Evaluation of Alternative Propulsion Concepts for Mobile Machinery:

A Modelling Approach using the Example on LNG-powered Port

Handling Equipment

Patrick Driesch, Kai Horwat, Niko Maas and Dieter Schramm

Chair of Mechatronics, University of Duisburg-Essen, Lotharstraße 1, 47057 Duisburg, Germany

Keywords:

Semi-physical Modelling, Evaluation of Alternative Propulsion Concepts.

Abstract:

The prevailing climate goals and related emission limits for ﬂeets are forcing industry to consider alternative

fuels alongside established fuels such as diesel and gasoline. A conversion of a ﬂeet of vehicles to another

source of power comes along with an investment of money and a risk that the day-to-day operations of the

new drive technology do not provide the expected effects regarding the fuel consumption and the exhaust

gas emissions. In order to assess these barriers in advance, this paper presents an approach of a simulation

tool based on a semi-physical modelling of the powertrain of mobile machinery to predict the impact of a

conversion of a ﬂeet from diesel to liqueﬁed natural gas (LNG) in case of fuel consumption and exhaust gas

emissions. For the semi-physical model, a combination of a physical model of the vehicle and powertrain

dynamics and a black box modelling of the internal combustion engine by artiﬁcial neural networks is chosen.

The simulation tool will be used in the future to assess the feasibility of converting the drive system from port

transhipment facilities to LNG. The work was carried out as part of the research project LeanDeR ﬁnanced by

the European Regional Development Fund (ERDF).

1 INTRODUCTION

The climate targets of the European Union pose ma-

jor challenges for the development and the use of

vehicles. By the year 2030 greenhouse gas emis-

sions have to be reduced by 40 % compared to the

level of 1990 and by 2050 climate neutrality is to

be achieved, (European Commission, 2018). In ad-

dition to the reduction of greenhouse gas emissions,

it is also necessary to reduce the emission of air pol-

lutants. Beside the industry and energy sector, the

trafﬁc and transport sector is a signiﬁcant producer of

emissions, (International Energy Agency, 2019). In

Germany, e. g., the trafﬁc and transport sector emit-

ted about 18.4 % of the energy-related greenhouse

gas emissions in 2017, (Federal Ministry for the En-

vironment, Nature Conservation and Nuclear Safety,

2019). The term transport is often conﬁned to the cat-

egory of road vehicles, as these are used in the imme-

diate vicinity of the population and their living envi-

ronment and can therefore be directly observed. An-

other segment of vehicles that differs in this respect

is the so-called mobile machinery. Unlike road vehi-

cles, which are primarily used to transport goods from

one place to another, these machinery have the task

to perform mechanical work in addition to the trans-

port task, (Geimer and Pohlandt, 2014). A study by

(Helms et al., 2017) shows that these machinery emit

a non-negligible proportion of greenhouse gas emis-

sions and air pollutants. Achieving the mentioned

goals will require a conversion from established fossil

fuels towards alternatives which provide less exhaust

gas emissions. Suitable alternatives are, e. g., natural

gas or an electriﬁcation or hybridisation of the pow-

ertrain, (Milojevi

c et al., 2018; Lajunen et al., 2018).

All these energy sources have in common, that there

is a lack of sufﬁcient tank or charging infrastructure

to enable a changeover at short notice. In addition,

due to the fact that most of the mobile machinery is

diesel-driven and many different types of vehicles are

included in this section, there is an insufﬁcient re-

search on whether and how everyday operation with

alternative fuels can be performed. In order to coun-

teract this, the operation of a multimodal liqueﬁed

natural gas (LNG) station infrastructure is tested in

the port of Duisburg, called duisport, as part of the

research project LeanDeR. To investigate the suitabil-

ity of LNG for everyday use, two mobile machinery

from the entire duisport ﬂeet were tested with natural

gas and compared with a similar diesel-powered ve-

Driesch, P., Horwat, K., Maas, N. and Schramm, D.

Evaluation of Alternative Propulsion Concepts for Mobile Machinery: A Modelling Approach using the Example on LNG-powered Port Handling Equipment.

