`Measurement and Control
`
`http://tim.sagepub.com/
`
`
`Integrated microprocessor control of a hybrid i.c. engine/battery-electric automotive power train
`P.W. Masding and J.R. Bumby
` 1990 12: 128Transactions of the Institute of Measurement and Control
`
`DOI: 10.1177/014233129001200303
`
`The online version of this article can be found at:
`
` http://tim.sagepub.com/content/12/3/128
`
`Published by:
`
`http://www.sagepublications.com
`
`
`
`On behalf of:
`
`The Institute of Measurement and Control
`
`
`
`
`
`(cid:160)(cid:160)
`(cid:160)(cid:160)(cid:160)(cid:160)
`
`
`
`Additional services and information for can be found at:Transactions of the Institute of Measurement and Control
`
`
`
`
`http://tim.sagepub.com/cgi/alertsEmail Alerts:
`
`
`
`http://tim.sagepub.com/subscriptionsSubscriptions:
`
`
`
`Reprints:
`
`
`http://www.sagepub.com/journalsReprints.nav
`
`
`
`http://www.sagepub.com/journalsPermissions.navPermissions:
`
`
`
`Citations:
`
`
`http://tim.sagepub.com/content/12/3/128.refs.html
`
`>>
`
`Version of Record
`
`- Jan 1, 1990
`
`What is This?
`
`
`
`Page 1 of 20
`
`FORD 1909
`
`(cid:160)
`(cid:160)(cid:160)
`(cid:160)
`(cid:160)
`(cid:160)
`(cid:160)
`(cid:160)
`(cid:160)
`
`
`Integrated microprocessor control of a
`hybrid i.c. engine/battery-electric
`automotive power train
`
`by P.W. Masding, BSc, PhD and J.R. Bumby, BSc, PhD, CEng, MIEE
`
`This paper describes the development of a fully
`integrated microprocessor control system for a hybrid
`i. c. -enginelbattery-electric automotive power train.
`Torque control systems for the internal-combustion
`engine and the electric-traction motor are designed using
`digital transfer functions and indirect methods of torque
`measurement. Root-locus methods are used in all
`designs to provide fast, critically damped closed-loop
`response. In all cases simple proportional-plus-integral
`control proved sufficient to achieve this. An overall cycle
`speed controller allows the laboratory test system to be
`exercised over any test driving cycle and offers the ability
`to carry out sophisticated power sharing and transmis-
`sion shifting strategies.
`
`Keywords: Hybrid vehicles, automotive power train
`control, microprocessor control, electric vehicles, i.c.-
`engines
`
`Symbols
`
`a
`f~
`It
`
`Zero of g~(w’ )
`Counter value from flywheel speed probe
`Number of teeth on the flywheel speed probe
`gear
`Gain of g~(w’ )
`g
`g~(w’) P+I Controller in w’-plane form
`Bilinear discretisation of g~(w’ )
`g,(z)
`Armature current, A
`ia
`Field current, A
`if
`Flywheel inertia, kg m2
`J
`Constant relating dynamometer
`K
`speed to load
`Flywheel count to roadspeed conversion
`factor
`gaTj’2
`Equivalent vehicle mass, kg
`Speed, rev/min
`Inlet manifold depression, mbar
`Final drive ratio
`Vehicle wheel radius, m
`Motor torque, Nm
`Torque in gearbox output shaft, Nm
`Engine torque, Nm
`Controller design criteria, rise time, s
`Control system base sampling period (20 ms)
`
`K,
`
`ki
`Me9
`N
`PM
`r f
`rw
`Tem
`T f
`7~
`t,
`T,
`
`School of Engineering and Applied Science, University of
`Durham, South Road, Durham DH1 3LE, England
`
`128
`
`Page 2 of 20
`
`6
`0~
`()m
`ç
`a
`Wd
`
`Engine throttle position, 0.9° steps
`Demand throttle position, 0.9° steps
`Motor accelerator demand
`Controller design criteria damping factor
`Real part of closed-loop pole
`Damped frequency, rad/s
`
`1. Introduction
`
`In this paper some of the control problems encoun-
`tered in designing and operating a ’drive-by-wire’
`hybrid internal-combustion (i.c.)
