Hydrailic Systems Linkoping

Linkping Studies in Science and Technology.Thesis No. 1595

Fluid Power Systems for Mobile

Applications

with a Focus on Energy Eciency and DynamicCharacteristics

Mikael Axin

LIU-TEK-LIC-2013:29Division of Fluid and Mechatronic Systems

Department of Management and EngineeringLinkping University, SE581 83 Linkping, Sweden

Linkping 2013

ISBN 978-91-7519-600-8ISSN 0280-7971LIU-TEK-LIC-2013:29

Copyright c 2013 by Mikael AxinDepartment of Management and EngineeringLinkping UniversitySE-581 83 Linkping, Sweden

Printed in Sweden by LiU-Tryck, Linkping, 2013.

To Jennie

Mycket f mnniskor lever i dag de

esta gr frberedelser fr att leva i

morgon.

Jonathan Swift (1667-1745)

Att inse att man r okunnig r ett bra

steg mot kunskap.

Benjamin Disraeli (1804-1881)

Abstract

This thesis studies an innovative working hydraulic system design formobile applications. The purpose is to improve the energy eciency andthe dynamic characteristics compared to load sensing systems withoutincreasing the complexity or adding additional components.

The system analysed in this thesis is referred to as ow control. Thefundamental dierence compared to load sensing systems is that thepump is controlled based on the operators command signals rather thanfeedback signals from the loads. This control approach enables higherenergy eciency since the pressure dierence between pump and load isgiven by the system resistance rather than a prescribed pump pressuremargin. High power savings are possible especially at medium ow rates.

Furthermore, load sensing systems suer from poor dynamic charac-teristics since the pump is operated in a closed loop control mode. Thismight result in an oscillatory behaviour. Flow control systems have nostability issues attached to the load pressure feedback since there is none.

Pressure compensators are key components in ow control systems.This thesis addresses the ow matching problem which occurs whenusing conventional compensators in combination with a ow controlledpump. Flow sharing pressure compensators solve this problem since thepump ow will be distributed between all active functions. A novel con-trol approach where the directional valve is controlled without aectingthe cylinder velocity with the objective of optimizing the damping isproposed.

In this research, both theoretical studies and practical implementa-tions demonstrate the capability of ow control systems. Experimentsshow a reduced pump pressure margin and energy saving possibilities ina short loading cycle for a wheel loader application.

i

Acknowledgements

The work presented in this thesis has been carried out at the Divisionof Fluid and Mechatronic Systems (Flumes) at Linkping University.There are several people who have made this thesis possible and towhom I would like to express my gratitude.

First of all I would like to thank my supervisor, Prof. Petter Krus, forhis support, supervision and valuable inputs to my work. I am also verygrateful to Prof. Jan-Ove Palmberg, former head of division. Thankyou for giving me the opportunity to be a part of this division. I wouldlike to give special thanks to Dr. Bjrn Eriksson for his great support inmy work. I would also like to thank all my other colleagues for makingthe university a fun place to work at.

Thanks go to Parker Hannin AB for their nancial involvement andtheir help with hardware and other resources.

Most of all, I would like to thank my family and friends for alwaysbeing there for me. My greatest gratitude goes to you Jennie, my won-derful love, for sharing life with me.

Linkping, April, 2013

Mikael Axin

iii

Papers

The following three appended papers are arranged in chronological orderof publication and will be referred to by their Roman numerals. Allpapers are printed in their originally published state with the exceptionof minor errata and changes in text and gure layout in order to maintainconsistency throughout the thesis.

In papers [I], [II] and [III], the rst author is the main author, respon-sible for the work presented, with additional support from the co-writers.A short summary of each paper can be found in chapter 8.

[I] M. Axin, B. Eriksson, and J.-O. Palmberg. Energy Ecient LoadAdapting System Without Load Sensing - Design and Evalu-ation. In: The 11th Scandinavian International Conference onFluid Power (SICFP09). Linkping, Sweden, June 2009.

[II] M. Axin, B. Eriksson, J.-O. Palmberg, and P. Krus. DynamicAnalysis of Single Pump, Flow Controlled Mobile Systems. In:The Twelfth Scandinavian International Conference on FluidPower (SICFP11). Vol. 2. Tampere, Finland, May 2011, pp. 223238.

[III] M. Axin, J.-O. Palmberg, and P. Krus. Optimized Damping inCylinder Drives Using the Meter-out Orice - Design and Exper-imental Verication. In: 8th International Fluid Power Confer-ence (IFK). Vol. 1. Dresden, Germany, Mar. 2012, pp. 579591.

v

Papers not included

Paper [IV] is not included in the thesis but constitutes an importantpart of the background. The two rst authors are the main authors,responsible for the work presented, with additional support from theco-writers.

