Integration av en CR-insprutare i enforskningsdieselmotor och undersökning avmultipelinsprutning

(1)

Integration of a CR Injector in a

Research Diesel Engine and

Investigation of a Multiple Injection

ARTIOM LAMADRID FERNANDEZ

Master of Science Thesis

Stockholm, Sweden 2010

(2)

Integration of a CR Injector and

Investigation of a Multiple Injection

Artiom LaMadrid Fernandez

Master of Science Thesis MMK 2010:17 MFM134

KTH Industrial Engineering and Management

(3)

Examensarbete MMK 2010:17 MFM134

Integration av en CR-insprutare i en

forskningsdieselmotor och undersökning av

multipelinsprutning

Artiom LaMadrid Fernandez

Godkänt 2010-04-21 Examinator

Hans-Erik Ångström

Handledare

Hans-Erik Ångström

Uppdragsgivare

LVK TU München

Kontaktperson

Simon Thierfelder

Sammanfattning

I utvecklingen av den encylindriga forsknings-diselmotorn Hatz 1H30 på Technische

Universität München har tidigare en ottoinsprutare använts i kombination med ett

common rail-system som en temporär lösning. Ett common rail-system ger möjligheten

att separera regleringen av insprutningen från de mekaniska komponenterna av motorn.

En common rail-insprutare gör det möjligt att göra flera insprut varje arbetscykel och

tillhandahåller möjligheten att elektroniskt variera timingen av insprutningen.

Efter att en marknadsundersökning har gjorts bland huvudtillverkarna av injektorer

inköptes en Bosch CRI2 injektor från en BMW M47N-motor. Det förra cylinderhuvudet

var anpassat för en ottoinjektor och var därför tvunget att designas om. En konstruktion

gjordes för att fästa injektorn och för att försluta förbränningskammaren.

För att styra injektorn behövs ett flertal signaler för att utvärdera vevaxelpositionen och

betämma insprutstimingen. Behandlingen av dessa signaler och mjukvaran för styrning

var först testade med en AFT PROtroniC motorstyrenhet för sig själv i en provrigg med

en signalgenerator. Senare utfördes en installation av styrenheten med de riktiga

signalerna från själva motorn. Funktionaliteten kunde verifieras genom klickandet från

injektorn.

För att använda fördelarna med common rail-systemet utfördes tester för att utvärdera

förbränningsegenskaperna med en multipelinsprutsstrategi. Dessa tester gjordes med

den tidigare otto-injektorn som en förstudie för att kartlägga effekterna av en sådan

strategi hos 1H30-motorn. På grund av förlängda tillverkningstider kunde inga tester

göras med CRI2 injektorn inbyggd i motorn.

Försöken med en för- och en huvudinsprutning visade att en multipelinsprutsstrategi

erbjuder en viss optimeringspotential men löser fortfarande inte problemen med

emissioner vilket inte tillåter motorn godkännas av EPA4 cykeln. De största problemen

finnes vid låga vridmoment.

(4)

Master of Science Thesis MMK 2010:17 MFM134

Integration of a CR Injector in a Research Diesel

Engine and Investigation of a Multiple Injection

Artiom LaMadrid Fernandez

Approved 04/21/2010 Examiner

Hans-Erik Ångström

Supervisor

Hans-Erik Ångström

Commissioner

LVK TU München

Contact person

Simon Thierfelder

Abstract

In the development process of the Hatz 1H30 one-cylinder diesel research engine at

Technische Universität München has an otto-adapted injector in combination with a

common rail system been used as a temporary solution. A common rail system provides

the possibility to separate the injection control from the mechanical components of the

engine. A common rail injector is capable of multiple injections during each combustion

cycle and offers the possibility to electronically change the injection timing.

After performing a market research among the main manufacturers of injectors a Bosch

CRI2 injector from a BMW M47N engine was purchased. The previous cylinder head

was adapted for the otto injector and had to be redesigned. A construction was made to

attach the new CRI2 injector and seal off the combustion chamber.

To control the injector multiple different signals are needed for an evaluation of the

crankshaft position and injection timing. The treatment of these signals and the control

software were first tested with an AFT PROtroniC ECU alone in a test bed using a

signal generator. Later a installation of the ECU using the actual signals from the engine

was made. The functionality was verified through a clicking injector.

To use the advantages of the common rail injector tests were made to evaluate the

combustion characteristics with a multiple injection strategy. These tests were still made

with the Otto injector as a pre study to map the impacts of the multiple injections on the

1H30 engine. Due to delayed manufacturing times was the new design not done during

the project and no tests with the CRI2 injector assembled in the engine could be made.

The testing of one pre- and one main injection shows that a multiple injection strategy

provides some optimization potential but does not solve the emission problems that the

engine has and keeps it from passing the EPA4 testing cycle. The largest problems can

be found at low loads.

(5)

Preface

This Master Thesis in Vehicle Engineering was written at the Lehrstuhl für Verbrennungskraft-maschinen, LVK (The Chair in Internal Combustion Engines) at Technische Universität München, TUM in Munich, Germany for a project with HATZ Motorenfabrik GmbH & CO.KG.

I would like to thank the following: Dipl.-Ing Simon Thierfelder for assisting me through the entire project, Dipl.-Ing Johann Wloka, for technical expertise on injection systems and HiL, Dipl.-Ing. Hans-Philip Walther for CAD-drawings of the injector as well as ECU software and help on electrical problems, Prof. Dr.-Ing. Georg Wachtmeister for helping providing the nozzles, Dipl.-Ing. Alexander Preis for expertise on nozzle design and manufacturing, M.Sc. Reinaldo LaMadrid, for help on design, constructive and manufacturing expertise, M.Sc. Jo-hannes Scharpf, for technical expertise on control unit programming and Dipl.-Ing. Michaela Benedikt for practical help on sensor systems.

(6)

Definitions and abbreviations

Abbreviation Full name Description

Bosch Robert Bosch GmbH A German manufacturer of

automotive components

CA◦ Crank Angle Degrees A usual timing measurement unit

CAD Computer Aided Design A usage of a computer to design

CAN Controller Area Network A communication bus

CPU Central Processing Unit The part of the computer that carries out software instructions

CR Common Rail A modern diesel injection system

DIN Deutsche Institut für Normung The German organization for standardization

DMU Digital Mockup A virtual prototype of a finished

product

DoE Design of experiments A process to optimize testing

ECU Engine Control Unit An electronic control unit that

manages the engine functions

EPA U.S. Environment An organization that writes and

Protection Agency enforces environmental laws FPGA Field Programmable Gate Array An integrated circuit that allows

to be configured by the user Hatz Hatz Motorenfabrik GmbH A German engine manufacturer

& CO.KG

HELICOIL R Helical thread coil insert A bushing insert to create threads

Hole/shaft fit ISO 282-2 tolerance matching

ISO International Standard Organization A non governmental organization that manages standards worldwide

MATLAB A numerical computing environment

and programming language PCB Printed Circuit Board A board to mechanically support

and integrate electrical components Piezo Piezoelectricity/Piezoelectric

The ability of materials to generate an electric field as a response to mechanical stress

