Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

~.A Wi!
The Engineering
~ Resource For
~ AdvancingMobility 400 COMMONWEALTH DRIVE WARRENDALE,PA 15096

MO~
~#

850174

A Digital Control Algorithm

for Diesel Engine Governing
Daniel C. Garvey
Woodward Governor Co,
Engine and Turbine Controls Division
Fort Collins, CO

International Congress
& Exposition
Detroit, Michigan
February 25 - March 1, 1985
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

The appearance of the code at the bottom of the first page of this paper indicates
SAE's consent that copies of the paper may be made for persona! or internt:1 use, or
for the personal or inlernal use of specific clients. This consent is given on the con~
dition, however, that the copier pay the stated per article copy fee through the
Copyright Clearance Center, Inc., Operations Center, P.O. Box 765, Schenectady,
N.Y. 12301, for copying beyond that permitted by Sections 107 or 108 of the U.S.
Copyright Law. This consent does not extend to other kinds of copying such as
copying for general distribution, for advertising or promotional purposes, for
creating new collective works, or for resale.

Papers published prior to 1978 may also be copied at a per paper fee of $2.50
under the above stated conditions.

SAE routinely stocks printed papers for a period of three years following date of
publication. Direct your orders to SAE Order Department.

To obtain quantily reprint rates, permission to reprint a technical paper or per~

mission to use copyrighted SAE publications in other works. contact the SAE
Publications Division.

ISSNOI48-7191
Copyright © 1985 Society of Automotive Engineers,lnc.

This paper is subject to revision. Statements and opinions ad·
vanced in papcrs or discussion arc the author's and are his
responsibility. not SAE's; however. the paper has been edited
by SAE for uniform styling and format. Discussion will be
printed with the paper if it is published in SAl:: Transactions.
ror pennL~sion to publish this paper in full or in part, contact
the SAl: Publications Division.
Persons wishing to submit papers to be considered' for pre·
sentation or publication through SAl:: should send the manu·
script or'a 300 word abstract of a proposed manuscript to:
Secretary. En!c!inecring Activity Board, SAE.
Printed in U.S.A.
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

850174

A Digital Control Algorithm

for Diesel Engine Governing
Daniel C. Garvey
Woodward Governor Co.
Engine and Turbine Controls Division
Fort Collins, CO

ABSTRACT necessary to implement the algorithm are identified in terms of

The performance of a microprocessor based precision
engine speed control system was investigated. A sample rate
selection criteria is presented along with a procedure to
implement a high performance digital PID control algorithm.
Thealgorithm requires a digital speed sensorof 12 to 14 bits to
minimize excessive fuel rack motion at steady state due to
digital quantization effects. Computer simulation and
experimental test results of the algorithm are presented for an
1800 RPM, 125 Kilowatt engine generator set.
INTRODUCTION
an operating range and number of bits. Computer simulation
and experimental engine test results obtained with the digital
controller are presented.
ANALYSIS
A block diagram descriptive of the engine control system
is shown in Figure 1. To aid in the design and evaluation of
different digital control algorithms, a sample rate selection
criteria was determined and a mathematical model descriptive
of the system response was developed. Analytical expressions
describing each of the blocks shown in Figure 1 are developed
in the mathematical model. Mathematical symbols are identi~
fied in the Nomenclature tabulation:

The use of microprocessors (IotP) in industrial control z

systems has become very popular for several reasons. A ,uP ,----... FUEL RACK
ACTUATOR
DIESEL
ENGINE r--- n
can reduce the number of circuit components and cost in a
design and increases system reliability and control system
design flexibility, When a pP based control system is applied to
an industrial diesel engine propulsion system or electric power
generation system, several performance requirements must I
be satisfied. The control system transient response must be
DIA CLOCK AID
rapid to minimize engine speed deviations due to sudden CURRENT SAMPLE SPEED
changes in load. During steady state operation, the control DRIVER RATE SENSOR
system must maintain precise control of speed with a minimum
of fuel rack motion. These requirements must be satisfied in
the presence of engine torsional oscillations in speed which
t F = lIT t
can be severe.
These high standards impose constraints upon the design DIGITAL
requirements of a J,tP based control system. A digital control B.. CONTROLLER B.
system which meets these standards requires the selection of