DOI: 10.5220/0009872502250232

In Proceedings of the 10th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2020), pages 225-232

ISBN: 978-989-758-444-2

225

hicle. For this purpose, measurement data from the

LNG and the diesel-driven vehicles were recorded by

autonomous data-logging systems whilst their every-

day operation. One of these vehicles, a terminal trac-

tor (see ﬁgure 1), is used to manoeuvre trailers around

the port area. The terminal tractor considered in the

course of the research project is powered by LNG in

a mono-fuel operation.

Figure 1: Diesel- and LNG-driven terminal tractors at the

duisport, (Duisburger Hafen AG, 2019).

Since the operation of a LNG ﬁlling station be-

yond the end of the project at the end of May 2020,

only for the refueling of two vehicles, appears uneco-

nomical, it must be examined whether a ﬂeet-wide

switch from diesel to natural gas would bring both

economic and ecological advantages. Because a fail-

ure of the machinery directly extends to further pro-

cess steps in the form of process follow-up costs, a so-

lution is needed to weigh up a ﬂeet-wide changeover

with low risk and economic expenditure simultane-

ously. In the course of this paper, a simulation tool

will be presented using the example of the terminal

tractors. This tool enables the upscaling for a com-

plete changeover of a ﬂeet of mobile machinery to

an alternative fuel. Based on the detailed database

of the individual measured vehicles, semi-physical

models of the vehicles' powertrains need to be devel-

oped. These models shall predict the fuel consump-

tion and exhaust gas emissions of the other vehicles

of the entire ﬂeet in the respective powertrain. For

the rest of the ﬂeet only the knowledge of the vehicle

speed, its payload and the current ambient tempera-

ture is required. The vehicle speed can be recorded

during daily operation without a great effort. To es-

timate the payload of the port handling equipment,

an analysis of the performed manoeuvring orders can

be achieved from the terminal operating system. In-

formation about the ambient conditions can be taken

from accessible weather databases. By a fusion of

these information depending on the respective time

stamp the required model input is provided.

In section 2 the methodical approach to the de-

velopment of the simulation tool is described. After-

wards the modelling is explained in section 3. Finally,

section 4 provides a summary and outlook for future

work.

2 METHODICAL APPROACH

The objective of a simulation environment for predict-

ing a ﬂeet-wide changeover from diesel-driven termi-

nal tractors to natural gas propulsion as described in

section 1 requires a systematic approach. For this pur-

pose, the following steps based on (Verein Deutscher

Ingenieure e.V., 2016) are fulﬁlled:

1. Formulation of tasks and objectives

2. Structural and functional analysis

3. Data collection and analysis

4. Determination of the relevant model aspects

5. Problem decomposition

6. Determination of the model type

7. System and process description

2.1 Formulation of Tasks and

Objectives

The goal of the simulation tool is to estimate unknown

process variables of the remaining ﬂeet vehicles in

everyday operation with conventional and alternative

fuels. For this purpose, the operation of the vehicles

must be simulated with one diesel and one natural gas

powertrain each, so that the resulting fuel consump-

tion and exhaust gas emissions for both drive types

can be concluded. Afterwards the sum of the fuel con-

sumption and exhaust gas emissions of the whole ﬂeet

have to be calculated.

2.2 Structural and Functional Analysis

The total fuel consumption m

F, p

of the vehicle ﬂeet

for p different powertrains as well as the correspond-

ing masses of exhaust gas emissions m

, p

of chemi-

cal compounds α like CO

occur as unknown process

variables and thus as output variables of the simula-

tion tool. As mentioned before, the velocity v and

the payload m

Trailer

of the terminal tractors as well as

the current ambient temperature T

Amb

during every-

day operation are used as input variables. These daily

operations can be subdivided into several rides from

one shutdown of the engine to the next. Each ride

consists of a time-dependent vector of inputs, so that

for an assumed ﬂeet of n vehicles with m rides each

∑

i=1

m(i) time-dependent input vectors are available.

Thus, for an estimation of the fuel consumption and

SIMULTECH 2020 - 10th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

226

exhaust gas emissions of a whole ﬂeet, the simulation

process has the structure illustrated in ﬁgure 2.