[IV] M. Axin, R. Braun, A. DellAmico, B. Eriksson, P. Nordin, K.Pettersson, I. Staack, and P. Krus. Next Generation SimulationSoftware using Transmission Line Elements. In: Fluid Power andMotion Control (FPMC). Bath, UK, Sept. 2010, pp. 265276.

vi

Contents

1 Introduction 1

1.1 Motivation and Needs . . . . . . . . . . . . . . . . . . . . 1

1.2 Aims . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Delimitations . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.4 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Mobile Working Hydraulic Systems 3

2.1 Valve Controlled Systems . . . . . . . . . . . . . . . . . . 4

2.1.1 Open-centre . . . . . . . . . . . . . . . . . . . . . . 4

2.1.2 Constant Pressure . . . . . . . . . . . . . . . . . . 6

2.1.3 Load Sensing . . . . . . . . . . . . . . . . . . . . . 7

2.1.4 Individual Metering . . . . . . . . . . . . . . . . . 9

2.2 Valveless Systems . . . . . . . . . . . . . . . . . . . . . . . 13

2.2.1 Secondary Control using Transformers . . . . . . . 13

2.2.2 Pump Controlled Actuators . . . . . . . . . . . . . 14

2.2.3 Electro Hydraulic Actuators . . . . . . . . . . . . . 15

2.3 System Summary . . . . . . . . . . . . . . . . . . . . . . . 17

3 The Flow Control Concept 19

3.1 Pressure Compensators . . . . . . . . . . . . . . . . . . . 21

3.1.1 Traditional Compensators . . . . . . . . . . . . . . 21

3.1.2 Flow Sharing Compensators . . . . . . . . . . . . . 22

3.2 Pump and Valve Control Approaches . . . . . . . . . . . . 24

3.2.1 Flow Control using Traditional Compensators . . . 24

3.2.2 Flow Control using Flow Sharing Compensators . 26

3.3 Energy Eciency . . . . . . . . . . . . . . . . . . . . . . . 28

vii

4 Dynamic Analysis 314.1 Mathematical Model . . . . . . . . . . . . . . . . . . . . . 324.2 Pump Stability . . . . . . . . . . . . . . . . . . . . . . . . 34

4.2.1 Load Sensing Systems . . . . . . . . . . . . . . . . 344.2.2 Flow Control Systems . . . . . . . . . . . . . . . . 35

4.3 Damping . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.3.1 Active Control of the Inlet Orice . . . . . . . . . 374.3.2 Design and Control of the Outlet Orice . . . . . . 40

5 Experimental Results 435.1 Energy Eciency Improvements . . . . . . . . . . . . . . 43

5.1.1 Hardware Requirements . . . . . . . . . . . . . . . 435.1.2 A Demonstrator System . . . . . . . . . . . . . . . 44

5.2 Improved Damping . . . . . . . . . . . . . . . . . . . . . . 47

6 Summary and Conclusions 49

7 Outlook 51

8 Review of Papers 53

Appended papers

I Energy Ecient System Without Load Sensing 61

II Dynamic Analysis of Flow Controlled Systems 85

III Optimized Damping Using the Meter-out Orice 109

viii

Nomenclature

The quantities used in this thesis are listed in the table. Capital lettersare used for linearized and Laplace transformed variables.

Quantity Description Unity

Ac Cylinder area m2

Ac1 Compensator area exposed to control pressure m2

Ac2 Compensator area exposed to control pressure m2

As Directional valve opening area m2

Bp Viscous friction coecient Ns/mCq Flow coecient -Fs Compensator spring stiness NKca Flow-pressure coecient for the inlet orice m

3/Pa sKcaopt Kca which gives the highest damping m

3/Pa s

Kcb Flow-pressure coecient for the outlet orice m3/Pa s

Kcbopt Kcb which gives the highest damping m3/Pa s

Lp Pump inductance Pa s2/m3

mL

Load mass kgPa Pressure on the piston side of the cylinder PaPamax Maximum pressure on the piston side PaPb Pressure on the piston rod side of the cylinder Pap

LLoad pressure Pa

pLmax

Maximum load pressure PaPp Pump pressure Papr Reduced pressure Paps Supply pressure PaQa Flow into the cylinder m3/sQb Flow out of the cylinder m3/sq

LLoad ow m3/s

ix

Qp Pump ow m3/sQpref Pump ow demand m

3/ss Laplace variable 1/sU Mechanical gear ratio -Va Volume at the piston side of the cylinder m3

Vb Volume at the piston rod side of the cylinder m3

Vp Pump hose volume m3

Xp Piston position me Bulk modulus Pai Parameter for the inlet orice -o Parameter for the outlet orice -hmax Maximum damping -pp Pump pressure margin PaPp Pump pressure margin PaPpref Pump pressure margin demand Pa Cylinder area ratio - Density kg/m3

Go Open loop transfer functionGpFC Pump transfer functionGpLS Pump transfer functionGva Inlet valve transfer functionGvea Inlet valve transfer functionGvb Outlet valve transfer functionHs Pump hose transfer functionZ

LLoad transfer function

x

1Introduction

Fluid power systems are used in a wide range of applications, mobile aswell as industrial. In mobile machinery, such as construction, forestryand agricultural machines, uid power is used for both propulsion sys-tems and working hydraulics. An example of working hydraulics is thesystem controlling the bucket motion of an excavator. This thesis coversthe area of working hydraulics in mobile machinery. An innovative sys-tem design is presented and discussed in relation to both existing andnot yet commercially available mobile hydraulic systems.

