Professional Documents
Culture Documents
Diagnostic Methodology in Four-Stroke Marine Diesel Engine by Identifying Operational Parameters
Diagnostic Methodology in Four-Stroke Marine Diesel Engine by Identifying Operational Parameters
Diagnostic Methodology in Four-Stroke Marine Diesel Engine by Identifying Operational Parameters
net/publication/328007937
Article in Journal of the Brazilian Society of Mechanical Sciences and Engineering · October 2018
DOI: 10.1007/s40430-018-1416-x
CITATION READS
1 253
4 authors, including:
Ulisses A. Monteiro
Federal University of Rio de Janeiro
26 PUBLICATIONS 45 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Optimisation of marine energy systems for liquefied natural gas carriers View project
All content following this page was uploaded by Ricardo Gutierrez on 19 October 2018.
TECHNICAL PAPER
Abstract
For the purpose of assisting in the maintenance of diesel engines installed in ships and offshore platforms, a methodology
was developed to detect, identify and quantify faults of its components. A thermodynamic zero-dimensional model of a
6-cylinder marine diesel engine was constructed taking into account the characteristics of the engine operation. The fuel
burn rate was modeled using the double function of Wiebe, and because the performance was of interest, the emissions
were ignored. The thermodynamic model was validated with data obtained experimentally. A methodology has been imple-
mented that allows simulating four types of faults, based on the developed thermodynamic model. A sensitivity analysis
was performed to identify which failures significantly affect the performance parameters. Finally, the diagnostic model was
based on parameter estimation using the Levenberg–Marquardt method. Results showed a high degree of confidence in the
identification of all the simulated faults.
13
Vol.:(0123456789)
500 Page 2 of 10 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500
md Chamber shape factor in the diffusive combustion parameters for which it was designed, since any change in
phase (–) performance can worsen the toxic emissions levels [7–9].
n Polytropic coefficient of air, 1.35 In diesel engine faults identification, the use of thermo-
nc Combustion efficiency, 0.99 dynamic models, experimentally calibrated, is an important
rAFR Air/fuel ratio real mass (–) tool in the development of methodologies that allowed to
vp Average speed of the piston (m/s) diagnose components failures during their operational life
x Fraction of burned mass of fuel (–) [5, 8, 10].
xi Molar fraction (–) This work is specifically aimed at the diagnosis of 4 (four)
xp Fraction of fuel burned in the premixed combus- thermodynamic faults in a 6-cylinder marine diesel engine.
tion phase (–) Thus, in order to know the actual operating conditions of
xd Fraction of fuel burned in the diffusive combustion the diesel engine, it is necessary to develop methodologies
phase (–) that allow detecting, identifying and quantifying the severity
βe Emissivity of the implanted faults, even when they are still incipient.
λ Air/fuel equivalence ratio (–)
𝜃 Crankshaft angle (°)
𝜃i Injection start angle BTDC (º) 2 Simulation of the marine diesel engine
θig Ignition start angle BTDC (º) operation
Δθp Premixed combustion duration (º)
Δθd Diffusive combustion duration (º) The engine performance was based on zero-dimensional
Δθ Total combustion duration (º) modeling, which takes into account the manufacturer’s
τ Ignition delay (°) operating variables, which can be found in the diesel engine
σ Boltzmann constant (W/m2 K4) operating manual, to provide maximum torque and power
[11, 12].
2.1 Thermodynamic modeling
1 Introduction
The zero-dimensional model ranges from the opening of the
In the offshore industry, where the daily cost of operation inlet valve to the closure of the exhaust valve.
rises to exorbitant amounts, unexpected production outages
can mean huge economic losses. In addition, the unexpected 2.1.1 Analysis of gas inside the cylinder
equipment’s failure can produce accidents, causing structure
damage and also serious risks for the crew, as well as the As the goal of this work is to simulate the performance only,
resulting environmental impacts. the mixture of gases inside the cylinder during the process
In this context, it is important that companies invest in of admission and exhaustion was considered at constant
a good predictive maintenance plan for the engines, since pressure. Then, the compression stage can be considered
this allows them to follow up on the operation, thus avoiding as a polytropic process, Eq. 1, and other processes can be
sudden production halts. considered as that of an ideal gas [11, 12], Eq. 2:
The performance of diesel marine engines depends on the
operation of their components, which are determined by the dP P dV
= −n (1)
manufacturers to provide the maximum torque and power d𝜃 V d𝜃
for which they were designed [1, 2].
