Co-Simulation of Engine Model and
Control System with focus on
Turbocharger Model
Martin Wadner
Mechanical Engineering, master's level 2020
Luleå University of Technology
This thesis was conducted at the group for Gas Exchange Performance (NMGP) at Scania CV AB in S¨odert¨alje, Sweden, during the spring and summer of 2020. This thesis concludes five years of studies at Lule˚a University of Technology for a degree in mechanical engineering with a master in computational mechanics. I am utterly grateful for the experiences and everything that I have come to learn during this time. I feel privileged to have had the opportunity to meet many interesting people and for the lifelong friends I have gained during the course of these five years.
I would like to thank my supervisors at Scania CV AB, Tobias Gezork and Joakim Rodeb¨ack for the crucial guidance and support during this work. I am very grateful for their patience and time spent during the difficult circumstances that the COVID-19 pandemic brought. A big thanks to NMGP for welcoming me with open arms, for the fun times and for have given me this the opportunity to learn about your interesting and challenging work. Many thanks also to the group of NCFG for taking the time to answer all my questions.
I would also like to thank my supervisor at Lule˚a University of Technology, Jan-Olov Aidanp¨a¨a for all the guidance and support.
Finally i would like to thank my family and friends for pushing me and supporting me during this time. This would simply not have been possible without you.
Abstract
The demands on heavy duty vehicles is constantly raising with government legislations on CO2
Symbol Description
EMS Engine management system pco Pressure at compressor outlet
pci Pressure at compressor inlet
pto Pressure at turbine outlet
pexh Pressure at turbine inlet and pressure in exhaust manifold
pexh,OS Outer scroll inlet pressure
pexh,IS Inner scroll inlet pressure
pexh,ave Average pressure between turbine inlets
pco,ref Reference compressor outlet pressure
Tco Temperature at compressor outlet
Tco,ref Reference temperature at compressor outlet
Tci Temperature at compressor inlet
Tto Temperature at turbine outlet
Texh Temperature at turbine inlet and temperature in exhaust manifold
Texh,IS Temperature at inner turbine inlet
Texh,OS Temperature at outer turbine inlet
Texh,ave Average temperature between turbine inlets
˙
Wc Actual work consumed by compressor
˙
Wt Actual work produced by turbine
ωc Compressor speed
ωc,ref Reference compressor speed
ωt Turbine speed
γ Ratio of specific heat
cp Specific heat at constant pressure
cp,t Specific heat at turbine inlet
cp,c Specific heat at compressor inlet
cv Specific heat at constant volume
˙
mc Mass flow through compressor
˙
mt Mass flow through turbine
Πc Pressure ratio over the compressor
Πc,ref Reference pressure ratio over the compressor
hci Enthalpy at compressor inlet
hco,ideal Ideal enthalpy at compressor outlet
hco Enthalpy at compressor outlet
ηc Isentropic compressor efficiency
ηc,ref Reference compressor efficiency
Πt Pressure ratio over the turbine
ηt Isentropic turbine efficiency
ηt,T −S Isentropic turbine efficiency, total to static
ηm Mechanical losses
ωt Turbine angular velocity
ωt,ref Reference turbine angular velocity
Jtc Turbocharger inertia
pref Reference pressure
Symbol Description ˙
mr Reduced mass flow
ωr Reduced speed
˙
mcorr Corrected mass flow
ωcorr Corrected speed
˙
mwg Waste gate mass flow
˙
mexh Exhaust gas mass flow
˙
me,ref Reference engine mass flow
˙
mt Turbine mass flow
˙
mt,ref Reference turbine mass flow
˙
mwg Demanded waste gate mass flow
Ψ Psi-function R Ideal gas constant CD Discharge coefficient
Awg Waste gate area
Mc,ref Reference compressor torque
SOI Start of injection ne Engine speed
δ Delta value prail Rail pressure
Pt Total produced power from the turbine
dwg Waste gate opening
OS Outer scroll IS Inner scroll
SP R Scroll pressure ratio S Scaling value
Ptop Maximum pressure
Pavg Average pressure
∆t Time step
1 Introduction 1
1.1 Turbocharged diesel engine . . . 2
1.2 Problem description . . . 3
1.3 Research aim and delimitations . . . 3
1.4 Literature study . . . 4
2 Theory 6 2.1 Compressor and turbine modelling . . . 6
2.1.1 Compressor model . . . 6
2.1.2 Turbine model . . . 8
2.1.3 Turbo Shaft . . . 9
2.1.4 Performance parameters . . . 9
2.1.5 Map interpolation . . . 10
2.1.6 Twin scroll turbine . . . 10
2.2 Turbo Control Logic . . . 12
2.3 Engine Simulation Models . . . 15
3 Method 17 3.1 EMS implementation . . . 17
3.1.1 Communication interface . . . 18
3.1.2 Simulink model adaptations . . . 18
3.1.3 Transient reference data . . . 19
3.1.4 Engine simulation model adaptations . . . 20
3.1.5 Convergence study . . . 22
3.2 Turbine modelling improvement implementation . . . 23
3.2.1 Outer scroll inlet pressure estimation . . . 24
3.2.2 Turbine efficiency estimation. . . 25
3.2.3 Pulse estimation for inlet pressure . . . 26
3.2.4 Implementation of pulse estimation algorithm . . . 29
3.2.5 Averaging for overall turbine performance . . . 31
4 Results 32 4.1 Outer turbine inlet pressure estimation . . . 33
4.2 Turbine efficiency estimation . . . 34
4.3 Pressure pulse estimation for inner turbine inlet . . . 35
4.4 Turbo control performance analysis . . . 36
Introduction
The heavy duty vehicle industry is today ruled by government legislations on CO2emissions becoming
stricter and customer demands for performance of combustion engines getting higher. It is therefore essential to continuously develop new tools and methods within the product development that enables the industry to meet these demands. Today it is standard within the industry to apply advanced simulation methods as a complement to the conventional product development process. This enables companies to simulate product performance in a model based design approach. The developing trend is to create more comprehensive models where a wider variety of functionality is included in the simulations.
New regulations for emission control have recently been set for heavy duty vehicles in Europe where a reduction of 15 percent is required by 2025 in relation to the current emission levels [1]. It is critical as part of the development process to ensure that new combustion engines will operate within the required limits of the regulations while continuously guarantee their safety and performance. It is also important that this has to be done for the large variety of environmental conditions that the vehicle will operate in. Testing of newly developed engines will therefore, at some point during the development process, be conducted at locations where the conditions are more demanding. These conditions include, but are not limited to, very cold and very hot climates and high altitudes. A cost effective way to prepare the new engines for these conditions is to use a simulation model where the operation of the engine in different conditions can be simulated. The engine parameters can then be tweaked in such a way that the engine is meeting the demands.
It is essential that the models are accurate enough so that the results show a true representation of the engine performance. When building the model, the development team is therefore not only required to consider the accuracy of the engine model itself but also the surrounding systems such as the engine management systems, EMS. The purpose of the engine management system is to control the engine during operation based on driver input and data from sensors in the engine. The opportunity to perform co-simulations between the engine simulation model and the engine monitoring system will enable better predictions for engine operations and help ensuring that the engine operates in a desired manner. The process of building a co-simulation platform will benefit the engineers, linking software engineers responsible for the EMS and the combustion engine engineers together, creating good prerequisites for building better products.