1.1 Motivation and Needs

There are several dierent reasons for preferring uid power systemsto other technologies. Fluid power components have a superior powerdensity compared to other technologies, for example electrical compo-nents [1]. Furthermore, uid power systems have the ability to handleforce impacts, which makes it more robust than for example mechanicaltransmissions. Fluid power components are generally available at lowercost compared to other technologies, especially for high power applica-tions [1]. Another property of uid power systems is their good heattransfer capability.

Fluid power systems also present some challenges. The most impor-tant one concerns their energy eciency [2] [3]. Much progress has beenmade in making the individual components more ecient [4] [5]. How-ever, each component has its own optimum working condition, whichoften leads to poor overall system eciency [5].

When improving energy eciency in uid power systems, the trend

1

Fluid Power Systems for Mobile Applications

is to use additional components and more sophisticated control algo-rithms [2] [6]. Meanwhile, less attention has been paid to the dynamicproperties. A hydraulic system with poor dynamic properties has a ten-dency to oscillate, which has a negative impact on both the productivityof the application and the comfort of the operator.

1.2 Aims

The purpose of this thesis is to investigate and analyse how the energyeciency and the dynamic characteristics of the working hydraulics inmobile machinery can be improved. Dierent valve concepts are studied.The hypothesis is that there exist valve controlled systems which im-prove the energy eciency and the dynamic properties compared to loadsensing systems, without increasing the complexity or adding additionalcomponents. The solutions presented in this thesis are demonstratedthrough both simulation and experiments.

1.3 Delimitations

This thesis concerns the energy eciency and dynamic characteristicsof mobile uid power systems. Other aspects, such as manufacturingand marketing are not taken up. Industrial hydraulics and propulsionsystems are not included in this work. The thesis is also limited to thehydraulic system; the combustion engine powering the hydraulic pumpis therefore not included. The eld of digital hydraulics is not includedin this thesis.

1.4 Contribution

The most important contribution of this thesis is a deeper understandingof the dynamic characteristics of ow control systems. Novel ways ofdesigning and controlling the directional valves in order to optimize thedamping are proposed and demonstrated. Energy measurements, whereow control and load sensing systems are compared in a wheel loaderapplication, are performed analytically and veried by experiments.

2

2Mobile Working

Hydraulic Systems

Mobile hydraulic applications distinguish themselves from other hy-draulic applications, such as industrial hydraulics, because the pressureand ow demand varies greatly over time and between dierent func-tions. Unlike other hydraulic applications, several functions are oftensupplied by one single pump. This means that the total installed poweron the actuator side is generally considerably higher than the installedpump power. This is possible because the actuators almost never requiretheir maximum power at the same time.

Fluid power systems have successfully been used in mobile machinesfor several decades. Because of the machines versatility, dierent hy-draulic systems have been developed for dierent applications. Impor-tant properties for hydraulic systems are energy eciency, dynamic char-acteristics and complexity. However, the order of these properties variesfor dierent applications. The following sections give an overview ofthe most commonly used working hydraulic systems of today. It alsopresents some interesting system designs that have not yet been com-mercialized but are attracting considerable attention both in industry aswell as academia. Energy eciency, dynamic characteristics and systemcomplexity are discussed and compared.

3


2.1 Valve Controlled Systems

The most common hydraulic systems in mobile machines are systemsbased on valve control. Common to these systems is that they can besupplied by one single pump, which gives a cost eective and compactsystem solution. Four dierent hydraulic system designs are here cate-gorized by open-centre, constant pressure, load sensing and individualmetering.

2.1.1 Open-centre

Today, most hydraulic systems in mobile machines are of the open-centretype. In such systems, the directional valves are designed so that theentire pump ow is directed to tank when no valve is activated. Thisis commonly achieved by providing the directional valve with a chan-nel in the centre position connecting the pump port and the tank, seegure 2.1a. By means of this open-centre channel, the system pressureis kept at a low level while the system is idle and the valve is closed.These systems are designed for use with xed displacement pumps andare therefore often called constant ow systems.

(a) Simplied schematic of anopen-centre system.

ow

useful wasted powerpre

ssure

power

load demand

system operation point

(b) Pressure and ow characteristicsin an open-centre system.

Figure 2.1 Schematic and eciency of an open-centre system.

4

Mobile Working Hydraulic Systems

When a valve is shifted from its centre position, the open-centre chan-nel begins to close and the pump pressure increases. Simultaneously,the pump port is connected to either of the load ports, depending onthe direction of spool movement, while the other load port is connectedto tank. When the pump ow is restricted so that the pump pressureis higher than the load pressure, the check valve opens and there willbe a ow to the load. The rate of this ow is thus not only dependenton spool displacement, but also on load pressure. This is called loaddependency.

If several valves are activated simultaneously, the ow to each ac-tuator will not only be dependent on its own load, but also on otheractivated loads. This means that the pressure level at one load canheavily inuence the speed of another actuator, a phenomenon calledload interaction.

Another disadvantage of open-centre systems is that the ow is loaddependent. For heavy loads, the major part of the lever stroke is used torestrict the pump ow in order to obtain a high pump pressure. Only aminor part of the stroke is then left for controlling the speed. This mightbe a serious problem if a heavy load is be positioned with accuracy, asis often the case for instance with mobile cranes.