`engine/battery-
`electric vehicle are examined. With two power sources
`in the drive train, considerable flexibility in design and
`control of the complete system is possible. Various
`drive train arrangements have been investigated in
`previous computer-aided-design studies (Willis and
`Radtke, 1985; Burke and Somuah, 1980) but most have
`favoured the parallel hybrid arrangement illustrated by
`Fig 1. This mechanical configuration consists of an
`i.c.-engine and an electric traction motor connected
`mechanically in parallel so that both power sources are
`capable of driving the road wheels directly. The
`advantages of such a hybrid drive system stem from its
`versatility in being able to operate in pure electric mode
`in urban areas yet retaining an i.c.-engine for high-
`speed operation and long-range capability. By correct
`design, such a drive arrangement not only has the
`potential to reduce exhaust emissions in the urban
`environment substantially, but also of substituting up
`to 70% of the petroleum fuel used by the average road
`user (Forster and Bumby, 1988; Sandberg, 1980).
`Precisely how much petroleum substitution is achieved
`depends on the individual vehicle use pattern.
`To realise the full potential of the hybrid drive,
`integrated control of both the prime movers and the
`common transmission is required. The problems asso-
`ciated with the development of such an integrated
`control system can be divided into two parts: mode
`selection; and component control. Mode selection is
`concerned with deciding whether the vehicle should
`run in an electric mode, an i.c.-engine mode or whether
`the i.c.-engine and the electric motor should provide
`propulsion torque together. Selecting which of the
`many possible operational modes to use under given
`operating conditions is a complex problem and inter-
`acts strongly with the basic design of the hybrid power
`train. An optimisation study of these problems based
`on a computer simulation of different hybrid-vehicle
`
`FORD 1909
`
`
`
`Fig 1
`
`Parallel hybrid electric vehicle drive train
`
`power-train configurations, component ratings and
`control strategies is discussed in some detail in
`Bumby and Forster (1987). The end result of the
`optimisation process is a mode controller which
`receives, as input, the driver’s brake and accelerator
`signals and then adjusts the torque demand to the
`engine and motor to meet the total demand. In addi-
`tion, it controls the gearbox, since selection of the
`correct gear ratio, to define engine/motor speed, has
`a critical effect on their efficiencies.
`Once the mode controller has decided on a gear
`ratio and torque demand to be met by each of the
`prime movers, it is necessary then to design individual
`components controllers which operate the engine,
`motor and gearbox so that they meet the appropriate
`demand as quickly as possible and, when necessary,
`also allow smooth transition between modes. Earlier
`work has examined the control problems relating to
`automation of discrete ratio transmission units
`(Masding et al., 1988). Conventional discrete ratio
`transmissions are ideal for this purpose since they offer
`the highest efficiency of any transmission system, and
`for this reason automation of such transmissions is
`attracting considerable attention (Main et al, 1987;
`Busca et al, 1979).
`In this paper the additional component control
`problems relating to engine and motor torque control
`and smooth engine starting are addressed. One earlier
`system which tackled these problems was built in the
`USA by The General Electric Co. during the Electric
`and Hybrid Vehicle initiative and resulted in a
`microprocessor-controlled prototype
`hybrid
`car
`(Trummel and Burke, 1983; Somuah et al, 1983). On
`the basis of preliminary design studies, this vehicle
`used an i.c.-engine and an electric traction motor con-
`nected mechanically in parallel. Control of the electric
`traction system was achieved by a chopper in the field
`
`Page 3 of 20
`
`circuit and series/parallel battery switching to vary the
`armature voltage. Starting resistances and clutch slip
`were thus necessary to move the vehicle from rest. In
`the present work, power electronic armature and field
`choppers are used to give smooth, efficient motor per-
`formance over the whole operating range and to
`remove the need for a clutch system. This same electric
`drive system has been successfully used in an opera-
`tional all-electric van produced by Lucas Chloride and
`Bedford (Manghan and Edwards, 1983). Bose et al
`(1984) describe the control methods adopted in the
`HTV-1 but concentrate solely on transfer functions
`developed in the s-domain. In contrast, the control
`systems for the engine and motor presented in this
`paper make extensive use of digital models which have
`been previously developed to describe their dynamic
`characteristics (Masding and Bumby, 1990a; 1990b).