( )
In this perspective, the good functioning of the engine dP dT dV 1
= mR −P (2)
depends directly on the good functioning of its components, d𝜃 d𝜃 d𝜃 V
and any failure that occurs in one of them can modify the
performance or, in the worst case, break the machine [3]. 2.1.2 Thermodynamic analysis
The fuel used in the combustion itself can cause damage,
such as piston corrosion, corrosion of the injector nozzle Applying the first law of thermodynamics and consider-
[4], and deposits of combustion residues can also clog the ing that the mass flow during the process of admission and
injector nozzle. In addition, there are the electronic compo- exhaustion is stationary, Fig. 1, we have
nents that control the injection system, and due to the highly
dU 𝛿Q 𝛿W
severe environment, its calibration can be impaired [5]. = − (3)
d𝜃 d𝜃 d𝜃
Finally, due to requirements of environmental standards
[6], it is sought that the diesel engine operates within the
13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500 Page 3 of 10 500
( )
dT 𝛿Qt 𝛿Qw dV dc 1
= − −P − mT v (10)
d𝜃 d𝜃 d𝜃 d𝜃 d𝜃 mcv
and the term d𝜃w is due to the transfer of heat through the
𝛿Q
Δ𝜃p = − 13.2 − 1.64 × 10−2 Pr
walls of the cylinder and can be divided into convection and
radiation heat transfer [11, 12] and can be calculated using + 3.82 × Pi + 7.3 × 10−2 × Ti (14)
Newton’s law of cooling: +
0.22
+ 3 × 10−3 N
( ) 𝜆
𝛿Qc hA T − Tw
= (7)
d𝜃 6N The duration of the premixed and diffusive combustion
( ) results in the total combustion process duration and can be
𝛿Qr 𝛽e 𝜎A T 4 − Tw4 estimated by the following correlation [16]:
= (8)
d𝜃 6N
50
The coefficient 𝛽e was adopted as 0.56 [13], and the coef- Δ𝜃 = 30 +
rAFMR × 0.06691 × 0.7 (15)
ficient h was calculated from the Eichelberg correlation [14]:
then,
h = 7.8 × 10−3 P0.5 Ti0.5 v1∕3 (9)
i p Δ𝜃d = Δ𝜃 − Δ𝜃p (16)
Finally, by algebraic manipulation, one arrives at the dif- In Eq. (11), the values of a, mp, md, xp and xd were con-
ferential equation which calculates the gas mixture tempera- sidered constant and equal to 4.605, 3, 1, 0.18 and 0.82,
ture inside the cylinder: respectively [13]. Therefore, by knowing x, one can calculate
the rate of heat released by the burning fuel, Eq. (17):
13
500 Page 4 of 10 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500
2.2 Faults simulation
3 Diagnostic methodology development
There are several factors that influence the diesel engine per-
formance and may lead to deviation from its normal operat- The main contribution of the present work is found in this
ing condition [3–5, 7]. In addition, faults can be classified in item. This methodology can be applied in online systems
two ways: global, when the fault affects all cylinders simul- for monitoring and diagnosing of diesel engines installed in
taneously, and local, when the fault occurs in one cylinder, the naval and offshore industries, helping, in this way, the
while the others operate normally. predictive maintenance plain.
It should be noted that malfunction of some engine’s With this aim, in order to develop a model that allows
components causes variations in performance, often with to detect, identify and quantify the failures, the Leven-
similar effects [4]. So, the diagnostic methodology must be berg–Marquardt method [21] was used. This technique cal-
sufficiently robust to be able to identify the faults correctly. culates the “best estimate” of the changes that occurred in
The only independent variable of the thermodynamic the components or subsystems of the diesel engine, which
model is the crankshaft angle; however, performance resulted in the measured parameters.
13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500 Page 5 of 10 500
13
500 Page 6 of 10 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500
0-D Model
Geometric Thermodynamic
Characteristics Intake Pressure Analysis
Operation State Equation
Intake Temperature
Condition of Gas Mixture Solve
Fuel Wall Temperature Model of
Consumption Wiebe Parameters Combustion
Exhaust
Dinamometer Pressure VS Crankshaft Angle
Tachometer 150
Fuel Sensor Sim.
Inlet
Pressure (bar)
Exp.