- a IJP sample rate FIGURE 1 DIESEL ENGINE DIGITAL SPEED CONTROL SYSTEM,
- a control algorithm
- a precision speed sensor MICROPROCESSOR SAMPLE RATE SELECTION - A
- a ,uP computation method measure of the performance of an analog control system is
- a particular J,tP given by mspection of the open loop transfer function

This paper presents a sample rate selection criteria and Gis) • His) (1)
describes the design of a J.tP based PIO control algorithm
which yields rapid transient response and satisfies the steady to obtain measures of crossover frequency (Fa) and phase
state requirements of precision speed control with a minimum margin (0). In a high performance regulator loop, a phase
fuel rack motion. The requirements of the speed sensor margin of

0148,7191/85/0225,0174$02.50
Copyright 1985 Society of Automotive Engineers, Inc.
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

2 850174

o • 30 (degrees) Let
ilF
is desirable at a high open loop crossover frequency (Fa). a •
Values of phase margin much less than 30 degrees are F"
undesireable and cause stability problems in the control
system. A 0 degree phase margin demonstrates instability. the phase shift at crossover frequency is given by

180 F- (1-al
C(.) ilB • (Degreel (51
F

The ratio of sample frequency to open loop crossover

frequency is given by

FIGURE 2. SAMPLING IN AN ANALOG CONTROL F 180 (1-a) (61

F- ilB
The effect of sampling on this system is evaluated by
100
considering an equivalent digital system shown in Figure 2. F" SAMPLE FREQUENCY; Fo" OPEN LOOP CROSSOVER FReQUENCY
(1)- Time domain plots of a signal and its sampled and II " 0:.1:.2:.3 REDUCTIONS IN FO (HERTZ)
90
clamped resultant are shown in Figure 3. The boxcar output of
the zero order hold follows the input curve delayed by T/2
so
seconds, with T the sample period. The open loop transfer
function with sampling is given by
70

G(S) • HIS) e -(T/2IS (21

60
U.
where ~
0
~ 50
~
T ,;l sample time
'" 40
F =: 1fT ~ sample frequency
"
20

10

0
0 2 4 6 8 10 12 14 16 18 20 a".3

DELTA OPEN LOOP PHASE MARGIN

o T 2'1" 3'1' 4'1' 5T FIGURE 4

TIME RATIO (F/F") AS A FUNCTION OF PHASE MARGIN LOSS
AND REDUCTION IN CROSSOVER FREQUENCY (1-a).
FIGURE 3. SKETCH SHOWING THE DELAY OF "1"/2
SEC INTRODUCED BY THE ZERO-
ORDER HOLD. This equation is plotted in Figure 4 as a function of

- the reduction in phase margin liB

In an open loop Bode plot, the additional phase shift at
crossover frequency (Fa) due to sampling is given by - the reduction in the open loop crossover
frequency (1~a)
ilB • FaT 180 (degree) (3)
A small high speed diesel engine is characterized by an open
This phase shift is a reduction in phase margin at the crossover loop crossover frequency of
frequency Fa. If the open loop crossover frequency is reduced
by AF to improve the phase margin in the presence of F- • 4.0 Hertz
sampling, then
Allowing a minor phase margin reduction of

Fa . F" (1 - ~)
F-
(41
AO = 5°

and no reduction in bandwidth, a =0

"Numbers in parentheses designate references at the end of L . 36

the paper. F"
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

850174 3

and DIESEL ENGINE - The steady state performance of a

typical diesel engine is shown in Figure6.ln this figure, engine
F 144 Hertz Sample Frequency power is plotted versus engine speed at several different
positions of the fuel rack. The engine speed governor forces
or operation along the dashed lines for either droop or isochro-
nous speed control. Droop operation is required during parallel
T = 1/F = .0069 seconds operation of two mechanical drive units or with the commercial
utility. Speed droop requirements of 5 to 10 percent establish
the required ,u.P sample time for a small high speed diesel the minimum range necessary for the governor speed sensor.
engine. If a sample time is selected for a small high speed The dynamic response of a diesel engine is described by a
engine,T = .016 second then normalized torque balance on the engine flywheel.