`Satisfactory performance of the completed con-
`trollers is confirmed by using an extensive laboratory
`test facility. The test facility is a full-scale version of
`a parallel hybrid drive train using a 35 kW i.c.-engine
`and a 32kW D.C. traction motor as prime movers.
`Both the engine and motor are coupled to a 4-speed
`synchromesh gearbox via a toothed drive belt. To the
`rear of the gearbox a flywheel-and-dynamometer com-
`bination provide a simulation of the loadings due to
`vehicle inertia and aerodynamic/tyre drag, respectively.
`Control of the laboratory system is carried out by an
`M68000 microprocessor system which is responsible for
`receiving data from the extensive range of transducers
`round the rig and responding with appropriate control
`signals to the throttle servosystem, power electronics
`and gearbox. This system allows the control algorithms
`developed in this paper to be fully tested under opera-
`tional conditions as well as in simulation. A complete
`description of the test bed facility is given in Bumby
`and Masding (1988).
`
`129
`
`FORD 1909
`
`
`
`TABLE 1: Possible operating modes for the parallel hybrid vehicle
`
`The final two modes are the regenerative braking
`mode and the accelerator ’kick down’ mode. The latter
`provides the driver with full power from both the
`engine and the traction motor and is intended mainly
`for use in emergency conditions when all economy
`considerations are overridden. Finally, regenerative
`braking is used whenever the vehicle is braked, in order
`to recover some of the kinetic energy of the vehicle and
`return it to the batteries. Having the motor connected
`to the drive train permanently means that regenerative
`braking is always immediately available.
`All the above operating modes pose common control
`problems in that, after a particular mode has been
`chosen, it must be possible to schedule and control
`the torque output of both the engine and motor. In
`addition, to provide smooth transition between modes
`it is necessary to start and synchronise the engine
`with a moving drive train accurately. Torque schedul-
`ing is the responsibility of the overall vehicle-mode
`controller on the basis of a strategy arising from the
`optimisation study mentioned earlier; however, in this
`paper the secondary problem of individual component
`control to achieve the desired torques and to start
`the engine is addressed.
`
`3. Controller design
`Experience has shown that robust controllers suit-
`able for all the applications in the hybrid vehicle can
`be produced using proportional-plus-integral control.
`Such controllers can give satisfactory performance not
`only for torque control of both prime movers but,
`in addition, for engine speed on no-load and overall
`speed control through a cycle. An advantage of these
`low-order controllers is their speed of execution: during
`a typical driving cycle the main computer takes only
`3 ms to carry out cycle speed and prime-mover torque-
`control calculations. High speed and accurate compu-
`tation is encouraged by the use of 32-bit integer
`arithmetic throughout. All controller design is carried
`out in the w’-plane using root-locus pole placement
`methods. Z-transfer functions are mapped into this
`plane by the transformation pair
`~- 2 Z-1
`w = T, (z - + 1)
`~&dquo;
`T s ‘z + 1/
`
`,
`
`2 (z -1)
`
`
`
`...(1)...(1)
`
`...(2)
`
`_ w ’ + 2/T~
`z = w’ -21Ts
`where T, is the sampling period. Owing to the similarity
`between the s and w-planes, the proportional-plus-
`integral controller retains its usual form:
`w’+a
`
`...(3)
`
`, w ’W +a
`g~(w~)-gB w’ l
`Acceptable closed-loop performance is defined in terms
`of the rise time, t&dquo; and the damping factor 6. These
`are defined for a second-order system by the equations:
`w’ _ ~ . j wd
`... (4)
`cos-’(-~)
`wd = cos 1 tr (-~)
`*r
`
`...
