Evaluation of an Electric Fuel Feed Pump as a Virtual Sensor

IN THE FIELD OF TECHNOLOGY DEGREE PROJECT

DESIGN AND PRODUCT REALISATION AND THE MAIN FIELD OF STUDY MECHANICAL ENGINEERING, SECOND CYCLE, 30 CREDITS STOCKHOLM SWEDEN 2016 ,

Evaluation of an Electric

Fuel Feed Pump as a Virtual Sensor

Utvärdering av elektrisk bränslematarpump som virtuell sensor

AMANDA EL RAYES

KTH ROYAL INSTITUTE OF TECHNOLOGY

Evaluation of an Electric Fuel Feed Pump as a Virtual Sensor

Utv¨ ardering av elektrisk br¨ anslematarpump som virtuell sensor

Amanda El Rayes

Master of Science Thesis MMK 2016:169 MDA 567 KTH Industrial Engineering and Management

Machine Design

SE-100 44 STOCKHOLM

Examensarbete MMK 2016:169 MDA 567 Utv¨ ardering av elektrisk br¨ anslematar- pump som virtuell sensor

Amanda El Rayes

Godk¨ant Examinator Handledare

Hans Johansson Mikael Hellgren

Uppdragsgivare Kontaktperson

Scania CV AB Tommy Albing

Sammanfattning

Denna rapport unders¨ oker den potentiella anv¨ andningen av en elektrisk konstantvolym- spump som en virtuell sensor i ett l˚ agtrycks-br¨ anslepumpsystem som ett s¨ att att elim- inera fl¨ odes- och tryckgivare i systemet.Det ¨ ar m¨ ojligt att f˚ a en korrekt uppskattning av tryck och fl¨ ode givet hastigheten och str¨ omf¨ orbrukningen av en ideal pump, men i verkligheten finns det m˚ anga fler parametrar kan variera, s˚ asom l¨ ackage, br¨ ansletem- peratur och verkningsgrad.

En modell har skapats i Simulink f¨ or att simulera tryck och fl¨ ode i hela systemet och en rigg byggdes f¨ or testning och validering av resultaten fr˚ an modellen.

Efter valideringen manipulerades parametrar i modellen f¨ or att fastst¨ alla noggrannheten av simuleringarna i de fall d¨ ar v¨ ardena avviker fr˚ an m¨ atdatat, till exempel vid ¨ okan- de/minskande br¨ ansletemperatur och mellan ett nytt och ett igensatt br¨ anslefilter. Ab- soluta felet ber¨ aknades sedan f¨ or att visa omfattningen av dessa avvikelser.

Den erh˚ allna data indikerar att en elektrisk pump kan anv¨ andas som en virtuell sensor, om ¨ an inom vissa gr¨ anser. Modellen har visat sig vara anv¨ andbar f¨ or att unders¨ oka skillnader mellan ett flertal elektriska pumpar, effekterna av ˚ aldrande samt andra sys- temparametrar.

F¨ orslag p˚ a n¨ ar och hur diagnostik ska k¨ oras f¨ or att uppt¨ acka igens¨ attning av br¨ ansle- filter i systemet ges. Avsikten ¨ ar att minska sl¨ oseriet vid byte av filter som idag sker efter specifika tidsperioder ist¨ allet f¨ or n¨ ar de uppt¨ acks att vara igensatta.

Nyckelord: virtuell sensor, elektrisk pump, br¨ anslefilter , filterigens¨ attning , diagnos-

tik

Master of Science Thesis MMK 2016:169 MDA 567

Evaluation of an Electric Fuel Feed Pump as a Virtual Sensor

Amanda El Rayes

Approved Examiner Supervisor

Hans Johansson Mikael Hellgren

Commissioner Contact person

Scania CV AB Tommy Albing

Abstract

This thesis analyses the potential use of an electrical fixed displacement pump as a vir- tual sensor in a low pressure fuel pump system as a way to eliminate flow and pressure sensors in the system. It is possible to obtain an accurate estimate of the pressure and flow given the speed and the current consumption of an ideal pump but in reality there are many more parameters that must be taken into consideration, such as leakage, fuel temperature and efficiency.

A model was created in Simulink in order to simulate pressure and flow throughout the system and a rig was built to test and validate the results from the model.

Once validated, parameters in the model were manipulated to determine the accuracy of the pressure and flow estimation in cases where the values deviate from the measured data. For example increasing/decreasing fuel temperatures and differences between a new filter and a clogged filter. The absolute error was then calculated to show the extent of those deviances.

The resulting data indicates that an electric pump can be used as a virtual sensor albeit within certain limitations. The model is proven to be useful in examining discrepan- cies between multiple electric pumps and the effects of ageing as well as other system parameters.

Suggestions of when and how diagnostics should be run in order to detect clogging of the fuel filters in the system are given. The intent is to reduce waste when replacing a filter which currently takes place after a specific time period instead of when it is discovered to be clogged.

Keywords: virtual sensor, electric pump, fuel filter, filter clogging, diagnostics

Acknowledgements

I would like to thank Mikael Hellgren for great input throughout the thesis, Hans Johansson for taking time to read this thesis,

and Ruoyu Zhang for his comments and questions during the opposition.

I would also like to mention,

Patrick Fogelberg, Zohreh Pakdaman, Carina Forsberg and Ola Stenl˚ a˚ as for taking the time to answer my many questions.

I especially want to thank Tommy Albing for all his help, patience and dedication.

On a more personal note I would like to thank my parents Vera and George for their continuous support and love.