100
Fuel Pressure
Rail Sensor Processing
50
Fuel 0
Outlet 6-Cylinder 300 400 500
Common Rail Diesel Engine
Fresh Crankshaft Angle (°)
Air
13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500 Page 7 of 10 500
13
500 Page 8 of 10 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500
-5
-10
Variation (%)
-25
-30 Power
Torque
-35
Faults
Global faults were simulated as 10% of pressure drop 4.3 Diagnostic model results
in the intake manifold and 15% of common rail pressure
drop, separately. It can be observed in Fig. 5 that the first Different fault severities were simulated, generating the
fault affects the pressure curve more significantly, due to vector Z of Eq. (24), and then, the diagnostic model was
the lower pressure at the beginning of the compression, and used to estimate the fault vector, f̂ . Figure 7 presents a
when reaching the end of this process it does not reach the schematic of the procedure from the faults simulation to
ideal pressure conditions to start the combustion, despite the faults identification. Tables 2 and 3 present the results
having the same amount of fuel injected. In addition, the obtained in diagnosis of failures that occurred in cylinder
second simulated global failure has a smaller effect on the 1 and in six cylinders simultaneously, respectively.
pressure curve, but an increase in the maximum pressure {{ } { } { } }
is observed, maybe due to the fact that the ignition delay Z = Pmi,j , Pmax,j , Pj (𝜃) , Pw , Tq (31)
and premixed combustion duration, Eqs. (13) and (14), are
smaller than normal, resulting in an increase in maximum In the results presented in Tables 2 and 3, it can be
pressure. observed that the methodology developed has a high
When simulating local faults, 15% compression ratio degree of accuracy having errors smaller than 5% when the
reduction and 40% injected fuel mass reduction were con- failures occur in a single cylinder and less than 10% when
sidered in all cylinders, separately. In Fig. 5, it can be seen the failures occur in several cylinders simultaneously.
that the compression ratio reduction has a greater effect on Regarding the error increase in the diagnosis, it is
the pressure curve, due to not reaching the ideal pressure emphasized that this is due to the Levenberg–Marquardt
at the end of the compression process. With respect to the method that estimates the vector of faults, f̂ , in the least
second simulated local fault, it is observed that the maxi- squares sense, to be closest to the actual vector of faults, f.
mum pressure presents slight variation, while the width of It should be noted that the vector Z in Eq. (31), which
the pressure curve is smaller. This is because the smaller in the present work was simulated by the thermodynamic
amount of fuel present in the cylinder leads to a shorter time model, can be obtained experimentally in diesel engines
of total duration of combustion (Eq. 15), and consequently, installed in the naval and offshore industries, provided that
lower heat releases. they have the proper instrumentation as they are mainly
The changes in the pressure curves due to failures lead pressure sensors installed on cylinder heads, cooling sys-
to variations of the Pmax and Pmi producing changes in the tem temperature sensors, tachometer to measure engine
power and torque of the motor, as shown in Fig. 6. Note speed and torque and power measurement system (e.g.,
that depending on the source and severity of the failure, the strain gages installed on the propeller shaft). It should be
effects on the performance parameters can be similar and noted that some of these data can be obtained directly from
may make the diagnosis difficult. the monitoring system installed onboard the units.
13
Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500 Page 9 of 10 500
Pressure (bar)
100
Variation of Parameters
Operating Identified
Parameters 50
ΔP i, Δ Pr, Δ T w, Δ r { f } + error
Δm i, Δm f, Δ θinj 0
300 400 500
Crankshaft Angle (°)
Variation ofPerformance Parameters
P(θ), P mi, Pmax
Table 2 Results of the Case Simulated fault severity (%) Diagnostic fault severity (%)
diagnostic methodology applied
in cylinder 1 ΔPi ΔPr Δr Δmf ΔP̂ i ΔP̂ r Δ̂r Δm
̂f
Table 3 Results of the Cylinder Simulated fault severity (%) Diagnostic fault severity (%)
diagnostic methodology applied
to the six cylinders ΔPi ΔPr Δr Δmf ΔP̂ i ΔP̂ r Δ̂r Δm
̂f
5 Conclusions and recommendations a situation that can worsen when global failures occur
simultaneously.
The methodology used for the development of the thermo- The developed methodology proved to be robust enough
dynamic model satisfactorily represented the experimental to detect, identify and quantify multiple failures in several
data, allowing to make of performance predictions of the cylinders simultaneously.
engine in conditions not tested experimentally. It is also recommended to include other operational
It is noteworthy that global failures influence all cyl- parameters in the fault vector in order to have a better
inders simultaneously, i.e., their occurrence influences diagnosis of the diesel engine.
engine performance more significantly. The inclusion of a model for pollutant emissions esti-
On the other hand, local failures influence performance mation would be an important extension of the present
to a lesser extent when they occur in isolation. However, work, since the influence of the injection advance delay
if more than one local fault occurs simultaneously, in one on both the performance and the emissions of the diesel
or more cylinders, performance is significantly influenced, engine could be evaluated.