1.1. TURBOCHARGED DIESEL ENGINE CHAPTER 1. INTRODUCTION
1.1
Turbocharged diesel engine
Turbocharged diesel engines are today very common in heavy duty vehicles. Supercharging is a collective name for methods that increase the density of the intake air into the combustion chamber. Since the produced power and torque is limited to the amount of air mass in the combustion chamber, more air allows for more efficient engines. Supercharging is hence a means for boosted performance without increasing the fuel consumption which thereby also allows for decreasing the engine volume, i.e. the fuel consumption, without loosing performance. This is what is called downsizing and is key to meet both stricter legislations and higher customer demands at the same time. One common method for supercharging diesel engines is using a turbocharger. A turbocharger is compressing the intake air by using a compressor. The compressor is connected to the turbine which runs by the expansion of the exhaust gases flowing out from the combustion chambers [2].
The gas exchange system of an internal combustion engine is referring the volumes where intake air and exhaust gases travel through the engine. The gas exchange system includes the pathways through the intercooler and also the turbocharger in the case of a turbocharged engine. Important to the system is also the valves controlling the gasflow [3]. A general description of a gas exchange system setup in a turbocharged diesel engine is presented in Figure 1.1.
Intercooler Cylinders Wastegate Compressor Turbine Intake manifold Outlet manifold EGR Outlet Intake Shaft
Figure 1.1: Representation for the gas exchange system in a turbocharged diesel engine. The arrows in the figure show the direction by which the flow passes through the different volumes. The engine block consists of mainly three components; the cylinders, the intercooler and the turbocharger. The turbocharger consists of a turbine and the compressor connected by a turbo shaft. Three valves are controlling the gas flow through the system; the throttle, the exhaust gas recirculation valve, EGR, and the waste gate.
A consequence of downsizing is that the flow exhaust gas can become insufficient at running the turbine during lower engine speeds. A common method to prevent this phenomena is to isolate the pulses of exhaust gas flow that results from the individual combustions in the cylinders. This is done by having multiple exhaust gas manifolds that keeps the flow separated up to the point of turbine entry [4]. This method is called pulse turbocharging and the two most common types of pulse tur-bocharging turbines are the twin scroll turbine and the double entry turbine. The difference between these types of turbines is how they receive the separated exhaust gas flows along the circumference of the turbine wheel.
1.2
Problem description
Engines are commonly calibrated by running the engine in engine test cells. The test cell has the functionality to simulate different environments. However the engines also have to be tested in real trucks. These tests are done in extreme conditions where the altitude and air temperatures are far outside conditions replicable in test cells. The gas exchange performance group at Scania CV AB is responsible for preparing the engine calibration for these tests. To ensure that the engine will operate within the desired limits the group calibrates the engine by using simulation models. These engine models allows for predictions on engine behaviour to be made and thereby also the opportunity to alter the engine parameters if necessary. The group at Scania CV AB uses the 1D-simulation tool GT-POWER developed by Gamma Technologies to build their models [5].
It is important not only to consider the accuracy of the engine model, but also the engine management system in order to make accurate predictions. Previous versions of the engine models have not included the EMS used for controlling the engine in the truck. As a result, a gap has been observed in engine behaviour between on the one hand, the real tests and the simulation predictions on the other hand. These problems are related to the engine simulation models not taking the behaviour of the EMS into account.
1.3
Research aim and delimitations
This study is aims at improving the current turbocharger modelling and control strategy in GT-SUITE engine simulation models, developed by the gas exchange performance group at Scania CV AB.
The contents of the thesis work are:
• To study and improve turbocharger modelling.
• To study and improve control strategy for turbocharger control. • To test control strategy using simulation scenarios.
The motivation for the study is to obtain engine simulation models that can make more accurate predictions and to have a turbo control strategy that is sufficient for the control goal. The goal with the turbocharger control is to control the speed of the turbocharger for maximum engine performance without exceeding its safety limits. The idea is to achieve this by implementing the part of the EMS that handles the turbo control into the current simulations models.
1.4. LITERATURE STUDY CHAPTER 1. INTRODUCTION
FRM, built with a larger degree of simplifications allowing the simulations to be solved faster. The other simulation model, in relation to the FRM, includes more details giving more accurate results in exchange for being solved in a longer period of time. The logic system that handles the turbo control in the EMS is given in the form of a MATLAB Simulink model.
The improvements will partly focus on the possibilities to implement the current EMS turbo control strategy into the given simulation models. The work also aims at implementing improvements to turbocharger modelling within the current EMS turbo control logic. The type of turbocharger modelled in the engine simulation models are fixed geometry twin scroll turbines.
Due to the nature of this thesis work a number of delimitations are set as follows:
• The results from this study will not be validated against engine cell tests but rather against existing test data.
• The suggested improvements in this study is limited to the modelling of the turbocharger and excludes other physical modelling that also might affect the results such as the wastegate actu-ator.
1.4
Literature study
Eriksson and Nielsen [2], Zander [3] and Hiereth and Prenninger [6] all present, in various degrees physical modelling of a turbocharger. Eriksson and Nielsen [2] give an introduction and an in-depth understanding of physical modelling and control of engines and drivelines. The book explains the critical aspects of turbocharger modelling and the tools used for describing the turbocharger performance.
Zander [3] is centred around the topic of gas exchange systems and the related topic of supercharging for internal combustion engines. It discusses different parts of the common gas exchange system and the distinctive differences between different types of engines. The book specifically discusses different methods for supercharging 4-stroke engines.
Hiereth and Prenninger [6] focus on supercharging and its many topics. The book presents the fundamentals and objectives of supercharging and its interaction with the combustion engine from a fluid mechanical and thermodynamic stand point. It also presents different types of supercharging including an in depth explanation of exhaust gas supercharging, turbocharging, and its different versions. Hiereth and Prenninger [6] also discuss different methods for designing optimal engine to turbocharger performance.
As described in Hiereth and Prenninger [6] and Zander [3] the flow conditions in an internal combus-tion engine is of an unsteady nature as pulses resulting from valve openings propagate throughout the gas exchange system. Many studies such as Copeland et al [7], [8], [9], Costall et al [10] and Hajilouy et al [11] have been investigating the impact of unequal admission conditions in multi en-try turbocharger turbines where they concluded that a significant performance penalty is occurring during the unequal admissions in relation to the equal admission. This effect is especially significant for the type of multi-entry turbines called twin scroll turbines as concluded in Romagnoli et al [12]. According to the paper this could be a result of the leaking of flow between the two inlets occurring during states of unequal admission.
interaction map method for describing the performance variation of the turbine during operation. Further measurements are performed at different operating conditions such as different scroll pressure ratios and unequal scroll temperature admissions. The paper concludes that the scroll pressure ratio and scroll temperature admissions are of high relevance when it comes to describing turbine performance and that the flow interaction maps proved to be a representative way to describe the flow conditions.
Newton et al [14] presents a method of map extrapolation considering unequal and partial admission in a double entry turbine. The presented method uses a full admission turbine map and a number of unequal admission data points from which three constants are determined. The paper then presents how to estimate the turbine performance following a series of steps. According to the paper this method is giving satisfactory results in predicting the turbine performance during any unequal admission state.
Chiong and Rajoo [15] considers the problem related to neglecting pulsating flow conditions in 1D simulations models that turbines in a turbochargers are subject to. It is common to rely on steady state flow measurements when building these types of simulation models which according to the paper prohibits acceptable accuracy in the results. The paper specifically investigates twin entry turbines and presents the geometrical effects under full admission pulsating flow conditions. Five different geometries are presented as subjects for investigating the performance under these conditions. Marelli and Capobianco [16] is presenting an experimental investigation of a turbocharger turbine and waste gate performance. The paper evaluates the turbine efficiency under different flow conditions including unsteady pulsating conditions occurring in common automotive combustion engines. It is concluded that a thermodynamic method for estimating the turbine power can differ largely from other mechanical estimation methods. It is also concluded that opening of the waste gate has a significant reduction of the turbine efficiency. The final conclusion states that the results from considering unsteady flow conditions reduces the turbine efficiency on a cycle average in relation to the resulting efficiency during the steady state conditions.