ProE Pro/ENGINEER A parametric 3D CAD/CAE/CAM

software

Simulink A tool for modeling, simulation and

(7)

Formula sign Unit Description

B1 [-]

Factor regarding decrease in screw strain due to subsidence

CO [g/kWh] Carbon monoxide emission concentration

CO2 [%] Carbon dioxide emission concentration

dD [mm] Mean diameter of sealing

Ddef [N] Deformation load on gasket

E Residuals matrix

ETmain [µ s] Main injection duration

Fop [N] Operaing load on the sealing

Fs,0 [N] Pre-tensioning force on each individual screw

Fs,max [N] Maximal force on each individual screw

Fs,tot [N] Total pressure on the sealing

Funl [N]

Reducing seal pressure force

due to combustion chamber pressure

HC [g/kWh] Hydrocarbon emission concentration

I [A] Electric

i [-] Amount or index

k0 [mm] Effective width of sealing

k1 [mm] Fictive operational width of the seal

KD [-] Temperature based shaping resistance

n [-] or [min−1] Amount or Engine speed

N OX [g/kWh] Nitrogen oxides emission concentration

pamb [bar] Ambient pressure

pcyl [MPa] Inner pressure of the cylinder

P M [g/kWh] Blackening rate (Particulate Matter)

Qpost [CA◦] Post injection amount

Qpre [CA◦] Pre injection amount

RM [%] Relative ambient humidity

SD [-] Safety factor for seals

SOImain [CA◦] Main injection timing

SSE Sum of squared residuals

Tamb [◦C] Ambient temperature

(8)

Formula sign Unit Description

X Matrix with the factors

x Regression factor

x Measured factor value

x Element of the measured factors matrix

x Vector with factors or combination of factors

x±α Corner points of the star experiment design

x± Corner points in the cube

Y Matrix with the target variables

y Target variable

y Measured target variable

β Regression coefficient

ˆ

β Calculated regression coefficient

Residual

λ [-] Combustion-air-ratio

ϕ [◦crank angle] Crank angle degree ∆ϕpost [CA◦] Post injection timing

∆ϕpre [CA◦] Pre injection timing

(9)

1 Introduction

During the 20th century the combustion engine has been the primary source of energy conversion from fossil fuels both in stationary and mobile applications such as the automobile. Recent environmental changes and pollution increase has changed the public opinion and forced many nations to enforce stricter laws regarding the consumption of fossil fuels. This development has forced many engine and automobile manufacturers to drastically decrease the consumption and emissions and increase the degree of efficiency to keep up with current legislation and market trends. New engines has to minimize toxic and green house gas emissions, noise emissions and simultaneously provide a high reliability and low cost [1] [2].

In order to improve the properties of the combustion engine, the chair of internal combustion engines at TUM is developing a research engine to evaluate the properties of the combustion and optimization possibilities, mainly regarding emission exhaust. The aim is to develop a low cost CR system for industrial applications and series production. The engine is based on the 1B30 engine from HATZ Motorenfabrik GmbH & CO.KG and is a 1 cylinder diesel 4 stroke engine which is usually connected to an electrical generator, and may in the future also be integrated as a range extender in hybrid vehicles. The research engine 1H30 will be used to research the combustion properties of small stationary engines with CR injection systems.

1.1 Background

In the development of the 1H30 a temporary fuel injection design has to date been used. The former design had an otto-adapted injector connected to a high pressure rail. This type of injector is designed to operate with a fuel injection pressure of a few bar only and is controlled electromagnetically only and not hydraulically. A diesel injection in itself defines a mayor part of the combustion cycle characteristics. A diesel engine doesn’t use a spark plug and ignites purely because of the high pressure. Due to a higher compression rate than an otto cycle the cylinder contains a higher pressure at the point of injection and brings the fuel to self ignition. Thus increases the requirements of the diesel fuel pressure, by certain engines as high as up to 2000 bar to function properly. The former injector is as previously mentioned not suited to operate at such high pressures and is therefore heavily overloaded. Due to improper application this setup can cause unwanted effects such as cavitation and fluctuations in injection pressure. Furthermore the high pressure of the injection spray wears the nozzle bore, which leads to decreased durability and robustness of the design.

1.2 Problem description

To solve the problematic mentioned above a common rail injector was integrated into the engine. The project can be divided in 4 parts as follows:

(12)

1. A market research to find a suitable common rail injector for a direct injection setup, of which the dimensions should suit the existing cylinder head. Furthermore should the injection-, thermodynamical and control properties be adapted for the previous setup. 2. Redesigning of the cylinder head to fit the new injector.

3. Control unit programming to bring the injector into service by connecting it to the control unit and programming it.

4. Investigation of the optimization possibilities with a multiple injection strategy1_.

Each part are furthermore divided in subtasks that will be described in detail later.

1.3 Project course

The table 1 and figure 1 shows the Project course in a table resp. a Gantt chart. To ensure the quality of work a milestone after each project phase was achieved.

Phase Description Milestone

Pre-1

Pre-study of diesel injectors,

their design and potentials Requirements specification of investigation of improvement possibilities the new injector

in respect to current design

1 Market research of available products Purchase of injector Pre- 2

Pre-study of cylinder head designs,

Requirements specification for design/CAD-, product development

the new design and manufacturing techniques

2 Design of new cylinder head Complete drawings and DMU Pre-3

Pre-study of engine management

Requirements specification of the systems and controlling strategies

control software Simulink and AFT software

3 Program the control software Complete software model Pre-4 Pre-study of multiple injection and DoE An experiment plan

4 Realization and evaluation of experiments Present optimization possibilities

Pre-5 Report writing Preliminary report

5 Completion of report and Completed report and presentation preparation for oral presentations

Table 1: Project course

(13)

(14)

1.4 Delimitations

The available time window for a master thesis at KTH is normally set to 20 labour weeks. The original project plan contained a phase where a design of a new nozzle that fits the combustion chamber was supposed to be made. The assumption that the injector nozzle could be duplicated from the former injector and produce an almost identical spray appeared to be false. Due to differences in inner geometry and needle appearance the flow through a similar bore with another injector nozzle would not create the same fuel spray. Therefore a hydraulic simulation of the motions of the injector components and a CFD simulation of the flow are required to be able to compare the the injectors. Because special nozzle design know-how is needed, was the former unit injector nozzle from the Hatz 1B30 (series produced) engine specially designed by and purchased from Robert Bosch. Thus the project was delimitated to finding a suitable injector with a modifiable injector nozzle without nozzle bores and hence leaving the nozzle design for a separate project.

During the last part where the investigation of a multiple injection strategy was made there ware complications in the manufacturing process at Hatz and the parts for the attachment of the new injector was not ready on time. Hence made it impossible to do any testing with the injector assembled in the engine. The multiple injections investigation was therefore made as a pre study for decision of an injection strategy and an investigation of the optimization potential at the operating points that the EPA4 cycle dictates.