L = 15.625 t - L :; ~ (8)
F' y

and from Figure 4, the phase margin is reduced by 12" at a = O. with engine acceleration coefficient
This is a major reduction in phase margin and yields a ringing
type of response.
y
In a large bore, low speed marine propulsion engine

0.2 " F' " 0.6 Hertz The engine developed torque is proportional to fuel rack
position but is delayed by engine dead time
and is a function of the engine speed range. With F/F- = 36
then t :; z e~T,S (91
21.74 " F " 7.25 Hertz The engine dead time is characterized by
for a,uP sample rate on a low speed marine engine.

These results demonstrate that in a high performance digital

control system it is desireable to satisfy the design criteria
where

(Sample Frequency) < 36 NA :; engine RPM

15" i.- (7)
Q :; number of cylinders firing per revolution
F' {Open Loop Crossover Frequency}

to minimize control system loss in phase margin or reduction in The position of the fuel rack is proportional to the applied
crossover frequency. Figure 5 depicts a plot of the required current but delayed by the electrohydraulic actuator time
digital control samplfng frequency (F) versus control loop constant (T~)
crossover frequency (F*) satisfying the above constraint. __
1_ (101
z = T,s+1

Representative values of engine and fuel system constants

for a small high speed engine are given by T:; .51, T, :; .015
and T2 :; .08.

ENGINE
-+- ~ 15 POWER
I ,0

100%~\
ISO[HRONOU~
/I \ ./ DROOP

50%
~-----!---\--
! \,
2S%~ I \
Y-+-+-+-t-l-+-+-+-.J--+.,. N
, ••.• 0."". • , • • 0 fO'
ENGINE SPEED 1,0 I +P
F' CONTROL LOOP CROSSOVER FREQUENCY (HERTZ)

FIGURE 5. MICRO PROCESSOR SAMPLE FREQUENCY (F) FIGURE 6: DIESEL ENGINE

AS A FUNCTION OF OPEN LOOP CROSSOVER PERFORMANCE MAP WITH
FREQUENCY (F'). SPEED GOVERNOR.
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

4 850174

SPEED SENSOR - A magnetic pickup or proximity The measured value of the speed must be scaled to the
sensor is commonly used to obtain a measure of engine speed. accuracy of the computational algorithm
The magnetic pickup output is a sine wave at a frequency
proportional to engine speed. This frequency is distorted by Bo = 2M - N 0 B 1141
engine torsional oscillations at crankshaft speed and firing
frequency. A frequency to voltage converter with an analog to
digital convertor provides a digital measurement of speed. The CONTROL ALGORITHM SELECTION - The ~P control
measurement accuracy of a digital speed sensor is proportional algorithm selected follows the format of an analog PI or ?:D
to the operating speed range of the engine and the digital word controller. This algorithm was selected because of its familiarity,
leng~h. The output of an ideal digital speed sensor is given by applicability to the majority of industrial control problems and

B "" Integer -n-n

-'-
[ ( "2- 0 ,
)0 N
2 ] 111 i proven results, Several different forms of the PlD control
algorithm exist. For tuning and analytical purposes. a non-
interacting form of the PID algorithm in the frequency domain
is deSIrable. (2) The PID control algorithm selected is described
where
by a speed sensor,
n actual speed per unit
", minimum speed per unit V=Kon (15)
"2 :;:: maximum speed per unit
and N:::: digital word length and the control
i = I 0 (Ils+ 1) 0 (os+ 1) _V_] (16)
The per unit gain of the digital speed sensor is given by S (ps+l) ps + 1

AB The product (10K) represents per unit gain on a Bode diagram,

0 _1_ = Ko 1121
1/8 and 1//3 are adjustable zero locations and 1/p characterizes
2N iln
the roll off filter pole location,
with
An equivalent digital controller algorithm can be obtained by
_1_ 1131 using one of several different methods to convert a continuous
Ko = n2-n , transfer function to its discrete form. (3) A simple method is
given by the backward difference approach, which substitutes
The minimum per unit gain of the digital speed sensor can be
defined in terms of the incremental measurement accuracy at s = l-Z-l
steady state. For an incremental accuracy requirement of -r
An = .0003 in the analog transfer function expression. This substitution is
applicable to systems which satisfy Shan nons Sampling
and a 12 bit speed sensor, Theorem or the Nyguist criteria

IKolmin = .B13 p ~ Pnyquist = 1..