Chapter 2
Theory
2.1
Compressor and turbine modelling
A turbocharger mainly consists of a turbine, a compressor and a turbo shaft that connects the turbine and the compressor. A turbocharger is designed to allow the engine to make use of the otherwise lost enthalpy in the exhaust gases. The turbocharger converts fluid work in the exhaust gases to mechanical work that is then used to compress the intake air. To model the performance of the turbocharger a thermodynamic analysis is made with respect to the transfer of power [2]. Figure 2.1 presents the model boundaries for the compressor and the turbine along with the associated notations for pressure, temperature, power and angular velocity.
C T Tco Tci Tto Texh pto pexh pco pci ωc ωt ˙ Wc ˙ Wt
Figure 2.1: The parts marked with a C and T represents the compressor and the turbine respectively. The figure presents the current temperatures and pressure at the inlets and outlets of the turbine and the compressor along with the respective angular velocities and resulting power.
2.1.1
Compressor model
where Πc = pco/pci and is called the pressure ratio. The efficiency is a critical parameter which is
needed for describing the performance of a compressor or turbine. The efficiency for a compressor, ηc,
can be calculated as the ratio between the actual difference of enthalpy over the compressor and the difference of enthalpy over the compressor in an ideal process [6]. This ratio is, under the assumption that the medium in question is an ideal gas with constant specific heat, cp, equal to the ratio between
the actual consumed power ˙Wc, and the consumed power in an ideal process, ˙Wc,ideal
ηc= hco,ideal − hci hco− hci = W˙c,ideal ˙ Wc (2.2)
where the power consumptions for these respective processes are
˙
Wc= ˙mccp,c(Tco− Tci) (2.3a)
˙
Wc,ideal = ˙mccp,c(Tco,ideal− Tci) (2.3b)
where ˙mcis the compressor mass flow and cp,cis the specific heat for the ideal gas at constant pressure
[2]. For isentropic processes the following expression
Tco,ideal Tci =pco pci γ−1γ (2.4)
is true for the relationship between the pressure and temperature in a compressor [2][6]. Where γ is the ratio of specific heat, γ = cp
cv. Combining equations (2.2) to (2.4) the isentropic compressor
efficiency can then finally be expressed as
ηc= pco pci γ−1γ − 1 Tco Tci − 1 . (2.5)
The resulting efficiency is considered to be of total-total nature, further noted ηc,T −T, if all the
temperatures and pressures in equation (2.5) are of total nature. In relation to the total-total compressor efficiency it is common to also consider the total-static compressor efficiency [2]. In that case the compressor outlet pressure of is static nature, pco,S giving the following expression for
total-static compressor efficiency
ηc,T −S = pco,S pci γ−1γ − 1 Tco Tci − 1 . (2.6)
Solving for the compressor outlet temperature, the following final expression is obtained
2.1. COMPRESSOR AND TURBINE MODELLING CHAPTER 2. THEORY
2.1.2
Turbine model
The generic turbine model is defined according to the following four equations
˙
mt = fm,t˙ (1/Πt, Texh, ωt) ηt= fη,t(1/Πt, Texh, ωt)
˙
Wt= fW ,t˙ (1/Πt, Texh, ωt) Tto= fT ,t(1/Πt, Texh, ωt)
(2.8)
where 1/Πt = pexh/pto and is called the expansion ratio. The isentropic turbine efficiency can be
expressed in a similar manor as to equation (2.5). Rather than power consumption, the isentropic turbine efficiency is expressed in terms of power delivery
ηt= hexh− hto hexh − hto,ideal = ˙ Wt ˙ Wt,ideal (2.9)
where the power delivery for these respective processes are
˙
Wt= ˙mtcp,t(Texh− Tto) (2.10a)
˙
Wt,ideal = ˙mtcp,t(Texh− Tto,ideal) (2.10b)
and where ˙mt is the turbine mass flow and the cp,t is the specific heat [2][6]. Combining equations
(2.9) to (2.10b) and the principles of equation (2.4) gives the isentropic turbine efficiency as
ηt= 1 − Tto Texh 1 − pto pexh γ−1γ . (2.11)
Similarly to the total-total compressor efficiency, is turbine efficiency of total-total nature, ηt,T −T, if
the all temperatures and pressures in equation (2.11) have total properties. In case of total-static turbine efficiency, is the turbine outlet pressure of static nature, pto,S, which gives the following
expression for total-static turbine efficiency [2],
ηt,T −S = 1 − Tto Texh 1 − pto,S pexh γ−1γ . (2.12)
Isentropic efficiencies does not take into account the heat transfer from the turbine or compressor housing to the respective gases flowing through. Due to the large temperature differences over the turbine, the heat exchange can have a large impact on the resulting power delivery. These losses can be seen as mechanical losses between the turbine and the compressor. Mechanical losses, ηm, can
therefore be expressed as the compressor power over the turbine power, ˙Wc/ ˙Wt [2]. Solving for the
turbine power according to
one obtains an expression for the turbine power that is taking the heat exchange in consideration. A lumped efficiency combining isentropic and mechanical efficiency can be expressed as
ηt,T −Tηm = ˙ mccp,c(Tco− Tci) ˙ mtcp,tTexh(1 − ppto exh γ−1γ ) . (2.14)
2.1.3
Turbo Shaft
Newton’s second law is used to model the dynamic behaviour of the connection between the turbine and the compressor. The angular acceleration as a result from the power imbalance between the turbine and the compressor can be expressed as
dωtc dt = 1 Jtc W˙t ωtc ηm− ˙ Wc ωtc (2.15)
where ωtc is the angular velocity of the turbocharger and Jtc is the inertia [2].
2.1.4
Performance parameters
Compressor and turbine maps are interpolation tables used to describe the performance of a tur-bocharger during its operation. The maps are experimentally measured in gas stands that allows for operating the turbocharger in a very controlled manner under steady flow conditions. The maps describes the relation between a number of performance parameters. These are rotational speed, ω, mass flow, ˙m, total to static isentropic efficiency, ηt,T −S and ηc,T −S, and total to total pressure ratio,
Π, for compressor maps and total to total expansion ratio 1/Π for turbine maps. Reduced properties are often used for mass flow, ˙mr, and rotational speed, ωr. The reduced performance parameters
and can be expressed as
˙ mr = ˙ m√Tin pin (2.16a) ωr = ω √ Tin (2.16b)
where Tin and pin are the inlet temperature and pressure respectively. It is also common to use
corrected properties for mass flow and speed which are in relation to the reduced properties also normalized against reference pressure and temperature. Corrected mass flow, ˙mc, and speed ωc is
expressed as ˙ mcorr = ˙ mpTin/Tref pin/pref (2.17a) ωcorr = ω pTin/Tref (2.17b)
2.1. COMPRESSOR AND TURBINE MODELLING CHAPTER 2. THEORY
2.1.5
Map interpolation
In order to extract the correct value from within performance maps and in-between performance maps, an appropriate interpolation strategy is needed. A commonly used interpolation strategy is regular linear interpolation. Two dimensional linear interpolation is expressed according to
y − ya
yb− ya
= x − xa xb− xa
(2.18)
where x is the input value, y is the output value and ya, yb and xa, xb are the surrounding values in
the y-vector and x-vector respectively.
