p- • "

I Jr" .... ", ' . ... I -~, -, ~ • ~ ,

r 1 L)L·'- _I..': U[ 1 1':':'1 \~ j: \ Ii. '

1ih International Conference

Military Technical Colle~Je
on Applied Mechanics and
Cairo , Egypt
Mectlanical Engineering (AMME;

RECONSTRUCTION OF DIESEL ENGINE CYLINDER PRESSURE

USING ENGINE VIBRATION AND ACOUSTIC EMISSION
~.

C'~-..._""'".
.. ~, il !\
"D" ;-:.cU...J;:::"'v'\ ': I;
...........
1\
h ..
n r.
U., t-\. CI-.-.._ : !-......
L;..IIIIClI,;y , 0ClI;IX
1\
r\ , 0C' . , vilO"",
C"h-.I"", ; ,'"I
: ~ \01 i . ,-, .

and Moh2med K. L

ABSTRACT
In this paper the diesel engine cylinder pressure is reconstructed in the crank angle
do """'''''''''' Ileinro the> ,..,..,""~C,
II I U H . .'IoJL4 ....., .....,rpd
"-<I"""'ll ~ ,_ e>nginc
_. • _ \I1·
'1.1 \; br~f,or.
_ ......
• I '::)nd arous tirv .em'lc::sinn
, . . . . ,_'. . Th'IS
. , tl"lchniqu p
L _ I _ .. . L _ ,. _

depends on frequency analysis of engine vibration and acoustics. The frequency

response functions (FRF's) are determined from the measured cylinder pressure,
engine vibration and acoustic emiSSion at specified engine speed and loads. The
cylinder pressure is reconstructed from the measured engine vibration and acoustics
at similar conditions by applYing the inverse filtering technique uSing the already
determined FRF's. The measured and reconstructed cylinder pressures are
compared to each other for vibration and acoustics at different loads. Comparison
between the reconstructed cylinder pressures from acoustic and vibration signals at
the similar speeds and loads with the measured ones show good results as they are
predicted. Slight differences in results for measured and re constructed cylinder
pressure for both Vibration and acoustics IS due to non-linearity of the engine system.
The measured and reconstructed cyl inder pressures from vibration signal are
compared to that ones from acoustics for similar conditions and they have the same
results .

KEYWORDS

Cylinder pressure, vibration, acoustics, frequency response function , fast Fourier

transform and inverse filtering ,

PhD Postgraduate candidate , Dept. of Mech. Power and Energy, Military Technical
College, Cairo , Egypt.
::.~gyptian Armed Forces.
As ";s."''''''+
vi lOl l l
D"OfI . , IC...,,..
I I
, , i l.l r.f
CAvUl l Y VI
Cnroi,...,oor;,..,ro
L...I ::::f11H...,d .... "II'~1

Technology, Cairo, Egypt

NOMENCL,'},TURE

A(t) Acoustic slgn,ll Iii lime domain

A«i)) Acoustic signal In crank angle domain
A((I) ) Acoustic sign81 In frequency domain
AE Acoustic emission
BDC Bottom dead centre
c Damping constClnl
uFi LJisGrete t=ourlE::!1 Tldll~;U i iil
Force amptitude
F (t) Force vector
FFT Fast Fourier Transform
fft MATLAB function for cmnputlng fast Fourier transform
FRF Frequency Response Function
f Frequency
f~ Sampling frequency
H(w) Frequency response function
I Sampling index
Ifft MATLAB function for computing inverse fast Fourier transform.
J Imaginary unit
k Stiffness constant
m Mass
N Number of samples
Pmeas Measured cylinder pressure
Precon Reconstructed cylinder pressure
P{t) Cylinder pressure signal in time domain
P(~) Cylinder pressure signal in crank angle domain
P(cu) Cylinder pressure signal in frequency domain
RMS Root mean square
RMSE Root mean square error
TDC Top dead centre
Time
Xo Response amplitude
X{t) Fourier transform of a continuous signal
x Displacement
x(t) Response vector
x Velocity
x Acceleration
Xk Direct Fourier transform
Xn Indirect Fourier transform
V(t) Vibration signal in time domain
V(¢) Vibration signal in crank angle domain
V(w) Vibration signal in frequency domain
q) Crank angle [degree J
CD Frequency
 · e li h ~ f ~ I \ -

1. INTRODUCTION

Cylinder pressure proviaes comprehenslve Informatloll about engIne performance

and operating conditions. It is directly related to heat analysIs, pollutant Information,
vibration and mechanical events [1] It can be used for heat transfer analysIs [1 L fleat
release analysis [2], engine control [3J, and fault detection [4J . The traditional meihod
for n1eaSLJrlng the In-c'y'linder pressur e is by fitting a pressure transducer In the
combustion chalT,bcr through a !lole In the engIne cylinder head . ThiS method IS
Inconvenient In small engines v·/ith small space for fixation of the pressure
transducer E:ven In large engines With enough space If IS 100 expensive to fiX a
pressure Lidllsuucer in each cy linde! in lhE: engir·le. lh8 pressure transducers haVE: d
limited lifetime in the combustion chamber environment of high heat and carbon
deposits. Nowadays alternative possible method for reconstructing the ill-cylinder
pressure through eas}l-to-use and low cost tran~ducers IS looked for.