(15)

2 Frame of reference

The figure 2 shows the former design of the cylinder head. Here the cylinder head and the cylinder head cover can be seen assembled with the former injector mounted and the high pressure rail attached on top.

Figure 2: Previous design

2.1 Description of a common rail system

2.1.1 System architecture

The pump supplies fuel to a high pressure fuel container, a rail, where the fuel is stored at the op-timal pressure for the engine’s instantaneous operating conditions in preparation for subsequent injection.

If an engine has more than one cylinder each one is equipped with an injector, the figure 3 shows a common rail system with one injector. The injector valve opens and closes and allows fuel to flow through the injector into the combustion chamber.

(16)

Figure 3: System configuration for the Bosch Common Rail system

The engine ECU registers load demand. In an automobile it is registered by the signal from the gas pedal. For the available test rig the engine control is made through a speed control through an electrical motor and the torque output is controlled through the amount of injected fuel (injection duration) which can be set through the ECU.

2.1.2 The fuel compression and control

The CR pump supplies fuel from the tank through the low pressure fuel inlet 2 to a high-pressure reservoir, the rail 5. There are different controlling strategies. The Hatz 1H30 engine has a high pressure side control, which is visualized better with figure 4.

(17)

Figure 4: Rail pressure control - 1. CR Pump, 2. Low pressure fuel inlet, 3. Return line, 4. Blow off valve, 5. Rail, 6. Pressure sensor, 7. CR injector connections

The 1H30 engine rail pressure control occurs at the high pressure side, which means that the pump 1 works with its maximum capacity at all operating points. The rail pressure is controlled through the blow off valve 4. It lets excess pressure out from the system when needed. The advantage with this system is that the engine becomes more dynamic. If the load is rapidly decreased and the rail pressure needs to drop it can be done quickly as opposed to the low pressure side control. A low pressure side control controls the fuel flow through the pump and therefore the pressure inside the rail 5. This is however better for the efficiency. If a low pressure is required, usually by low load operating points the pump does not have to compress excessive fuel. The 1H30 engine with its high pressure side control suffers therefore high energy losses at low load operating points.

A high pressure control system would have the same pump performance and the pressure would be controlled by a pressure control valve that lets out excessive pressure. This system is more dynamic but suffers higher energy losses at the CR pump. It is more expensive because it requires high performance valve controls on the high pressure side.

(18)

2.1.3 The common rail injector

Figure 5: A common rail injector 1. Fuel return, 2. Electrical connection, 3. Control unit of the solenoid, 4. Fuel inlet from rail, 5. Valve ball, 6. Drain restrictor, 7. Feed restrictor, 8. Valve control chamber, 9. Control piston, 10. Nozzle inlet channel, 11. Nozzle needle

The figure 5 shows a generation 2 common rail CRI2 injector from Bosch. The fuel flows through the inlet 4 from the rail at a high pressure into the valve control chamber 8. The control piston 9 is therefore pressed down and the injector is closed. If the solenoid 3 induces a current the valve ball 5 lifts from its position and lets fuel from the control chamber 8 to flow up through the outlet 1 back to the tank. This reduces the pressure in the control chamber. When it is lower than the pressure in the nozzle inlet channel 10 and the pressure form the spring the the control piston 9 moves up, the needle 11 lifts from its seat and the fuel is injected into the combustion chamber.

(19)

2.1.4 The engine control unit

The engine control unit reads all incoming signals, evaluates the crankshaft position and controls the injection using a user input. More on engine control units will be mentioned in section 5.

2.2 Technical information of the 1H30 and 1B30 engines

The tables 2 and 3 present technical data on the reference engine and on the research engine. Type Aircooled 1 cylinder 4 stroke Direct Injection Diesel Injection Mechanical with 1 unit injector

Rotation speed control Mechanical

Stroke/Bore 69mm/80mm

Displacement 347cm3

Power 5kW @ 3000rpm

Compression ratio 21.5:1 Emission certification EPA 4

Table 2: Reference 1B30 engine

Type Air/oilcooled 1 cylinder 4 stroke Direct Injection Diesel Injection 1 low pressure common rail injector

Rail pressure control Simplified control on pressure side only

Common rail pump 1 piston pump, mechanically controlled over camshafts Engine control unit AFT/Magneti Marelli (INCA)

Stroke/Bore 78mm/74mm

Displacement 335cm3

Power (aim) 5kW @ 3000rpm Compression ratio 20.3:1

(20)

3 Selection of a proper injector

The requested injector was an injector adapted for a small diesel engine equipped with common rail and that can be operated with a low rail pressure. Injectors that are adapted for bigger engines hence bigger bore must inject more fuel and are therefore adapted for greater flow and a deeper penetration of the injected fuel. Engines with higher rail pressure requires more sophisticated rail pressure control, expensive valves and smaller nozzle bores to throttle the amount of fuel flowing into the combustion chamber. Such injectors usually have a higher needle opening pressure. Therefore the designated injectors where searched for in smaller automobiles with engines, whose displacement, bore and power is roughly similar to the 1H30 engine. Another requirement was the accessibility to blank nozzles to the injector i.e. nozzles without drilled holes that was adapted for the original combustion chamber. A problem that occurred is that automobile engines usually have their injectors vertically mounted. The 1H30 engine has due to the packaging a shortage of space in the cylinder head (see figure 6 and therefore has the injector mounted with a declination of 7◦.

Figure 6: Injector packaging in the cylinder head from the 1H30 engine

Hence the spray from an unchanged injector would hit the cylinder head and not the piston bowl. This would cause fuel to stick to the cylinder liner and/or cylinder head, poor air-fuel mixture and lead to high hydrocarbon emissions. Therefore it was decided that a nozzle could not be taken from any series produced engine.

(21)

The original idea was to get one of the major manufacturers of injectors to supply a suitable setup with an injector and a blank nozzle. Due to secrecy and patents specific technical information from the mayor injector manufacturers was not handed out to a research project without a direct interest in mass production. Therefore it was chosen to focus on obtaining so much information as possible from CR injectors used in automobiles. There exists an enormous amount of different piston bowl types and sizes and therefore almost equally many injectors. Due to the many variations, and the cost of the injectors none of the contacted retailers had injectors in stock. The injectors are by retailers ordered when needed. Some retailers had a few single either broken injectors or injectors waiting to be mounted in customers cars. The companies contacted are shown in table 4 below.

Company name Country Company profile

MALI Common Rail Technologies AG CH Manufacturer of automobile parts Stanadyne Corporation US, IT Manufacturer of inter alia fuel systems

Robert Bosch GmbH DE Manufacturer of automobile parts

United Diesel Fuel Injection Services Ltd UK Retailer of injection systems DTN Diesel Technology Aguicu GmbH DE Service Centre

AED Diesel Centre GmbH DE Service Centre and retailer Denso Europe B.V. JP,NL,DE Manufacturer of automotive parts Köller + Schwemmer GmbH + Co. KG DE Retailer and service centre Autohaus Luzzi & Luzzi GmbH DE Retailer

MB Offroad parts GmbH DE Importer and retailer

Autohaus Lies GmbH DE Retailer

Magneti Marelli S.p.A. IT Manufacturer of automobile parts

DUAP AG CH Manufacturer of inter alia fuel systems

BMW Niederlassung Frankfurter Ring DE BMW retailer and service

Table 4: Companies contacted The considered injectors are presented in table 5 below.