The fact that the ow is load dependent is from a dynamic point ofview actually an advantage. It gives the system a naturally high damp-ing, which means that the system is less prone to oscillations. To obtaindamping from a valve, the ow has to increase when the pressure dropacross the valve increases and vice versa. Damping is a preferred prop-erty when handling large inertia loads, for example the swing functionof a mobile crane.

The most important disadvantage of open-centre systems is that itmay have poor energy eciency. High energy losses accur when liftingheavy loads slowly; the pump pressure needs to be high but only a minorpart of the ow is directed to the load, see gure 2.1b. Most of the owis then directed through the open-centre channel to the tank with a highpressure drop, resulting in high energy losses.

To summarize, open-centre systems have the following advantages anddisadvantages:

Advantages The system is simple and robust. It has high damping,which makes it suitable for heavy mobile applications.

Disadvantages Poor overall eciency and interaction between simul-

5


taneously operated loads. The actuator velocity does not corre-spond to a specic lever displacement but is also a function of theload pressure.

2.1.2 Constant Pressure

A constant pressure system can be realized using a pressure controlledvariable displacement pump or a xed pump working against a pressurerelief valve. In this section, the pressure controlled pump solution willbe discussed because of its higher eciency, see gure 2.2a. When thesystem is idle, each directional valve has a closed pump port and thevariable pump is de-stroked to a small displacement, compensating forits own losses and thus keeping the pressure constant. The directionalvalves are of closed centre type.

(a) Simplied schematic of aconstant pressure system.

useful power

wasted power

ow

pre

ssure

load demand


(b) Pressure and ow characteristicsin a constant pressure system.

Figure 2.2 Schematic and eciency of a constant pressure system.

There is a ow to the actuator when its directional valve is shiftedfrom neutral position. Simultaneously, the pump controller increases itsdisplacement in order to maintain a constant system pressure. The owrate is dependent on both spool displacement and load pressure. Conse-quently, constant pressure systems suer from load dependency. How-ever, the controllability of these systems is better than in open-centre

6


systems as far as interaction between actuators is concerned. This is be-cause there is no dependency between the load pressure and the pumppressure. From a dynamic point of view, constant pressure systems havesimilar characteristics to open-centre systems due to their load depen-dency. The damping is therefore high.

Regarding energy eciency, constant pressure systems are a goodchoice if the present loads tend to be constant. The pump pressureis then matched against the mentioned constant load. However, if theload situation alters, high losses might occur. This is especially truewhen raising a light load with a high velocity, see gure 2.2b. The mainpart of the entire pressure drop then occurs across the directional valveand only a minor part is used to lift the load. The major fraction of thetotal power is therefore spent in heating the oil. Consequently, these sys-tems not only have large energy losses but also often need extra energyto cool the oil.

Advantages No interaction between simultaneously operated loads anda high damping.

Disadvantages Poor eciency for light loads and the actuator velocitydoes not correspond to a specic lever displacement but is also afunction of the load pressure.

2.1.3 Load Sensing

Load sensing systems use a variable displacement pump and closed cen-tre valves, similar to constant pressure systems. However, the pump con-troller is designed in a dierent way. Instead of maintaining a constantpressure, the pump pressure is continuously adapted according to thehighest load, see gure 2.3. Another load sensing system design wouldbe to use a xed displacement pump and a pressure relief valve, adapt-ing its cracking pressure according to the highest load. That solution,however, is not discussed in this thesis because of its lower eciency.An early review of load sensing systems was made by Andersson in [7].

When all directional valves are closed, the pump is de-stroked, main-taining a low system pressure. When a valve is shifted from neutralposition, the pump controller senses the load and increases its pressure,thereby allowing a ow to the actuator. Since the pump pressure con-tinuously adapts to the load, a specic lever displacement results in acertain ow, independent of the load pressure.

7


Figure 2.3 Simplied schematic of a load sensing system.

Load sensing systems have no load dependency as long as only oneload is controlled. However, when several loads are operated simultane-ously, only the heaviest load will be load independent. All lighter loadswill suer from both load dependency and load interaction. In applica-tions where controllability is an important feature, the valves are oftenequipped with pressure compensators. The pressure drop across eachdirectional valve is then kept at a constant level and all functions arethereby load independent and there will be no load interaction. Pressurecompensators are studied in detail in section 3.1.

One weakness of load sensing systems using pressure compensatedvalves is the hydraulic damping. The primary design endeavours toachieve low inuence on the ow from the load pressure. This decreasesthe damping capability of the valve. The dynamics of load sensing sys-tems are studied in more detail in chapter 4.

Load sensing systems have high energy eciency since the pump con-tinuously adapts its pressure just above the highest load. A pressuredierence, usually around 20-30 bar, between pump and load is necess-ary to overcome losses in hoses and valves. This pressure margin is oftenset substantially higher than necessary to ensure it is high enough at alloperational points. More details regarding the pressure margin can befound in section 3.3. When several functions are operated simultane-

8


ously, high losses might occur at lighter loads. An example is when alight load is operated with a high velocity and a heavier load is activatedat the same time, see gure 2.4.

pre

ssure

ow

useful powerwasted power

useful power

pump pressure margin

load 1

load 2

system operationpoint

Figure 2.4 Pressure and ow characteristics in a load sensing system.