Brwon et al. [3] and Citron et. ai. [4] reconstructed cylinder pressure from crankshaft
speed flUCtuation. The cOI-lditioilS in the cylinder of all engine, ii-Ie extemai 10Cld and
the control parameters completely determine the flywheel angular velocity. If the
conditions in the cylinders, the external load and the control parameters are known ,
then, at least in principle, the flywheel angular velOCity can be calculated . The
converse is only partly true. Given the angular velocity of the flywheel, it is only
possible to determine some of the conditions in the cylinders . Therefore, the angular
velocity of the flywheel is used for estimation of cylinder torque and in-cylinder
pressure. Angular velOCity fluctuation is measured with an interval timer and a
magnetic sen sur, by timing the flywheel gear teeth as they pass t' Ie sensor.

Zeng et. al . [1-2] used the reconstructed cylinder pressure for heat release and heat
transfer analysis. The reconstruction of cylinder pressure is based on the concept of
a dimensionless pressure curve in the frequency domain. It is shown that cylinder
profiles acquired over a wide range of engine speeds and loads exhibit simJiarity.
Hence, cylinder pressure traces collapse into a set of dimensionless curves within a
narrow range after normalization In the frequency domain The dimensionless
pressure traces can be described by a curve-fit family , which can be used for
reconstructing pressure diagrams back into time domain at any desired condition
relative to known engine speed and load .

The relationship between the engine speed fluctuations and the pressure inside the
cylinder was investigated by Moro et al [5] using the FRF. The FRF is de.fined as the
ratio between the FFT of the cylinder pressure and the FFT of the corresponding
speed fluctuations. From many FRF values, each one corresponding to a different
operating condition, FRF map was created for estimation of the in-cylinder pressure.

EI-Ghamry et. al. [6] used acoustic emission (AE) for indirect measurement of the
cylinder pressure from diesel engines. Raw AE was utilized for modeling and
reconstruclil1g the t->ressure wave form in the tIme domain during expansion but not
during compression due to lack of signal energy during compreSSion To overcome
this p lO~~\ lern tlK; :, iglldi 'v\i3S dl\lid~; d n~~) tv.'o SC .::l;ons leve::-.t~ 1 :'UI18 i:iC; ccrn;.;'c:.,~:c: :
part of the signa l :-JIld rllo fu ei I"Jt:ctlonlc xp.Jn :,lon slr(Jk,c; T;le cornpression par! ()f q--I':'
pressure sIgn8i W8:o r8consti'ucled by uSing polynonLial flUing . An aul or8grf:;SSlve
technique WDS used dUring the Injection/expansion stroke . The lOot Inean squar e
(RMS) acoustic emission signal is we! 1correlated with the pressure signal In the time
and frequency domain The complex cepslrum analysis was then used to model the
pressure signal for the complete combustion phase (compression, inl8clion cl11d
expansiofl' ThR main advanlaqe of using cepstral analysis IS due to the utilization
th p frequenev conlen t OJ the RMS At:: signal, which gives ull 20v8IilGiQi::: \,I/hen th e
signu: h.:JS b'.ver ~'!r:er:;;"1 C'ontAn t. dunng the cornpressll)('1prucc:,v.

In this paper an Indirect method to reconstruct in-cylinder pres sure of single cylinder
diesel engine is attempted . The method depends on tile analYSIS of engine vibration
and acoustic emission . Tr,e present technique depends on the rela tionship between
the in-cylinder pressure, the engine acoustics and the engine surface vibrations.
Firstly the FRF is considered as the ratio of the fast Fourier transform (FFT) of the
cyiinder pressul'e to FFT of the engine sound pressure . Secondly FRF is considered
as the ratio of FFT of cylinder pressure to FFT of the engine surface acceleration.
The frequency response functions are determined from the measured cylinder
pressure, engine surface acceleration and engine sound pressure at specified engine
speed and loads as signatures . Assuming that the engine is a linear system, the
cylinder pressure is reconstructed from the measured engine vibration and acoustics
for similar conditions by applying the inverse filtering technique to the FRF
signatures ,
2. FREQUENCY RESPONSE FUNCTION

The frequency response function FRF relates the output (displacement velocity or
acceleration) per unit of input (excitation force) at each frequency [7]. For an engine
which is modeled as a linear single degree of freedom system the equation of motion
is given by:

mx + ex + let ~ F(t) (1 )

The steady state solution is obtained by assuming that the excitation force F(t) in Eq.
(1) is a phasor described by:

(2)

where f- " IS the force amplitude.

According to (7-9) , the response phaser is assumed to have the same frequency UJ so
that:
...,- (' .I \1 --
-
(e, )
where \ IS tne response amplitude . Substitution of Eq, (2) and (3) Into (1) gives

.r (4 )
Ii. k 111 flJ" · .Ie ( !}

3. DISCRETE FOURIE:R .\ kAN~r-Ok.rvi

According to Ziemer et. al. [10], the Fourier transform of a continuous signal, x(t) is :

.x· (/ ) = J.\" ( I ) c " ~" II til (5)

The discrete Fourier Transform (OFT) is described as a double sampled version of

the continuous-time Fourier transform (i,e., sampled in buth time and frequency), The
DFT pair is:
\' - ,
X ,,-- -J2r.A1l .',v ,k=1,2,. "." .. ,N, (6)
J: = ~ xn e
Ii = 0

For the direct transform, and

1 t. .' I
X
JJ
= -_" ~
X
£
('.I 2 :. k/J IN
J n= 1, 2, .". " .",N (7)
N ! .. ()

for the inverse transform .