Every ride needs to be simulated with p different

powertrains. Therefore a simulation framework pre-

dicts the mass ﬂow of the fuel ˙m

F, p

and the mass ﬂows

˙m

, p

of the corresponding exhaust gas emissions of

chemical compounds α for every ride with p power-

trains.

Ride

of rides

of Veh

i,j

Trailer,i,j

Simulation

tool with

powertrains

Fleet of

vehiclesn

Veh

while ≤j m

while ≤i n

E ,p, ,ji

F,p,i,j

∑

i=1

∑

j=1

. .

∫ td

E ,p,i,j

F,p,i,j

E ,p

F,p

Postprocessing

Amb,i,j

Figure 2: Structure of the simulation process.

In total p ·

∑

i=1

m(i) simulations must be carried

out. To estimate the total mass of fuel and exhaust

gas emissions for the whole ﬂeet for p powertrains,

the mass ﬂows are integrated and added up for every

powertrain in a postprocessing.

2.3 Data Collection and Analysis

The development of simulation models requires

knowledge about the system which shall be described.

This is gained from the analysis of the everyday

operation of the reference vehicles during a mea-

surement period from January 2019 to May 2020 in

course of the research project LeanDeR. The vehi-

cles were equipped with self-sufﬁcient data-logging

systems which permanently recorded measurements

of the ambient temperature, the GPS data set, the ac-

celerations in three space dimensions as well as data

from the CAN bus of the vehicles with a measure-

ment frequency of 1 Hz. Thus measurements like the

engine speed, its torque, the temperature of the en-

gine coolant, the massﬂow of exhaust gas recircula-

tion, the massﬂow of intake air and the massﬂow of

consumed fuel were available. Furthermore, the vol-

ume concentration of the chemical compounds CO,

, NO

, CH

and SO

of the exhaust gas emissions

were measured at the exhaust pipe with an exhaust

gas emissions measurement system of type J2KNpro

from the manufacturer ecom. Since permanent opera-

tion of the exhaust gas analysis device is not possible

due to protective mechanisms in the device against the

poisoning of individual sensors and a renewed cali-

bration of the sensors to prevent measurement drift,

measurements were performed at speciﬁed times. To

determine the mass ﬂows of the respective exhaust

gas component, a mass-based calculation according

to (European Parliament and the Council, 2017) was

carried out. At the end of February 2020 the data base

consisted of 11,460 km driving distance and 1,244 h

driving duration of the diesel-driven terminal trac-

tor. For the LNG-driven terminal tractor 11,909 km

driving distance and 1,237 h driving duration were

recorded. A distance of 261.3 km at a driving dura-

tion of 27.3 h were driven by the LNG-driven termi-

nal tractor while an exhaust gas emissions measure-

ment was executed. For the diesel-driven terminal

tractor 138.9 km driving distance at 14.1 h of driv-

ing time during exhaust gas emissions measurement

were recorded. In addition to the measurements of

the motion of the vehicles, information regarding the

load of the vehicles at speciﬁc timeslots can be taken

by the terminal operating system. This system orga-

nizes the cargo handling and provide the gross weight

of handled container which are placed on the trail-

ers. These gross weight of the container includes the

weight of the container itself and the goods inside the

container, (Zhao et al., 2020).

2.4 Determination of the Relevant

Model Aspects

Internal variables are required to implement the rela-

tionship between the input and output vectors shown

in the system structure in ﬁgure 2. An analysis of the

physical relationship between the inputs and outputs

points out, that the combustion engine of the vehicles

acts as the essential interface. It converts the chem-

ical energy of the fuel by internal combustion into

mechanical power on the crankshaft, causing exhaust

gas emissions. This mechanical power is transferred

through the powertrain to provide the required driving

force at the wheels. Accordingly, the engine speed

Engine

and its torque M

Engine

are deﬁned as internal

variables of the simulation tool.