2.1.6
Twin scroll turbine
The turbocharger in question for this thesis uses a twin scroll inlet to the turbine. A twin scroll turbine receives the exhaust gases from two separate exhaust gas manifolds. In a six cylinder engine, for example, the two exhaust gas manifolds, could receive exhaust gases from three cylinders each. A representation of a six cylinder engine configuration with two exhaust gas manifolds and a twin scroll turbine is presented in Figure 2.2. The figure shows how the two exhaust gas manifold are feeding one scroll each. The waste gate is only connected to one of the two manifolds. How the respective scrolls in turn are feeding the turbine wheel is presented in Figure 2.3. The two scrolls in a twin scroll turbine are feeding the turbine blades around its whole circumference [6]. The flow from the two inlets then mixes in the channel created by the turbine wheel.
T Engine
Inner scroll
Outer scroll
WG
Figure 2.2: Six cylinder engine configuration with two exhaust gas manifolds receiving exhaust gases from three cylinders each. The two manifolds are providing the exhaust gases to the inner and outer scroll of a twin scroll turbocharger turbine respectively. The waste gate, noted WG, is only connected to the exhaust gas manifold feeding the outer scroll.
Unsteady flow conditions, as a results of the combustion in the engine, are creating variations in scroll admission [10, 11, 13, 14]. This means that the magnitude of flow travelling from the respective scrolls to the turbine blades are varying relative each other during the engine cycle. The two scrolls, referred to as the inner and the outer scroll, therefore have different pressure and temperature inlet conditions. Pressure and temperature for the outer and the inner scrolls are noted with an OS and
anIS respectively. The twin scroll turbine can consequently be described using two sets of equations
Inner scroll Outer scroll
Figure 2.3: Visualization of a twin scroll turbine configuration. The two scrolls are both feeding the turbine around the its whole circumference. The red arrows represent the direction of flow through the turbine.
One could imagine a series of admission states in-between which the admission from the two scrolls are varying. Full admission is when the two scrolls are feeding the turbine equally and unequal admission is when an imbalance between the scroll admissions is occurring. Single admission is when only one scroll is feeding the turbine. One could also imagine a so called back flow when leakage of flow is occurring from one scroll to the other [4, 13]. Back flow however is not considered in the thesis.
A method for describing the performance of multi entry turbines is to apply multiple maps describing different admission levels for each scroll. This in addition to the regular map valid for multi entry turbines at the point of full admission [9, 13, 14]. Using multiple maps therefore requires some interpolation strategy for obtaining the correct current performance from the scrolls since the level of admission is constantly varying in between the levels of admission described by the maps.
If one turbine is described using two sets of performance maps for the two respective scroll inlets, it is important to ensure that the same reduced speed is used. To ensure the same reduced rotational speed, according to equation (2.16b), for scroll performance estimation, Chumpsty and Horlock [17] suggest a mass averaged inlet temperature between the two inlets. This method is valid when an ideal gas is assumed and aims to preserve the enthalpy flux. The mass averaged turbine inlet temperature Texh,m is expressed as Texh,ave= Texh,OSm˙ OS+ Texh,ISm˙IS ˙ mOS+ ˙mIS . (2.19)
When the ideal gas assumption cannot be made the specific heat capacities at the two scroll inlets also has to be considered. Chumpsty and Horlock [17] also suggest a work averaged inlet pressure between the two inlets. This pressure is important for determining the overall turbine performance. This suggested averaged pressure is calculated as
2.2. TURBO CONTROL LOGIC CHAPTER 2. THEORY
The combined total-static isentropic efficiency for the scrolls is commonly calculated by adding the isentropic power contribution from the two scrolls, [9, 13, 14], in equation (2.9) according two
ηt,T −S = ˙ Wt ˙ Wt,ideal,IS + ˙Wt,ideal,OS . (2.21)
2.2
Turbo Control Logic
The EMS turbo control logic is built in a MATLAB Simulink model. The purpose of the EMS turbo control logic is to determine the demanded waste gate actuator position based on engine measurements and estimations of physical quantities and reference inputs. The waste gate actuator is what controls the waste gate valve on the engine. The model takes in a total of 20 inputs. Two of the inputs are reference parameters that dictates the demanded output. The rest of the inputs are measured or estimated physical quantities, such as temperature, pressure, torque and mass flow, along with system status data. The input parameters are listed in Table 2.1. The layout of the EMS turbo control logic is presented in Figure 2.4.
Reference value conversion
Reference Speed and Reference Torque estimation Turbine model PID regulator Reference Inputs Measured/Estimated Inputs
Demanded waste gate mass flow and actuator position
Figure 2.4: Layout of the MATLAB Simulink model with the EMS turbo control logic.
Table 2.1: Listed used input parameters to the EMS turbo control logic.
Type Parameter Notation Reference Engine mass flow m˙e,ref
Compressor outlet pressure pco,ref
Measured/Estimated Compressor inlet pressure pci
Temperature at compressor outlet Tco
Air inlet temperature Tci
Turbo speed ωt
Turbine outlet pressure pto
Exhaust gas temperature Texh
Exhaust gas pressure pexh
Exhaust gas mass flow m˙exh
Based on the inputs listed in Table 2.1, the model is calculating a demanded waste gate mass flow from the difference between the exhaust gas mass flow, ˙mexh, and a reference turbine mass flow ˙mt,ref
˙
mwg = ˙mexh− ˙mt,ref. (2.22)
The demanded waste gate mass flow is the amount of exhaust gas mass flow passing through the waste gate valve that, according the turbo control logic system, will results in optimal turbocharger performance at the current time step. The turbine torque is thereby limited by reducing the turbine mass flow if the exhaust gas mass flow from the engine is to significant. When the demanded waste gate mass flow has been determined, a demanded effective waste gate area, Awg, can be calculated
using the throttle equation. The throttle equation describes the flow through the throttle opening, ˙ mth, according to ˙ mth= p √ RTACDΨ(Π( pds pus )) (2.23)
where pus and Tus is the upstream pressure and temperature respectively, R is the ideal gas constant,
CD is the discharge coefficient for the throttle opening and A is the effective throttle area. Ψ(Π(ppusds))
is the psi-function expressed according to
Π(pds pus ) = max(pds pus , 2 γ + 1 γ−1γ ) (2.24a) Ψ(Π(pds pus )) = r 2γ 1 − γ Π2γ − Π γ+1 γ (2.24b)
where pds is the down stream pressure and γ is the ratio of specific heats [2]. Using the throttle
equation (2.23) to solve for the area, the demanded effective waste gate area is calculated according to Awg = ˙ mwg √ RTexh pexhCDΨ(Π(ppexhto )) . (2.25)
The effective waste gate area is then used to estimate the waste gate actuator position. The reference turbine mass flow is essentially derived from the two reference input parameters according to Table 2.1. Combining the shaft equation (2.15), the isentropic turbine efficiency according to equation (2.9) and the actual produced turbine power according to equation (2.10a), can the reference turbine mass flow be expressed as
˙ mt,ref = ωt,ref(Jtcdωdttc + (Mc,ref)) cp,tηmηt,T −STexh 1 −Π1 γ−1 γ (2.26)
where ωt,ref is the reference turbine speed and Mc,ref is the reference compressor torque. These
are derived from the input reference parameters. (2.22) to (2.26) refers to the Turbine Model in Figure 2.4. The reference engine mass flow, ˙me,ref is first corrected using equation (2.17a) giving
the corrected reference engine mass flow, ˙mcorr,e,ref. A reference compressor pressure ratio, Πc,ref, is
obtained using the reference compressor outlet pressure, pco,ref, and the compressor inlet pressure,
pci. These steps refers to the Reference value conversion according to Figure 2.4. The compressor
2.2. TURBO CONTROL LOGIC CHAPTER 2. THEORY
performance map to interpolate a reference compressor efficiency, ηc,ref and a reference compressor
speed, ωc,ref. After compensating for a possible difference between turbine and compressor speed,
can the reference compressor speed finally be used to estimate the reference turbine speed, ωt,ref.