To derive the OFT from the continuous-time Fourier transform, the time-sampling
interval S taken as T/N=~t, where N is the total number of samples taken in a T-
second interval (time window), and the frequency-sampling interval is taken as 1fT.
Thus, both relationships imply periodicity: the first in frequency with period fs
=NfT=1/~t, and the second is in time with period T=Nj,t=N/fs. In terms of the sample
number, the first is periodic in k with period N, and the second is periodic in n with
period N.

Fast Fourier transform (FFT) is another version of discrete Fourier transform requires
samples of power 2, N samples that accept 1092 (N= 2L092N) calculations, The FFT
must be computed at all N samples at once, while the DFT can be computed at just
few samples of interest. There are many algorithms in different programming
languages for computing the FFT Here a MAT LAB algorithm is used for computing
the FFT FFT algorithm is a function exists in the MATLAB package it is written in the
 · ' -'

I'"~ \ J •
; -' : i l .{" ...1" '..

cornrr~<..jild Vy}ll~dc, ·:·\ ;:::' ttt;>~) cH l0 its :: ,', ,, ':~ ',-' ; ., ~.£!i>::: \",Ih (-~:-e ), !s Hle tit"H? dC',' ~'i~JI i ::.~~~~?nl-ll
The FFT !s faster than thE: OFT and thai CL~n be proved by an example l:::qu8tlon (tJ )
2
shows that for 8 samples the DFT can be performed by N or 64 complex
mu ltiplications That means for each Xk, lI'lere IS a SUIT] of eight cOlllpie x pi oducls fur
n goes from 0 to 7. On the other ha :":d the number of complex multiplications by FFT

is appl oXllnalely !~:

\
log, ;\' . Therefore, th e number of comple );
'
,,~ultiplications by FFT

for 8 5C:ir.·lpi8S arc 12 vVlii ch is appro x iiT,atci/~' ~I;nc Ic ~;s thun tl'13t by Ut-- i

4. TEST SETUP

In the present paper tests have been carried out on a single cylinder direct injection
diesel engine Deutz Fl L-511 ''vith 7.7 Kw power at 1.500 rrm Appendix (A) shows
the engine specifications. The engine load is measured by an ELZE/Herman
hydraulic dynamometer of type AN5F . The measurement of engine speed and digital
triggering of sampling are encountered by a shaft encoder of type GI 3281. The in-
cylinder pressure is measured by a Kistler 601A piezoelectric pressure transducer.
The engine vibration is measured by a Bruel & Kjaer accelerometer of type 4367
which is fixed on the engine head cover. The engine noise is measured by a Bruel &
Kjaer microphone of type 4190 at a distance of 1 meter (according to SAE handbook
[11]) from the right side of the engine, Bruel & Kjaer conditionir.g amplifier of type
2626 and Kistier 50108 charge amplifier are used for conditioning of vibration and
pressure Signals respectively. A Bruel & Kjaer 8 channel microphone power supply of
type 5697used for conditioning the acoustic Signal.

A National Instrument PCI-MIO-16E-1 card and LabVIEW 7.1 software are used for
data acquIsition. A P-1I1 Dell computer with 933 MHz processor, 128 MB RAM and 80
GB HD is used. The data is acquired at speed 700 rpm at no load, 40% load and full
load with sampling frequency 8400 Hz.

5. EXPERIMENTAL RESULTS

The cylinder pressure, engine vibration and acoustics are acquired at engine speed
700 rpm and different loads: zero load, 40% load and full load. The time domain
signals are converted into crank angle domain by multiplying the time by 6n where n
is engine speed . The measured engine acoustic signals in crank angle domain A(q))
are shown in subfigure (a) of Fig.1, Fig.2 and Fig.3 for zero, 40% and full loads
respectively. The measured engine vibration signals in crank angle domain V(!)) are
shown in subfigure (b) of Fig. (1). Flg.2 and Flg .3 tor zero , 40% and full loads
respectively. The measured engine cylinder pressure P(q)) signals and acoustic
signals in crank angle domain are shown in subfigure (c) of Fig. (1), Fig.? and Fig 3
for zero, 40%) and fuli loads respectively. 'The measured engine cylinder pressure
signals and vlbr'::'ltion signals ill crank angl(c, domain are shown in subfigure (d) oi
Fig. 1, Flg2 and Fig.3 for zoro, 40 % and fuil loads respectively. The exact speeds
700, 700 and 711 rpm correspond to zero, 40% and full loads respectively All the
acoustic and Vibration signals at different loads are filtered by a bandpass filter 1000-
3000 Hz, Vdilch is the bandwidth that contains the combustion induced noise and
vibration

6. CALCULATION OF FRF FROM THE MEASURED ENGINE SIGNALS

The FFT is obtained for the filtered acoustic signals A(t) and Vibration signals V(t) at
the different loads . The FRF's are calculated from the measured in-cylinder pressure
and the FFT of acoustic and vibration signals by the fOliowing formulae .