`
`...(5)
`
`2. Hybrid-vehicle control modes
`The different operating modes available with a
`hybrid drive system are summarised in Table 1. In
`general, the electric mode can be used in urban areas,
`for short journeys and when the engine load would be
`small giving rise to low engine efficiency. It is always
`used for moving the vehicle away from rest, since a
`conventional clutch system is not included. When the
`drive-train speed exceeds 1000 rev/min, the engine can
`be started and synchronised with the moving drive
`train to provide additional power if required. Such
`operation is possible owing to a free-wheel unit in the
`engine drive line which allows the engine to remain
`stationary when the rest of the drive train is in motion.
`Primary i.c.-engine mode is used when vehicle speed
`and loading are both high, which gives high engine
`efficiency. When necessary, the engine torque can be
`augmented by the motor for rapid acceleration or hill
`climbing. Typically, the motor will be used to provide
`extra power if the engine output would otherwise
`exceed 90% of maximum, since this leads to ineffi-
`ciency. Over journeys with an exceptionally large
`amount of acceleration or hill climbing, the battery
`state of charge may become very low, but this can not
`be allowed to continue until the batteries are com-
`pletely depleted, since the vehicle would then be
`unable to move away from rest. To counter this
`problem, a negative torque may be scheduled from
`the motor so that the engine both drives the wheels
`and charges the traction batteries. As discussed in
`Bumby and Forster (1987), this mode is necessary but
`has low overall efficiency and so should be avoided if at
`all possible.
`
`130
`
`Page 4 of 20
`
`FORD 1909
`
`
`
`Fig 2 Engine speed control block diagram
`
`Wd
`
`tan (cos 1 (-~))
`
`
`
`(6)... (6)
`
`By choosing a suitable rise time for a specific controller
`and adopting ~ = 0.707 for critical damping in all
`cases, the position of the required closed-loop poles
`is defined. These pole locations can only be used as
`an initial guide, however, because in reality the plant
`does not produce a second-order closed-loop system.
`Fine tuning of the controller design is achieved by an
`iterative process. Eqn (3) can be transformed back into
`the z-plane by the reverse mapping of Eqn (1) to give
`g~(Z) _ (g+k;)z+(k;-g)
`z-1
`where kl = gaT,12. From this equation comes the
`discrete direct realisation for the controller output Uk
`... (8)
`uk = Uk-1 + ( g + ki) ek + (ki - g) ek-1
`When referring to plant transfer functions for the
`control of both engine and motor torque, the coeffi-
`cients that are quoted apply to the following general
`discrete transfer function
`y(Z~ -IJO+biZ 1+... -I-tJmZ’n
`1-Q1Z 1- ... -ClnZ n
`u(z)
`
`
`
`(9).
`
`
`
`. (7)-(7)
`
`4. Engine starting and speed synchronisation
`Whenever the hybrid vehicle is operating in an
`all-electric mode or is stationary, the i.c.-engine
`will be uncoupled from the drive train by means of the
`one-way clutch. Since in either of these situations the
`engine is not required to provide torque, the most
`obvious strategy is to shut it down entirely in order
`to conserve petroleum fuel. Adopting this strategy
`means that the next time the engine is required it must
`be started and synchronised with the moving, and pos-
`sibly accelerating, drive train, before it can replace or
`augment the torque supplied by the electric traction
`system. Consequently, a starting system is required
`which has fast response and no tendency to overshoot
`the prevailing drive-train speed, thus avoiding a shock
`torque in the drive shaft as the one-way clutch is
`engaged. Design of such a control system uses the
`transfer function relating throttle position to engine
`
`Page 5 of 20
`
`speed identified in Masding and Bumby (1990a) and
`repeated as
`A N(z) = 0, 838 - 1 ,5 10 z~~ ~ + 1 ,922 ~ z~~
`0 N(z) _ 0.838-1.510~+1.922~
`1 -1.790 z-~+0.795~
`A 0(z)
`
`(10),..(10)...(10)
`
`=
`
`
`
`
`
`When this is connected to the required control algor-
`ithm and throttle servo-system, the block diagram of
`Fig 2 is produced.