With a 14 bit speed sensor and the same incremental "

With the above substitution, the analog PID algorithm was
measurement standard converted to a digital PlD form.

(Kolmin = .203 The measured speed signal applied to a digital filter

Similarly. a maximum value of per unit gain in the speed sensor

can be defined in terms of an operating range
1
61=61 Z- + (P~T) 1
0(6 0 -6 1Z- )
(17)

The digital measure of error

62 = 6 REF - 61 (18)
The minimum acceptable range is given by
The-lead and lag or integration given by
n2 - n, = ,1
64 = 62 + (/lITI 0 (62 - 62Z-1) (19)
such that for either a 12 or 14 bit speed sensor
66 = 66Z-1 + (IT) 0 64 (20)
IKolmax = 10.0
The lead and lag combination
The requirements of a minimum accuracy measurement and a 68 = 66 + (OfT) 0 (66 - 66Z-1) (21 )
minimum speed range define a per unit gain range for either a
12 or a 14 bit digital speed sensor 69 = 69Z-1 +(p IT) 0(69 - 6gz -1)
(22)

.813 oS Ko oS 10. for 12 bits CURRENT DRIVER - The performance of the digital to analog
convertor and current driver is given by
and

.203 oS Ko oS 10. for 14 bits 612 = Integer (6 9 ~~) (23)

 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

850174 5

I AID - - -11- - - - - - - - - - - - - MiCRO PROC:O:- - - - - - - - - --l

I S~:..gR II B" I
I II I
I INTEGERf,(O'O'),J II • B, B, B. B. I
~o. 0, ] II I
IL--------~II I
L JI I
I I
I I
I I
L , I
o
-OIESELENGIN-;-- ------ -I '---",-:-- i I I
I I I I I I
I r;:~l-.lC__~1--f IB" INTEGER(~") I
I
L _ I I I I 2"
________ -.1 I I L J
~/~CURRENTOR~ J

fIGURE 1. MATHEMATICAL MODEL OF A DIESEL ENGINE DIGITAL CONTROL SYSTEM

(241 DIGITAL CONTROL SYSTEM RESPONSE, The quantization

effects of

and when normalized input speed sampling (N = 12 or 14 bits)

i =A
AR
The above system of equations presents a mathamatical
model which can be used in the evaluation of digital control
system concepts. A mathematical block diagram depicting
this model is shown in Figure 7.
Typical values of adjustment range for the analog control and
the equivalent digital control with a sample time of 8 milli-
seconds are shown in Tables 1 and 2. With an 8 millisecond
sample rate. the time required for the microprocessor to
execute a control algorithm must be minimized. A minimum
algorithm execution time along with a wide computational
range is achieved with 16 bit fixed point mathematics. To avoid
overflow, computations were performed with a 24 bit word.
The adjustment ranges and typical settings for a small high
speed engine are shown in Table 2.
output current sampling (R = 12 bits)
fixed point mathematics (M = 16 bits)
(25)
Lead to errors in the control system which cause low
frequency (.1 to 4 Hertz) limit cycle oscillations in fuel rack
motion and engine speed,
Large amplitude low frequency fuel rack oscillations are
objectionable due to excessive linkage or fuel pump wear and
the possibility of exciting power transmission system natural
frequencies. A limit cycle amplitude of less than .005 per unit
fuel rack motion is desirable. To achieve the design goal of
.005 per unit fuel rack motion with a 1 biterrorin measurement
accuracy at the speed sensor, it is necessary to properly
distribute total digital controller gain (K o 0 loT) between the
digital speed sensor IK o ) and the digital integrator liT) for
both the PI and the PID algorithms. The design approach
implemented satisfies the total controller gain
(K II A = KD •

TABLE 1: Typical Adjustment Range of Analog Control Algorithms

ANALOG CONTROL SMALL HIGH SPEED ENGINE

Parameter Range PID PI

IK I)A 2.0 to 500. 60. 16.