F or acoustics:
'."',Yf' ( ,,( )\
I-NI' .= - - , ., rJ (8)
IFr(J'(I))

For vibration ;

I-RI' = Fvr(l '(I )) (9)

Ff.T(P(I))

Equations (8) ::3nd (9) can be rearranged In other forms as functions of frequency;

_. ;I(w}
ll((O ) - - - (10)
I>(w)

(11 )

Where H(OJ) is the FRF, A(m) is the FFT of the acoustic signal, V(m) is the FFT of the
vibration Signal and P(ro) is the FFT of the pressure signal. The calculated FRF,s
from the cylinder pressure signals and acoustic signals are shown in subfigure (e) of
Fig. (1) , Fig.(2) and Flg.(3) for zero , 40% and full loads respectively. The FRF,s
which are calculated from the cylinder pressure signals and vibration signals are
shown in subfigure (e) of Fig. (1), Fig .(2) and Fig.(3) for zero, 40% and full loads
respectively .
 .. . . 1

I I : ~' I
".
',;,'

::'

( , Oi,
o

" I ,'t ., 1' ./1 0 ) h~ /:'(l

\
~
;J,
lJ
;;:
·0
r, 1

[)
l} ..

'. !
:~.
I

I)
fir.
f"
11l/}
~

t
: Ib "
(eI)
.. ....
' ) ,H)
.. 4.....,I

:.' , I

t ' . . c':; ' ~" " I

'\ ·1 r :~ I r, . ! '" '. IIII'

t.: ,Ii
~
,l " L __
;.;. f ""
,i
u 1 LV \ :)GI) =)1'10 o lll O 3GO :" ,!(i
C tflllll An~)lc \..:1eq[('(~ ] C II\l l ~ )l.. t1qll' (U'-~!-J JI!(: J
GOO , . , GO D, _.•
(")
i: .

l~J
\
, ~ 40°1
~
I I ;!OO
.

I
u
-~ .) -3
L 2 -, 0 ,
J
;! ~-i -l
01
~
[)
.!."
.
'I ·3 ·2 ·1 () :> 3 .1
l-It'!qIH"rH.;V [KHz.1 ~rr.que'ncy ~ KH7]

Fig . 1.Measured engine cylinder pressure, vibration and acoustics, and calculated
FRF at engine speed 700 rpm and zero load, (a) acoustic signal , (b) vibration signal,
(c) cylinder pressure and acoustic signal, (d) cylinder pressure and vibration signal,
(e) FRF from acoustic signal , (f) FRF from vibration signal.

(n)
L\ "\
,,' 0 /. f -- . -. . I
""
9~
~
005

to: ~ 1... t........,........... t. . .~t.i

r
0

r
I o o . i. . . . .
_ _. . . . . . .. . . . . . . . . . . . . . . . . . . .

<-0':
... .1 -0 ;> I
'1<O 300 540 720 o lao 3GO r?o
G 1_' '- .' . .. _!~D

II -
(Co)
I f;-,\
---0"

j r~
ACQU5l'C5
~: 4 . ... . Prl~ss u'n I
:i : ~

i :t=-= I
- __ I _ . __ _
i \, " .......... .. ...... 1
I ~
: .:-....... ......-........=---...----,'
\ ...• ~._.~
........ j -.......
--l

, {)O :"\60 ~40 720 1 UO ,'10 0 ~ ) AI)

Cmnk Ang,e 10"91 Cr,.,nk Al1gle {(l~y]

GO ---.. . .
~ GO!
j

r-requcncy I KHz]
$

l:l...
40,

-:. -4 -3
.1

JJ... -2 -1
'- feqW"· n cy [KII7]
o
" L - -_.-
j

Fig.2. Measured engine cylinder pressure, vibration and acoustics, and calculated
FRF at engine speed 700 rpm and 40% load , (a) acoustic signal, (b) vibration signal
(c) cylinder pressure and acoustic signai, (d) cyiinder pressure and vibration signal,
(e) FRF from acoustic signal, (f) FRF from vibration signal.
 -

i : ~~,._,u~~I'_I••~
0 ~
I
I
~ u ,' I
i
".~ CI ..... ••
I

?- (l; '!
J ' "

o~
1t:(: : H.)l\ (' I A( , ', .1\(/

(, )
1 0. 1(} ~
- , .... (."'.. J ' ,(I (:·, i __ 0- vlh ",: .O I.
.. .. .. f', fl ~ . ' . 11" \ '0 ! P "' ~.:·.Ultl

... ,.,=
. =~~,.,..-"---~~~~-
.;:
. ~, :-)
(, 1n ( 1 ."In (l ~1.1 l J I ;,:U ,-: 1f l r j ;.I(.C ~. ,,1 (j 7 2 fJ
C . ~ H)h. AnU l r. [ d U ~. l rfu) Crrlnk " nulG I d O >iiOOI
: I ()j :"'iO
I (f' ! ( I)
~ I:l
I E :'0 i
;!!)