Amanda El Rayes

S¨ odert¨ alje, October 2016

Contents

Contents

1 Introduction 1

1.1 Background . . . . 1

1.2 Purpose . . . . 1

1.3 Problem Description . . . . 3

1.3.1 Research Questions . . . . 6

1.3.2 Definitions . . . . 6

1.3.3 Delimitations . . . . 7

1.4 Method Description . . . . 8

1.5 Report Outline . . . . 8

2 Frame of Reference 11 2.1 Low Pressure Fuel Circuit (LPFC) System Components . . . . 11

2.1.1 Gerotor Pump . . . . 11

2.1.2 Fixed-Displacement Pump . . . . 12

2.1.3 Filters . . . . 14

2.1.4 Tech-tank . . . . 16

2.1.5 Fuel . . . . 16

2.1.6 Piping . . . . 16

2.2 Clogging . . . . 17

2.3 Diagnostics . . . . 17

2.3.1 Condition Based Maintenance System . . . . 18

2.4 Simscape . . . . 18

2.5 Pressure . . . . 19

2.6 Current consumption vs. pressure difference over the pump . . . . 20

3 System Description 23 3.1 Simulink Model . . . . 23

3.1.1 Lookup Tables . . . . 23

3.1.2 The filter and pipes . . . . 25

3.1.3 Fluid and Fluid Temperature . . . . 25

3.1.4 The Solver . . . . 26

3.1.5 The Calculation of I . . . . 26

3.2 The Rig . . . . 28

3.2.1 Components . . . . 28

3.2.2 Software . . . . 29

4 Validation 33 4.1 Sensor Resolution . . . . 35

4.2 Current calculation . . . . 35

Contents

5 Analysis and Results 37

5.1 Plot setup . . . . 37

5.2 Pump Parameters . . . . 37

5.2.1 Displacement . . . . 38

5.2.2 Efficiency . . . . 39

5.2.3 Normal variations between pumps . . . . 40

5.3 System Parameters . . . . 42

5.3.1 Temperature . . . . 42

5.3.2 Pipe Diameter . . . . 44

5.3.3 Pipe Lenghts . . . . 45

5.4 Sensor Resolution . . . . 46

5.4.1 Speed Sensor . . . . 46

5.4.2 Current Sensor . . . . 47

5.5 The impact of Clogging on Pressure and Flow . . . . 49

6 Discussion 51 6.1 Limitations . . . . 52

6.2 The Rate of Accuracy . . . . 52

6.3 Worst-Case Scenario . . . . 52

6.4 Sustainability . . . . 52

7 Conclusion and Future Work 55 7.1 Conclusions . . . . 55

7.2 Future Work . . . . 55

7.2.1 Both Subsystems Together . . . . 55

A Pump Data 59 A.1 Setup . . . . 59

A.2 Data and Results . . . . 59

B Fuel Viscosity Experiment 61 B.1 Introduction . . . . 61

B.2 Setup . . . . 61

B.3 Data and Results . . . . 61

C Plots 63

List of Figures

List of Figures

1.1 A map of the current relating to the pressure difference and rpm over the pump. The dip in the corner is explained by lack of relevant data for

those data points and is therefore set to zero. . . . . 2

1.2 The estimation map in Simulink. . . . . 2

1.3 A map of the ideal flow of the reference pump vs. its speed. . . . . 3

1.4 A simplified overview of the fuel pump system currently in use. . . . . 3

1.5 A simple overview of the new low pressure fuel pump system. The figure is a modified version of the low pressure fuel pump system used in a previous thesis at Scania [1]. . . . 6

2.1 Gerotor [6] . . . . 12

2.2 Input/Output overview of fixed-displacement pump in Simulink model 13 2.3 Fuel filter . . . . 14

2.4 The difference between puller and pusher filters. . . . 15

2.5 Pressure overview . . . . 17

2.6 The steps in a CBM program [10] . . . . 18

2.7 The relationship between different types of pressure [12]. . . . 19

2.8 Overview of how the comparison is done between the reference pump and manipulated pump model. I

is unknown and must be calculated. . 20

2.9 How power is translated through the pump. η

is the overall pump efficiency. [13] . . . . 21

3.1 The differences between the transfer and the feed subsystems. The pres- sure sensor marked with a red box is the pressure estimation that is simulated with the help of the electric pumps. . . . 23

3.2 Input/Output overview of lookup table in Simulink model . . . . 24

3.3 A lookup table where the lines represent pump speed at varying pressure drop. . . . 24

3.4 The filter and the pipes in Simulink. . . . . 25

3.5 Overview of main tank including solver and custom fluid. . . . 26

3.6 The equation for estimating I and the lookup table implemented in Simulink. . . . . 27

3.7 Simplified rig overview where P denotes a pressure sensor and q denotes a flow sensor. The ball valve is there to simulate the High Pressure Pump (HPP). . . . 28

3.8 Components that make up the filter in rig. . . . 29

3.9 Engine Control Unit (ECU) setup for Controller Area Network (CAN) communication. . . . 30

4.1 Pressure difference over pump at 3000 rotations per minute (rpm). . . . 33 4.2 Comparison between measured signal from rig and Simulink model where

the dotted (red) line is the model and the straight (blue) line is the

List of Figures

4.3 Ideal estimated pressure (red signal) vs. Ideal pressure (blue signal). . . 35

4.4 Ideal estimated current (red signal) vs. Ideal current (blue signal) . . . 36

5.1 Absolute error for pressure at varying displacement. . . . 38

5.2 Absolute error for flow at varying displacement. . . . 39

5.3 Estimation of pressure for 5 different pumps as rpm varies. The plot shows the absolute error between the reference value and the estimated value in bar. . . . 41

5.4 Estimation of flow for 5 different pumps as rpm varies. The plot shows the absolute error between the reference value and the estimated value in bar. . . . 41

5.5 Absolute error for pressure at varying temperatures. . . . 43

5.6 Absolute error for flow at varying temperatures. . . . 43

5.7 Absolute error for pressure at varying pipe diameters. . . . 44

5.8 Absolute error for flow at varying pipe diameters. . . . 45

5.9 Absolute error for pressure at varying pipe lengths. . . . 45

5.10 Absolute error for flow at varying pipe lengths. . . . 46

5.11 Absolute error for pressure at varying rpm sensor faults. . . . . 47

5.12 Absolute error for flow at varying rpm sensor faults. . . . 47

5.13 Absolute error for pressure at varying current sensor faults. . . . 48

5.14 Absolute error for flow at varying current sensor faults. . . . 48

5.15 Zoomed in plot of flow vs. pressure difference over the filter at different clogging points. Note that the orifice diameter is give in mm. The line with the stars is a new (unclogged) filter. . . . 49

B.1 The data points collected from the viscometer. The x-axis shows the temperature range in °C and the y-axis shows a combined scale for den- sity, kinematic- and dynamic viscosity. . . . 62

B.2 A graf showing the viscosity of different fossil fuels at temperature rang- ing from 10 to 140 °C where the line for diesel is the one closest to the x-axis. The image is copied from [20]. . . . 62

C.1 Displacement Pressure . . . . 63

C.2 Displacement Flow . . . . 64

C.3 Pumps Pressure . . . . 65

C.4 Pumps Flow . . . . 66

C.5 Efficiency Pressure . . . . 67

C.6 Efficiency Flow . . . . 68

C.7 Temperature Pressure . . . . 69

C.8 Temperature Flow . . . . 70

C.9 Pipe Diameter Pressure . . . . 71

C.10 Pipe Diameter Flow . . . . 72

C.11 Pipe Length Pressure . . . . 73

C.12 Pipe Length Flow . . . . 74

C.13 Speed Sensor Pressure . . . . 75

C.14 Speed Sensor Flow . . . . 76

List of Figures

C.15 Current Sensor Pressure . . . . 77

C.16 Current Sensor Flow . . . . 78

List of Tables

List of Tables

1.1 The advantages and disadvantages of the new system . . . . 5

2.1 Pump Parameters . . . . 13

3.1 Feedback from the pump . . . . 30

3.2 Feedback from the sensors . . . . 31

5.1 Clogging Levels . . . . 37

A.1 Measured parameters . . . . 59

A.2 rpm = 700 . . . . 60

A.3 rpm = 3000 . . . . 60

A.4 rpm = 5000 . . . . 60

B.1 Measured parameters . . . . 61

1 Introduction

1 Introduction

This section presents a short description of the fuel system as well as research questions that will be investigated and answered. This section also includes a set of delimitations for the thesis and a method description for how to go about solving the problem.