2.5 Problem Decomposition

By considering vectors of internal variables, a deﬁned

connection exists between the input and output vec-

tors of the simulation tool. Thus the overall simula-

tion model can be subdivided into two subsystems for

every powertrain p, as ﬁgure 3 illustrates, where the

internal variables are the output of the ﬁrst and the in-

put for the second subsystem. Thus both subsystems

can be developed in parallel and optimized speciﬁ-

cally with the available database. The ﬁrst subsystem

includes the modelling of the dynamics of the vehicles

Evaluation of Alternative Propulsion Concepts for Mobile Machinery: A Modelling Approach using the Example on LNG-powered Port

Handling Equipment

227

and their powertrains to determine n

Engine

and M

Engine

by v, T

Amb

and m

Trailer

. To estimate the mass ﬂows

of the fuel consumption ˙m

and the chemical com-

pounds of the exhaust gas emissions ˙m

, the second

subsystem represents a model to describe the com-

bustion process of the engine of the respective pow-

ertrain. This model uses the outputs of the ﬁrst sub-

system and also some of the global inputs, e. g., the

ambient temperature T

Amb

as inputs.

Figure 3: Problem decomposition into two subsystems for

every powertrain.

2.6 Determination of the Model Type

The subsystems can be developed with different mod-

elling methods. According to (Schramm et al., 2018),

the dynamic behaviour of real processes is carried out

either by theoretical modelling based on physical laws

or by experimental modelling based on measurements

of inputs and outputs. A modelling by physical laws

grants the advantages that the results of the simulation

are easier to interpretate physically, (Dubois, 2018).

Thus a better understanding of the model is given

but unknown physical parameters have to be identi-

ﬁed. Experimental modelling, also called black-box

modelling, based on a mathematical description of the

relationship between the input and output measure-

ments by, e. g., an artiﬁcial neural network (ANN) al-

lows modelling of the system without sound process

knowledge, (Xu, 1997).

For the development of the simulation tool a semi-

physical modelling, i. e. a combination of both mod-

elling forms, is chosen. Because some physical pa-

rameters of the vehicles' mechanical powertrain are

provided by the manufacturer, for the ﬁrst subsys-

tem a physical modelling approach is applied. Since

many different drivers with a variety of driving styles

are working with the terminal tractors, a backward-

facing quasi-stationary model of the vehicles' longi-

tudinal dynamics and powertrain is used. This type

of model does not require a driver model and the nec-

essary torque and speed of the engine is calculated

backwards through the powertrain based on the power

requirement at the wheels to overcome the driving re-

sistances, (Mohan et al., 2013; Wipke et al., 1999).

The mathematical modelling of internal com-

bustion engines is complex and requires knowledge

about thermodynamic conditions, e. g., in the com-

bustion chamber and intake as well as exhaust man-

ifolds, (Schramm et al., 2020; Guzzella and Onder,

2010; Serikov, 2010). Due to the fact that no mea-

surements of thermodynamic conditions of the com-

bustion engine were available, a black-box modelling

approach like in (Serikov, 2010) is used. The estima-

tion of the mass ﬂow of fuel ˙m

and corresponding

mass ﬂows of the chemical compounds ˙m

of the

exhaust gas emissions shall be done by ANNs using,

e. g., n

Engine

and M

Engine

as inputs.

2.7 System and Process Description

The system and process description represents the de-

velopment of the system structure shown in ﬁgure 2

and the subsystems illustrated in ﬁgure 3 for p power-

trains. Since working with ANNs with frameworks

like Tensorﬂow in the high level programming lan-

guage Python is very common, the implementation of

the physical models as well as the development of the

ANNs is performed in Python. This means that both

subsystems can be connected in the same software en-

vironment. In the following section 3 the modelling

of both subsystems is described in detail.

3 MODELLING OF THE

SUBSYSTEMS

The simulation tool consists of two subsystems which

are based on different modelling approaches. After-

wards both subsystems are explained. In section 3.1

the physical modelling of the vehicle and powertrain

dynamics is shown. The black-box modelling of the

combustion engine is presented in 3.2.