Manufacturer Originally found in Comment

Bosch Volvo D5 engines Physically measured

Bosch Alfa Romeo JTD engines Bosch Daimler CDI engines

Bosch Puegot HDi/Ford DLD engines Bosch Fiat Multijet engines

Bosch BMW M47N engines CAD drawings available

Duap No industrial use, prototype injector CAD drawings available

Bosch VW D24 engines Physically measured

Table 5: Considered Injectors

The preferred injectors were the injectors from the Volvo D5, BMW M47N, VW D24 engines and the Duap prototype injector, because geometrical data was available and proven to fit the 1H30 engine. Finally the injector from the BMW M47N was chosen because it was the only injector to which blank nozzles and appropriate needles could be acquired. It fitted the previous cylinder head with only minor modifications and parts of the software to control this injector could be inherited. The injector from the Volvo D4 engine had another nozzle and needle which

(22)

could not be achieved as blanks. The injector from the VW D24 was a little too short to fit the cylinder head and make a good mounting construction and the Duap injector was a little too long and would not be as rigid as the BMW injector.

(23)

4 Redesigning of cylinder head

4.1 Requirements on the injector mounting

To be able to properly integrate the injector in the cylinder head the following design require-ments has to be fulfilled.

• The injector has to be mounted with a 7◦ _{angle in respect to the cylinder axis}

• The injector should be easily replaceable if for example experiments using other nozzles are to be made

• To assure identical spray at all times while mounted the injector should be rigid in all degrees of freedom (see figure 7)

• To assure a leak proof construction at all times a sealing should be integrated which resists the pressures, temperatures and chemicals it exposed to

• The fastening design should not interfere with other functions such as valves or lubrica-tion/cooling channels

• The fastening design should be as easy as possible to assemble and disassemble • The entire design should maintain lowest possible production costs

• Due to vibrations the design should not be sensitive for vibrations

• The design should prevent the injector from experiencing flexural stress while mounted • An additional point was to attach a sensor plate to the crankshaft

(24)

Figure 7: The spray change if the injector is rotated 180◦

4.2 Method

From the requirements above some desired functions could be identified:

• A sealing surface on the cylinder head where the sealing is placed. This surface must have a low surface roughness to assure sealing.

• A force that applies on the sealing to keep it pressed against a sealing surface • A control surface on the sealing to keep it from sliding into unwanted positions

• Two control surfaces that forbids the injector from moving in radial directions, one as near the nozzle as possible, and one as near injector head as possible

• Something that forbids the injector to rotate around its own axis.

The tool used to create the design is Pro/ENGINEER Wildfire 2.0 from PTC which is a para-metric, integrated 3D CAD/CAM/CAE software, and is commonly used within the internal

(25)

combustion engine industry for design and manufacturing [3].

The design was constructed using a bottom up approach or also called piece part engineering. Here each individual base element are first specified in great detail and are later linked to form subsystems in several levels until a top level system or Digital Mock-Up is achieved. Another possible approach would be to use a top down design. Thus starting from the top and defining a skeleton that holds general information on the overall design. Each subsystem and later on each part inherits parameters from the general structure in a hierarchy. This approach is more suited for designing new systems because it clearly illustrates how each module interacts with the entire system in forehand. However in this case many other functions were already defined in form of complete parts such as the injector itself or the cylinder head which is bought casted and after treated from Hatz. Hence the interaction has to be defined based on those parts. The weakness of the bottom up design approach is that much intuition is needed define the functionality of each module and more work effort is necessary in the DMU. This would make the bottom design approach less suited for more complex systems.

Due to the requirement to keep low production costs the amount of fine tolerances and variations in production methods were kept down at a minimum. All defined tolerances are functional dimensions to either assure the feature function or the interaction of the part with the system. Functions (for example sealing, aligning, fastening etc.) are also assigned separate features (surfaces, flanges, holes etc.) and if possible separate parts. For instance shouldn’t any feature both have to seal and control or any screw have to experience both share and tension stress. If one single screw cannot be placed in a major axis to eliminate share stress the force are divided into two features whereupon two screws are applied. All surfaces where two parts are mated there is an allocated control function otherwise a gap is always present to avoid unnecessarily accurate tolerances.

Tolerance chains (see figure 8 are kept as small as possible to minimize the largest possible error when summing up the tolerances. All measurements on each part are made from one origin unless there are special requirements. When mating two parts with alignment pins or screws the placement measurement origins from the same point on the mating surface (like in a top down approach).

(26)

Figure 8: Two ways of specifying dimensions with the same tolerance level ±0.1mm but due to the improper specification of the dimensions the total length of the right piece can vary by ±0.3mm

The regarded DMU in this project contains all the parts of the cylinder head subassembly. Because other subassemblies (cylinder block and down) were not affected by the changes the cylinder head subassembly was regarded as the top level system within this project and forms therefore the desired DMU to verify the design.

4.2.1 Sealing

To assure an isolated combustion chamber the injector has to be mounted in such a manner that it seals it off. When a gasket is pressed against surface with a fine surface roughness (sealing surface), the surface of the gasket deforms to to fit the small asperities of the sealing surface and therefore prohibits liquids to pass through the gap in between the gasket and the sealing surface. To chose the sealing the following was considered:

General properties Material and shape Constructive properties Type of sealing,

sealing parts, movability parts to one another, installation space

Chemical and Physical properties Pressure, temperature, medium

Installation and Maintenance Accessibility

Table 6: Criteria when selecting sealing

There are mainly two types of sealants for static applications: metal sealants and synthetic sealants. When selecting the material for the sealing multiple criteria has to be considered, such as temperature resistance, the hardness or resistance of deformation due to pressure and chemical

(27)

resistance. The synthetic sealants are usually more deformable and have a lower temperature and pressure resistancy. Therefore the selected sealant was a Cu-Washer according to DIN 7603 with the dimensions 8x13.5x1 to fit the injector seat (see figure 9) selected.