1.1 Background

The Exta-High Pressure Injection (XPI) system is a high pressure system with injection pressures up to 2400 bar. The system can be divided into a high pressure side consisting of the high pressure pump, the accumulator and the injectors and a low pressure side consisting of the fuel filter and the low pressure pump which supplies the high pressure side with fuel.

The current mechanical low-pressure pump is directly coupled to the engine’s operating point, resulting in the pump’s speed, pressure and flow being directly dependent on the engine speed. Examination of the possibility of replacing the existing mechanical low- pressure pump with an electronic system with only one additional pressure sensor, placed just before the high pressure side in the system is being done.

The new system will result in a number of benefits, but also technical challenges which are presented in section 1.3.

1.2 Purpose

The focus of this thesis is to use electronic pumps included in the system as virtual sensors by reading current and speed data from the inherent sensors in the pumps and transforming them into a pressure and flow estimation.

This is done in order to eliminate additional flow and pressure sensors in the system and thereby reducing costs. The purpose is to investigate whether estimating pressure and flow throughout the system will form a feasible and robust solution for a controllable and diagnosable system.

The issue with the new system is the minimal use of sensors in the proposed configu- ration (Figure 1.5). In order to collect the necessary data about the system so that it can be used to control and predict system behaviour, a model which includes virtual sensors must be developed and the results validated and verified. A study must be done to encompass non-ideal behaviours of the system.

The starting point is a pre-developed estimation mapping of an ideal pump which will

be called the reference pump going forward, shown in figure 1.1.

1 Introduction

Figure 1.1: A map of the current relating to the pressure difference and rpm over the pump. The dip in the corner is explained by lack of relevant data for those data points and is therefore set to zero.

This map is used in a Simulink model with the current [A] and the speed [rpm] of the pump acting as inputs and a pressure difference ∆P being the output as shown in figure 1.2.

Figure 1.2: The estimation map in Simulink.

To estimate the flow in the system a one dimensional map is used in the Simulink

model. The input is pump speed in rpm and the output is flow in liters per minute

(lpm). The map is shown in figure 1.3.

1 Introduction

Figure 1.3: A map of the ideal flow of the reference pump vs. its speed.

1.3 Problem Description

There are some disadvantages with the current low pressure fuel circuit. Those are listed

on page 4. A simplified system overview of the current system is seen in Figure 1.4.

1 Introduction

• It occurs that the low pressure pump transfers too much fuel to the high pressure pump and the motor which then has to go back to the tank thus wasting energy and fuel.

• When starting up the engine, there is a heavy load on the startup motor, which has to start the engine so that the mechanical fuel pump can transport fuel through the system.

• There are only two sensors in the system; a pressure transmitter placed right before the fuel injectors inside the High Pressure Circuit (HPC) and a level sensor in the main tank. This makes it difficult to run diagnostics on the system.

• The ability to have a Condition Based Maintenance (CBM), does not exist in this system. It is advantageous for the customer if there is a CBM system that can accurately detect when for example filters should be changed instead of having a set interval for changing the filters that coincides with the oil change as is done today.

The advantage to this approach is that this type of system has been in use for a long time and has fewer components compared to the new system which are prone to breakage or malfunction.

There is a bigger chance for the system with electrical pumps to break down since there are two pumps instead of one and more piping, which is why it is important to have proper diagnostics and make use of redundancy in the system.

The possibility of replacing the current low pressure fuel pump system with a new

electronic system which includes electrically controlled low pressure pumps has many

advantages. These are listed in Table 1.1.

1 Introduction

Table 1.1: The advantages and disadvantages of the new system

Advantages Disadvantages

Able to calculate and control pressure and flow thus minimising waste.

Higher cost due to change in produc- tion.

Pumping fuel when starting the vehi- cle does not require the motor to start.

This is especially good when the tem- perature is low.

Higher costs due to purchasing of new and more expensive components.

A better way of separating water using push-filters instead of pull-filters. This also eliminates the problem of air en- tering into the system.

Higher cost due to the addition of an- other tank and two pumps.

Easier to diagnose the system. More components that are prone to breakage.

The ability to implement a condition based maintenance (CBM).

The ability to pump fuel only into the tech-tank when testing out the trucks thus saving money.

Better usage of energy when accelerat- ing/decelerating.

Theoretically halving the troubleshoot- ing time by further separating the high- pressure system and the low pressure system

The new system (Figure 1.5), already developed by Scania consists of the main tank followed by a transfer pump and a pre-filter. The fuel is pumped into another, smaller tank called the tech-tank (more commonly known as a catch tank).

From there, another low pressure fuel pump, here called a feed pump, pumps fuel into

the high pressure circuit. The high pressure circuit consists of a high-pressure pump,

accumulator, fuel injectors and pipes.

1 Introduction

Figure 1.5: A simple overview of the new low pressure fuel pump system. The figure is a modified version of the low pressure fuel pump system used in a previous thesis at Scania [1].

The pressure and flow sensors in the figure above (marked out in red squares) shows the areas where the pressure and flow is crucial to know. Without these sensors the pressure and flow are unknown and must therefore be estimated with the help of the electric pumps.

What is not shown in Figure 1.5 is that the tech tank (≈ 0.025m

) is much smaller than the main tank (0.15m

– 1.5m

).

Since the model will include only one pressure sensor and one level sensor, not much is known about the rest of the system. Therefore the following research questions have been developed and will be examined in this thesis.

1.3.1 Research Questions

1. Can the electric pumps in the system act as virtual sensors for flow and pressure?

2. Can the electric pumps be used for diagnostics and for evaluating the level of clogging in the fuel filters?

1.3.2 Definitions

Throughout the rest of this thesis a few terms used. Those are defined in this section.

1 Introduction

Estimation Map is the map shown in figure 1.1 for pressure and in figure 1.3 for flow.

These maps include measured data from a reference pump running during ideal conditions.

Estimated Pressure/Flow is the pressure/flow that is estimated using an estimation map.

Simulated Pressure/Flow is the flow and pressure that are calculated in the Simulink model when simulating the system in Simulink using the Simscape library.

1.3.3 Delimitations

In order to be able to answer the research questions in a limited amount of time, the following delimitations have been set in place.

• Fuel Temperature The fuel temperature is assumed to be known throughout the system.