Using equation (2.7) together with the reference compressor efficiency and the reference compressor pressure ratio, a reference compressor outlet temperature is obtained according to
Tco,ref = Tci+ Tco ηc,ref n (Πc,ref) γ−1 γ − 1 o . (2.27)
The compressor reference torque is finally obtained by dividing the actual consumed compressor power according to equation (2.3a), with the reference compressor speed according to
Mc,ref =
˙
me,ref(Tco,ref − Tci)cp,c
ωc,ref
. (2.28)
A feed forward PID-regulator is used to control speed of the turbocharger. This is done by regulating a demanded angular acceleration, dωtc/dt, to minimize the difference between the reference turbine
speed, ωt,ref and the actual turbine speed ωt.
Important to the MATLAB Simulink model is the constant library containing all the constant values needed for the equations in the model. Some of the quantities stored in the constant library are actually variable, but they are for simplicity assumed constant for this application. The constants are presented in Table 2.2. Apart from the constants listed in Table 2.2 the library also contains information about the compressor performance map and PID-regulator parameters.
The MATLAB Simulink model is mainly built using blocks with basic mathematical operations to solve its equations. No time integration is performed and therefore the model can be viewed as discrete. All operations are performed using SI units and input signals that are not in SI units are converted. Important is also the fact that the model expects time averaged input signals and is therefore not taking the pulsating conditions in the exhaust gas manifold into account.
Table 2.2: Constants in the EMS turbo control logic library.
Parameter Notation Turbine efficiency ηt
Specific heat at turbine inlet cp,t
Specific heat at compressor inlet cp,c
Ratio of specific heats at turbine inlet γ Turbocharger inertia Jtc
2.3
Engine Simulation Models
As previously mentioned, two engine simulation models are subject to this thesis. These models are built using GT-POWER, which is the engine simulation software package of GT-SUITE developed by Gamma Technologies. GT-POWER solves the flow of gas through the engine using the 1D-solution of the fully unsteady and non-linear Navier Stokes equations [5].The two models are simulating the same engine with different levels of accuracy. The difference in accuracy is determined by the level of detail by which, relative the physical engine, the flowing gas volumes, are described. The FRM is roughly simulating in real time speed which means that it takes 1 second to simulate 1 second whilst the detailed model is roughly ten times slower than real time. The models are built in GT-POWER by combining objects describing the mechanics and fluid mechanics for a specific part in the gas exchange system. The type of objects used in the model are pipes, cylinders, crankshaft, compressor, turbine, intercooler and turboshaft. A representation of an engine model layout is presented in Figure 2.5. Crankshaft Intercooler Cylinders Turbine Compressor Turboshaft
Figure 2.5: Representation of engine model layout. Representative for both GT-POWER engine models. The lines connecting the objects represent the pipe objects through which gas is flowing. Observe the two inlets for the twin scroll turbine.
To control the engine simulation models, a number of reference signals are needed. The crankshaft object requires a signal for demanded engine speed, ne. Each cylinder object is controlled by an
injector object that models the fuel combustion and is dependent on three variables. Information about when to inject the fuel is called start of injection, or SOI, and tells at which crank angle the injection each stroke is to be started. A rail pressure reference signal is needed for information about the profile of fuel flow which is to be injected into the cylinder. Finally the amount of fuel per engine cycle, delta, is also needed for the injector object in mg/cycle.
2.3. ENGINE SIMULATION MODELS CHAPTER 2. THEORY
The reference signals can be either transient or stationary defining the type of simulation desired. Stationary simulations with constant reference signals are automatically terminated once steady state has been reached.
Method
To succeed with the goal of this thesis a co-simulation setup had to be built. The purpose of the co-simulation setup is to allow the engine simulation model to communicate with the turbo control logic of the EMS turbo control model through a communication interface. The aim is to setup a communication that provides the engine simulation model with the appropriate waste gate diameter for the current time step.
This chapter is divided in two main parts. The first part, called ”EMS implementation”, describes the method for developing the simulation strategy. The second part, called ”EMS testing and im-provement implementation”, describes the method for testing and investigating the functionality of the EMS turbo control logic and the suggested improvements.
3.1
EMS implementation
This thesis work has been conducted using MATLAB version R2019b along with the GT-SUITE version v2019. To make the EMS turbo control logic accessible to the engine simulation model, a communication interface between MATLAB Simulink and GT-POWER was required. This would allow the models to exchange data during the co-simulation. A co-simulation approach utilizes the strengths of MATLAB Simulink and GT-POWER in the types of modelling they are designed for, rather than building the turbo control logic directly in GT-POWER. Two options of communication interface tools were considered for this application.
• Simulink Harness: A built-in function in GT-POWER that allows for a variety of options regarding co-simulation together with MATLAB Simulink models.
• Functional Mockup Unit, FMU: A method for combining multiple simulation models in the form of functional mock-up units in an interface called functional mock-up interface. The method is applied for other adjacent applications at the company.
3.1. EMS IMPLEMENTATION CHAPTER 3. METHOD
3.1.1
Communication interface
To setup the co-simulation using the Simulink Harness tool, the GT-SUITE Control Coupling and Real Time manual was followed. The GT-POWER Simulink Harness functionality was made avail-able in MATLAB Simulink by adding a path in MATLAB so that the MATLAB directories point to the folder in which the functionality is stored. This makes a Simulink Harness block available in the MATLAB Simulink Library browser. A code generation target file was also added which is respon-sible for generating a .dll file during model compilation. This file is needed to run the co-simulation with GT-POWER as master and contains a c-code version of the compiled Simulink model. A visual representation of how the co-simulation setup allows communication between the EMS logic and the engine simulation model is presented in Figure 3.1.
MATLAB Simulink GT-POWER .dll -file Simulink Harness Compilation Data Data (Master) (Slave) EMS logic
Engine simulation model
Figure 3.1: Visual representation of the co-simulation setup.
The solver time step was set to fixed. The fixed solver time step, later determined by convergence study, dictates the communication interval between the MATLAB Simulink model and the GT-POWER model. A discrete solver was chosen for the model since no continuous states are occurring, which means that no time integration is performed.
3.1.2
Simulink model adaptations
When the interface had been prepared in MATLAB the process of adapting the baseline Simulink model containing the EMS control logic could start. First, a so called Simulink Harness-block was added to the model. The block contains one input and one output that represent all the signals that are being sent to and from the GT-POWER model. Figure 3.2 describes the Simulink model when the Simulink Harness block was added. The Simulink Harness output signal was split into its components using the MATLAB Simulink standard block demux. The demux-block splits the input signal by a specified number. Therefore it is important that the specified number is the same as the number of signals being sent from the GT-POWER model. The order by which the signals are sent, is also of importance since the demux outputs are not labelled. The output signals from the Simulink model were combined for the input to the Simulink Harness block using the MATLAB Simulink standard block mux. Similarly to the model inputs, the order of model outputs needs to correspond to the order of expected inputs in the GT-POWER model. Note that the number of outputs and inputs presented in Figure 3.2 is not representative for the actual model.
phys-Reference value conversion Reference Speed and Reference Tourqe estimation Turbine model PID regulator Demux Mux Simulink Harness
Figure 3.2: A representation of the MATLAB Simulink model with the Simulink Harness added.
ical engine tests as input. The simulation results could be matched against the expected outputs, also stored in the engine test results. When the data matched, the isolated model was working correctly.
3.1.3
Transient reference data
Transient reference data was required to run the co-simulation. Transient reference data contains information about how to run the engine during the simulation and consists of a number of reference signals. The required reference signals are presented in Table 3.1.
