 Structure of DEMON SMR47 software
`
`Part F: Journal of Rail and Rapid Transit
`
`IC) !MechE 1995
`
`Downloaded from
`
`pif.sagepub.com
`
` by guest on June 3, 2013
`
`6
`
`
`
`DIESEL LOCOMOTIVE RELIABILITY IMPROVEMENT BY SYSTEM MONITORING
`
`5
`
`been at full power for a set time, or is idling, or has just
`been turned off.
`If the results of the analysis require a message to be
`sent from the vehicle this is instigated by creating a file
`containing the message. This file forms the interface
`between the analysis program and the software control(cid:173)
`ling the communications link. A separate program iden(cid:173)
`tifies that a fault file has been created, establishes
`communications, via cellular telephone, with the infor(cid:173)
`mation display system and passes over the information.
`The complexity of on-board analysis varies with the
`parameter or vehicle system in question. There are a
`number of factors that lead to the task being more
`complex than is initially apparent, such as the effects of
`varying duty, the need to normalize for ambient condi(cid:173)
`tions and the requirement to provide prognostic infor(cid:173)
`mation.
`The use of out-of-limit alarms is in general too sim(cid:173)
`plistic for many parameters and will not provide the
`level of decision support needed if the system is to
`succeed. It is far more useful to know that failure is
`likely at a particular time as this then allows the con(cid:173)
`troller to make a decision based upon the remaining
`duty requirement and the best point of repair. In other
`words the primary requirement is not one of simple
`diagnostics but of prognostics.
`Prognostic estimates will always have a level of
`uncertainty. The approach that has been taken through(cid:173)
`out has been to base all time to failure estimates on 95
`per cent confidence limits. The statistics are obviously
`transparent to the information receiver who just sees an
`upper and lower time to failure. The information is
`given in absolute time so that it is not affected by any
`communication delays, it does not require the receiver
`to make mental calculations of likely failure time and it
`compares easily with operating schedules.
`
`3.2.1 System development and testing
`The first stage in the development of data analysis pro(cid:173)
`grams was the collection of data to see how the oper(cid:173)
`ation of the vehicle affected the results. It is easy to be
`'swamped' in data by condition monitoring systems, so
`to avoid this it was decided that data collection would
`be targeted specificaJly for each of the systems of inter(cid:173)
`est. Using past experience of condition monitoring data
`collection and analysis, specific data coJlection pro(cid:173)
`grams were written for each monitored system. These
`stored, at the appropriate time or engine condition and
`at an appropriate sampling and averaging rate, only the
`measurements considered relevant.
`This resulted in completely different data coJlection
`programs for each monitored system, all running
`together at the same time. The data collected was auto(cid:173)
`matically downloaded from the vehicle each night to a
`micro VAX computer network where it was archived.
`The data collected were stored in files compatible
`with standard spreadsheet programs. This meant that
`downloaded data could be placed straight into a
`spreadsheet where they could be plotted and manipu(cid:173)
`lated with ease. Spreadsheet functions and macros were
`used to experiment with analysis possibilities before
`they were finally programmed. When a suitable method
`of analysing the data was decided upon, a program was
`written for inclusion on the vehicle.
`
`~ IMechE 1995
`
`A comprehensive software test environment was
`created. The physical outputs from the sensors on the
`vehicle were replaced by a program that wrote data from
`a file into either of the data buffers in Fig. 3. The data
`file format used to seed the buffer was the same as that
`used for data storage. This gave an easily controlled
`and repeatable means of testing the response of the
`analysis programs to either previously collected real
`data or simulated fault data, designed to test the
`analysis under a particular set of circumstances.
`
`3.2.2 Coolant monitoring
`The level of water in the header tank is measured by a
`potentiometer attached to a pivoted float arm. During
`periods when the engine is off or idling the level stays
`fairly constant, provided there is no leak. However,
`when the engine is running under power, the level fluc(cid:173)
`tuates due to vehicle movement. The level also changes
`due to expansion and contraction of the water with
`temperature. As the temperature of the water is mea(cid:173)
`sured and the capacity of the system is known, the level
`can be corrected for temperature. Prognostic analysis
`predicts the time at which the tank will be empty and
`sends out a message when this time is less than two
`hours away.