3.1 Modelling of the Vehicle Dynamics

As mentioned before, a backward directed model of

the vehicles' longitudinal dynamics and its power-

train shall be applied to estimate the engine speed

and the torque from the vehicle's velocity, its payload

and the ambient temperature. First, based on a time-

dependent input vector

x(t) =



v(t) T

Amb

(t) m

Trailer

(t)



(1)

as well as vehicle and environmental parameters

like the vehicle mass m

, its front cross-sectional

SIMULTECH 2020 - 10th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

228

area A

, its drag coefﬁcient c

, the air density ρ,

the acceleration due to gravity g and the dynamic

rolling radius of the wheels r

dyn

, the acting driving

resistances are determined according to (Schramm

et al., 2020). It is assumed that the vehicle is driving

straight forward and the wheels roll slip-free. Since

the vehicles are moved on a surface without any

signiﬁcant gradient, the road inclination and thus

the slope resistance is ignored. Also not considered

is the wind speed, which is more or less evenly

distributed in the spatial directions over time. In total,

the acceleration resistance F

Acc

, the wind resistance

Air

, the rolling resistance F

Roll

and the resistance

by pulling the trailer F

Trailer

counteract the vehicle's

movement as shown in ﬁgure 4.

Air

Acc

Roll,F

Roll,R

Trailer

Drive

Figure 4: Acting driving resistances upon the terminal trac-

tor in longitudinal direction, (Terberg Benschop, 2015).

The required drive force F

Drive

sums up the driv-

ing resistances. Thus, the corresponding drive power

Drive

can be calculated by equation (2).

Drive

= F

Drive

· v

= (F

Acc

+ F

Air

+ F

Roll

+ F

Trailer

) · v

(2)

The acceleration resistance F

Acc

describes the

translational and rotational inertia of the vehicles

which counteracts the change in motion and can be

determined by equation (3).

Acc

= m

· (1 + λ ) ·

(3)

Here λ represents the rotational mass surcharge

factor, which expresses the rotational inertia J

in the

powertrain in the form of a translatory force according

equation (4).

λ =

· r

dyn

)

(4)

The air resistance F

Air

represents the aerodynamic

resistance of the vehicle and is considered with equa-

tion (5).

Air

· c

· A

· ρ · v

(5)

According to (Mitschke and Wallentowitz, 2014)

the wheel resistance results mainly from the rolling

resistance F

Roll

. Taking into account, that the wheels

of the front and rear axis are similar, the rolling resis-

tance can be described by equation (6) with a rolling

resistance coefﬁcient f

R,V

Roll

= f

R,V

· m

· g (6)

The pulling resistance required to move a trailer

is considered by a separate modelling of the

trailer, (Haken, 2015). The rotational inertia of the

trailers' wheels and the trailers' air resistance can be

assumed to be negligible compared to the transla-

tional inertia of the trailers' mass and rolling resis-

tance of its wheels. Thus the total resistance by

pulling the trailers can be described by equation (7)

with a rolling resistance coefﬁcient f

R,Trailer

Trailer

= f

R,Trailer

· m

Trailer

· g + m

Trailer

(7)

In order to determine n

Engine

and M

Engine

, a mod-

elling of the vehicles' powertrains as shown in ﬁgure

5 is carried out. The powertrains include the combus-

tion engine, an automatic transmission with a torque

converter and an axle drive. Besides the mechanical

power to drive the vehicle, auxiliaries, e.g., the alter-

nator and the air conditioning are also driven.

Engine

Torque

Converter

Automatic

Transmission

Auxiliaries

Axle

Drive

⁄

Drive

⁄

AUX

Figure 5: Overview of the terminal tractors' powertrains.

The drive power P

Drive

is used as the input of the

powertrain model to estimate the engine speed and the

torque backwards through the entire powertrain. Be-

ginning with equations (8) and (9) the speed and the

torque of the tires, which equals the output speed and

the torque of the axle drive, is calculated.

Tires

2 · π · r

dyn

(8)

Tires

Drive

2 · π · n

Tires

(9)

By taking into account the gear ratio i

and ef-

ﬁciency η

of the axle drive, its associated input

speed and torque can be determined by the equations

(10) and (11). Those values equal the output speed

and the torque of the automatic transmission.

Evaluation of Alternative Propulsion Concepts for Mobile Machinery: A Modelling Approach using the Example on LNG-powered Port

Handling Equipment

229

= n

Tires

· i

(10)

Tires

· η

(11)

The input of the automatic transmission is calcu-

lated with a gear ratio i

which depends on the se-

lected gear G

and an efﬁciency η

by equations

(12) and (13).