To summarize, pressure compensated load sensing systems have thefollowing advantages and disadvantages:

Advantages Energy eciency is high although pressure and ow de-mands vary greatly over time and between dierent functions. Thesystem has excellent controllability since there is no load interac-tion and no load dependency.

Disadvantages Low damping, meaning that the system can show anoscillatory behaviour in certain points of operation. High losses atlighter loads when several functions are operated simultaneously.A needless pressure loss in most points of operation due to anexcessive pressure margin.

2.1.4 Individual Metering

A step forward from load sensing systems using conventional spool valvesis to decouple the inlet and outlet orices in the valve, see gure 2.5.Numerous congurations for individual metering systems have been de-veloped, both in academia as well as in industry [8]. These conceptsprovide a higher degree of freedom as all four orices are separated

9


and can be controlled individually. The main benet of this increasedfreedom is that the ow paths can be changed during operation. Fourdierent operational cases can be identied [9], see gure 2.7.

Figure 2.5 Simplied schematic of an individual metering system.

Normal operation The load is operated as in a conventional system;oil is withdrawn from the pump and the return oil is fed to tank. Inconventional systems, the outlet orice opening area is determinedby the spool position. For an independent metering system, theoutlet orice can be separately controlled with the objective of, forexample, reducing metering losses.

Regenerative operation When operating a light and a heavy load si-multaneously, it is possible to use the cylinder as a discrete trans-former. This is done by connecting both load ports to the pumpline, thereby increasing the pressure level and decreasing the owlevel. As can be seen in gure 2.6, this might result in substantiallylower power losses.

Energy neutral operation This mode is benecial when, for exam-ple, lowering a load while the system pressure level is high. Insteadof taking high pressure oil from the pump and throttling it acrossthe control valve, the oil could instead be withdrawn from tank.

10


No power is then needed from the system pump. The return oil isfed to tank.

Recuperative operation This mode is similar to energy neutral op-eration. However, instead of feeding the return oil to tank, it isdirected into the pump line. The cylinder thus works as a pumpand can be used to drive other loads or operate the system pumpas a motor.

pre

ssure

ow

useful powerload 1

load 2

wasted power

useful power


Figure 2.6 It is possible to reduce the losses when using the regenera-tive operation mode in individual metering systems.

Advantages Possibility to optimize the eciency by means of changingthe ow paths. Recuperation of energy is possible. Active dampingmeasures are easy to implement because of increased exibility.

Disadvantages Complex controller and often sensor dependent for owpath selection.

11


(a) Normal operation, the load isoperated as in a conventional sys-tem.

(b) Regenerative operation, bothload ports are connected to thepump with the objective to even outpressure dierences between loads.

(c) Energy neutral operation, a loadcan be lowered without any powerfrom the pump.

(d) Recuperative operation, thecylinder works as a pump enablingenergy recuperation.

Figure 2.7 Flow paths for the dierent operational cases in individualmetering systems.

12


2.2 Valveless Systems

One hot research topic in the area of mobile hydraulics is systems inwhich the control valves are eliminated along with the metering losses,see gure 2.8. Three interesting concepts are here categorised by hy-draulic transformers, pump controlled actuators and electro hydraulicactuators. Such systems are not yet common commercially in mobileapplications but can be found in, for example, the aerospace indus-try [10].

pre

ssure

ow

useful power

useful power

load 1

load 2

Figure 2.8 Pressure and ow characteristics in a valveless system. Allmetering losses are ideally eliminated.

2.2.1 Secondary Control using Transformers

In a secondary control system, all actuators are connected to a commonpressure rail. This can be realized, for example, by a pressure controlledpump. An accumulator is often connected to the common pressure rail,allowing the pump to be downsized according to the mean ow over aduty cycle. It also allows the possibility of storing recuperated powergenerated by the load [11]. Controlling hydraulic motors is the mostcommon use of secondary control [12]. This technology can, however,not be used directly on linear cylinder drives since the piston area isxed. In that case, a hydraulic transformer is required, see gure 2.9.

A hydraulic transformer converts an input ow at a certain pressurelevel to a dierent output ow at the expense of a change in pressure

13


level, ideally maintaining the hydraulic power. One way of realizing atransformer is to combine two hydraulic machines, where at least one hasa variable displacement. However, the eciency is limited, mainly be-cause at least one of the machines will operate under partial loading [13].In recent years, an innovative transformer concept has been developedby the Dutch company Innas BV [14]. The conventional transformerwith 2 hydraulic machines has been replaced by one axial piston unit,thereby avoiding partial loading conditions. A mean eciency of 93%in a broad region of operation has been reported [5].

Figure 2.9 Simplied schematic of a constant pressure systemequipped with transformers and an accumulator, enabling energy recu-

peration.

Advantages High eciency, even when several loads are operated si-multaneously. Possibility to recuperate energy from the loads.

Disadvantages The number of hydraulic machines is increased, whichmeans more required space and a higher cost. Low damping sincethe ow is load independent.