{j .008 to 0.2 .128 0.0

{3 .008102.0 .35 .512

 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

6 850174

TABLE 2: Typical Adjustment Range of Digital Control Algorithms

DIGITAL CONTROL and T = .008 SECOND

INTEGER MATH SMALL HIGH SPEED ENGINE

Parameter Range PID PI

KDI. T 1/2 6 to 2' 60/128 16/128

51T 1 to 32 16 0

fliT 1 to 256 44 64

P .008 .008

R 12 12

M 16 16

IAmax/ARI 2 2

defined by the algorithm by first computing the digital and

integrator gain I which meets the minimum specified fuel rack
motion requirement when the speed sensor error is one bit. (I)PIO = 26. at N 14
The per unit gain of the digital speed sensor is then computed
as For the PI algorithm

(I)PI = 76 at N 12

and
and must lie within the constraints
Ill pl = 307 at N = 14

The gain of the digital speed sensor for the PIO algorithms is
The digital algorithm deviation equations for a 1 bit change in then given by
measured speed are obtained from Figure 7. The computational
control algorithm yields for a 1 bit error

dBg = 2M- N(.3.-\' 11 + 51TI ° (1 + fliT) ° II Tlo 1 1261

\f+TJ 10.0 at N 12

and the current driver is given by and

di = A max
A
R
0(_1_)
2R
Integer (~)
2M-R
1271 KO = 2.3 at N = 14

Both of these values lie-within the allowable range of the speed

Substituting representative values from Table 2, into Equation
sensor per unit gain but imply a narrow operating speed range.
(27), the constraint on fuel rack motion
If the minimum value of per unit gain for a 12 bit speed
sensor had been selected to maximize the speed range, then
di = .005
= ...§Q...
is satisfied with .813

dBg = 160 = 73.

The computational control algorithm requires that the gain on With this value in the PIO algorithm, the motion of the fuel rack
the digital integrator satisfy due to a 1 bit error obtained from equations (26) and (27) is
given by
I = (28)
(1 + fliT I] dBg 1745.

and
Solving the above equation with values from Table 2, for the
PID algorithm di = .053

(lipID 6. at N 12 This amplitude of fuel rack motion is unacceptable.

 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

850174 7

TABLE 3; Digital PID Control System Parameters for a Small High Speed Engine

Digital PID Algorithm Digital PID Algorithm

Parameter T = .008 Second T = .016 Second
Speed Sensor Speed Sensor

N = 12 Bit N = 14 Bit N= 12 Bit N = 14 Bit

..Ill 1 1 1 1

..IBg 160 160 160 160

;,li .005 .005 .005 .005

Bref 32768 42672 32768 53692

n2· n 1 .1 .43 .2 .83

..In 2.4X10- 5 2.6XlO- 5 4.0XlO- 5 5.0X10- 5

I 6. 26. 12. 50

KD 10 2.3 5. 1.2

fJ/T 44 44 18 18

OfT 16 16 8 8

To increase the operational speed range of the PID Both of these values are less than the minimum allowable per
algorithm, the ,uP sample time can be increased to T = .016 unit values of speed sensor gain. This result demonstrates that
second with some reduction in system transient response. the measurement accuracy requirement on speed cannot be
This allows the digital integrator gain to be increased for both met. Selecting the minimum allowable speed sensor gain to
the 12 and 14 bit speed sensors, satisfy the accuracy requirement for both the 12 and 14 bit
sensors
I = 12 at 12 bits
KO = .813
and
The PI algorithm then requires
I = 50 at 14 bits

The per unit gain for each speed sensor is given by

KO = 5.0 at 12 bits
I = 20
and
KO = 1.2 at 14 bits The resulting fuel rack motion due to a 1 bit error is given by
Equations (261 and (27),
Table 3 presents a summary of the digital PIG algorithm
at 12 bits
parameters for two different speed sensors and two different
sample periods. As the number of bits in the digital speed
..Illg = 40
.sensor increases or as the sample period T increases in the PID
algorithm, the gain of the digital integrator increases and
minimizes the per unit gain of the digital speed sensor.
..Ii = .0015
Employing a similar analysis for the PI algorithm, the gain
of the digital speed sensor is given by and at 14 bits
lK A 0 IAI