= n

· i

) (12)

) · η

(13)

Due to the fact that the selected gear is not in-

cluded in the input vector x(t), an ANN as shown in

ﬁgure 6 shall be used to predict the selected gear G

According to (Jeoung et al., 2020) the shifting of gear

in automatic transmissions is controlled by the veloc-

ity of the vehicle and the position of the accelerator

pedal. A change of the accelerator pedal position is

related to a change of torque and thereby to a change

of vehicles velocity and thus its acceleration. That is

why the vehicle speed v and acceleration

are used

as inputs, as ﬁgure 6 illustrates. The corresponding

gear ratio i

) is selected from a lookup-table

ﬁlled with information from the data sheet of the man-

ufacturer.

Figure 6: Neural network to predict the selected gear.

A torque converter is placed between the transmis-

sion and the engine to overcome speed differences

when starting from engine idling. Since sufﬁcient

information about the dynamics of the torque con-

verter are not available, a quasi-stationary modelling

approach based on characteristic values from the data

sheet of the torque converter according to (Luz, 2015)

is used. The desired engine speed and the torque cor-

respond to the input speed n

and the torque M

the torque converter taking into account an efﬁciency

as shown in equation (14) and (15).

In addition, the engine has to power auxiliaries,

which are considered by a sum of n different mechan-

ical loads M

AUX,n

∑

i=1

AUX,i

, (Schramm et al.,

2020).

Engine

= n

(14)

Engine

∑

i=1

AUX,i

(15)

In order to decide, whether the air condition-

ing is heating or cooling, an ambient-temperature-

dependent energy demand shall be considered.

3.2 Modelling of the Combustion

Engine

The combustion process of the engines measured dur-

ing the research project is modelled as a black-box.

In general, the model has to reply two questions for

every time step based on knowledge from its inputs:

• Which mixture of fuel, air and recirculated ex-

haust gas was burned in the combustion chamber?

• Which type and amount of exhaust gas emissions

are generated?

It was found that multiple factors can inﬂuence the

fuel consumption, exhaust gas emissions and the

performance of the engine as, e. g., the temperature

of the ambient air, the engine coolant temperature

or the rate of exhaust gas being recirculated to the

combustion chamber, (Abdullah et al., 2015; Abdel-

ghaffar et al., 2002; Hussain et al., 2012). Therefore

all the above-mentioned variables are considered in

the black-box modelling of the combustion engine.

To answer both questions, the engine model is

divided into two ANNs, using the outputs of the ﬁrst

one as input for the second ANN, as shown in ﬁgure 7.

Figure 7: Structure of the black-box engine model.

In the ﬁrst ANN the calculated n

Engine

and M

Engine

and the measured T

Amb

are used as inputs. The fourth

input, the engine coolant temperature T

, is neither

calculated nor measured during the operation of the

entire ﬂeet. This measurement value is just known

from the individual vehicles, but because of its im-

portance regarding the engines performance it is also

considered as an input. It is therefore up to the user

to deﬁne these temperatures in order to study engine

SIMULTECH 2020 - 10th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

230

performance under different engine conditions such

as cold start or engine warm-up. The outputs of the

ﬁrst ANN – the mass ﬂow of fuel ˙m

, the mass ﬂow

of air ˙m

Air

and mass ﬂow of recirculated exhaust gas

˙m

EGR

– answer the ﬁrst question.

The second ANN aims to answer the second ques-

tion by estimating the mass ﬂow of carbon diox-

ide ˙m

E,CO

, carbon monoxide ˙m

E,CO

, nitrogen oxides

˙m

E,NO

, methane ˙m

E,CH

and sulphur dioxide ˙m

E,SO

Therefore the inputs of the ﬁrst ANN as well as its

outputs are used as inputs to the second ANN.

4 SUMMARY AND OUTLOOK

This paper introduces an approach for a simulation

tool with which a complete conversion of a ﬂeet of

mobile machinery to an alternative fuel can be scaled

up using measurements of individual vehicles. The

simulation tool includes semi-physical models of p

different powertrains to estimate the fuel consump-

tion and exhaust emissions of the entire ﬂeet for ev-

ery powertrain setup. In this paper the simulation tool

is explained using the example of terminal tractors

and both, a diesel- and LNG-driven powertrain. Each

semi-physical model consists of a combination of a

physical modelling of the driving and powertrain dy-

namics and a black-box modelling of the engine by ar-

tiﬁcial neural networks. The development and param-

eterization of the models is fulﬁlled by measurements

of the everyday operation of two reference vehicles.