• Pipe dimensions All pipes in the system are assumed to have the same diameter by also assuming that any differences in pipe diameters are negligible.

• Verification Verification of the model will be done by comparing the results from the simulation to measured data and sensor data collected from a rig developed in this thesis. The assumption is that the sensor data from the rig is close enough to reality that it can be used to gather conclusions about the accuracy and robustness of the estimated parameters.

• Pumps in the model By using the SimHydraulics Library [2] to model the electric pumps, inherent assumptions and limitations present themselves. They are quoted here from [3]:

– Fluid compressibility is neglected.

– No loading on the pump shaft, such as inertia, friction, spring, and so on, is considered.

– Leakage inside the pump is assumed to be linearly proportional to its pressure differential.

• A fluid here called bench oil is used in the rig instead of diesel fuel. The properties of both fluids are very close and are here assumed equal.

• The modelling and the testing will focus on one subsystem at a time (described

in section 3.1) i.e. the complete system will not by investigated.

1 Introduction

1.4 Method Description

In order to further investigate and give an answer to the research questions the following methods are employed.

• Qualitative interviews to gain a deeper understanding of the system and devel- oping a basis for a quantitative data gathering. Interviews with those involved in the project in various ways are scheduled. In these interviews electrical pumps, general information about the current system and the new proposed system, fil- ters, diagnostics and requirements are discussed.

• Examination of previous work. Two previous theses done at Scania are the main interest. The first [1], is of particular interest for getting data on the be- haviour of a normal system compared to an analysis for a faulty system behaviour.

The second thesis [4] is interesting due to the modelling of a similar system’s hy- draulics.

• Modelling and Simulation. The modelling of the system will be done in MAT- LAB and Simulink [5] using the tool SimHydraulics [2] A quantitative method is employed to gather meaningful data about the system.

• Validation of the pump model will be done by comparing the results from the model to sensor data from a rig built for this purpose. The rig is described in detail in section 3.2. To get relevant results, experiments will be designed and tested to get an understanding of each phenomenon separately. The experiment setups can be seen in the Appendix.

• Analysis of the results will finally be done by comparing varying system be- haviours against a ”normal” case in order to answer the research questions.

1.5 Report Outline

Introduction: Section 1 on page 1 gives an introduction to the thesis and presents the research question that is to be answered.

Frame of Reference: Section 2 on page 11 describes the frame of reference for the thesis.

System Description: Section 3 on page 23 describes the Simulink model and the rig created.

Validation: Section 4 on page 33 describes how the Simulink model was validated and verified in comparison to the signals from the rig.

Analysis and Results: Section 5 on page 37 describes how diagnosis can be imple-

mented and presents results in the form of graphs.

1 Introduction

Discussion: Section 6 on page 51 discusses the results found in this thesis and how they can be utilised.

Conclusion and Future Work: Section 7 on page 55 draws conclusions from the results of the thesis and discusses improvements that the author has in mind.

Appendix: Starting on page 59 the appendices include recorded data and descriptions

of experiments made in connection with the thesis as well as various plots that

did not fit into the report.

1 Introduction

2 Frame of Reference

2 Frame of Reference

Almost all trucks today use a mechanical pump to transport fuel into the engine. Cars have long since moved on to electric pumps for fuel transportation but since cars have a lower need to transport high flows in their system and do not have the same require- ments in terms of longevity and robustness, these two implementations cannot be easily compared.

Today’s fuel system includes one single pressure sensor located on the high pressure side which makes it hard to know if a fault is on the low pressure side or the high pressure side of the system.

To solve this problem and to make it easier to diagnose the system the high pressure side and the low pressure side are separated. The objective is to used the new electric pumps as virtual sensors in order to evaluate the level of clogging and the flow in the filters.

2.1 Low Pressure Fuel Circuit (LPFC) System Components

The components described in this section make up the Low Pressure Fuel Circuit (LPFC) system as seen in figure 1.5.

2.1.1 Gerotor Pump

The name ”Gerotor” is an amalgamation of the words GEnerated and ROTOR. This is a type of rotary pump that is called a fixed positive displacement pump which means that the pump delivers the same amount of fluid for each gear rotation (i.e. the rate of flow is directly proportional to the speed of the motor).

There is also a leakage coefficient described for the pump since it realistically will have some leakage at each rotation.

The way a gerotor works is that it has two parts, a rotor and a stator (also called the

idler). The rotor has n number of teeth while the stator has n + 1 number of teeth as

seen in figure 2.1 which allows fluid that travels from the inlet to gather in the place of

the ”missing tooth” for each rotation. This volume of fluid is called the displacement

and given in the SI unit m

/revolution. That fluid is then pushed out through the

outlet as seen in the figure.

2 Frame of Reference

Figure 2.1: Gerotor [6]

Communication with the pump is done via CAN. The control variable is the rpm and the values that can be read are the:

• current consumption

• motor temperature

• the speed of the pump in rpm

The exact communication system is further described in section 3.2.2.

2.1.2 Fixed-Displacement Pump

The pump and all other hydraulic components in the model are from the Simscape hydraulic library [2]. Since the real pump is a gerotor pump and is of the fixed dis- placement type, the same type of pump was used in the Simulink model.

The equations that represent the fixed-displacement pump in Simulink are:

q = D · ω − k

· p (1)

where,

p = p

− p

(2)

k

= k

ν · ρ (3)

and;

k

= D · ω

· (1 − η

) · ν

· ρ

p

(4)

The most important parameters are; q which is the flow from the pump, p the pressure differential across the pump and k

which is the leakage coefficient which describes the the amount of leakage in the pump at every rotation.

The rest of the parameters are described in detail here [7]. The basic assumptions for

this component is that;

2 Frame of Reference

(a) Fluid compressibility is neglected.

(b) No loading on the pump shaft, such as inertia, friction, spring, and so on, is considered.

(c) Leakage inside the pump is assumed to be linearly proportional to its pressure differential. (This is seen in equations 1 and 2.)

The way the pump model is used in the Simulink model is that a control signal is sent to the pump; which in this case is angular velocity through an ideal angular velocity source block.

Figure 2.2: Input/Output overview of fixed-displacement pump in Simulink model The parameters that set for this block are shown in table 2.1 where the last two pa- rameters (k

and ρ

) are set to describe the fluid flowing through the pump.

Table 2.1: Pump Parameters

Parameter Name Unit

D Pump displacement m

/rev

E

Volumetric efficiency %

E

Total efficiency %

p

Nominal pressure bar

ω

Nominal angular velocity rpm

k

Nominal kinematic viscosity mm

/s

ρ

Nominal fluid density g/cm

These parameters, particularly pump displacement and volumetric efficiency are be

manipulated later on in section 5 in order to describe variations in performance as the

pump ages and for various discrepancies in individual pumps.