Ill p !
..Illg = 10

..Ii = .0005
.21 at N 12

and
A summary of the digital PI algorithm parameters for two
.05 at N 14 different speed sensors is presented in Table 4.
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

8 850174

TABLE 4: Digital PI Control System Parameters for a

Small High Speed Engine
.,,, I
i1UU t"
Digital PI Algorithm OUf. f·
T =.008 Second

~ ~ - Ir~I .:\·'- - - '. . .- J~~

Parameter Speed Sensor
=14 Bit
dB
N =12 Bit
1
N
1
::::,:fh~---
dllg 40 10
-.\104 !~
til .0015 .0005
-.Ilu[· i·
Bref 53280 53280 I
1.23 -.1101<' l..
"2- 0 , 1.23
dn 3.XlO-4 7.5XlO-5 - ill J~--"'---i-~;""_">"-~~-~"-'--'j~~_.-'---t--'--.{~:-- ..............!
TIME SECONDS
I 20 20 nETPf'Il DI[3Et 4~1I ~ 1';OO$;:H1 PIt> i.'"4 NT n.TF:G(R 0:-. [oI''lX.4''
l:;~.HHO 11'1)(12 IlIT'"""~··I.0'5)IVfVI211[h.-OTt>i.'(.,(ffi)t(;T~tl 04
,.TtlD~44 C')HP .If. IGflIN~I> PIII ['~NAt1ICS Pl":C'G lIH::AEI OCG
KD .813 .813
6IT 64 64
6/'1' 0 0 w r
z•
l.'.uml

o
•"«w lllHHI<i r
COMPUTER SIMULATION STUDY - A computer •Z 1l1l11llf. L
simulation of the mathematical model depicted in Figure 7 was i1-
performed to evaluate both the transient and the steady state
"«
w
I. 'IHII,,4
,
" l.llllfll12 ~
performance of the two control algorithms. The simulation
subroutine describing the PIO algorithm is described in
I
, I· ,"
Appendix 1. Figure B presents simulation results for a start to
rated speed and addition of 100% load, with the PIO control
• ~J"J:I,lC t.-----.
, J
I
algorithm, a 12 bit speed sensor and T = .008 seconds. The • ,.<!~.~,[. ~-

transient response of per unit engine speed in) and actuator

drive current (i) are depicted. The maximum speed deviation
.."" .. 1,
for the PID algorithm is 3.5 percent. Figure 9 shows the steady · ~'~<.'~'L· i
l
state performance at rated speed and no load of engine speed ;,~!t:l.~ ........ _~:;----'--.~;--~-'j~~-"-~-'(7----"._-.,,:".l'i~-·~'"· '-'.
and actuator drive current The limit cycle oscillation at steady TIME SECONDS
state is characterized by FIGURE' STEADY STATE RESPONSES AT A DIGITAL 1'10 CONTROt WITH A 12 BIT
SPEED SENSOR ANIH '.004 SECOND.

I n I = 2.5 X 10-5
I I I .005
at 1 Hertz. Figures 10 to 12 show similar transient results with
different speed sensors and sample times. When the control
algorithm is changed to PI with a 12 bit speed sensor and
T::: .008 seconds, the transient response of Figure 13 is
obtained. The results are summarized in Table 5 below.

_'~-'!

,.<
1. :.
I, nI
'~'; .~ __.__
1.<'

L~ ._J!"
-- I,
I

FIGURE' TRANSIENT RESPONSE OF A DIGITAL PIO CONTROL WITH A 12 BIT SPEED FlGURE 11) TRANSIENT RESPONSE OF A DIGITAL 1'10 CONTROL WITH A 14 OIT SPEED
BENSOR ANO T'.OO3 SECOND. SENSOR AND T".ooo SECOND.
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

850174 9

In each case the limit cycle oscillation at steady state satisfied

, the design criteria Ai ::; .005.
w,
w
Z l. 2
~ I I

::l
a: . 'J' TABLE 5: Digital Control Transient Response
~ .<:

"~ ALGORITHM SPEED SAMPLE PEAK SPEED SPEED

:~, SENSOR TIME T DEVIATION RANGE
N AN N2-N,

., PID 12 ,008 .035 .10

-thHWIf\f'd'---'--------. PID 14 .008 .035 .43

_;·~~,r.~J~J~~~-~~
".1
~. <. PID 12 ,016 ,044 ,20
PID 14 .016 .044 ,83
PI 12 .008 ,076 1.23
.. .~ l 1.5 2 ,,~ J ... 5 " '4." :. ~.~ (; b.~ 7
Tn£: S(CO«IJS
ntHWIl DltSel 4-71 eo 1(leEll'1f'tl ,",10 24 nIT tNTCGl:R f<% DROOP
T';i-.l:lIC, A/IHI? BIn...~a--1.10HI.-IVI?DlTt..-eTCeeftfl)HOV 50..
srnnq!, COl'll' ' £t lGAlt'-Fl<:: PIP DYNFI11CS f"F;:{)C INT SAE:4 lJCC

FIOURE 11 TRANSIEKT REllPONSE OF It. OIGITAl- PIC COHTllOL wmt II. 12 BIT $PE£D ENGINE TEST RESULTS - A real time simulation of the
SENSOR AND T',015 SECOND.
digital diesel governor has been developed using an HP 9836
computer and a HP 8940B multipr09rammer for AID and D/A
conversions. The real time model allows a rapid test of different
control algorithms and concepts before hardware is designed.
....
I . ~.

n
Control algorithms similar to Appendix 1 were implemented in
PASCAL and tested with an electrohydraulicfuel rack actuator
on a Detroit Diesel 4-71 engine generator set. The engine is
w
w
Z
.,
1.1,-
rated at 1800 RPM and 125 Kw,
0
~
w
w
.c'
~
~

z .,.,
"w , :'
~
,··1., , .•

~
,., I
.OJ 7 VOLT
, , I
, -'/*H~----.l._··--- ---
, ,....-7;~,....T'''.';"_;l_,,· o VOLT
" ,, -'T'~;f:"S~~:, l\ (.!:,
ltt'£. r....T WDS
DtTPOtl nrtSCl 4~71 e Ille8l'<f'H PIli 24 IJlT INTCr.t:k 0:' flRn)P +5%
TS-.~l& II-"U(14 lltn..Jc-l.lSJD-f1(12BlT"-BTr.ceDfl'NOV 7 IH
STfHI~IO ((IIiI' " [l IGflIl~'SO PIO D\1*'tlICS PI'tt1G INT $(£5 ICC

FIGURE 12 TRANSIENT RESPONSE OF A DIGITAL PIO CONmOl Wnlf A 14 lilT SPEEP

SENSOR AND T'.015 SECOND.

. -'-'-·1-5%
·~~"-·~'··~-·-~~"~~
..-~.~.""';;:;!..,~1 SECOND
PIO ANALOG CONTROL.

:±JmB7VOLT

o VOLT
+5%

-s%
1 SECOND

DIGITAL PIO CONTROL WITH 12 BIT AID SPEED SENSOR

AND 8 MILLISECOND SAMPLE RATE.

FIGURE" TRANSIENT RUPONIU! OF A DIGITAL PI CONTROL WITH A 12 8rt aPfED

5£H80A AND T'.OOI SECOND. FIGURE 14 EXPERIMENTAL ENGINE TEST DATA.
 Downloaded from SAE International by University of New South Wales, Saturday, August 25, 2018