As inputs for the simulation tool measurements of the

time-dependent velocity v, ambient temperature T

Amb

and payload m

Trailer

of the rides of the vehicles of a

ﬂeet are necessary. Based on these information the

simulation tool predicts the massﬂows of fuel ˙m

and

of the corresponding exhaust gas emissions ˙m

for

every powertrain p during the ride. For the calcu-

lation of the total mass of fuel and emissions of the

entire ﬂeet for p powertrains, in a postprocessing the

simulation results for every ride and every powertrain

p are integrated and added up.

At the time of writing this paper, the physical

models of the vehicle and powertrain dynamics are

implemented. The ANNs for the prediction of the

gear provide a sparse categorical accuracy of 95,4%

(LNG) and 96,4 % (Diesel). Unknown physical para-

meters like the rolling resistance coefﬁcient f

R,V

were

estimated by measurements of planed test drives in

which the terminal tractors did not pull any trailer.

The evaluation of the calculated and measured n

Engine

and M

Engine

during the test drives resulted in a mean

absolute error of 25.3 rpm and 21.9 Nm for the LNG-

and 30 rpm and 24.5 Nm for the diesel-powered ter-

minal tractor. To gain an understanding of the causes

of the errors, an analysis of the simulation results was

carried out. Two signiﬁcant causes of failure were

detected. Figure 8 presents a ﬁrst excerpt of the mea-

sured test drives of the LNG-powered terminal tractor

and demonstrates both types of failure. As the po-

sitions P

in ﬁgure 8 show, one cause of failure can

be vehicle standstill, where the engine load is inde-

pendent of the input data and deviates from the usual

loads. In addition, errors may occur if the estimated

gear differs from the measured gear (see P

), resulting

in a different gear ratio.

Figure 8: Measured and simulated n

Engine

and M

Engine

dur-

ing a test drive of the LNG-powered terminal tractor.

The next step is to identify the unknown rolling

resistance coefﬁcient f

R,Trailer

of the trailer model. To

do this, both, the measured data and the informa-

tion from the terminal operating system are linked

together. Then the internal structure of the ANNs

in the black box model of the combustion engine in

each powertrain has to be determined and data sets

for training, testing and validation have to be created.

Next, suitable hyperparameters of the ANNs must be

found by hyperparameter optimization. Afterwards

the accuracy of each ANN must be veriﬁed. Finally,

the physical model and the black box model for each

powertrain must be linked together and the accuracy

of each complete model must be examined. The com-

plete models can then be used to predict and discuss

the fuel consumption and exhaust emissions of the

ﬂeet of terminal tractors at duisport for the diesel- and

LNG-powered powertrain.

Evaluation of Alternative Propulsion Concepts for Mobile Machinery: A Modelling Approach using the Example on LNG-powered Port

Handling Equipment

231

ACKNOWLEDGEMENTS

The work presented here was carried out within the

framework of the research project LeanDeR. This

project was funded by the European Regional De-

velopment Fund (ERDF) and the state government of

North Rhine-Westphalia, Germany.

REFERENCES

Abdelghaffar, W., Osman, M., Saeed, M., and Abdelfatteh,

A. (2002). Effects of Coolant Temperature on the Per-

formance and Emissions of a Diesel Engine. Ameri-

can Society of Mechanical Engineers, Internal Com-

bustion Engine Division (Publication) ICE, 38.

Abdullah, N., Ismail, H., Michael, Z., Abdullah, A. R., and

Sharudin, H. (2015). Effects of air intake tempera-

ture on the fuel consumption and exhaust emissions

of natural aspirated gasoline engine. Jurnal Teknologi,

76:25–29.

Dubois, G. (2018). Modeling and Simulation: Challenges

and Best Practices for Industry. CRC Press.

Duisburger Hafen AG (2019).

European Commission (2018). A Clean Planet for all. A

European strategic long-term vision for a prosperous,

modern, competitive and climate neutral economy.