~ 1
I
I 2. 1
IJ
~
'v
I (J
§
rJ; uL -
0 ' --
-c, -4 -:3 -2 -, 0 1 2 ;j 4
-~

5 -ZJ -4 -3 -OJ -1 0 1 2 3 4 S
t-requen cv ,r: H,:] F r o~LJ~ I \ cy i KilLI

Fig.3. Measured engine cylinder pressure, vibration and acoustics, and calculated
FRF at engine speed 700 rpm and full load, (a) acoustic signal , (b) vibration signal,
(c) cylinder pressure and acoustic signal, (d) cylinder pressure and vibration signal,
(e) FRF from acoustic signal , (f) FRF from vibration signal ,

7. INVERSE FILTERING TECHNIQUE

By inverse filtering is meant the inversing of the FRF H(co), The inverse filter can be
expressed as I-I I (liJ) or _ ( I) ' Inverse filtering technique has many applications in
j., OJ

speech processing. It is used for estimation of pressure or airflow signal at the mouth
during voicing. The application of inverse filtering for estimation of in-cylinder
pressure is relatively new, although there are some authors utilized it according to EI-
Gamry et. al. [6). The reconstruction of cylinder pressure by inverse filtering is based
on the processing of measured acoustic or acceleration signals and the in-cylinder
pressure during combustion .

Equations (10) and (11) can be rewritten in forms of inverse filtering as follows:

J'(rv) =.. 1(w)fI I (m) ( 12)

(13)
·til -t
equaion~
I ~
\!L.);',
d!lu
,.., 1 I ~~
',i-:J} Ii
..... '1\ :( ~L
u18
' 1 _'
I:'V~ :~) '::;
- J . :.L.-,. -
. i'l(,·li.'
r I '
v"j ,
, r. _( ...... ", •. ;, ....
d·~.J",,,,·.i'"
~r:\""pOI'C:C
,e.:>
/,\(,'1
,-,; /. (",
~'lrl
<":' .)

vibratloll response V((u) a,e known, the force response ~;((I)) CC:Jn be ealct '!elled

Inverse filtering can be explained schematlca!ly as shovlm:

P(l) I-I~I-;l-' -'1 P(('lU

....-.. ~I
' - _ _ _ ',_ • __ ._ A . _.
1-·----1
•

I i H(( I)
!
i FIU~'
!
L- ,
1
I
A(t)
~I,
I
t;I;T P.,( u) ,j I
or or L I
V(t)
I V( u)
!

,
,i " ~(I))

A(t) rfT A(OJ P(cu) P(t)

or or IFFT
V(t) V(ro)

FigA . Schematic diagram shows the inverse filtering technique.

8. RECONSTRUCTION OF CYLINDER PRESSURE

The cylinder pressure is reconstructed from other measured acoustic and vibration
signals at similar conditions alternative to that ones already used in calculation of
FRF's (see Fig.S). The new acoustic and vibration signals whic.h are acquired at zero,
40% and full loads correspond to speeds 700, 695 and 715 respectively. This
acoustic and vibration signals are also filtered with the same bandpass filer 1000-
3000 Hz. The new reconstructed cylinder pressure signals from the new measured
acoustic and vibration signals are calculated by applying the inverse filtering
technique . The amplitude of the measured and reconstructed pressures is converted
from volt to bar according to pressure transducer calibration curve that is shown in
appendix (8). Figure.6 .a, Fig.7.a and Fig.8.a show the new measured cylinder
pressure and the reconstructed one from the new measured acoustic signal at zero
load, 40% load and full load respectively. Figure.6.c, Fig.7.c and Fig,8.c show the
new measured cylinder pressure and the reconstructed one from the new measured
acoustic signal at zero load, 40% load and full load respectively at a range of 30')
before the TOe during compression stroke and 180" after the TOe during expansion
stroke. Figure .6.b, Fig .7.b and Flg.8.b show the new measured cylinder pressure and
the reconstructed one from the new measured vibration signal at zero load, 40% load
and full load respectively. Figure 6.d, Fig.7d and Flg.B.d show the new measured
cylinder pressure and the. reconstructed one from the new measured vibration signal
at zero load. 40% load alld full load respectively at 3 range of 30" before the TOC
during compression stroke and 1(30" aftel Hie TOe dUring expanSion stroke A
comparison betvveen the measured cylinder pressures, estimated ones from the new
measured acoustic signals and reconstructed ones from the new measured vibration
signals at zoro load, 40 l % load and full load are shown in Flg.9.

Figure9 a, Fig.9.c and Fig.9.e show the new measured cylillder pressures,
reconstructed ones from the acoustic signals and reconstructed ones from the
Vibration signals respectivelv Figure 9 b, F!g9.d 2nd FigS.f show 111e:8sured cylinder
pressures, reconstructed ones fro;T. the acoustic signal and reconstructed ones fiorn
the vibration signal respectively at a range of 30" before the TDC during compression
stroke and 180" after the TOe during expansion stroke of the engine cycle

In) (")

. , ..
004 .- .-. -- - 02 .
1
_.- -
"g
:ij
a
002

0
~
E
1
0 'I
o[ t ... ~ . I
1

f 002
<
-0
'I
,
0 :·: '
180 3GO 540 '; ' 0 0 1110 3eo ~/,0 720

0 ,,- (<.:)
.- ~ --- ----
·1 02
«I)