Figure 9: The injector seat and the sealing To assure the sealing capability the following calculations were made: The deformation load Fdef be defined as:

Fdef = π · dD· k0· KD (1)

where

dD Mean diameter of sealing

k0 Effective width of sealing

KD Deformation resistance

When the deformation load Fdef is sufficient to form the sealing through adaptions of the

roughness of the surfaces of the gasket, it assures a leak proof construction at normal atmospheric pressure. When the pressure in the cylinder increases this force has to be sufficient to maintain leak proof at higher cylinder pressures as well. The deformation load Fdef has to be greater

than the operating load on the sealing Fop

Fop= π · pcyl· dD· k1· SD (2)

(28)

pcyl Inner pressure in the cylinder

k1 Fictive operational width of the seal

SD Safety factor (usually set to 1.3 for metal seals)

To assure the deformation of the gasket by mounting the total pressure on the sealing Fs,tot has

to be

Fs,tot= i · Fs,0≥ Fdef (3)

and maintain leak proof by operation

Fs,tot= i · Fs,max≥ B1· Funl (4)

where

i Amount of screws that fastens the construction Fs,0 Pretensioning force on each individual screw

Fs The force on each individual screw after deforming the gasket and assuring sealing

B1 Factor regarding decrease in screw strain due to subsidence

Funl Unloading force from combustion chamber pressure

For gaskets the unloading force Funl is defined as:

FB = pcyl· π ·

d2 D

4 (5)

[4, p. 629]. A Cu 8x13x1 washer according to DIN 7603 A was chosen. With inserted numerical values the total force on the gasket has to be 10.7 kN to assure deformation by mounting and around 9 kN to assure safe sealing when the engine is in operation. The calculations are made with a static cylinder pressure of 100 bar which would usually only occur by a very early injection or knocking (the rest of the engine is designed to operate at 65 bar cylinder pressure, which corresponds to the maximal cylinder pressure by a normal combustion). The construction is fastened with 6 DIN ISO 1891 M6 screws with a strength category 8.8 or higher according to EN ISO 898-1. The yield strength of such a screw is minimum 800 MPa2[6, 2:23] which means that the construction can carry a static force of

Fs,tot= i · Rm· π(d/2)2≈ 140 kN (6)

before the screws plasticize. The unloading gasket force from the maximum cylinder pressure on the injector (the force from the cylinder pressure at 100 bar) is

(29)

Funl= π ·

dnozzle

2 2

· pcyl≈ 380 N. (7)

The maximum load on each screw is:

Fs,max=

Fs,tot+ Funl

i ≈ 1.9 kN (8)

If the screws are dynamically loaded, which they are in this case, the design has to carry the dynamic loads as well. Screws normally have a lower dynamic yield strength. According to [7, p. 1073] is a DIN ISO 1891 M6 screw with a strength category 8.8 suited to carry a dynamic load of up to 2.5 kN if it is tightened with a torque wrench or high precision screwdriver that uses dynamic torque or screw elongation measuring and if the torque is controlled after tightening. Hence are suited for the design.

The screws has to be tightened with a recommended torque of 9.2 Nm which gives a pre ten-sioning force around 9.2 kN per screw according to [6, 2:39]. The total force with this design on the injector and the gasket will be 55 kN and is enough to seal off the combustion chamber.

(30)

4.3 Fastening of injector and assuring sealing

Figure 10: Exploded view of the fastening construction assembly - 1. DIN ISO 7984 M6x20 Screw, 2. Bosch CRI2 Injector, 3. Injector attachement block male, 4. Injector attachment block female, 5. DIN ISO 4014 M8x40 Screw, 6. DIN ISO 4014 M6x40 Screw, 7. DIN 6325 aligning pin, 8. DIN 7603 A 8x13x1 Cu sealing, 9. Cylinder head cover, 10. Cylinder head H30, 11. Inlet valve seat, 12. Outlet valve seat, 13. Inlet valve, 14. Outlet valve

Regard the exploded view of the assembly in figure 10 of the fastening construction. Complete design drawings can be found in Appendix B. The part 9 is the cylinder head cover. It is placed directly on the cylinder head 10. The injector 2 is placed in in a bore through both the cylinder head cover and the cylinder head. The bore is illustrated in the work drawing of 001-026-02. The blocks 3 and 4 are both injector fastening blocks and keep the injector mounted with screws attached to the cylinder head cover. The screws should be cross tightened. A rendered image of the assembled design in shown in figure 11.

(31)

Figure 11: Rendered image of the assembly

The cylinder head cover 9 (see drawing 022-02) is after the previous design (see drawing 001-022-01) heavily modified. Previous outer geometry was adapted for an exhibition and therefore had greater requirements on outer appearance. The previous cylinder head cover also included features that were used for previous concepts (alternative valve rocker arms, pressure compen-sating rooms for oil etc.) that were removed. Without the necessity to include these features the geometry could be reduced to the final design that are seen in drawing 001-022-02. Due to the far inferior complexity of the outer geometry and the exclusion of unnecessary features a great reduction of production costs could be made.

The cylinder head cover 9 is mounted to the cylinder head with two DIN ISO 4014 M8x40 screws 5 and one DIN ISO 4014 M6x40 screw 6 that run through three bores near the edges. The bore close to the middle of the cylinder head cover is made for sensor signal cables. The cylinder head cover are aligned with the cylinder head with using two DIN 6325 _{4x10 h6 alignment} pins 7. The alignment bores in the cylinder head cover has a tolerance N7 which creates a ISO 282-2 h6/N7 press fit. The pin can with this fit be pressed in or using a light force without any special instruments. This is a shaft based fit which means that the two aligned pieces can have different fits. To ease the assembly of the construction the fit with the other piece is running fit (this will be discussed shortly). The positioning measurement of the alignment pin bores has the finest possible tolerance according to DIN ISO 2768-1. On the top side there is an angled plate with 7◦ inclination that contains 4 similar bores as the bores for the alignment against the cylinder head. This is shown in section A-A of drawing 001-022-02. These four bores align the injector attachment blocks with the angled plate. The plate also has six bores with ISO DIN 13

(32)

T1 M6 windings. The cylinder head cover is made out of aluminum in which the threads alone are unsuitable for repeated assembling and disassembling. This occurs oft with lab equipment thus presenting a problem. Therefore these windings are made with the help of HELICOIL R

screw thread inserts, that are more wear resistant. These inserts are threaded steel bushings. They are used to thread a bore where either the thread is damaged or in this case the material is not suited for threading. All measurements on the angled plate is made from a centre point which where the injector bore is later made to minimize the largest possible deviation of the bores if measuring from another point. The bottom surface has to seal the oil channels and camshaft arm chambers from the outside and from the injector bore. Thus has a requirement of a surface roughness Rz of 3.2.3

The injector bore is made when the cylinder head and the cylinder head cover are assembled to minimize the offset of the injector bore axis between the two parts (see figure 12). The bore in the work drawing 001-026-02 contains only two controlling surfaces that are in direct radial contact with the injector (the bore surface in 3C with_{23.5 H7 and the surface in 7D with 7} H7). These surfaces has the function to prevent movement of the injector in radial direction. This could lead to bending stress within the injector for which it is very sensitive. Detail A illustrates the nozzle bore and the gasket placement surface. The surface with a roughness Ra

of 1.6 illustrated in C7 is a sealant surface. The radial surface with_{14 shown in C7 controls} the position of the gasket. A DIN 7603 washer has an outer diameter 13. The hole diameter is chosen as slightly bigger to be able to allow radial expansions of the sealing when force is applied and further expansion from when a thermal load is present.