European Parliament and the Council (2017). Commission

Delegated Regulation (EU) 2017/654 of 19 Decem-

ber 2016. Ofﬁcial Journal of the European Union, L

102:1–333.

Federal Ministry for the Environment, Nature Conservation

and Nuclear Safety (2019). Climate Action in Figures.

Facts, Trends and Incentives for German Climate Pol-

icy.

Geimer, M. and Pohlandt, C. (2014). Grundlagen mo-

biler Arbeitsmaschinen, volume 22 of Karlsruher

Schriftenreihe Fahrzeugsystemtechnik / Institut für

Fahrzeugsystemtechnik. KIT Scientiﬁc Publishing,

Karlsruhe.

Guzzella, L. and Onder, C. (2010). Introduction to Model-

ing and Control of Internal Combustion Engine Sys-

tems. Springer-Verlag Berlin Heidelberg.

Haken, K.-L. (2015). Grundlagen der Kraftfahrzeugtech-

nik. Carl Hanser Verlag.

Helms, H., Jamet, M., and Heidth, C. (2017). Renewable

fuel alternatives for mobile machinery. Technical re-

port, Institut für Energie- und Umweltforschung Hei-

delberg.

Hussain, J., Palaniradja, K., Natarajan, A., and Manimaran,

R. (2012). Effect of exhaust gas recirculation (EGR)

on performance and emission characteristics of a three

cylinder direct injection compression ignition engine.

Alexandria Engineering Journal, 51:241–247.

International Energy Agency (2019). CO2 Emissions From

Fuel Combustion 2019.

Jeoung, D., Min, K., and Sunwoo, M. (2020). Automatic

Transmission Shift Strategy Based on Greedy Algo-

rithm Using Predicted Velocity. International Journal

of Automotive Technology, 21(1):159–168.

Lajunen, A., Sainio, P., Laurila, L., Pippuri-Mäkeläinen, J.,

and Tammi, K. (2018). Overview of Powertrain Elec-

triﬁcation and Future Scenarios for Non-Road Mobile

Machinery. Energies, 11:1184.

Luz, R. (2015). Simulationsbasierte Methode zur Zerti-

ﬁzierung der CO

Emissionen von schweren Nutz-

fahrzeugen. PhD thesis, TU Graz.

Milojevi

c, S., Skrucany, T., Milosevic, H., Stanojevic, D.,

Panti

c, M., and Stojanovic, B. (2018). Alternative

Drive Systems and Environmentaly Friendly Public

Passengers Transport. 3:105–113.

Mitschke, M. and Wallentowitz, H. (2014). Dynamik der

Kraftfahrzeuge. Springer Fachmedien Wiesbaden.

Mohan, G., Assadian, F., and Longo, S. (2013). Compar-

ative analysis of forward-facing models vs backward-

facing models in powertrain component sizing.

Schramm, D., Hesse, B., Maas, N., and Unterreiner, M.

(2020). Vehicle Technology - Technical Foundations

of Current and Future Motor Vehicles. De Gruyter

Oldenbourg, Berlin / Boston.

Schramm, D., Hiller, M., and Bardini, R. (2018). Vehicle

Dynamics. Springer, Berlin / Heidelberg, 2 edition.

Serikov, S. (2010). Neural network model of internal com-

bustion engine. Cybernetics and Systems Analysis,

46:998–1007.

Terberg Benschop (2015). YT222 LNG, 4x2. Technical

Speciﬁcation.

Verein Deutscher Ingenieure e.V. (2016). VDI 4465

Blatt 1: Modellierung und Simulation - Modellbil-

dungsprozess.

Wipke, K., Cuddy, M., and Burch, S. (1999). Advisor 2.1:

A user-friendly advanced powertrain simulation using

a combined backward/forward approach. Vehicular

Technology, IEEE Transactions on, 48:1751 – 1761.

Xu, X. P. (1997). Experimental Modeling of A Hydraulic

Load Sensing Pump using Neural Networks. PhD the-

sis, University of Saskatchewan.

Zhao, N., Liu, Y., Mi, W., Shen, Y., and Xia, M. (2020).

Digital Management of Container Terminal Opera-

tions. Springer Singapore.

SIMULTECH 2020 - 10th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

232