`For large changes in throttle demand - that is,
`greater than four steps per sample period - the throttle-
`position transfer function, GT(z), is non-linear as
`explained in Masding and Bumby (1990a). However,
`for design purposes, small variations in throttle demand
`are assumed when GT(z) reduces to 1/z producing
`a linear system which can be transformed to the w’
`plane for controller design. In order to produce an
`acceptably short synchronisation time for the engine,
`a system rise time of tr = 0.5 s and critical damping are
`chosen as the design criteria. By Eqns (4)-(6) this
`suggests closed-loop poles w’ = 4.71 ± j4.71. Fig 3
`shows the compensated system root locus with the
`controller
`
`g~(w’) = 0,012(~&dquo;’ w’ + 1.1) I
`
`B
`
`/w
`
`/
`
`...(11)
`
`With this controller the presence of the closed-loop
`
`Fig 3 Compensated root locus for control of engine speed
`on no-load
`
`131
`
`FORD 1909
`
`
`
`pole on the real axis modifies the system response so
`that the performance criteria are not achieved with
`exactly the calculated imaginary pole locations given
`above. The simulated and experimental closed-loop
`response of the system being shown in Fig 4. On this
`diagram the experimental throttle trace shows the
`step-rate non-linearity which was not included in the
`design. This is a consequence of the high gain needed to
`meet the fast system-response requirement, and the
`large errors present at the beginning of the step
`demand. Actual engine response is delayed by the
`throttle-step-rate limitation causing it to lag behind the
`simulation; however, satisfactory damping is retained.
`
`4.1 Engine starting and load transfer
`
`When required, the warm engine will fire in, typic-
`ally, 250ms using the conventional electric starter
`motor, but there is a further delay while the engine
`accelerates up to the drive-train speed. Inertia starting
`used in the HTV-1 project (Trummel and Burke, 1983)
`allowed the engine to be completely coupled into the
`drive train in 300 ms, but the cost was the need for an
`additional clutch between the engine and the engine
`flywheel. A time analysis of the starting process is
`shown in Fig 5. In this experiment the motor was
`
`Fig 4 Step test for the engine speed
`control system
`
`initially accelerating under load, as illustrated by
`the motor speed and torque traces. At time t = 0.45 s,
`however, the computer receives the start command:
`immediately it turns on the ignition and engages the
`starter motor. At the same time the throttle is
`opened 9° and the computer then waits for the engine
`to fire. This is adjudged to happen when the engine
`speed passes 490 rev/min. Above this speed the starter
`motor is turned off and the speed control algorithm is
`entered to run the engine up to the drive-train speed.
`Synchronisation is deemed complete when the engine
`speed is within 45 rev/min of the drive-train speed
`which in this case is achieved within 0.7 s of the original
`command to start. At this stage, torque control is
`transferred to the engine which continues to accelerate
`the load. Total times for starting, speed synchronisa-
`tion and transfer of load are consistently about 1 s, as
`demonstrated by Fig 5. Starting a cold, and perhaps
`damp, engine is still an unreliable feature of modem
`cars; consequently the software must be ready to cope
`with failure to start. In the event of the engine failing to
`start after 5 s, the starter motor is disengaged, to allow
`battery recovery, before a second attempt is made.
`With a very hot engine, better starting is often achieved
`with full throttle opening and this might be a useful
`strategy for the computer to adopt on the second
`
`132
`
`Page 6 of 20
`
`Fig 5 Analysis of the engine starting
`and load transfer process
`
`FORD 1909
`
`
`
`attempt at starting if the engine had been operated very
`recently.