.... .. t~ .. ..
i
~ ~ 0 1,
I
~ 0
~ o• jI • •
f -005

_.
f -0 1 I
- 02 1_. ... --
-_. - 180 360 $40 720 0
- lHO
-_.-
360
-- - '-f;.1rJ
- --- --720
-
(c)
--- .--- - - .- C1 4 .- .- _. _.. - -- (r)
- .- ---- ·1
oj

~
~
JI)
d 0
J
~
I <

o·
I
+-. • ..
-e-
-t
, _. . . .
f .02
I
-0
0
-- - _.- .~-

'80
---- 360
"- - --",
540
.-
720
-0 -4
0
-- 100
.- ... ..
360 540 720
en'lnk Af'\OIC [t.JC!..tl Crnnk Anutc fdc':)]

Fig.S. The another measured acoustic and Vibration signals, (a) acoustic signal at
speed 700 rpm and zero load, (b) vibration signal at speed 700 rpm and zero load,
(c) acoustic signal at speed 695 rpm and 40% load, (d) vibration signal at speed 695
rpm and 40% load, (e) acoustic signal at speed 715 rpm ·and full load, (f) vibration
signal at speed 715 rpm and full load

9. DISCUSSION OF RESUL T5

The root mean square error (RMSE) percentage is evaluated for the reconstructed
cylinder pressure from acoustic and vibration signals as follows :
 1- - '\
I I '>~ (II 11)-
Vi -\ ..:.- ' I , . . 'I' "I",",

ii,I/S/-'-- ,'1 ',IOl) ( 1-'1 )

,'"
')-- ,iJ
L- ' l/J I'd .\

; I

The RMSE percentage between the measured and reconstructed cylinder pressure
from the acoustic and vibration signais are shown ill Tabie ~I ,
1 able 1 Root mean square error (RiviSE) ~eI1,81-lta8f; bet\,vee;, t!, c: :-nc::;::.u:-~)d ,?;:-.r:i
reconstructed cylinder pressure 21 different loads

------.. -- r---- --Root Me-an-- -Square Error (RMSE) -----I

Load I Percentaoe I
f--- - Acoustics ---- -- - I- .... -Vibration -- ------1
I,--- Zero - - , ----2 ,60/-;;------ --/---- -----2,6%--- -1
1----,-- -,---I - - ----- 1 ------ I
/ ___400/~ ____ 'I--'-~4~--~-- ___?:_!O!u _______ J
; _~u~ ______,_3_?.% _____ ~ ___ __~,8% ____ --'

The measured cylinder pressure and the reconstructed one from acoustic and
vibration signals at zero load, 40% load and full load, as shown in Fig ,6, Fig,7 and
Fig ,8, may be considered coincident. It is shown in Fig,6 ,a and Fig_6,c that the
reconstructed cylInder pressure from acoustic Signal at load zero is less than the
measured one with an error of about 2,6%, Similarly, Fig ,6,b and Fig ,6,d show that
the reconstructed cylinder pressure from vibration signal at zero load is the same as
that reconstructed from the acoustic signal with the same error 2,6%,

Figure7,a and Fig,7,c show that the reconstructed cylinder pressure from acoustic
signal at 40% load is less than the measured one with an error of about 2,4%,
Similarly, Fig,8,b and Fig,8,d show that the reconstructed cylinder pressure from
vibration signal at zero load is the same as that reconstructed fro the acoustic signal
with the same error 2.4%,

The reconstructed cylinder pressure from acoustic Signal at full load differs from the
measured one with an error of about 3 8%, see Fig,8,a and Fig,8,c, Similarly, Fig,8,b
and Fig ,8,d show that the reconstructed cylinder pressure from vibration signal at
zero load is the same as that reconstructed fro the acoustic signal with the same
error 3,8%,

Figure .9,a and Fig,9.b show a comparison between the measured cylinder pressures
at different loads, zero, 40% and full. The measured cylinder pressure increases with
the increase of load at constant speed _ The reconstructed cylinder pressure at 40%
load is the greatest and that at full load is the smallest, and that al zero load is in
between which is abnormal criteria , The reconstructed cylinder pressure from
acoustic signal at 40%, load IS the greatest and that at zero load and full are the
smallest and cOincided, see Fig 9 c, and Fig ~ -j T: k iGCOnstrllcted cy!:nder pressure
from vibration signal at 40%) load !s the greater at the curve beak and the smaller at
the ends of the strokes than at zero and full loads which are coincident, see Fig ge
and Fig g.t. Also the coincident of full load and zero load, and increase of 40 110 load
at the curve beak IS abnormal. This abnormality may be due the linearity assumption
of the engine vibrating system, 111 fact the engine system IS nonlinear, because the
Inertia of the engine vanes with the crank angle , Therefore, the validity of the
8ssulrl~J tlun or lirleal ily is iirniip.d for low "peed engliics . AIsD the assumption of single
The 3::;::; U;-'liJtiO;-, of ::>,j",gie
degree of freedom IS invalid for a muitr-cyirnder engme r he degrees oj treedom
should be at least equal to the number of cylinders
(.1)
l.tt

.
7.'

-
~
.11 1

."J
"I
II tHO
J CI
: h,j) ~; 4IJ
!
I
I
-,-;.!.(,
0

~
~
~
;
~
""
,'/ 1 .