2 Frame of Reference

2.1.3 Filters

Fuel filters are used in the system to separate any dust or particles that might contam- inate the fuel. Filters play an important role in fuel pump system since any contami- nation in the fuel might lead to erosion and decreased performance in the pump motor.

The filters will naturally increase the flow resistance in the system especially as they become more clogged. It is therefore important to include them in the model in order to give an accurate representation of the system. Shown in figure 2.3 is a drawing of a fuel filter.

Figure 2.3: Fuel filter

There are two types of filters, puller filters and pusher filters. The type of filter used

depends on which side of the pump the filter is placed. As the name indicates a puller

filter works when a pump pulls fluid through the filter. Similarly a pusher filter works

when a pump pushes fluid through it as seen in figure 2.4.

2 Frame of Reference

(a) Puller Filter

(b) Pusher Filter

Figure 2.4: The difference between puller and pusher filters.

The main reason for wanting to use pusher filters is that the pressure drop limit is much higher, it reduces the risk of air being pumped into the fuel, it can handle higher flows and it is easier (i.e. it requires less energy) for the pump to push the fluid through the filter(Oral reference: Carina Forsberg, Scania, february 2016 ).

The maximum pressure drop over a puller filter before it needs to be replaced is approx-

imately 0.4bar but can go up to 0.7bar. In contrast, the pressure drop over a pusher

filter is between 1 and 2bar when measured at a flow rate of 5.5 lpm.

2 Frame of Reference

hard to pin-point the exact flaw

. The advantage of the new system is that it separates the low and high pressure system, thereby theoretically cutting troubleshooting time in half.

2.1.4 Tech-tank

A new component in the system is the tech-tank (more commonly known as a catch tank). Schematically the tech-tank is placed between the pre-filter and the feed pump (refer to figure 1.5 on page 6).

The advantages of having a tech-tank are numerous. They are listed below.

• For driving short periods, i.e. testing; only the tech-tank needs to be filled instead of the main tank which requires a lot more fuel.

• With a tech tank, the vehicle will be less sensitive when for example driving in slopes.

• An additional factor that saves money is if the trucks do not need to be filled with as much fuel during assembly.

Today the largest tanks have to be filled with at least 190 litres [8] for the trucks to be able to drive up the ramps that are needed to transport the trucks to the customer.

A smaller catch tank reduces the risk of air being sucked in to the system when driving uphill, which means that each truck needs less fuel.

2.1.5 Fuel

There are many types of fuel that can be used in trucks. The most common is Diesel, but Fatty Acid Methyl Esters (FAME) (better known as biodiesel) and B100 which is FAME in pure form can also be used.

In this thesis all calculations and simulations are based on data from bench oil which is an oil that is closely related to diesel fuel in density and viscosity.

2.1.6 Piping

Pipe lengths in a truck were obtained by simply measuring the distances that the pipes went along on a truck. The pipe lengths are just a rough estimation of how it might look like in the finished product but there will be differences between various implementations.

The impact of pipe length on the system is analysed further in section 5. The pipes are assumed to have the same diameter and are made of hard plastic (Oral reference:

Patrik Fogelberg, Scania, february 2016 ).

1

Be it either with the pump itself, a clogged filter or a leak somewhere in the system.

2 Frame of Reference

2.2 Clogging

Clogging of the filters is measured as the pressure difference over the filter when the flow through the system is at 5.5l/min.

In order to assess the level of clogging in the filters, P

(as seen in figure 2.5) must be calculated. This calculation is done with the help of equation 6.

Figure 2.5: Pressure overview

∆P

= P

− P

(5)

P

= ∆P

+ P

(6)

 = e P

− P

(7)

P e

is the pressure measured on the rig and P

is the simulated pressure.  shows the error margin between the two.

P

is assumed to be known. P

is also assumed to be known in the transfer subsystem and it is measured by a pressure sensor if it is in the feed subsystem (refer to figure 3.1).

2.3 Diagnostics

Usually when diagnosing the filters of trucks on the road preventative maintenance is executed in order to reduce time spent in a workshop.

This is important since it is an economic loss for a truck to stand idle in a workshop.

Preventative maintenance means that maintenance is scheduled to occur at certain time intervals. In the case of filter changes, they coincide with oil changes in the trucks which are schedules every 150 000 km in Europe (Oral reference: Zohreh Pakdaman, Scania, february 2016 ).

This is however wasteful since there is no check to see if the filter are actually clogged

2 Frame of Reference

workshop and saving costs by only changing components when they are detected to be close to breakdown.

2.3.1 Condition Based Maintenance System

A CBM software subjects the system to continuous inspection to identify the level of deterioration (in this case percentage of clogging up in the filters) to determine if the system needs maintenance.

Amari and McLaughlin [9] define a deteriorating system as: “a process where the im- portant parameters of a system gradually worsen, and if left unattended, the process leads to deterioration failure.”

Figure 2.6: The steps in a CBM program [10]

Figure 2.6 describes the process. The first step (Data Acquisition) is done by saving the data sent from the pumps via CAN such as voltage, current and motor temperature.

This leads to the next step which is Data Processing. In this step those signals are processed in a way that they can be compared to recorded data or a dynamic model that will declare if the parameters indicate a faulty or non-faulty system.

In the next step, a course of action is decided upon. If the parameters indicate a faulty system or at least that the maintenance threshold has been reached repairs are done.

Otherwise no action or alert is performed.

2.4 Simscape

Simscape is an extension of Simulink where instead of using mathematical equations, the blocks in the Simscape library represent physical components

[11].

The simscape library in Simulink offers a variety of modelled physical components that facilitate the modelling process.

The hydraulic components were used to model the system and some mechanical com- ponents were used for the gerotor pump. This type of physical modelling makes it easy to model complicated systems without spending too much time on the implementation.

This gives more time for analysis and comparison of results. The complete Simulink model is presented in section 3.1.

2

Hydraulic, pneumatic, electrical, thermal and mechanical components.

2 Frame of Reference

2.5 Pressure

Figure 2.7 illustrates the many different ways of denoting pressure. Three of them are gauge pressure, barometric pressure (also known as atmospheric pressure) and absolute pressure.

Figure 2.7: The relationship between different types of pressure [12].

In this thesis all references to pressure are meant to denote absolute pressure and the

units used are bar or P ascal.

2 Frame of Reference

2.6 Current consumption vs. pressure difference over the pump

In order to simulate the pressure difference over the pump and filter in the Simulink model an estimation map (described in section 1.2) for the pressure estimation is utilised. The current will behave differently than in the mapping table due to manufac- turing tolerances, operation of the pump, ageing and hardware changes in the system.

Therefore the estimated pressure difference will deviate from the ”normal case” and an error will be present between the reference value and the estimated value (∆P

).

Parameter changes in the pump and the system are analysed and conclusions drawn about the extent of the absolute error.