3_{This unit is inherited from the old cylinder head cover, and is therefore defined as R}

z and not Ra. Rz= 3.2

(33)

Figure 12: What can happen when two bores that were made separately before assembly are aligned

The injector attachment blocks 3 and 4 (see drawings 001-060-03 and 001-061-03 in Appendix B) have the function to prevent the injector from rotating and to apply the force necessary to assure a safe sealing of the washer described above. They are made out of heat treated steel because of high its durability, resistance for mechanical fractures, toughness and high accessibility. The blocks are held together by two DIN ISO 7984 M6 screws 6. These screws are Allen screws R

with low head. The screws go through the horizontal bore presented in figure 13 which can also be seen section A-A in the drawing (see Appendix B drawings 001-060-03 and 001-061-03). The block 3 is the male block and contains two unthreaded clearance bores with countersinks for the screw heads. It also contains cylindrical flanges around the bores with a_{8 g6. They align the} male block to the female block 4 which contains countersinks for the flanges with _{8 H7 and} threaded bores for the screws 6. The flange is shorter than the countersink which assures the blocks to be pressed together from the screw torque instead of deforming the flange if it was as touching the bottom of the countersink (see figure 13).

Figure 13: Alignment of the two injector attachment blocks

The flange/countersink has a H7/g6 running fit with very small clearance. This fit is suited for accurate guidance of shafts. Section A-A also shows the bores for the alignment pins which

(34)

aligns the blocks against the cylinder head cover. It has a G7/h6 fit which also is a running fit with small clearance. The alignment pins are already pressed into the cylinder head cover. Therefore to facilitate assembling no additional force has to be applied to fit the alignment pins into the blocks. The alignment pin bore pattern is unsymmetrical to only present one possibility to mount the injector (see figure 14).

Figure 14: Alignment pin bore pattern on the angled plate of the cylinder head cover The flange with the width 3.4 0

−0.1 slides into a notch in the injector. The blocks are attached

to the cylinder head cover by a total of six DIN ISO 7984 Allen M6x20 screws (three screwsR

per block) 1. If no torque is applied on the screws there is a clearance between the angled plate on the cylinder head cover and the injector attachment blocks (see figure 15).

(35)

Figure 15: The clearance between the injector attachment blocks and the angled plate of the cylinder head cover

If torque is applied to the screws the blocks deform and the gap decreases. The clamping force on the sealing gasket can therefore be controlled through the torque applied on the screws. The radius R 0.8 in detail A in the drawing (see Appendix B drawings 001-060-03 and 001-061-03) is made to increase the resistance against mechanical fractures. If the radius gets bigger the the flange cannot slide into the injector notch properly (see figure 16).

(36)

Figure 16: Injector and the female attachment block demonstrating the flange in the notch pressing down the injector on the gasket (The male injector attachment block is here invisible) The same applies to the radius R 0.8 in section A-A in the drawing (see Appendix B drawings 001-060-03 and 001-061-03). If this radius gets bigger the screw heads cannot lie flat on the surface and cause shear stress near the screw head. However, these two radiuses could be made 0.2 mm bigger if a negative tolerance was assigned to these measurements. If a finer tolerance than standard is set production costs rise. In this case it was however not necessary to optimize in respect to durability, so a solution with lower price was chosen.

(37)

5 Engine control

The aim in this case is to feed the injector with current at a desired timing. To understand how to do this knowledge on control unit architecture and electrical properties of the injector are needed.

The engine control unit controls and manages all activity of the engine. Along the main functions such as injection timing and duration, throttle valve, cam angle and valve lift adjustment the ECU also manages auxiliary functions. These functions could be adjustment of exhaust after treatment, variable turbine geometry or acquisition of sensor data which is vital for an accurate control. Additional to the ECU other control units are usually integrated such as the controller of the automatic transmission, undercarriage etc. All these control units must at all times be able to communicate with one another to guarantee safety and operation of the machinery. The communication of the single control units and the registration of sensor signals are carried out through CAN-buses. This system is developed by Bosch specially for automotive applications after recognizing the continuously rising amount of cross linkage. The advantage of this system is that micro controllers can communicate with one another without a host computer so if one micro controller brakes down the system can assure functionality of the unaffected parts [10]. The used ECU is an AFT PROtroniC which is an open platform micro controller and commonly used for research purposes because it allows the user to access and modify far more features than an ECU from a production model automobile. For instance can the engine speed sensor system be modified by a user after manufacturing without restrictions.

ECU’s are usually built up out of modules. The modules are categorized according to their function. Each module is represented on a separate PCB. The PCB’s that the AFT PROtroniC ECU contains are as follows:

Figure 17: PCB Modules of the AFT PROtroniC ECU

The carrying module contains switching outputs: 4 banks with 6 channels, digital I/O: 3 banks with 6 channels each, communication, power supply and diagnostics. The switching outputs can control for example throttle valve and other actuators.

The heart of the ECU, the CPU module is placed on top of the carrying module directly. The CPU module contains a µC Motorola PowerPC MPC565 processor and is responsible for all calculations made. PowerPC stands here for Performance optimization with enhanced RISC (Reduced instruction set computer), and PC stands for Performance Chip.

(38)

The carrying module is a module of Field Programmable Gate Array type or short FPGA. FPGA modules are freely programmable modules where an internal structure can be created by logical commands (AND, OR, NOT etc.).

Injectors, especially in diesel applications have very high electrical requirements. They need to be controlled with a high current (sometimes up to 30A) and that at a very precise timing. Therefore the injector control function is embedded in its own module. The injection module contains high voltage outputs for maximum 6 vents which feeds electrical current to the injectors and calculates the timing. The

CPU module "Net+50" is an add-on interface. In a series produced ECU would this module not be programmable. The bypass module contains an interface to connect to other bypass modules on other control units. The CAN-buses are connected to this module.

The sensor module 1 contains λ- and knocking sensors as well as communication drivers. The sensor module 2 however contains analogue and digital inputs and outputs where for exam-ple signals from the rotation speed measurement sensor and the camshaft sensors are recognized. These are in series produced ECU’s not alterable. Thus all sensors would have to be adapted for the specific ECU.

5.1 Architecture of the ECU software

It has now clear how the hardware is built up. To be able to program this ECU it is also necessary to understand how the modules function and how they process signals.

The function software model contains information on all processes taking place within the ECU. It controls how the signals are received, processed and transmitted to actuators such as the throttle or injectors. This model is built up in simulink. To make changes, for example change the engine speed sensor signal type it has to be done in the PCU configurator software first. The PCU configurator software is a user friendlier interface that allows the user to change certain settings. When this is made the changes has to be implemented in the simulink model before the model can be transferred to the ECU. When the ECU is running operating settings like injection timing or amount can be changed from the MARC software without flashing the ECU.