"
: ' (1
(, ll h) : H .l } -.2(,

(rJ)
)
" 'l lI l

l·· ....-_· ...M"""-

_ Ih !.O.\:.
. -- "'J
- I'

- 2u t ____
; ).30
.> _ _ _

_1,,(,
_ ______

_i" I )
-

4::11 04!'"oll
__ _ _ ~ ___ • ___ j
.... HI) r,10
-!"i(l .. _.
, 1, \0 .\j ,(I -'!W)
-l ..tHO
I
!"i t O

Fig.6, Measured and reconstructed engine cylinder pressure at zero load, (a)
measured and reconstructed from acoustic signal, (b) measured and reconstructed
from vibration signal, (c) and (d) zoomed figures for (a) and (b) respectively,

f
f
!
no

4 C.\ :
- - X'. . . el\)

_.~_!J:!( ~"~
,c..-..-,,, --=-- 1
_:
~
(h )

~ ] 0
~-----.--.~ -.
"H)

(, . ) (.1)

... ,,,
_
"" .. " . 1._ ".
_
M,IO"
I h" ' ull
\ '
I'\(J
\ l\f~ ' O"
L_ . • __ . - • I

j
S
~
;J"
a : ul

Fig.7. Measured and reconstructed engine cylinder pressure at 40% load, (a)
measured and reconstructed from acoustic signa!. (b) measured and reconstructed
from vibration signal , (c) and (d) zoomed figures for (a) and (b) respectively .
 ,'I} ~" \
l,u 'Il l,
'v~IHI "

I:
M(lOf'.

wiI ., -_.- r-h~("_ o r\

(.; 0 - ~ , ,~, c. f.)I'",

" Jl "

~---------
.",1I i
~
~
~ ·HJ
0' r:
~ :' ()
C. :: '(1

~
i; ll_ -- ....-.~ ~-.-,--- E. ( ' ~- J
_ ~ .() i
I
"
,,_C)
. "'" : . _1....1 _. ,'0
.. ~ I' <J . 1,1 7 ~'()

\ /d ,
fl()
" 1,/llu • .oo:: --, t. Ji4 \ ,I" :i
11"'1'""'1 r \
0 _... 1\':'" 1 ' 1\

t., I
I ', E no
llU
,
~
!I
~
~o
r
!
!
,.!
~
~
'.'
a:.'
~
~!l!
;
r\
:""\.
J

~() ! 20~
~
~
~
L.
/ ,,
I "
' '
, .
o !...--
r . ,' .. - _.. <) .. , - _._- ' !
~ 1 , 1() :JGO :.1$() 420 4!lO "UO GI() :J30 ::)60 ::\ !=)o 'I2u 4"U 4AO ~'O

(_,I ' f)r-IK 1\.11~lo iO(.lc;J'I:.Iul (;"'Hnlo. 1\11~JIH Idu ~ ,rAH,

Fig,B. Measured and reconstructed engine cylinder pressure at full load , (a)
measured and reconstructed from acoustic signal, (b) measured and reconstructed
from vibration signal, (c) and (d) zoomed figure for (a) and (b) respectively.

(U) (il)

x-
(10,
B' r ....... No 101\,1- -·"1
~
(10
-_.-; ''---~. . ~~,~otu' ,I
tr
~

~
GO
40!
40'\, 1u..<.1
lull '()dd ....J
I
j
e'
~
a.
:1
20,
i - .-::: __~II Iood I·
§ ~
~ ::1 __ . .
0 ltlO
_._-_ •.. _ - -
360 540
,_._1
720
~ -20
0.
1
:l3O 360 :1no 420 4~~ 400 510 540

1!
l
GO i

40
,
(e)

-:I--=::'=:~o
-
loa" '-=-'1
40" ·0 1000d 0
j
~
60~
40
-
(<.J)

-.--:~~.- . -:/::d]
[ 2 0·
L:-'-' ~~ olRd .
[ 20 ·
-.-. full load
,---- ._-- - ,

~ 0 '.
-..::------ i
~ -'. -. ......- ....
., -.-~-.-

a- -20 --- ----_. -. - _ _ _ 1

~ -2 0 ..... ,
0 HIO :)(iO (AO 1:20 330 3GO 300 4 20 11 .~,O AUO ,010 !:.1<10
(c) (n
60 , nO
~. , .... .-.-.-:- No-loo,j " -r, ~ lo~(j II
~.

~
40:

20 !
_ 40 a ·CI load
I.,....-------,.!
luiliolld ,
~
~
~
40

20 ·
I 1=
._-
No
40" .. lonn (
lull I<lDd

~ 0
I f
~ -_._ ---_.. ,_ .. _. ,_. ~._..J ~ ·;>0 ' '- "'- - - , -- ----
HlD 3Ril !>40 no 330 :160 390 ~20 450 100 !J10 540
Cronk Angle- [<i"gree[ C,arlk Angle [degree I

Fig.9. The measured cylinder pressures and reconst:-ucted ones at zero load, 40%
load and full load, (a) measured , (b) Zoom of measured, (c) reconstructed from
acoustic signals, (d) Zoom of reconstructed from the acoustic signals, (e)
reconstructed from vibration signals, (f) Zoom of reconstructed from vibration signals .
 -