`Refills need to be detected in order that predictions
`can be reset. Data have shown that sudden increases in
`level can occur normally when the engine is running,
`and the engine is not always stopped in order to refill
`the tank, so the refill detection method used depends on
`whether the engine is running at the time.
`
`3.2.3 Fuel monitoring
`is a very rare
`A locomotive running out of fuel
`occurrence so analysis simply
`involves sending a
`warning message if the level in the tank drops below ten
`percent.
`Fuel efficiency information is also calculated. The
`power output of the main generator is integrated to give
`the total energy provided and this is divided by the
`calorific value of the fuel used, over a reasonable time
`period.
`Measurement errors occur while the vehicle is in
`motion due to acceleration or deceleration. Addi(cid:173)
`tionally, because of the arrangement of the fuel supply
`system, track gradients or cant can cause the engine to
`draw fuel from only one tank while the vehicle is sta(cid:173)
`tionary. Fortunately these effects are not significant
`when measuring engine efficiency over a tank of fuel
`since the amount of fuel burnt is quite large in compari(cid:173)
`son with the errors.
`One problem with efficiencies calculated in this way is
`that the measured power output from the vehicle
`includes only that provided by the main generator.
`Therefore, power produced for auxiliary equipment
`(battery charging, electric train heating, etc.) is ignored.
`
`3.2.4 Battery monitoring
`There are two problems that occur with the battery
`system:
`(a) flat batteries due to poor battery health;
`
`Proc Jnstn Mech Engrs Vol 209
`
`Downloaded from
`
`pif.sagepub.com
`
` by guest on June 3, 2013
`
`7
`
`
`
`6
`
`K N FRY
`
`(b) flat batteries due to excessive discharge over a
`period.
`Battery health. Battery health and state of charge can
`be determined by measuring the voltage drop under
`load. The load needs to be of reasonable magnitude and
`should last for some time. For the Class 47 locomotive
`the necessary conditions are met when the elec~rically
`powered fuel, oil and water triple pump runs wtth ~he
`engine turned off. This situation occurs before engme
`start-up and also after shutdown for a sho~t time. .
`It is anticipated that battery health Wlll dete~10rate
`slowly so that occasions when insufficient runmng to
`fully dharge the battery happen (thus prohibiting the
`determination of condition) should not severely affect
`the prediction of battery maintenance need.
`The fact that battery health and state of charge cause
`the same effect on the voltage drop under load means
`that the effect of battery deterioration on voltage drop
`can be determined from laboratory tests. Tests have
`been conducted to determine the voltage drop under an
`equivalent triple pump load for a good battery at
`known states of charge. A voltage drop equivalent to a
`value of health below 70 per cent generates a fault
`message.
`.
`Discharge monitoring and ability to start engme. The
`system predicts the time when the available ene~gy will
`just meet the energy required to start the engme and
`then sends out a warning one hour before. The battery
`capacity available depends upon the amount o~ charg(cid:173)
`ing before engine shutdown, the am.ount of dtscharge
`since shutdown
`the rate of the dtscharge, and the
`battery state of 'health. The amount of energy. required
`to start the engine depends upon the engme tem(cid:173)
`perature and condition.
`As mentioned, the initial capacity of the battery can
`be determined when the engine is turned off by measur(cid:173)
`ing the voltage drop due to the triple pump load. From
`the known initial charge it is possible to calculate the
`theoretical charge at any point in time by counting out
`the ampere-hours discharged. By assuming that the
`present discharge rate will remain unchanged, the
`battery capacity at any time in the future can be pre(cid:173)
`dicted.
`The battery capacity required to start the engine at
`different temperatures is well defined as it is a value
`used in the vehicle design for battery sizing. These data
`relate only to the severest design case, where a !~co
`motive is to be started "from cold after an overmght
`stand. They do not cover the case where a vehicle has
`stood for a short time and is then required to start. In
`this case the energy required to start the engine will be
`much less and using the data above will be very conser(cid:173)
`vative, resulting in unnecessary warning messages. To
`allow for warm start cases, the temperature-start capac(cid:173)
`ity relationship has been extrapolated to higher tem(cid:173)
`peratures and the engine oil temperature is used.