`In this work a warm engine has always been
`considered. The reasons for this are twofold: first the
`feasibility of the control algorithms were of principal
`importance and, if these would not work on a warm
`engine then cold-engine work is unnecessary; second
`much work on engine-management systems, including
`engine starting, is being carried out by the motor
`companies (Meyer et al, 1983).
`
`4.2 Fuel saving
`Although stopping the engine does save fuel, the cost
`is some delay in the availability of engine torque,
`however small. In addition, although having a station-
`ary engine saves fuel if the time between successive
`starts is sufficiently long, some fuel penalty must be
`associated with the starting process, making very short
`shutdown periods uneconomic. Potential fuel saving
`through stopping the engine under idling conditions
`prompted Volkswagen to incorporate this feature in its
`Formel E range of cars, after which fuel savings of up
`to 30% over the otherwise conventional i.e.-engine
`vehicle have been reported (Schmidt, 1981).
`
`5. Prime-mover torque control
`5.1 Engine torque control
`The engine torque controller is responsible for
`producing from the engine the torque demand
`requested by the main vehicle controller. Torque is not
`measured directly for control purposes, but is calcu-
`lated from measurements of speed and inlet manifold
`depression. This step is necessary because the strain-
`gauge transducers, used to measure torque directly, are
`too expensive and unreliable for use in an operational
`vehicle. During the theoretical analysis of the engine,
`carried out in previous work (Masding and Bumby,
`1990a), it was established that only one dynamic
`element is needed to describe the way that engine
`torque behaves when the input throttle angle is
`changed. This element, which is known as the manifold
`filling delay, can be described by a first-order digital
`transfer function which has speed-dependent coeffi-
`
`TABLE 2: Identified transfer-function coefficients and gain for the
`engine manifold filling delay
`
`cients. Identification experiments allowed the values of
`these coefficients to be determined at a number of
`engine operating speeds as set out in Table 2.
`As a consequence of this speed dependence, it might
`appear necessary to design a series of torque con-
`trollers, each valid over a small speed range, so that
`acceptable system performance is maintained at all
`times. Fortunately, the changes in engine gain and
`dynamics represented in Table 2 are not particularly
`great, and so it is possible to design one controller using
`the transfer function for 2000 rev/min which achieves
`good results at all speeds. In general, the linearised
`torque-control system appears as shown in Fig 6. Once
`the appropriate gains and transfer-function coefficients
`have been added for operation at 2000 rev/min, the
`uncompensated root locus of the system, appears as in
`Fig 7a. This locus has two open-loop poles, one due to
`the manifold filling delay and one nearer the origin due
`to a digital filter used to process the manifold-pressure-
`transducer reading. In designing the controller the aim
`is to achieve a fast, critically damped response; suitable
`design parameters for the engine are: rise time, t, = 0.2;
`and e = 0.707. By substituting these values into Eqns
`(4)-(6) the approximate pole locations are found to be
`~’ = ―11.78±y11.78. Using these locations as a guide
`the controller
`
`wl+7
`
`g~(w’)=0.4Cw +71
`gc(w/) = 0.4 -;;-
`was selected, which produces the modified root locus of
`Fig 7b. To test the completed system a step increase in
`demand of 10 Nm was applied with the results as shown
`in Fig 8a. As shown, the real system responded almost
`exactly as the simulation suggests it should. Away from
`
`
`
`...(12)...(12)
`
`Fig 6 Block diagram of the linearised engine model
`
`133
`
`Page 7 of 20
`
`FORD 1909
`
`
`
`Fig 7a Uncompensated locus for control of engine torque at
`2000 rev/min
`
`Fig 7b Compensated locus for control of engine torque at
`2000 rev/min
`
`Fig 8a Simulated and experimental
`performance of the engine torque
`control system at 2000 rev/min
`
`Fig 8b Simulated and experimental
`performance of the engine torque
`control system at 3000 rev/min
`
`the design speed of 2000 rev/min the gain and dynamics
`of the manifold-filling delay vary, causing some degre-
`dation of controller performance. At 3000 rev/min, the
`reduced manifold-filling gain increases system rise time
`as illustrated by Fig 8b.