Figure 2.8 shows how the current (I) approximation works. A motor model in Simulink can be used but since motor parameters for the pump are unavailable an approximation of the current is done analytically.

This is done so that the current I

in figure 2.8 can be known in order to be able to use the estimation mapping as seen in figure 1.1.

This section explains how the estimation map is used when deriving a pressure dif- ference over the pump for comparison to the ”normal” case

and how the necessary equations are derived.

Figure 2.8: Overview of how the comparison is done between the reference pump and manipulated pump model. I

is unknown and must be calculated.

The Simulink model with data of the reference pump is run where the current and rpm are known. From there a nominal pressure difference is obtained through the estimation map which is a function of current and speed (rpm). Then by manipulating one or more parameters in the model another value for the current I

(the rpm remains the same) is obtained and from that the respective pressure difference over the motor is obtained.

The two values (∆P

and ∆P

) are then compared and the absolute error shows

3

The ”normal” model is a simulation of previously measured data which is validated in the rig (see

section 4).

2 Frame of Reference

how far the estimate has drifted from the reference pump pressure difference. Figure 2.9 shows how the power is translated through the pump.

Figure 2.9: How power is translated through the pump. η

is the overall pump efficiency. [13]

Equations 8 through 12 describe the relation between the mechanical power of the motor W

and the hydraulic power of the pump W

.

η

= W

W

= η

∗ η

(8)

where η

stands for mechanical efficiency and η

stands for volumetric efficiency.

W

= T ∗ ω (9)

where T is torque and ω is the pump’s angular velocity.

The relationship between the torque of a DC motor and the current is described in the next equation. By calculating the k

Φ variable from available pump data, a motor model is created:

T = k

Φ ∗ I (10)

where k

Φ is the motor torque constant and I is current of a DC motor [14]. The angular velocity which is in rad/s

can be rewritten as rpm with the following equation:

ω = 2 ∗ π ∗ n

60 (11)

The hydraulic power out of the pump is a relationship of pressure difference over the pump and flow as seen below.

W

= ∆P ∗ Q (12)

To be able to calculate the current, equations 9, 10 and 12 are combined and the result is equation 13.

I = 60 ∗ ∆P ∗ Q k

Φ ∗ 2 ∗ π ∗ n ∗ η

(13)

2 Frame of Reference

The relationship in equation 8 is from [13]. By including equation 13 in the Simulink model it is possible to calculate the current and use it as an input to the estimation map (where the inputs are current and rpm and the output is an estimated pressure difference) it is possible to get an estimated pressure difference.

By looking at equation 10 it is seen that the torque T is proportional to the current I with a factor of k

Φ.

From equation 8 it can be concluded that a pump is ideal if η

= 1 and therefore W

= W

. By inserting equation 9 and equation 12 we get the following relationship:

T ∗ ω = ∆P ∗ Q (14)

By including the efficiency factor (η

) again, since no pump is ideal, the following equation is produced:

T ∗ ω ∗ η

= ∆P ∗ Q (15)

For an ideal fixed displacement pump with no leakage the flow is given by the following equation:

Q = D ∗ ω (16)

where D is the displacement of the rotor per revolution and ω is the angular velocity of the motor. With leakage included the following equation is produced:

Q = D ∗ ω ∗ η

(17)

where η

stands for volumetric efficiency. Combining equations 8, 15 and 17, results in:

T = ∆P ∗ D

η

(18)

This shows that ∆P is also proportional to the torque at a fixed displacement.

T ∝ I ∝ ∆P (19)

In conclusion it is shown that ∆P is proportional to the current I when the efficiency and the displacement are constant. This means for example that a 10% increase in pressure difference will result in a 10% increase in current.

For a non-ideal pump the efficiency of the pump must be taken into account as the

pressure changes.

3 System Description

3 System Description

In order to model the behaviour of the system (figure 1.5), it is split up in two parts.

The first part will be called the transfer system and the second part will be the feed system.

Both the rig and the Simulink model are made up of only one of the subsystems but are made to represent both. The only difference being that the limit of clogging for each of the filters is defined differently.

This chapter describes how the rig and Simulink model were constructed and how they will be used in order to answer the research questions presented in section 1.3.1.

3.1 Simulink Model

As mentioned before, the low pressure fuel system is divided into two very similar subsystems; the transfer system and the feed system. The differences between them are shown in figure 3.1.

Figure 3.1: The differences between the transfer and the feed subsystems. The pressure sensor marked with a red box is the pressure estimation that is simulated with the help of the electric pumps.

The pressure differences over the pre-filter should not exceed ∆p = 1bar while the fuel filter can go up to ∆p = 2bar. These limits indicate that the filters are fully clogged.

This pressure difference is measured when the flow is ≈ 5.5liter/min.

Another difference is that pressure is measured after the filter in the feed system.

In the transfer system the pressure after the pre-filter is assumed to be 1bar, i.e. at- mospheric pressure since it is connected with the tech-tank which is open to the atmo- sphere.

3.1.1 Lookup Tables

3 System Description

Figure 3.2: Input/Output overview of lookup table in Simulink model

The data is collected at certain rpms at seven different pressure differences between 1 to 8 bar. This data makes up each line in figure 3.3.

Any data points between or outside of these lines is either interpolated or extrapolated in Simulink.

Figure 3.3: A lookup table where the lines represent pump speed at varying pressure drop.

Figure 3.3 is used in the Simulink model to describe the actual current output of the electric pump and is used in the simulations as a means to get the pressure difference over the pump using current.

The estimation map presented in section 1.2 on page 1 is also embedded in the Simulink

model as a lookup table. It is an inverted map of the one described above and is further

discussed in section 3.1.5 on page 26.

3 System Description

3.1.2 The filter and pipes

Simulating the filter in the model is done by using a variable hydraulic orifice.

Instead of having small holes distributed throughout the filter they are grouped together into a single orifice as seen in figure 3.4 where the ”orifice area” is calculated as a circle.

Figure 3.4: The filter and the pipes in Simulink.

The fuel filter is crucial for the model but the pipes are also important in order to get as close to reality as possible. Therefore pipes are included in the model and the parameters set to resemble the rig as closely as possible.

Note that these parameters (i.e. length, inner diameter and type of pipe) may vary significantly in the final product.

3.1.3 Fluid and Fluid Temperature

3 System Description

The variations in viscosity and density for the test oil were tested in an experiment.

For the complete setup of the experiment and the results gained see appendix B on page 61.

Figure 3.5: Overview of main tank including solver and custom fluid.

Since this is a low pressure system, the bulk modulus of the fluid, which is the fluid’s resistance to uniform compression, has little to no impact on the system.

It is therefore neglected and set to 0.8e9P a in the model which is the default parameter value of the bulk modulus of the custom in the model.