`
`3.2.5 Oil monitoring
`Experience has shown that idling is the best conditio~
`under which to analyse oil system data because th1s
`condition is the most critical with respect to possible
`engine shutdowns. Oil pressure is af!'ect~d by the oil
`temperature and engine speed, comphcatmg the detec-
`
`Part F: Journal of Rail and Rapid Transit
`
`tion of low-pressure faults since a reduction in pressure
`can be caused by a reduction in engine speed or a
`reduction in oil viscosity (due to either fuel dilution or
`an increase in oil temperature).
`Neural networks can be envisaged as a 'black box'
`that can be trained to provide an output related to
`inputs. Work conducted by British Rail
`certain
`Research has shown that neural networks can be suc(cid:173)
`cessfully used to normalize oil pressure data for oil te~
`perature and the engine speed. The neural network 1s
`used to predict inlet and outlet pressures, and two new
`parameters are then produced, the difference between
`the measured and predicted inlet and outlet pressures.
`Using these new parameters a considerable increase in
`diagnostic sensitivity to genuine low-pressure fa~lts c~n
`be obtained. These parameters are used to tdentlfy
`potential low oil pressure while the engine is still
`warming up and correct for the filter pressure drop
`variation with temperature.
`A neural network was quickly and easily trained
`using data from the system in good condition. The
`network was then generated as computer code and
`embedded within the on-board analysis program for the
`oil system.
`
`3.2.6 Engine monitoring
`The engine monitoring program carries out two tasks.
`Firstly, it monitors the engine to provide a r~ord of its
`duty. Secondly, it identifies the cause of engme power
`loss due to isolation of the engine power control.
`Engine monitoring. When the engine is running, the gov(cid:173)
`ernor air pressure, effectively a measure of engine power
`demand is monitored and divided up into ten bands.
`The tot~l amount of time spent in each band is calcu(cid:173)
`lated and stored.
`Control relay monitoring. Power from the engine is con(cid:173)
`trolled by the power control relay coil. Only when 110
`V is supplied to this coil can power be provided by the
`main generator. When the engine is started the coil is
`directly supplied with 110 V. After this, when the engine
`is running normally, the 110 V is supplied via a number
`of other contacts. These contacts are controlled by the
`correct operation of the power control relay, the power
`earth fault relay, the load regulator, the auto air gover(cid:173)
`nor, the equipment governor and the co~trol governor.
`When intermittent faults on these dev1ces occur, the
`supply to the power relay coil is broken and the trac(cid:173)
`tion power is cut off. If this happens while the engin~ is
`running under power, power is cut off to the tractiOn
`motors and cannot be reapplied for at least 30 seconds.
`This ty'pe of occurrence is obviously very inconvenient
`for the driver:
`When any of these devices has an intermittent fault it
`is very difficult to detect which one of them it is,
`because once power has been lost all of the healthy
`systems will react by opening thei! contacts as we.ll. ~y
`the time the driver is able to examme the contacts tt will
`not be possible to determine which contact opened first.
`It may also be difficult for the maintenance staff to iden(cid:173)
`tify the causes of such faults.
`The analysis program monitors the voltage at each of
`the contacts, which control the supply to the power
`control relay to see which contact opens first. When the
`state of any ~f the contacts changes the new condition is
`C IMechE 1995
`
`Downloaded from
`
`pif.sagepub.com
`
` by guest on June 3, 2013
`
`8
`
`
`
`DIESEL LOCOMOTIVE RELIABILITY IMPROVEMENT BY SYSTEM MONITORING
`
`7
`
`latched in hardware just after the change. The computer
`then reads the status of the contacts and the analysis
`program decides whether the change was due to a
`normal set of circumstances, such as the driver shutting
`down the engine, or a fault on one of the interlocking
`systems.
`
`3.2.7 System self-monitoring
`In order to provide reliable information it is important
`that the monitoring system is able to identify faults with
`itself and ensure that such faults do not result in mis(cid:173)
`diagnosis. Experience has shown that the automatic
`identification of mon