10 850174

A simple digital speed governor was developed using an 8 n engine speed per unit
bit pP, an 8 millisecond sample time and a 12 bit digital speed ~n incremental speed measurement accuracy per unit
sensor. Figure 14 depicts a comparison of the transient and
steady state response between an analog PID control and the n, minimum engine speed per unit
digital control on a diesel engine. On each chart the upper n2 maximum engine speed per unit
trace depicts voltage applied to an actuator which positions the T sample period of digital control
fuel rack and the lower trace measures engine speed. The
digital governor was set with I = 4, WT =25, OfT = 16. The t engine developed torque per unit
maximum speed deviations were 4.2 percent for the analog Z-1 sample period delay of digital control
control and 4.6 percent for the digital control. The transient z fuel rack position per unit
response and the steady state response of the two controls
compare favorably. i engine acceleration coefficient per unit
CONCLUSIONS - A ,uP sample rate selection criteria 71 engine dead time seconds
has been presented and the theoretical performance of two 72 actuator time constant seconds
digital control algorithms for diesel engine governing has been
described. The effect of a 12 bit and a 14 bit digital speed .uP microprocessor abbreviation
sensor upon the operating range and steady state performance {j digital contrallead compensation time constant
of each algorithm has been shown. A high performance digital [3 digital control lead compensation time constant
PIO algorithm provides rapid transient response with minimum
fuel rack motion at steady state but imposes constraints on the
digital speed sensor operational range, The speed sensor
range can be increased by increasing the number of bits (N) in
the speed sensor and the sample period T. For a small high REFERENCES
speed engine, minimum PlD algorithm requirements are a 12
bit speed sensor, a .008 second sample period and 0.1 per unit 1) Ragazzini, J.R. and G.F. Franklin, "Sampled-Date
speed range. A 14 bit speed sensor would allow the per unit Control Systems", McGraw Hill Book Company, New
speed range to be increased to .83 at .016 second sample York, New York, 1958, pp 123-125.
period.
The digital PI algorithm does not provide the rapid 2) DeBolt, R.R. and B.E. Powell, "A Natural 3-Mode
transient response of the PIO algorithm but does yield minimum Controller Algorithm for DOC", ISA Journal, Sept, 1966.
fuel rack motion with a wide digital speed sensor range for both
the 12 and 14 bit sensors. 3) Franklin, G.F. and J.D. Poweil, "Digital Control of
Dynamic Systems", Addison-Weley Publishing, Reading,
MA,1960.
NOMENCLATURE

A actuator drive current

APPENDIX 1
AR current range for rated engine load
COMPUTER SIMULATION PROGRAM FOR THE PIC ALGORITHM.
a a measure of the reduction in crossover frequency
PID_cenl: l PID CONTROLLER SUBROUTINE
B .uP integer value I WOROSIZE IN COMPUTATIONS 24 8ITS
8b'( (N_pm·.95)/(1.05-.95))·4096 !SPEED SENSOR 12 8,T
Bj ,uP integer value j =0, 1, ... 12 8b'INT(Bb)
IF 8b>4096 THEN 8b,4096 ICLAMP ON
F sample frequency of digital control system IF Bb<O THEN Bb'O lRANGE 12 8IT
80~16'Bb
Fo final open loop Bode plot crossover frequency Bl'BH'( (BO·811)/2l IFILTER TIME CONST .006
Bl~INT (Bl)

F' initial open loop Bode plot crossover frequency B2~8_rel-Bl

B3'INT {82HNT (62~) lSTABILITY
84,INT {B_I·63j·INT (62)
~F reduction in open loop Bode plot crossover frequency B5'INT (B4'lgllin'2/256j
IF B5>B_5111 THEN 85'8_5al ICHECK FOR OVERFLOW
I control integral gain per unit IF 5"-B_sal THEN 65~B_sat
!INTEGRATION
actuator drive current per unit 66,INT {B5j'INT (B6:) INTEGRATOR OUTPUT
IF 6IPB_maK THEN 86'8_maK lCLAMP ON
J flywheel moment of inertia IF B&<B_min THEN 66'B_mln 1 INTEGRATOR
67~INT (86HNT (B6~)
KA analog control speed sensor gain per unit 88~INT \B6)~INT (B7'D_I)
IF 66>8_max THEN 88,6_mllK !CLAMP ON
Ko digital control speed sensor gain per unit IF B8<8_min THEN 86'S_min !OUTPUT
89~{69l)~{B8·B9lj/2 1.008 FILTER@ OUTPUT
B12'INT (B91l6j 116 BIT TO 12 8IT CIA
L E rated engine torque Ma=200·{BI2I4096·S0.2j !OUTPUT CURRENT 50 MA BIAS (O.200MA)
l_llCI~Mllllr !NORMALIZED OUTPUT CURRENT
L engine load torque per unit SAVES
Bll,INT {Blj l FILTER
M computational word length 82l"INT (B2) lSTABllITY
B6:'INT (BS) IINTEGRATOR
NR rated engine speed B91'lNT (89j IFILTER
RETURN
N digital speed sensor word length ! END PID DIGITAL CONTROLLER LOOP········· ..• .