13
500 Page 10 of 10 Journal of the Brazilian Society of Mechanical Sciences and Engineering (2018) 40:500
An important extension of this research is the devel- 9. Bahri B, Aziz AA, Shahbakhti M, Muhamad MF (2013) Said,
opment of a dynamic model to incorporate the torsional Understanding and detecting misfire in an HCCI engine fuelled
with ethanol. Appl Energy 108:24–33. https://doi.org/10.1016/j.
vibration signal into the fault vector, f, in order to identify apenergy.2013.03.004
a greater number of failures, mainly in several cylinders 10. Tinaut FV, Melgar A, Laget H, Dominguez JI (2007) Misfire and
simultaneously. compression fault detection through the energy model. Mech Syst
Signal Process 21:1521–1535. https://doi.org/10.1016/j.ymssp
Acknowledgements This work was possible due to the sponsorship .2006.05.006
of the Human Resources Program (PRH-35), through its scholarship 11. Heywood J (1988) Internal combustion engine fundamentals.
program. The experimental data were performed and provided by Eng. Mcgraw-hill, New York
Alexandre Schalch Mendes, M.Sc. 12. Stone R (1992) Introduction to internal combustion engines.
Springer, Berlin. https://doi.org/10.1007/978-1-349-22147-9
13. Souza Junior GC (2009) Thermodynamic modeling of diesel
engines working with diesel oil and biodiesel to check the perfor-
References mance parameters and emissions. Masters dissertation. Federal
University of Rio de Janeiro
1. Mollenhauer K, Tschoke H (2010) Handbook of diesel engines. 14. Rousseau S, Lemoult B, Tazerout M (1999) Combustion charac-
Springer, Heidelberg. https: //doi.org/10.1007/978-3-540-89083- 6 terization of natural gas in a lean burn spark-ignition engine. Proc
2. Dempsey P (2008) Troubleshooting and repairing diesel engines. Instit Mech Eng Part D J Automob Eng 213:481–490
McGraw-Hill, New York. https://doi.org/10.1036/0071493719 15. Maroteaux F, Saad C (2013) Diesel engine combustion modeling
3. Benvenuto G, Campora U (2007) Performance prediction of a for hardware in the loop applications. Energy 57:641–652
faulty marine diesel engine under different governor settings. In: 16. Ghazal OH (2013) Performance and combustion characteristic
2nd International conference on marine research transportation, of CI engine fueled with hydrogen enriched diesel. Int J Hydro-
pp 35–44 gen Energy 38:15469–15476. https://doi.org/10.1016/j.ijhyd
4. Hountalas DT, Mavropoulos GC, Kourbetis G (2006) Experimen- ene.2013.09.037
tal investigation to develop a methodology for estimating the com- 17. Merker PG, Schwarz C, Teichman R (2009) Combustion engines
pression condition of DI Diesel engines. Energy Convers Manag development. Springer, Berlin. https: //doi.org/10.1007/978-3-642-
47:1–18. https://doi.org/10.1016/j.enconman.2005.04.001 14094-5
5. Hountalas DT (2000) Prediction of marine diesel engine perfor- 18. Kuo KK (2005) Principles of combustion. John Wiley
mance under fault conditions. Appl Therm Eng 20:1753–1783 19. Rakopoulos CD, Hountalas DT, Tzanos EI, Taklis GN (1994) A
6. IMO, Marpol 73/78 (2009) MARPOL - International convention fast algorithm for calculating the composition of diesel combus-
for the prevention of pollution from ships tion products using 11 species chemical-equilibrium scheme. Adv
7. Alfieri E, Amstutz A, Guzzella L (2009) Gain-scheduled model- Eng Softw 19:109–119
based feedback control of the air/fuel ratio in diesel engines. 20. Gutiérrez RHR (2011) Study on the performance of an ottolized
Control Eng Pract 17:1417–1425. https://doi.org/10.1016/j.conen diesel engine working with natural gas through thermodynamic
gprac.2008.12.008 simulation and experimental analysis. Federal University of Rio
8. Grados VD, Uriondo Z, Clemente M, Espadafor FJ, Gutiér- de Janeiro, Rio de Janeiro
rez JM (2009) Correcting injection pressure maladjustments to 21. Ozisik MN, Orlande RBH (2000) Inverse Heat Transfer. Rout-
reduce NOX emissions by marine diesel engines. Transp Res ledge, New York
Part D Transp Environ 14:61–66. https: //doi.org/10.1016/j.
trd.2008.09.012
13