3.1.4 The Solver

The solver type that is chosen for the Simulink model is a ”backward euler” solver with a sample time of 0.02 which is good enough for accuracy but also fast enough for simulations to run quickly. The system is assumed to be ”stiff”, meaning that the numerical method must take small steps to obtain satisfactory results [15].

3.1.5 The Calculation of I

Using the measured data, the current was mapped out to the pressure difference and rpm in the measured points. The result is seen in figure 1.1.

This map is inserted in the Simulink model as a 2-D lookup table where the inputs are

current and rpm and the output is a pressure estimation over the pump.

3 System Description

The equation for calculating current that was derived in section 2.6 is also included in the model as seen in figure 3.6.

Figure 3.6: The equation for estimating I and the lookup table implemented in Simulink.

3 System Description

3.2 The Rig

The rig described in this section is build specifically for this thesis. It is constructed in order to test the pump at different operating points and to record data from the pump and flow/pressure sensors that are included.

The parameters are listed in tables 3.1 and 3.2. The rig represents one subsystem as described in section 3.1 on page 23.

3.2.1 Components

Figure 3.7 shows an overview of all the components of the rig and how they fit put together. The bold black lines represent pipes and the fuel-filter is constructed by using a configuration of ball valve and electric valve (figure 3.8).

Figure 3.7: Simplified rig overview where P denotes a pressure sensor and q denotes a flow sensor. The ball valve is there to simulate the HPP.

There are three pressure sensors and one flow sensor included, which are connected to a CAN line in order to get relevant data throughout the system.

By calculating the difference between the pressure sensors it is possible to get the total

pressure difference (∆P ) over the pump and the ”filter”.

3 System Description

3.2.1.1 Pump The pump is a hydraulic fixed-displacement gerotor pump from the manufacturer Parker [16].

3.2.1.2 Valve The filter is modelled using two valves and a third valve to simulate the high-pressure system. The fuel filter in figure 3.7 is further modelled as seen in figure 3.8 with an electric valve in parallel with a regular ball valve.

Figure 3.8: Components that make up the filter in rig.

Since the electric valve is always a little bit closed and cannot by itself generate a pressure difference as low as 150mbar (which is the pressure difference over a new filter), another manual ball valve is placed in parallel to get the pressure difference over that system down to 150mbar.

The electric valve is helpful when the pump is running at high rpm’s. Then the electric valve can be tuned to give just the right pressure drop. Data from one of the experiments done can be seen in appendix A.

3.2.1.3 Pipes The pipes through which the fluid flows are made of hard plastic with an inner diameter of d = 9mm and the lengths were cut to fit.

3.2.2 Software

As seen in figure 3.9, the setup needed to collect significant data from the rig requires several components. First, all the sensors and the pump from the pump system (figure 3.7) are connected to a voltage supply and the ECU.

A CAN line is then connected to the pump in order to request different rpm levels.

Another CAN line is connected to an adaptor and then a to the Vehicle Communication

Interface (VCI) which is a communication dongle for connecting a PC to CAN.

3 System Description

Figure 3.9: ECU setup for CAN communication.

3.2.2.1 Sensors From the sensors included in the rig it is possible to calculate the pressure difference over the pump and the pressure difference over the filter (valves).

These parameters are useful to get a comparison between the pressure differences recorded in the rig and the pressure differences recorded in the Simulink model.

In tables 3.1 and 3.2 the signals that are recorded from the rig are presented.

Parameter Unit

Current A

Rotations per minute rpm

Temperature °C

Voltage V

Table 3.1: Feedback from the pump

3 System Description

Parameter Unit

Pressure sensor 1 hP a Pressure sensor 2 hP a Pressure sensor 3 hP a

Flow sensor lpm

Table 3.2: Feedback from the sensors

3 System Description

4 Validation

4 Validation

This section describes how the validation process for the Simulink model and the calcu- lated pressure estimator is done.

To verify that the model is working, signals measured from the rig are compared to the output data of the model. The inputs are rpm and pressure difference over the filter (clogging) and the outputs are current, flow and pressure over the pump.

The model simulates how the parameters (current, rpm, flow and pressure) compare to rig data in ”normal” working conditions.

The way it is done is that previously recorded sensor data signals from the rig are sent in as a signal into the model and are there compared to the signals simulated by the model. The data is plotted in Simulink and is shown in figure 4.2.

Then the pressure estimation that was described in section 2.6 is compared to the signals given in the Simulink model in order to get an idea if the estimation calculations are correct. The results are shown in section 4.2.

Figure 4.1: Pressure difference over pump at 3000 rpm.

Figure 4.1 shows how the signal from the rig (blue) and the signal from the model (red)

showing pressure difference over the pump while it is running at 3000 rpm.

4 Validation

Figure 4.2 shows comparison data for pressure difference, current and flow at varying rpms.

700 rpm

(a) ∆P (b) Current, note that the signal is 0 due to sensor resolution

(c) Flow

3000 rpm

(d) ∆P (e) Current (f) Flow

5000 rpm

(g) ∆P (h) Current (i) Flow, note that the highest

flow that can be detected by the flow sensor is 5 lpm.

Figure 4.2: Comparison between measured signal from rig and Simulink model where the dotted (red) line is the model and the straight (blue) line is the measured signal.

Note that the y-axis has been resized for each plot in order to better represent the data.

4 Validation

4.1 Sensor Resolution

At low rpms (between 700 and ≈ 1000 rpm) when receiving feedback from the pump the current remains 0 A until pressure difference over the pump reaches a certain level.

This can be seen in subplot 4.2(b) i.e. the sensor resolution is not high enough to get relevant data. Pressure estimation at such low rpms is therefore disqualified.

In subplot 4.2(i) it is seen that at 5000 rpm the flow signal is stagnant at 5 lpm, this is due to the flow sensor in the rig measures the flow through a voltmeter where 0 V is equal to 0 lpm and 5 V is equal to 5 lpm.

Since the voltage source to the flow sensor does not go above 5 V it cannot in its present configuration detect a flow above 5 lpm.

4.2 Current calculation

The reference parameters are measured data from the exact pump that is modelled in the Simulink model. This is done as described in section 2.6. Since the pump turned off in the beginning the model needs some time to saturate and therefore the time period is longer in these plots.

The interesting points are when the calculation is saturated.

Figure 4.3: Ideal estimated pressure (red signal) vs. Ideal pressure (blue signal).

4 Validation

Figure 4.4: Ideal estimated current (red signal) vs. Ideal current (blue signal) The figures above show the estimation at 3000 rpm and the clogging value (i.e. the orifice area of the filter) is set so that pressure difference over the pump ends up at approximately 1.5 bar.

To collect data for the estimator, analysis of how parameters in the pump and system

can change due to ageing, friction and other circumstances can begin. The results are

presented in the next chapter.