`A small change in demand of 10 Nm was chosen for
`initial testing so that the non-linear effects of throttle
`step rate would not affect the system. In practice the
`
`control system must be able to cope with larger changes
`in demand. An example of when such a change does
`occur is the transition between acceleration and cruise
`in a driving cycle. Fig 9 shows that for a sudden 25 Nm
`drop in demand, which is representative of such a
`transition, the system response does not deteriorate
`significantly. In this diagram the torque-transducer
`trace is included to illustrate the accuracy of the derived
`
`134
`
`Page 8 of 20
`
`FORD 1909
`
`
`
`Fig 9 Engine torque control system:
`comparison of indirect and direct torque
`measurements during a large step
`disturbance
`
`torque measurement based on manifold pressure and
`speed. During the transient interval, direct comparison
`of the two torque readings is difficult since the torque
`transducer requires heavy filtering to remove noise
`effects and therefore demonstrates a delayed response.
`
`5.2 Electric motor torque control
`As explained in Masding and Bumby (1990b), the
`Lucas Chloride power electronics controller uses three
`control modes to cover the complete speed and loading
`range of the electric traction motor. The control unit
`contains both field and armature chopper circuits which
`allow it to provide full closed-loop control of both field
`and armature current. Each of the three control modes
`is characterised by the way the field current is
`controlled in response to changing brake or accelerator
`signals. In brief, relatively low loads and speeds give
`rise to the field-boost mode, whereby field current
`increases rapidly with accelerator demand. Once full
`
`rated field current is reached, then the full-field mode
`ensues with all control achieved via the armature. Full-
`field mode gives way to field-weakening mode at
`relatively high speeds. At speeds above 2000 rev/min
`there is a direct transition from field-boost mode to
`field-weakening mode, with rated field current not
`being reached.
`Each of the control modes results in a different
`transfer function relating accelerator demand, 0~ to
`torque output, Tem. In addition the gain and dynamics
`of the field-boost and field-weakening modes vary with
`the operating point defined in terms of initial speed,
`current values and accelerator setting.
`Fortunately, these variations in the gain and
`dynamics of the field-boost and field-weakening modes
`are not particularly great; hence it is not necessary to
`solve the problem of designing a controller which
`continuously adapts to the changing system. In fact it is
`possible to design a single fixed controller for each
`mode which maintains adequate performance over the
`
`Fg 10 Motor torque control system
`
`Page 9 of 20
`
`135
`
`FORD 1909
`
`
`
`whole range of gains that are encountered. A simple
`software algorithm allows the computer to distinguish
`between the operating modes, and so the appropriate
`controller gains can be selected.
`The block diagram for the electric-motor torque-
`control loop is shown in Fig 10. As with the engine,
`indirect torque measurements, this time based on field
`and armature current, are used for control purposes. In
`this instance the pole-placement method is used to
`design a controller giving a critically damped response
`and a rise time of 150 ms. Three controllers are
`produced, each tuned to a transfer function representa-
`tive of an individual mode. Table 3 shows the three sets
`
`TABLE 3: Electric-motor transfer-function coefficients and torque-
`control parameters for all three operating modes
`
`Fig 11 a Uncompensated locus for motor torque control in
`the field-boost mode
`
`nrai w
`
`Fig 11 b Compensated locus for motor torque control in the
`field-boost mode
`
`nrai w
`
`136
`
`Page 10 of 20
`
`of model parameters and the corresponding controllers
`in both the w’-plane and the final z-plane form after
`bilinear discretisation. Part of the design is illustrated
`by Figs lla and llb which show the uncompensated
`and compensated root-loci for the field-boost transfer
`function.