Table 3.1: Required reference signals.
Parameter Notation Unit Usage
Engine Speed ne [RPM] Engine Simulation Models
Delta δ [mg/str]
SOI SOI [-]
Rail Pressure prail [bar]
Reference compressor outlet pressure pco,ref [Pa] EMS control logic
Reference engine mass flow m˙e,ref [kg/s]
3.1. EMS IMPLEMENTATION CHAPTER 3. METHOD 0 20 40 60 80 100 40 50 60 70 80 90 100
Percent of totalt test duration [%]
P ercen t of maxim um test en gine sp eed [%]
Reference Engine Speed, ne
0 20 40 60 80 100 40 50 60 70 80 90 100
Percent of totalt test duration [%]
P ercen t of maxim um delta [%] Reference Delta, δ
Figure 3.3: Engine simulation model reference signals
0 20 40 60 80 100 40 50 60 70 80 90 100
Percent of totalt test duration [%]
P ercen t of maxim um Mass Flo w [%]
Reference Engine Mass Flow, ˙me
0 20 40 60 80 100 40 50 60 70 80 90 100
Percent of totalt test duration [%]
P ercen t of maxim um Reference Pressure [%]
Reference Compressor Outlet Pressure, pco,ref
Figure 3.4: EMS logic reference signals
The reference data presented in Figures 3.3 and 3.4 were later divided into a number of shorter transients. This enabled shorter simulations to be performed which are more suitable for the detailed simulation model. Reference data for SOI and rail pressure were not logged in the engine test results and the information had to be obtained using another method.
3.1.4
Engine simulation model adaptations
simulations.
To implement the EMS logic in the engine simulation model, a so called GT-POWER Simulink Harness object was added. This object replaced the original waste gate controller and outputs a demanded waste gate diameter to the turbine object. The Simulink Harness object extracts the required signals, according to table 2.1, from the engine simulation model and delivers them to the EMS turbo control logic. The MATLAB Simulink model must be compiled to a .dll file beforehand. This file is called upon in the GT-POWER Simulink Harness object. Each signal that is sent to the EMS logic needs to be listed in the same order as they are expected. The Simulink Harness object also receives the calculated outputs from the EMS logic, e.g. the demanded waste gate diameter. Similarly, the received signals had to be listed in the same order as they were sent from the EMS logic. A representation of the engine simulation model with the Simulink Harness object is presented in Figure 3.5. Crankshaft Intercooler Cylinders Turbine Compressor Turboshaft Texh pexh pto ˙mexh Tco pci Tci ωt Simulink Harness pco,ref ˙ me,ref ne δ
Figure 3.5: Representation of the adapted GT-POWER engine simulation model including a Simulink Harness object providing the turbine object with the EMS logic output.
3.1. EMS IMPLEMENTATION CHAPTER 3. METHOD 0 20 40 60 80 100 0 20 40 60 80 100 0 100 Engine Speed [%] Brake Torque [%] SOI [%] SOI 0 20 40 60 80 100 0 20 40 60 80 100 0 50 100 Engine Speed [%] Brake Tourqe [%] Rail Pressure [%] Rail Pressure
Figure 3.6: Interpolation tables for SOI and Rail Pressure.
3.1.5
Convergence study
When all preparations with the simulation setup were completed a convergence study was performed. The convergence study aimed at finding a suitable fixed time step for the MATLAB Simulink model with the EMS logic. This time step dictates the communication interval between the models and is critical for the co-simulation to converge properly. Multiple simulations were performed using varying times steps ranging from 0.01 seconds down to 0.0001 seconds. For each simulation a new .dll -file had to be compiled containing the correct model version using the desired time step. The model was found to be converging at a time step of 0.001 seconds and lower.
3.2
Turbine modelling improvement implementation
Analysis of the EMS logic showed that the turbine modelling had room for improvements. These improvements were mainly related to three areas:
• The turbine efficiency in the original turbine model is considered constant.
• The original turbine model is not taking into account the differences in twin scroll inlet condi-tions.
• The original turbine model is not taking into account the pulsation of the turbine inlet tem-perature and pressure.
The work proceeded by investigating the possibilities of implementing different improved versions of the turbine model with respect to the areas of improvements listed above. One of the advantages with the new co-simulation setup is that it allows for investigating how the performance of the EMS turbo control logic is influenced by different turbine models of varying accuracy. By utilizing the ability to make quick adaptations to the co-simulation models, one can easily change the premises by which the EMS turbo control logic is calculating the waste gate diameter. Improving the turbine model is preferably done with more, otherwise non-available data, that can be used for better describing the conditions in the turbocharger. The co-simulation platform has the unique ability to extract this otherwise non-available data directly from the engine simulation model. The turbine modelling in the EMS logic is essentially summarized by the denominator in equation (2.26) according to
cp,tηmηt,T −STexh 1 − 1 Π γ−1γ = Pt ˙ mt (3.1)
which is equal to the total produced power from the turbine divided by the total turbine mass flow. The accuracy of the turbine modelling is thereby dictated by the accuracy of each individual component in this equation.
Enabling more accurate modelling of the twin scroll turbine required a couple of assumptions to be made. One of the assumptions made is to view the two scrolls of the twin scroll turbine as two separate turbines. The implications of this assumption are discussed in Newton et al [14]. This assumption neglects any influence between the two inlets. Each of the scrolls are thereby assigned separate pressure ratios, efficiencies and inlet temperatures. To then obtain the overall performance of the turbine an averaging scheme is needed to calculate the average scroll inlet temperature, inlet pressure and efficiency. This averaging scheme is presented later in this chapter.
To deal with the assumption of viewing the two inlets to the turbine as two separate turbines, an ad-ditional set of turbine inlet conditions parameters were introduced. These differentiated temperature, pressure, specific heat, gamma and efficiency between the two exhaust gas manifolds. Parameters specifically assigned to the outer or inner manifolds are noted with OS and IS respectively.
Further-more, three versions of improved turbine models were implemented to investigate their effect of the EMS turbo control performance.
• Version 1: A turbine model where previously non-available input parameters, needed for an improved turbine model, are estimated based on original inputs. Estimation schemes are only created for the outer turbine inlet pressure and the turbine efficiency.
3.2. TURBINE MODELLING IMPROVEMENT IMPLEMENTATION CHAPTER 3. METHOD
• Version 3: A turbine model considered to be true or ideal, calculated by the engine simulation model in GT-POWER according to Pt/ ˙mt, which is the total produced power from the turbine,
Pt divided by the total turbine mass flow, ˙mt.
A non-available parameter is any quantity that is not included in the original set of inputs provided to the EMS turbo control logic. For this thesis any parameter that can add to the description of the conditions in the turbine is of interest to help develop the turbine model.
The turbine model Version 1 is aimed at exploring the possibilities of creating a more accurate turbine model using only the original inputs and parameters previously available to the EMS. This requires estimations schemes to be created for the parameters describing the turbine inlet conditions at the outer inlet. These are parameters not included in the original inputs and are needed for the new model to take both inlets into consideration. An attempt at performing a turbine efficiency estimation will also be made since also this parameter is not originally provided to the EMS.
The turbine model Version 2 is based on the same assumption made for Version 1 but is not dependent on any estimations to me made. The turbine model Version 2 instead extracts these otherwise non-available parameters directly from the engine simulation model in GT-POWER. This model therefore investigates the potential of improving the EMS performance with a turbine model according the Version 1, without being dependent on possibly insufficient data sets required for the parameter estimation.
Turbine model Version 3 investigates a total potential performance gain by directly utilizing the results from the GT-POWER turbine model rather than the using the turbine model in the EMS turbo control logic.