(39)

5.2 Rotation speed measurement and TDC recognition

To time the injection the ECU has to exactly know the position of the crankshaft and the camshaft4_{. This is made through the following sensor system.}

Figure 18: The rotation speed sensor

The reluctor ring is magnetic and attached to the crankshaft. It contains 360 small gaps. When it rotates it changes the magnetic flux through a solenoid and therefore induces a current. It gives one 5 V square pulse per crank angle degree as figure 19 describes:

4_{The valve timing on the 1H30 engine is controlled by a single camshaft and not two like on most automobile}

(40)

(41)

To recognize if the engine currently is in a high pressure phase or in a gas exchange phase a sensor is placed on the camshaft. It gives the following signal:

(42)

The combustion phase is recognized by a single pulse from an incremental sensor as follows:

Figure 21: The 0 pulse signal from the incremental encoder on the camshaft

It has a -90 degree offset from the combustion top dead centre. The offset is necessary to recognize the exact location of the TDC to time the injection. If the pulse appears at 360 crank angle degrees sharp the next injection would not occur until around 700 degrees later which can give inaccurate control. This offset has to be set in the MARC software during operation.

5.3 Injector control

The input signals mentioned above are vital for the recognition of the crank angle and the top dead centre. Here the output signals that control the injector current are discussed.

(43)

Figure 22: Solenoid injector in open and closed state

type. To open the valve of a solenoid injector electric current has to be fed through the solenoid as explained in 2.1.3. In piezo injectors the solenoid is replaced by a series of Piezo elements. These elements are crystals that generate electric potential in response to mechanical stress, this is called a direct piezoelectric effect. A reverse piezoelectric effect is that mechanical stress is caused when an electric field is applied which results in a shape change of the crystals. The change by these crystals is smaller that 1% of their original size and the crystals are therefore often connected in a series to one another to be able to exhibit sufficient change. This type of valve needs a higher current to function and requires a complex electric feed curve due to the complex behavior of the piezo crystals. A solenoid valve is in this case simpler. The electric feed that the Bosch CRI2 injector requires over time is shown in 23 below.

(44)

Figure 23: Electric feed diagram for the Bosch CRI2 injector

The first phase 0 opens the injector solenoid valve. Due to the inertia of the system a boost voltage has to be applied to open the valve. During phase 1 the boost current is still applied but only requires an operating voltage to maintain the high current. After the solenoid has opened phase 2 begins and the high current is no longer required so the voltage is therefore shut off to decrease the current. After the current has reached an operating level it is held constant as long as required during phase 3. The load is controlled through change in this phases duration. Phase 4 shuts off the injection with a negative boost voltage. The two later phases are pre configured can not be altered in the PCU configurator or MARC.

(45)

6 Multiple injections

A common rail injector is capable of making multiple fuel injections every combustion cycle. Pre-and post injections are used to reduce emissions, noise Pre-and create a smoother combustion. With the former design with an otto-adapted injector and a single injection strategy the engine had problems passing the EPA4 cycle5. With a multiple injection strategy combustion properties could be optimized to pass the EPA4 cycle. The combustion characteristics are usually non linear and very complex, thus makes it difficult to describe using simulations so testing is required. To cover the entire operating map a great amount of testing is required. If the rail pressure is held at a constant level over the entire operation map, has a single injection strategy 3 variables

injection duration ETmain

injection timing SOImain

and engine speed n.

Table 7: Variables with a single injection strategy

If two injections are made - one main and one pre injection, then the amount of variables will be 5

-injection duration ETmain

pre injection timing ∆ϕpre

pre injection amount Qpre

and engine speed n

Table 8: Variables with a main- and pre injection strategy

and with 3 injections main, pre and post injection the amount of variables will be 7 -injection duration ETmain

pre injection timing ∆ϕpre

pre injection amount Qpre

post injection timing ∆ϕpost

post injection amount Qpost

and engine speed n.

Table 9: Variables with a multiple injection strategy

To understand the problematic with testing of multiple variables a quick calculation is made. It shows that if a series of testing is about to be made where 5 variables are to be varied in two steps each a total amount of 25 _{= 16 measurements and with 7 variables it would be 2}7_{= 64}

so with each variable the amount of tests increase very rapidly. If each point is measured one time only no information about the deviation from the real value is known which can lead to false results. Another issue to regard is that a test where each variable is tested in two steps is

(46)

only good to describe linear effects. If information is required about a non-linear phenomena 3 or more points have to be included for each variable. Such testing increases the amount of test points. The table below illustrates this problematic when testing multiple parameters npar and

multiple steps nsteps.

npar/nsteps 2 3 4 5

2 4 9 16 25

3 8 27 64 125

4 16 81 256 625

5 32 243 1024 3125

Table 10: Amount of measurements needed with various parameters and steps

The table 10 above only goes to 5 variables and 5 steps and still results in an amount of measurements which cannot be carried out and still may not provide sufficient information about the system. When such testing is carried out costs, time and wear of the engine has to be concidered. If too many tests with knocking or a generally improper combustion is made the prototype engine can suffer sever damages and may result in a permanently damaged prototype. With this in mind a strategy for the investigation has to be made to avoid an unnecessary amount of experiments.

6.1 Defenition of DoE terms

In a complex experiment process multiple items has to be defined before the testing itself can be carried out. First the aim with the experiment and the target variables must be known, then the influencing variables and later the design for a successful regression analysis to describe the studied phenomena has to be defined.

6.1.1 Target variables

As mentioned above it is first of all important to identify the target variables. These are the variables that either can be measured, calculated or evaluated from the measured data. From each test can multiple target variables be measured. The target variables in this case are:

• Nitrogen oxides N OX • Carbonmonoxide CO • Particulate Matter P M • Unburned hydrocarbons HC • Fuel consumption be • Combustion-air-ratio λ

(47)

• Noise, represented by the pressure gradient δp/δϕ

The emissions and pressure were measured with respective sensors. The particle emissions P M were measured through a AVL smoke meter. The λ-value was measured with a Horiba MEXA 7070 emission analyser. The measurements were saved through INCA in MATLAB as a .matR

file.

6.1.2 Influencing variables

The influencing variables are the variables that (possibly) influences the experiment outcome and influences the target variables. The influencing variables can be divided into two groups: the control factors and the confounders. The control factors are the ones that can be alternated freely (within certain boundaries). The confounders are either variables which can not be affected or factors that are not considered as influencing variables. If a factor that has great impact on the outcome of the experiment is not considered and varies much during the experiment false conclusions can be made. Therefore it is important to list all variables that can possibly have an effect on the outcome. If all the variables are to be investigated the total amount of measurements rapidly becomes very high or maybe all variables can not be altered due to technical or economical reasons. For instance can certain factors be held as constant as possible if information about their impact is not wanted or if it they are suspected to have a insignificant impact on the outcome. Certain variables can therefore be excluded from the experiment and held constant. A graphical interpretation of this division of influencing variables are shown by figure 24 below.

Figure 24: Graphical interpretation of the breakdown of the influencing variables The influencing variables and their breakdown in control factors, constants and confounders for this investigation are presented by the table 11

(48)

Engine speed n control factor

Load Me control factor

Amount of pre injected fuel Qpre control factor

Timing of the injection SOImain control factor

Timing of the pre injection ∆ϕpre control factor

Ambient temperature Tamb confounder

Ambient (relative) humidity RH confounder Ambient air pressure pamb confounder

Engine wear confounder

Oil quality confounder

Table 11: Influencing variables

Note that the main injection is determined through the injection duration and the pre injection through the amount. They are chosen in this matter because the input in the INCA control unit is defined respectively. The user defines the pre injected amount Qpre in INCA. The pre

injection duration is determined from the pre injected amount through a characteristic map which is integrated in the control unit.