5 Analysis and Results

5 Analysis and Results

Once the model has been verified, an attempt at manipulating the parameters that the model is based on begins.

This section describes how the analysis was done and plots of pump performance and other parameters are presented. The analysis is done on one subsystem as previously described in section 3.

All plots show the difference between the reference system behaviour (denoted as black stars in the plots) and manipulated behaviour.

5.1 Plot setup

The effects each manipulated parameter has on the flow and pressure is shown in this chapter. For each parameter two plots are shown. The first describes the changes in pressure estimation due to parameter manipulation. The x-axis shows the speed of the pump and the y-axis shows the absolute error of pressure (i.e. the difference between the reference values and the estimated values). The absolute error is calculated thusly:

error = value

− value

(20) The starred black line is actual measured data while all other lines show simulated data.

How much the parameters are manipulated is shown in the legend.

The second plot is similar in construction to the first with the difference being that the y-axis shows the absolute error for flow through the system instead of pressure.

Three different clogging limits are chosen to display the changes that occur in pressure and flow estimations as the clogging increases. These limits are chosen in order to best demonstrate how the system behaviour changes as the filter deteriorates and are shown in table 5.1.

Table 5.1: Clogging Levels

Clogging Level Description

1 ∆P

= 0.15bar New fuel filter

2 ∆P

= 1bar Half clogged fuel filter

3 ∆P

= 2bar Clogged fuel filter

All plots in this chapter show a clogging level of ∆P

= 0.15bar. Plots for the two remaining clogging levels are placed in appendix C starting on page 63.

5.2 Pump Parameters

5 Analysis and Results

The method described in section 2.6 and especially figure 2.8 on page 20 is used in this chapter to evaluate the electric pump.

5.2.1 Displacement

A pump’s displacement is relatively constant for the duration of its lifetime. It is set by the manufacturer. It can also vary depending on the manufacturing tolerances of each separate pump and the wear and tear it might be subjected to during its lifetime.

To illustrate what might happen if the displacement changes and the same model is still used to approximate the pressure and flow in the system, a series of different levels

of displacement were set in the Simulink model and the behaviour was plotted.

Figure 5.1 shows the estimation of pressure difference over pump as displacement and rpm varies. The plot shows the absolute error between the reference value and the estimated value in bar.

Figure 5.1: Absolute error for pressure at varying displacement.

Figure 5.2 shows that the flow will change if the displacement get higher or lower so it is important to take that into account.

4

Increase/decrease by up to 5% of the original displacement.

5 Analysis and Results

Figure 5.2: Absolute error for flow at varying displacement.

Figures C.1 and C.2 in Appendix C beginning on page 63 shows that the error gets larger the more clogged the filter gets. In these cases the displacement has been changed by up to ±5% increase/decrease.

The increase shown here is merely an example of what the change in displacement may look like as the pump ages. Again the differences increase as the speed of the pump increases. The point is to illustrate that the worst case scenario is when the filter is fully clogged and the speed of the pump is at its highest.

5.2.2 Efficiency

Depending on the pump, pump efficiency will change to varying degrees as it ages. There are also slight differences in each individual pump due to manufacturing tolerances.

Changes in efficiency in a pump might also be due to friction, normal wear and tear and can also depend on the purity of the fluid that flows through the pump.

Mechanical Efficiency (E

) (i.e. pump wear) has many different sourcees according to

an article by LaBour Pump Company [17].

5 Analysis and Results

Erosion: due to impurities in the fuel.

Corrosion: which can be caused by air entering the system or a rapid temperature change.

Fluid velocity: where the higher the fluid velocity, the faster the pump will wear down.

Turbulent flow: that can cause uneven wear.

Water hammer effects: if any valves are closed/opened suddenly this may wear out the pump due to major changes in pressure differences.

Volumetric Efficiency (E

) on the other hand can change due to leakage in the pump which increases over time. As the pump ages more/less friction and more leakage might occur.

The efficiency of course varies with the speed the pump and at pressure differences across the pump.

The relations between E

and E

are shown in equations 21, 22 and 23.

E

= E

∗ E

(21)

E

= q

− q

q

(22)

E

= perf ormance

perf ormance

(23)

where E

must be higher than Overall Efficiency (E

).

E

≥ E

(24)

Due to the way the Simulink model was constructed, relevant data about the efficiency of the pump could not be extracted and so analysis of the ageing of the pumps is left for future work.

5.2.3 Normal variations between pumps

In order to evaluate how much the efficiency varies in new pumps, several pumps with different pump data

were tested and compared to the first one tested.

The documented efficiency levels of five new pumps are inserted into the Simulink model and a plot showing the level of error between those pumps is seen in figures 5.3 and 5.4.

5

Data is sent from the supplier.

5 Analysis and Results

Figure 5.3: Estimation of pressure for 5 different pumps as rpm varies. The plot shows the absolute error between the reference value and the estimated value in bar.

Figure 5.4: Estimation of flow for 5 different pumps as rpm varies. The plot shows the

5 Analysis and Results

when compared to pump one but the error is less than 0.5 bar.

In comparison the same test was done but with a filter that is fully clogged. Then the error margin was between -1.2 and 0.2 bar.

The error grows as the pump speed increases; a finding that is consistent with the findings in the following sections.

The pressure estimation error increases as the filter becomes more clogged. For the remaining plots, showing the errors at different clogging levels, refer to appendix C, figure C.3 and C.4 beginning on page 65.

The pressure difference changes significantly between pumps which raises the question if it is possible to have a pressure estimation that is based only on one pump data instead of an average of several pumps.

In future implementations it might be advisable to calculate an average of several pumps’ parameters in order to have a nominal value for simulations.

The flow remains the same for different pumps which is expected since the pump is a fixed-displacement pump and should have the same displacement (and therefore the same flow).

5.3 System Parameters

There might be differences in the final implementation of the LPFC system. Here are some parameters that might change the behaviour of the model.

5.3.1 Temperature

Using the same method as in section 5.2.1, the effect that fuel temperature has on the pressure estimation of the system is analysed.

For sub-zero temperatures the error becomes much higher. This is due to the charac- teristics of the fuel; the viscosity grows exponentially at those temperatures while it is quite stable at above zero temperatures.

The change in pressure estimation is significant when the fuel temperature goes below 0 °C. The pressure estimate might then vary up to 1.5 bar.

This means that for example when running the pump at a speed of 5000 rpm the dif- ference in pressure estimation between a temperature of 0 °C and -35 °C is around 1.5 bar.

The complete results of the simulations can be seen in Appendix C, figures C.7 and C.8 beginning on page 69.

For pressure and flow, the estimation results in the following plots.

5 Analysis and Results

Figure 5.5: Absolute error for pressure at varying temperatures.

Figure 5.6: Absolute error for flow at varying temperatures.