3.2.1
Outer scroll inlet pressure estimation
It is concluded in Marelli and Capobianco [16] that opening the waste gate has a great influence on turbine efficiency. Results from a co-simulation using the FRM-model showed a significant pressure drop in cycle average outer scroll inlet pressure relative the cycle average inner scroll inlet pressure during waste gate opening. This is a result of the waste gate only being connected to one of the two scroll inlets and is something not considered in the original EMS turbine model where only one of the two scroll inlet pressures are received. The results also showed different inlet pressures when the waste gate was not activated. This is due to differences in design between the outer and inner exhaust gas manifolds resulting in the pressure drop behaving differently. The resulting scroll inlet pressures along with waste gate diameter from the FRM-model simulation is presented in Figure 3.8.
Figure 3.8 shows a higher pressure in the outer scroll inlet when the waste gate is not activated. The outer pressure then deteriorates with the opening of the waste gate whilst the inner inlet pressure is seemingly unaffected.
An estimation scheme for the outer cycle average scroll inlet pressure was therefore created. The idea was to estimate the pressure, pexh,OS using the received inner inlet pressure pexh,IS and the
demanded waste gate opening, dwg. Simulation results of these parameters were gathered and put
40 50 60 70 80 90 100 40 60 80 100 120 Time [%] Pressure [%]
Scroll Inlet Pressure Outer Inlet Inner Inlet 0 20 40 60 80 100 120 W aste Gate Op ening [%] WG opening
Figure 3.8: Cycle averaged inner and outer scroll inlet pressure against waste gate opening.
pexh,OS(pexh,IS,dwg) = A + Bpexh,IS+ Cdwg + Dp
2
exh,IS + Epexh,ISdwg+ F d2wg (3.2)
where the constants A to F are chosen so that the function best fits the gathered data points. The performance of the function was then tested by comparing the results against the actual outer scroll inlet pressure. A similar behaviour was found for the inlet temperatures meaning that the outer scroll inlet temperature can be estimated using a similar method.
3.2.2
Turbine efficiency estimation.
To estimate the turbine efficiency correctly, turbine performance maps were implemented to the EMS logic. The performance maps were extracted from the detailed GT-POWER model where they are used to describe the behaviour of the turbine. The maps were implemented in the EMS logic as interpolation tables. They consist of an efficiency matrix and two corresponding vectors for turbine mass flow, ˙mt and pressure ratio 1/Πt. The scroll specific turbine mass flow, ˙mt,s, was estimated
by dividing the total turbine mass flow in half. The outer scroll specific mass flow was adjusted by subtracting the demanded mass flow. This was done since the waste gate only affects the gas flow in the outer manifold. The resulting scroll efficiency was linearly interpolated using the MATLAB function for 2-dimensional interpolation interp2. Each scroll has a separate set of performance maps describing the efficiency at different states of scroll admission. The results from each map are in turn linearly interpolated based on the scroll inlet pressure ratio according to
SP R = pexh,OS pexh,IS
(3.3)
which gives an indication of which type of admission is occurring at the current time step. Using this map based algorithm the efficiency for the respective scrolls was estimated during the simulation and compared to the efficiency calculated in GT-POWER.
3.2. TURBINE MODELLING IMPROVEMENT IMPLEMENTATION CHAPTER 3. METHOD
had to consider the pulsating exhaust gas flow for the efficiency estimation to be correct.
3.2.3
Pulse estimation for inlet pressure
To be able to perform accurate efficiency estimations with the turbine performance maps using the original time averaged inputs, a pulse estimation scheme was created. The idea was to estimate the pulses based on the cycle average exhaust gas pressure, pexh, engine speed, ne and engine delta,
δ. There were several motivations behind estimating the pulses using only engine speed and delta. The resulting pressure propagations in the exhaust gas manifold are of course mainly dictated by the actual design of the surrounding gas exchange system. However since the method is based on simulated data, the non-constant parameters that influences the characteristics of the pressure pulses are of interest. Engine speed and delta are non-constant and known to have influence on both the highest possible exhaust gas pressure Pmax and the current pressure peak, Ptop. By using only these
parameters for estimating the pulse characteristics, the model will neglect some aspects like the influence of the turbocharger and wave interference. However, using only engine speed and delta for pulse estimation is preferable when building the algorithm since these parameters are easily made available to the EMS and also reduce the complexity of the algorithm as compared to involving more parameters.
−1000 0 100 200 300 400 500 600 20 40 60 80 100 Crank angle [-] P ercen t of maxim um pressure [%]
Low engine speed
Low delta Mid delta High delta −100 0 100 200 300 400 500 600 0 20 40 60 80 100 Crank angle [-] P ercen t of maxim um pressure [%]
Mid engine speed
Low delta Mid delta High delta −100 0 100 200 300 400 500 600 0 20 40 60 80 100 Crank angle [-] P ercen t of maxim um pressure [%]
High engine speed
Low delta Mid delta High delta
Figure 3.9: Pressure pulses over one engine cycle for different engine speeds and delta. Each subfigure presents pressure pulses from equal engine speed but three different values of delta. The dashed lines represents the average pressure for each curve.
Figure 3.9 shows the large pressure variations occurring during the engine cycle, especially for higher engine speeds and delta values. At these circumstances the pulses differs largely from the cycle average values. As presented in the figure, the pressure curves have a periodic behaviour and is seemingly being scaled around its cycle average value for different engine speeds and delta.
To build the pulse estimation algorithm every one of these curves was first centred around their abscissa, meaning each data point in the curve was subtracted with the cycle average pressure. The curves were then normalized to have length and maximum height equal to 1. For each curve a scaling value, S, was calculated according to
3.2. TURBINE MODELLING IMPROVEMENT IMPLEMENTATION CHAPTER 3. METHOD
where Ptop is the highest data point on the curve and Pavg is the cycle average value. Examples of
centred and normalized curves are presented in Figure 3.10.
0 0.2 0.4 0.6 0.8 1 −1 −0.5 0 0.5 1
Figure 3.10: Normalized pulse curves
The idea was to estimate the pressure pulses using a general Fourier-series function. Note that each gathered pulse curve represents a specific engine speed and delta value. It was found that a Fourier-series with five terms according to
f = a0+ a1cos(tw) + b1sin(tw) + a2cos(2tw) + b2sin(2tw) + a3cos(3tw)+
b3sin(3tw) + a4cos(4tw) + b4sin(4tw) + a5cos(5tw) + b5sin(5tw)
(3.5)
has sufficiently many terms to describe an arbitrary normalized pressure curve. The variable t is the current time step and w is a weight constant. The constants a0-a5, b1-b5 and w in equation
(3.5) are fitted to each normalized curve using the MATLAB curve fitting tool cftool giving a set of constants valid for a specific engine speed and delta value. Arrays were created of fitted constant values with length equal to the number of gathered pressure curves. These arrays were then used build interpolation tables to describe how the constants change with engine speed and delta during the simulation.
0 0.2 0.4 0.6 0.8 1 −1 −0.5 0 0.5 1 Normalized curve Fit
Figure 3.11: Curve fitted fourier series function.
To obtain the correct amplitude for the curve, the previously calculated scaling values, according to equation (3.4), are used in an interpolation table to interpolate the current appropriate scaling value. Also this interpolation is based on engine speed and delta. Finally the input cycle average pressure, pexh,IS, is added to the curve to gain the right average pressure. This pulse estimation algorithm is
summarized in the following steps:
1. Approximate the normalized curve by interpolating each constant, according to equation (3.5), from interpolation tables based on previously gathered data, using engine speed and delta value.
2. Scale the approximated normalized curve with a scaling value obtained from interpolation table, also based on previously gathered data, using engine speed and delta value.