2.2.2 Pump Controlled Actuators

Instead of using one pump to supply all actuators, each actuator has adedicated pump in pump controlled actuator systems. To control the

14


speed, the pump displacement setting is used as the nal control element.All losses are thereby ideally eliminated. In reality, the losses are heavilydependent on the eciency of the system pumps. Pump controlled ac-tuator systems can principally be dierentiated into two dierent circuitlayouts, either with the pump arranged in a closed circuit [15] [16] or inan open circuit [17], see gure 2.10.

Since all actuators have their own dedicated pump, each has to besized to handle maximum speed. A typical example of a dimensioningmotion is the lowering bucket motion in a wheel loader. The loweringow can be several times higher than the maximum pump ow in asimilar valve controlled system. The dierence is that all ow has to behandled by the pump in pump controlled actuator systems. In singlepump systems, the pump can also be downsized since not every loadis actuated at full speed simultaneously very often. For these reasons,the total installed displacement tends to be high in pump controlledactuator systems.

Figure 2.10 Simplied schematic of a pump controlled actuator sys-tem. It can be realized in both closed- and open circuit.

Advantages High eciency, even when several loads are operated si-multaneously.

Disadvantages One hydraulic machine for each actuator, which meansmore required space and a higher cost. Low damping since the owis load independent.

2.2.3 Electro Hydraulic Actuators

The main component in electro hydraulic actuator systems, often re-ferred to as eha, is a xed displacement bidirectional hydraulic pump.

15


An electric motor is usually used to power the pump, enabling activecontrol of the rotational speed and thereby the ow to the actuator,see gure 2.11. A conventional eha requires a symmetrical actuator inorder to ensure ow balance. In that case, no additional oil reservoirs orcontrol valves are needed [10]. However, the pump in mobile applicationsis usually powered by and mechanically coupled to an internal combus-tion engine. Moreover, asymmetrical cylinders are predominantly usedin mobile applications. Some modications of the conventional eha lay-out are therefore needed. Solutions for handling asymmetrical cylindershave been proposed [18].

Figure 2.11 Simplied schematic of an electro hydraulic actuator sys-tem. An electric motor is controlling the rotational speed of a xed bidi-

rectional pump.

In eha systems, there are no metering losses and the pump only oper-ates when control action is needed. One disadvantage of using a speedcontrolled xed displacement pump is that the volumetric eciency is of-ten compromised at low pump speeds. However, the overall eciency ineha systems is still higher than in, for example, pump controlled actua-tor systems due to the pumps low eciency at small displacements [19].The eha technology has become well-established in the aerospace indus-try due to its eciency and compactness.

Advantages High energy eciency and does not require a variablepump.

Disadvantages Requires one hydraulic machine and an electric motorfor each actuator. Low damping since the ow is load independent.

16


2.3 System Summary

The systems described in this chapter are all used in dierent appli-cations. Open-centre and constant pressure systems can be consideredrather simple and often inecient system layouts. In such systems, thecomponent costs themselves are often important and eciency might beof less importance. Load sensing systems are often a good compromisebetween eciency and complexity, but suer from poor dynamic charac-teristics due to their closed loop pressure control. More details regardingthe dynamics of load sensing systems can be found in chapter 4.

When more than one load is actuated, often only the heaviest loadcan be operated eciently in single pump systems. This issue is solvedin valveless systems. When all loads have their own dedicated pump,the pressure can always be matched against the present load. It canbe realized with transformers, pump controlled actuators or eha, whereeha is the most ecient solution. However, one has to bear in mind thatvalveless systems might require several valves to handle, for example,asymmetrical cylinder actuation and safety requirements [17] [19].

This thesis proposes a system design that can be placed somewherebetween load sensing and pump controlled actuators, see gure 2.12. Itis called ow control system. It is similar to pump controlled actuatorsbecause the pump displacement setting is used to control the speed ofthe actuators. However, only one system pump is needed and it usessimilar valves to load sensing systems. A complete description of owcontrol systems is given in the following chapters.

17


Complexity

Ener

gye

cien

cy

Open-centre

Constant pressure

Load sensing

Traditional

Pump controlled actuators

Independent metering

Flow control

Hydraulic transformers

eha

Intelligent

Present

and

per

form

ance

hydraulics

hydraulics

Figure 2.12 The system design analysed in this theses, ow controlsystem, can be placed somewhere between load sensing and pump con-

trolled actuators in terms of energy eciency and performance.

18

3The Flow Control

Concept

In mobile hydraulic systems, the actuation of dierent loads is controlledby joystick signals. These signals pose either a ow or pressure demandfrom the operator. In applications where velocity control is important,the signals from the operator often correspond to ow demands. Anexample is load sensing systems equipped with pressure compensators.The compensators maintain a constant pressure drop across the direc-tional valves, which make the signals from the operator correspond toow demands. Nevertheless, the pump in these kinds of systems is stilloften pressure controlled.

In systems where the operators signals correspond to ow demands, itseems more natural to also control the pump by ow. This approach hassome benets regarding energy eciency, dynamic characteristics andincreased exibility compared to load sensing systems. It also presentssome challenges, for example the compensator design.

The idea of ow control is to use the joystick signals to control thepump ow and the valve opening simultaneously. This means that thepump software needs information about the ow demands for dierentfunctions. The pump displacement setting is controlled according to thesum of all requested load ows.