If experiments are carried out during a longer time especially at inappropriate operating condi-tions the engine wears and is unable to keep up the same performance as from the beginning of the testing. An example can be that blow by gases dilute the oil which makes the lubrication and cooling function worse which leads to higher temperatures. Another example could be that oil enters the combustion chamber which has a negative impact mainly on the HC emissions.

6.1.3 Factor steps

After it is clear which parameters are used and which confounders are to be regarded the steps with which the factors are varied in the experiment has to be defined. If the factors can assume more than two values (for example on or off, high or low etc.) and information is wanted on how the system behaves between the measured points an interpolation can be made. If the target variable y does not depend linearly on the parameter x the step size ∆x in which x is varied has to be considered so the measured points give a sufficiently good explanation of how the system depends on the parameter x. Regard the figure 25 below.

If a very large step size is used the phenomena illustrated by the top figure occurs - the maximum point is found but slightly displaced. Information on how the system behaves between the two measured points may not be correct from an interpolation. If the step size is too small a very large amount of experiments would have to be made. Another important matter to consider by small step sizes is the accuracy of the measurement. Each measured value y diverges from the real value y due to inaccuracies in the measuring- and control instruments as well as experiment procedures, adjustment errors, noise and confounder factors. If the standard deviation of the measuring σ is comparable to the step size it can be difficult to evaluate whether two closely to one another measured values differ due to a dependency on the varied factor x or due to inaccuracies.

(49)

Figure 25: Step sizes and their influence on experiments

assumed and the amount of points that were measured were little for example three it could be very difficult in pre hand to find an interesting area to investigate. An extrapolation outside the points would give unsatisfactory results as no information is known about extreme points outside the measured area. These effects would therefore not be presented in the model. In this case no variables can assume only discrete values. Nevertheless are testing at the discrete points that the EPA4 cycle dictates a good idea to provide accurate readings on the system behavior in these points.

6.2 Procedure

The testing was divided into two parts, both with the previous injector and cylinder head design. As mentioned before could no tests be carried out with the CRI2 injector attached to the rail or assembled in the engine due to delays in the manufacturing. The first part served to map the combustion properties. Second part contained additional testing at the points of the EPA4 cycle to investigate the possibilities and an injection strategy to pass the cycle. The testing are restricted to one main- and one pre injection only. One reason is the increasing complexity of variation of parameters of multiple injections and another is that the INCA software that controls the previous injector does not allow more injections than two. It will be seen below that the complexity of two injections are hard enough to evaluate as it is.

(50)

6.3 Experiment design types

There are several experiment design types. To investigate all the factors and their impact on the target variables they all have to be varied in several steps independently from one another. It can be done with several strategies, which are shown in figure 26.

Figure 26: Experiment designs

With a fractional factorial experiment design the amount of experiments are held at a very low level but would intrude on the complexity of the regression model. In the example in the figure 26 above three variables x1, x2 and x3 are varied. If a linear regression model has to be made

to investigate the behavior of the system can be described as

y = β0+ β1x1+ β2x2+ β3x3+ β4x1x2+ β5x1x3+ β6x2x3+ β7x1x2x3 (9)

(51)

be decided. Some effects has to be excluded from the model. A full factorial 2 step design has in this case the advantage that 8 points are measured and all linear coefficients can be decided. To describe non linear effects more complicated designs has to be used. Further examples will discuss non-linearities and factor interaction, and to visualize this they will be discussed with 2 independent variables x1 and x2only. It can also be seen as if the cubes in figure 26 are turned

in to 2 dimensional squares because x3 is held constant.

(52)

The figure above shows a linear regression with 2 variables. The deviation from parallelism indicates a presence of factor interaction effects.

Although it can provide sufficient information about a linear behavior and factor interaction it may still not be enough. Even if dependencies within a small area can be described with a linear model it may not be sufficient for larger steps sizes and especially if extreme points within the tested area are searched for. A center point helps to investigate non linearities (see figure 28).

Figure 28: How the center point can be used to detect non-linear effects with a 2 step full factorial design

If the center point deviates from the center of the linear model there are non-linear effects. If a second degree polynomial

(53)

is assumed some effects has to be neglected. The model contains 7 constants and 5 test points. Two different non-linear models can therefore be assumed. Model 1 is assumes linear dependency of x1and non-linearity in x2and model 2 assumes linear dependency of x2and non-linearity in

x1 (see figure 29).

Figure 29: Difference between assumed non linear models

The figure 29 makes it clear that the assumed non-linear model has an effect on the outcome. If a false model is assumed there is a small difference between the actual system behavior and the model.

To be able to decide all coefficients of a quadratic model at least 7 points has to be tested. A 3 step full factorial design provides this. It varies all factors in 3 steps uncorrelated to one another. The design could be further improved with a response surface (see figure 26). This design describes a square (2 variables) or a cube (3 variables or more) and a star in the middle and therefore covers a greater area than the 3 step full factorial design and gives more information about the factor interaction.

(54)

6.4 Part 1 - Pre-study and mapping

This part concludes a pre study with the former injector to map the combustion characteristics with a multiple injection strategy. Here it is investigated which parameters has the biggest impact on the combustion properties, which optimization potential could be expected from multiple injections.

The factors were varied as a response surface design suggests.

Factor xi Unit Min value xi,−α Max value xi,+α Name in regression

Qpre [mm3] 2.5 6 x1

∆ϕpre [CA◦] 4 9 x2

SOImain [CA◦ -TDC] 12 27 x3

Table 12: Variable values

The min and max points are the points of the star. To calculate the corner points of the cube x1,±, x2,±and x3,±the following algorithm was used.

xi= xi,+α− xi,−α (11) x1,+= xi+ xi,+α− xi 1.7 (12) x1,−= xi− xi− xi,+α 1.7 (13)

For each factor the center was tested 5 times to get information about the spread. All of this was done at 5 operating points as a 2 step full factorial design with center suggests. The engine speed factor is called x4and load x5 in the regression analysis.

Point nr Engine speed n Load Me Relative load

1 3000 rpm 14.6 Nm 100%

2 3000 rpm 7.3 Nm 50%

3 2300 rpm 14.6 Nm 100%

4 2300 rpm 7.3 Nm 50%

5 2600 rpm 11 Nm 75%

Table 13: Operating points

This resulted in a total amount of Npts = 100 test points which were randomized to eliminate

trends. If each factor is varied in order and some confounder gradually changes over the time it can falsify the results. The effect of the gradually increasing confounder cannot be separated from the actual effect of the factor. If for example the test are starting with a low engine speed and then gradually increase step wise, the final tests are performed with a considerably higher