3. Add the cycle average exhaust pressure to the curve to gain the inner average pressure.
3.2.4
Implementation of pulse estimation algorithm
The algorithm presented above had to be implemented into the EMS to test its performance. The most critical part of the implementation is to provide the algorithm with the appropriate time value, t, for the current time step. A new MATLAB subsystem block was added to the MATLAB Simulink model containing the EMS logic. This block had three inputs; cycle average inlet pressure, pexh,IS,
engine speed, ne, and delta value, δ. The block contains three sub-functions.
1. Calculation and storage of number of time steps in the current engine cycle. 2. A time step counter that provides a time value t between 0 and 1.
3. Pulse estimation.
The first sub-function calculates the number of time steps in the engine cycle given the engine speed stored at the start of the engine cycle. The number of time steps, N , is calculated by dividing the time it takes for an engine cycle to pass with the constant simulation time step,
N = 30 ne∆t
3.2. TURBINE MODELLING IMPROVEMENT IMPLEMENTATION CHAPTER 3. METHOD
where ∆t is the time step. The counter then uses the value N to count each time step until N is reached. The current count value is noted k. By dividing k with N a time value t is provided to the pulse estimation between 0 and 1 to fit the length normalization. When k is equal to or larger than N the process is repeated with N being updated with the engine speed of the current time step and k is reset to 0. A visual representation of this algorithm implementation is presented in Figure 3.12. ne pexh δ N = n30 e∆t N Memory k = k + 1 Memory k if k >= N k = 0, else k = k if k = 0 : Activate t = Nk Pulse estimation + pexh,new
3.2.5
Averaging for overall turbine performance
Having considered the twin scroll turbine as two separate turbines, an averaging method was im-plemented to obtain the overall performance of the turbine. Chumpsty and Horlock [17] presents averaging methods for temperature and pressure suitable for the type of flow occurring in multi entry turbomachinery. To obtain the average inlet temperature a mass flow averaging method was applied according to equation (2.19). The individual scroll specific mass flows was calculated by dividing the total turbine mass flow in half and subtracting the demanded waste gate mass flow from the outer scroll mass flow. This method neglects the possible difference in scroll specific mass flow due to differences in design between the two exhaust gas manifolds. However it does consider the decrease of mass flow through the outer turbine scroll when the waste gate is activated which is when the difference in mass flow through the two scrolls differs the most. The mass flow averaged temperature was Texh,ave = 1 2m˙exhTexh,IS + ( 1 2m˙exh− ˙mwg)Texh,OS ˙ mexh− ˙mwg . (3.7)
The same mass flow averaging method was applied for calculating, the average turbine efficiency. For turbine the models version 2 and 3 a mass flow averaging was also performed on the specific heat and gamma values where different values from the separate inlets were extracted from the GT-POWER model. To calculate a scroll average turbine inlet pressure Chumpsty and Horlock [17] suggests a work averaging method to maintain the enthalpy flux between the inlets. The work averaged pressure was applied according to equation (2.20) as
Chapter 4
Results
This chapter presents the results of the thesis. The thesis is investigating the possible improvements gained to the turbocharger waste gate control in an engine management system, EMS. This is done by implementing various estimations schemes for a set of variables required to improve a turbine model within the EMS - variables that previously were non-available to the EMS.
The performance from three estimation schemes are implemented and tested. The performance is tested by comparing the results against the considered true value with the objective of achieving a result that resembles the true value as much as possible. For this thesis the true value is the value calculated by an engine simulation model in GT-POWER to which the EMS waste gate control logic is connected via a co-simulation setup. Initially the performance from an estimation of the pressure in the outer turbine inlet is presented. This is followed by the performance from a turbine efficiency estimation.
The thesis also makes an attempt at estimating the pulsating conditions at the turbine inlet. The performance from the pulse estimation is presented in terms of turbine efficiency where pulsating conditions are estimated for the pressure at the inner turbine inlet.
4.1
Outer turbine inlet pressure estimation
The result from the pressure estimation scheme for the outer turbine inlet is presented in Figure 4.1. 20 40 60 80 100 60 70 80 90 100 110
Percent of total simulation duration [%]
P ercen t of maxim um pressure [%]
True inner pressure True outer pressure Estimated outer pressure
Figure 4.1: The estimated pressure for the outer turbine inlet in comparison with the true outer inlet pressure. The figure also presents the true inner turbine inlet pressure used for the estimation.
4.2. TURBINE EFFICIENCY ESTIMATION CHAPTER 4. RESULTS
4.2
Turbine efficiency estimation
The result from the turbine efficiency estimation is presented in Figure 4.2. The figure presents two different results from the efficiency estimation. These are compared to the true turbine efficiency calculated by the engine simulation model. For both results the same transient reference signals are used to perform the simulations. Each of the curves in the figure are averaged after the simulation to show the efficiency trends for each result. The grey curve shows the result of the estimation when the inputs to the efficiency estimation scheme are cycle averaged. Cycle averaged inputs are the type of inputs originally provided to the EMS. As can be observed this result differs quite a lot from the true efficiency. The black curve is the result from the scheme when pulsating (non-averaged) inputs are used. The pulsating behaviour of the inputs is provided by the engine simulation model. The resulting efficiency is afterwards averaged to create the results presented in the figure. As can be seen, pulsating inputs must be used for the efficiency estimation scheme to give what is considered a good result compared to the true value from the engine simulation model.
0 10 20 30 40 50 0.7 Time [s] T urbine efficien cy [-] True efficiency Pulsating inputs Averaged inputs
4.3
Pressure pulse estimation for inner turbine inlet
The results of the pulse estimation is presented in Figure 4.3. The purpose of the pulse estimation scheme is to investigate if it is possible to acquire the pulsating behaviour of the conditions at the turbine inlet when only cycle averaged data is available. To simplify the testing of the pulse estimation scheme, the pulse estimation is only applied for the pressure at the inner turbine inlet. Therefore the inner turbine inlet pressure input for this test is provided as a cycle averaged value. The rest of the pulsating inputs describing the conditions at the turbine inlet are provided by the engine simulation model. The performance of the pulse estimation is presented in terms of the resulting turbine efficiency which is compared against the true turbine efficiency calculated by the engine simulation model. The result is also compared against a turbine efficiency estimated using a non pulsating input for the pressure in the inner turbine inlet.
0 20 40 60 80 100 0.7 Time [s] T urbine efficie ncy [-] True efficiency Pulse estimation No pulse estimation
Figure 4.3: Comparison of estimated turbine efficiency with and without pulse estimation in the inner turbine inlet pressure. All other inputs to the turbine efficiency estimation are pulsating.
4.4. TURBO CONTROL PERFORMANCE ANALYSIS CHAPTER 4. RESULTS
4.4
Turbo control performance analysis
To investigate the possible performance gain to the EMS turbo control with improved turbine mod-elling, a turbo control performance analysis was made. In the previous chapter the following versions of improved turbine models were presented:
• Version 1: A turbine model where previously non-available parameters are estimated based on original inputs, like outer turbine inlet pressure and turbine efficiency.
• Version 2: A turbine model using ideal temperatures, pressures, efficiency, heat capacity and gamma extracted from GT-POWER
• Version 3: A turbine model considered to be true or ideal, calculated by the engine simulation model in GT-POWER according to Pt/ ˙mt, which is the total produced power from the turbine,
Pt divided by the total turbine mass flow, ˙mt.
This performance analysis is focusing on the performance using turbine models Version 2 and Version 3. A complete performance analysis is not possible with turbine model Version 1 since some required non available parameters, like the outer turbine inlet temperature, have not been estimated in this thesis.