`All three pole-placement designs have been tested by
`applying a step change of demand of 10 Nm to the
`system. A full-scale simulation is carried out simulta-
`neously with the step test in each case. Initial tests were
`carried out under similar operating conditions to those
`used to obtain the identification results given in Table
`3. As illustrated for all three modes by Figs 12a,
`12b and 12c, these tests produced good closed-loop
`response which agree closely with the simulations.
`Additional tests are also needed to confirm that the
`variations in motor characteristics that occur in the
`field-boost and field-weakening modes do not lead to
`unsatisfactory control. Fig 13a illustrates one such test
`for the field-boost mode, carried out well away from
`the original design conditions, where a decrease in
`system gain has caused the rise time to be longer than
`predicted by the simulation. A similar test for the field-
`weakening mode is illustrated by Fig 13b; in this case
`changes in system dynamics produce a slightly oscilla-
`tory response. There is no absolute law which can
`decide whether or not these results still represent
`acceptable system performance, but in no instance is
`system stability called into question nor is there a large
`departure from designed rise time.
`
`5.2.1 Determination of electric-motor operating mode
`Isolated step tests carried out entirely within one
`mode are not representative of working conditions for
`the torque controllers. Under normal conditions the
`motor will pass regularly between modes and so, in
`response, the torque controller must switch between
`parameters. There is no definitive signal from the
`power electronics to indicate the operating mode of the
`motor; consequently, a mode determination algorithm
`has been written into the software using measurable
`signals. As shown on the block diagram of Fig 10, the
`necessary inputs are speed, field current and acceler-
`ator demand. To avoid rapid switching between con-
`troller gains, a time constraint only allows a new mode
`to be selected after every ten sampling periods
`(200 ms). The mode-determination rules were tested by
`instructing the motor to follow an arbitrary torque
`profile and allowing natural speed variations to occur.
`As can be seen from Fig 14, the experiment succeeded
`in making the motor pass through all of its operating
`modes but, more importantly, smooth control was
`achieved at the transition points.
`When designing the mode-determination rules, due
`account was taken of the consequences of temporarily
`selecting the wrong mode and hence using the wrong
`controller parameters. Although most errors of this
`type would merely cause slight deterioration of con-
`troller performance it is important that the relatively
`high field-weakening controller gains are never used
`erroneously in the field-boost mode when plant gain,
`too, is relatively high.
`Tests show that should this condition arise the system
`is liable to go unstable whereas, should the reverse
`mistake be made, the use of low field-boost control
`
`FORD 1909
`
`
`
`Fig 12a Motor torque control test in
`the field-boost mode at design
`conditions
`
`Fig 12b Motor torque control test in
`the full-field mode
`
`Fig 12c Motor torque control test in
`the field-weakening mode at design
`conditions
`
`Page 11 of 20
`
`137
`
`FORD 1909
`
`
`
`Fig 13a Motor torque control test in
`the field-boost mode away from design
`conditions
`
`Fig 13b Motor torque control test in
`the field-weakening mode away from
`design conditions
`
`Fig 14 Motor torque control
`demonstrating satisfactory transition
`between operating modes
`
`138
`
`Page 12 of 20
`
`FORD 1909
`
`
`
`Fig 15 Motor torque control: effect of
`using field-weakening gains in the
`field-boost mode
`
`gains merely results in an extended rise time for the
`field-weakening mode. In Fig 15 the field-weakening
`gains were deliberately used under operating condi-
`tions similar to the design state for the field-boost
`controller. Although this particular step test does not
`illustrate unstable behaviour, overshoot is unaccept-
`able and there is evidence of amplification of noise in
`the controlled signal. Since this experiment took place
`near the design conditions for the field-boost mode, it
`should be possible to predict these oscillations by using
`the field-boost model in a simulation. When the field-
`boost model is combined. with the field-weakening
`controller gains, considerable oscillation is predicted as
`shown by Fig 15, but it is much less than that observed
`in experiment. Although this might, in