10. CONCLUS!O!I..JS

In this paper tile reconstructloll or ill -cylinder pressure using Inverse filterlllg
technique is presented. 11 is shown Ulat I'econ~\ructlng the in-cylinder pressure f'Tlin
acoustic and vibration measurements has been successfully achieved by the
mentioned technique

Reconstruc.tlc;'; of cyi;(;der p,-essure by Inverse filtG;-;J'ig tecil'-ilqu8 Cd('; be appilOcl for

,..""""1""'11.,; __ . . 1''''1 ............ l.-... ,....,,.... ... ,...J ,....r '_~ r/"'\I ,...,n""': .. "''1:,,+;,...,..... .....:J",t,.~,....~, ..... t""'< ;,- ......... , • • I~, , . . \,1,,, . . ......:,, \ ,· ;, .... ,,.... ............. ,.. ",h;I.··,
l·~c..JJ- · ~lll~"-' V,i-Uv.:.J,"" r......V,i ... ~VI UIIU i i "...)lliC •........ '.:,;\\.,.. __ .. i V i . Ii, , ,;·~t;~i-·'--'i. I.:""""''-'1 viI8IJIc..":,, \~'IJil·......

eilmlnatlng tne need tor in-cylinder pressure transducers

The reconstructed cylinder pressure from vibration Signal at constant speed and
different loads is the same as that reconstructed from the acoustic signal at that
speed and loads. In otller words the reconstructed cylinder pressures from acoustic
and vibration signals have the same root mean square error.

To apply reconstruction of cylinder pressure in engine diagnosis and control , it is

needed to evaluate a map of FRF 's that represents the correlation between engine
vibration and acoustics, and combustion pressure over a wide range of engine
speeds and loads.

REFERENCES
1. Zeng, P. and Deenis N. Assanis, "Cylinder Pressure Reconstruction and its
Application to Heat Transfer Analysis" , SAE paper, 2004-01-0922, (200 '~J .
2, Zeng, P., Robert G. Prucka, loran S. Filipi and Dennis N. Assanis,
"Reconstructing Cylinder Pressure of a Spark-Ignition Engine for Heat Transfer
and Heat Release Analyses", Proceedings of ASME, ICEF 2004-886, (2004) .
3. Citron S. J., John E, O'Higgins and Lillian Y. Chen, "Cylinder by Cylinder Engine
Pressure and Pressure Torque Waveform Determination Utilizing Speed
Fluctuations", SAE 890486, (1989).
4. Brwon, T. S. and W. Stuart Neill, "Determination of Engine Cylinder Pressures
from Crankshaft speed Fluctuations", SAE 920463, (1992).
5. Moro, D. and Enrico Corti, "Real Time In-Cylinder Pressure and Torque
Estimation In a 6 Cylinders Turbocharged Engine", Online,
www.machineunisa .itlMECA02/papers/MECA02_paper01 .pdf
6. EI-Gamry, M., J. A. Steel, R. L. Reuben and T. L. Fog, "Indirect Measurement of
Cylinder Pressure from Diesel Engines using Acoustic Emission", Mechanical
Systems and Signal Processing, Accepted 27 September 2004 , (2004).
7, McConnell, Kenneth G . Vibration Testing : Theory and Practice, John Willey &
Sons, INC ., First edition, New York, (1995)
8. Rao, Singiresu S " Mechanical Vibration, 41h edition, Prentice Hall, New Jersey,
(2004).
9
, LV\."i-
t:. -~ :)
I . ,~. , _, l
J I
1\1!(,,\(,1. ' T'.·.lC'~ -., ( ,!
!\f ....."J8 1 I \ . ..... IJII :..:; 1'~1. , -~i"- ' ,}
"" " ,1,
! ", r '":"I r-',I r"'. '
t.. · ........ '- , ·, ....... '" _",
",'
I
·' !~ \·II'r""~~j(··"
L1t'jl- I"'-~I." l".) ."
')nd
_ e-}dl'"lIG,l,I
I :'\s-.!
c>\.(J'- Ll,~ r"'!
.......,;
Studies Press LTO , England, (2000)
10. Ziemer R E., William H. Tranter and 0 Ronald Fannin : "Signal and Systems
Continuous and Discrete ", 4 111 edition, Prentice Hall, (1998).
11 SAE J10"74 FEB87, ''Engine Sound Level Measurement Procedure (A)". SAE
HandBook, Volume 3, (1989).

Appendix (A)

Engine Specifications

Engme model: F 1L-S11

Number of strokes 4
Bore. 100 mm
Stroke : 105 mm
Swept volume: 825 cm 3
Compression r8tio: 17: 1
Cooling ' Air cooling

Appendix (B)

Calibration Curve of Pressure Transducer

35[. ---, - . •.._, ..J. --...::..... . =-_- -r:::. -- "-',

, V01i v,; I-'ress I "
1-:111 I
--- -- -- ~

:I I: . / :~
, j ~~
L
?:) r .. 1 :,
I·

~ I: '-/- . :./
,:
::.<:>...
,J
1
I:
\ J.
I:
I
0[, ..

-lj
o lL. .. J ... _ ... . . __ I _
-.. . ..
(\ \I":, 00'
? 'J :!5