When no function is activated, the pump is de-stroked, delivering noow to the system, and all directional valves are closed. Activating ajoystick will simultaneously open a valve and increase the displacementof the pump. Pressure is built up in the pump hose and when the pumppressure becomes higher than the load pressure there will be a ow to

19


the actuator. When stationary, the ow delivered by the pump will go tothe load. The pump pressure will therefore adapt itself to a level neededby the system. This results in eciency improvements compared to loadsensing systems, which are described in more detail in section 3.3.

Figure 3.1 Simplied schematic of a ow control system. The pumpdisplacement setting and the valve openings are controlled simultaneously

by the operators joystick signals.

Flow control systems will suer from load dependency if more thanone load is activated simultaneously. This can be solved by introducingsensors into the system. Zhe [20] used the velocities of the actuatorsas the main feedback signals for pump and valve control. Jongebloed etal. [21] used pressure sensors at all load ports for the valve control. Tooptimize energy eciency, the valve at the highest load can be opened toits maximum while lighter loads are controlled by their valve openings.

Load dependency can also be solved by using pressure compensators.Since the pump is ow controlled, there will be dierent demands on thecompensator functionality compared to load sensing systems. However,it also opens up new possibilities regarding the control of the directionalvalves. Details regarding the compensator requirements and the controlapproaches are described in sections 3.1 and 3.2.

Flow control systems have many similarities with load sensing systems.Except for the pump controller, the two systems are almost equivalent.

20

The Flow Control Concept

The pump controller used in ow control systems could also be usedin pump controlled actuators. All components needed in ow controlsystems are therefore available on the market [22].

3.1 Pressure Compensators

In some mobile uid power applications, load dependency and load in-teraction are undesired system characteristics. An example is forestrymachines, where the operator wants to position the load with accuracy.Pressure compensators are commonly used in these kinds of applicationsto ensure good handling capabilities. Two dierent types of compen-sators can be realized, which are explained in section 3.1.1 and 3.1.2.In applications with less demand for accuracy, it is also possible to takeadvantage of ow forces for the pressure compensation functionality.

3.1.1 Traditional Compensators

The most common design is to place the compensator upstream of thedirectional valve. The reduced pressure is then working against the loadpressure and a preloaded spring, see gure 3.2a. The force equilibriumfor the compensator, equation (3.1), together with the ow equationgives the ow across the directional valve. According to equation (3.2),the compensator spring force sets the pressure drop across the directionalvalve, making the ow load independent.

Fs + Ac1pL = Ac1pr pr pL =FsAc1

(3.1)

qL

= CqAs

2

(pr pL) = CqAs

2

(FsAc1

)(3.2)

It is also possible to achieve the same functionality by placing thecompensator downstream of the directional valve. In that case, thesupply pressure is working against the reduced pressure and a springaccording to gure 3.2b. The force equilibrium, equation (3.3), togetherwith the ow equation gives the same result, equation (3.2) comparedwith (3.4).

Fs + Ac1pr = Ac1ps ps pr =FsAc1

(3.3)

21


ps pr pL

Ac1

Ac1Fs

As qL

(a) The compensator is placed up-stream of the directional valve.

ps

pr

pL

Ac1

Ac1Fs

As qL

(b) The compensator is placeddownstream of the directionalvalve.

Figure 3.2 Two dierent ways of realizing a traditional pressure com-pensator. The pressure drop across the directional valve is set by the

compensator spring force.

qL

= CqAs

2

(ps pr) = CqAs

2

(FsAc1

)(3.4)

These types of compensators are designed for use with a pressure con-trolled pump. In case of the pump being saturated, the supply pressurewill drop, resulting in the compensator spool at the heaviest load open-ing completely. That function will lose speed and possibly even stop.Functions operated simultaneously at lower pressure levels will, however,move normally.

3.1.2 Flow Sharing Compensators

Another design is to implicate the highest load pressure into the compen-sator. When the pressure is actively controlled, this design is equivalentto the traditional compensator design. However, its characteristics aredierent when the pump is saturated. All functions will then be giventhe same priority, which means that all functions will decrease in speed.This ow sharing functionality can be achieved by placing a compensatoreither downstream or upstream of the directional valve.

In case of the compensator being located downstream of the direc-tional valve, the reduced pressure is working against the highest loadpressure and a spring, see equation (3.5) and gure 3.3a. The pumppressure margin is dened according to equation (3.6) and the ow canbe calculated according to equation (3.7).

Ac1pr = Ac1pLmax + Fs pr = pLmax +FsAc1

(3.5)

22


pp = ps pLmax (3.6)

qL

= CqAs

2

(ps pr) = CqAs

2

(pp Fs

Ac1

)(3.7)

ps pr pL

Ac1

Ac1Fs

As qL

pLmax

(a) The compensator is placeddownstream of the directionalvalve.

ps

pr pL

Ac1

Ac1

As qL

pLmax

Ac2

Ac2Fs

(b) The compensator is placed up-stream of the directional valve.

Figure 3.3 Two dierent ways of realizing a ow sharing pressure com-pensator. The pressure drop across the directional valve is set by the pump

pressure margin.

The ow sharing pressure compensator placed upstream of the direc-tional valve is similar to its traditional equivalent. Instead of a spring,two pressure signals that constitute the pump pressure margin are act-ing on the compensator, see gure 3.3b. Equation (3.6) together withthe force equilibrium for the compensator, equation (3.8), gives the owaccording to equation (3.9). The spring in this type of compensator isnot required for the functionality. It can rather be used as a designparameter for, for example, prioritization [23].

Ac2ps + Ac1pL = Ac2pLmax + Ac1pr + Fs

(pr pL) =Ac2Ac1

(ps pLmax ) FsAc1

(3.8)

qL

= CqAs

2

(pr pL) = CqAs

2

(Ac2Ac1

pp FsAc1

)(3.9)

Flow sharing pressure compensators will distribute the entire pumpow relative to the individual valve openings also when the pump issaturated. A pressure controlled pump which has been saturated cannot

23


control the pressure and can therefore be seen as a ow controlled pump.These compensators are therefore appropriate to use together with a owcontrolled pump.

3.2 Pump and Valve Control Approaches

In ow control systems, the operators joystick signals control the pumpow and the valve opening simultaneously. For this to work properly,the system software needs knowledge about every ow consumer in thesystem. However, solutions for attaching auxiliary functions withoutknowledge about their ow demand have been presented in [24] and [25].Dierent control approaches are possible depending on whether tradi-tional compensators or ow sharing compensators are used.

3.2.1 Flow Control using Traditional Compensators

When all directional valves are closed, the pump is ideally de-strokedto zero, delivering no ow to the system. When the operator movesthe joystick, signals are sent to the pump and the valve simultaneously.The valve is shifted from neutral position and the pump starts to deliverow. Since the valve is traditionally pressure compensated, the springforce sets the pressure drop across the directional valve, and thereby theabsolute ow level that the valve is expecting, see gure 3.4. When thepump is delivering ow, pressure is built up in the hose connecting thepump and the valve. There will be a ow to the load when the pumppressure is higher than the load pressure. This works ne as long as theow sent by the pump equals the ow expected by the valve. If this isnot the case, two situations may occur.

The pump ow is too low This is the same case as when the pump issaturated in a load sensing system. The consequences will be thatthe compensator spool at the highest load will open completely,resulting in a decrease in speed for that load. It will possibly evenstop.

The pump ow is too high Both compensator spools will close moreand the pump pressure will increase until the system relief valveopens. The throttle losses will be huge and the system will emergeas a constant pressure system.

24


Figure 3.4 Simplied schematic of a ow control system using tra-ditional pressure compensators. The system can also be realized with

traditional compensators placed downstream of the directional valves.

The reason for this is that traditional pressure compensators controlthe absolute ow across the directional valve by reducing the pumppressure relative to the load pressure of its own load. This works neas long as the pump pressure is actively controlled, with for instance aload pressure feedback. Otherwise, the ow situation in the system isover-determined.

A lot of research solving this ow matching problem has been pre-sented. Djurovic and Helduser [26] introduce a position sensor placedon the directional valve. It allows precise knowledge of the ow expectedby the valve. It is also possible to equip the compensator with a positionsensor [27]. If no compensator is close to fully opened, the pump ow istoo high. In case of the pump ow being too low, the compensator atthe highest load would be completely opened. A bleed-o valve to tankis proposed by several authors [24] [26] [27]. A small overow is then ac-ceptable, which could be used in closed loop control if a position sensoris added. Fedde and Harms [28] discuss the pros and cons with overow

25


and underow when using a bleed-o valve. Grsbrink et al. [29] [30]propose a system design where the pump is pressure controlled for lowpump ows and ow controlled for high ow rates. It is also possible toshift from ow control to pressure control in case of an undesirable press-ure build up [31]. A review of solutions to the ow matching problem inow control systems using traditional compensators has been made byDjurovic in [32].

3.2.2 Flow Control using Flow Sharing Compensators

There are alternatives to address this ow matching problem withoutadding additional components or sensors to the system. The key is toimplicate the highest load pressure into the compensator and thus getthe ow sharing behaviour described in section 3.1.2. The compensatorsthan act as relief valves instead of reducing valves and all valve sectionswill work against the highest load pressure, see gur 3.5. This has beenstudied in, for example, [22] and [33].

Figure 3.5 Simplied schematic of a ow control system using owsharing pressure compensators. The system can also be realized with ow

sharing compensators placed downstream of the directional valves.

26


Using a ow controlled pump in combination with ow sharing press-ure compensators opens up new possibilities in terms of controlling thedirectional valves independently of the cylinder velocity. This can beexplained with a small example. Imagine that two loads are active, therst with 50% of the maximum velocity and the second with 25% ofthe maximum velocity. The directional valves have an opening area of50% and 25% and the ow delivered by the pump is constant. Bothdirectional valve openings are now increased, the rst to 100% and thesecond to 50%. The pump ow is still the same. Since the ow sharingpressure compensators will distribute the entire pump ow relative tothe individual valve openings, the velocities will be unchanged. Whathappens is that the absolute pressure drop across both directional valveshas been reduced, see gure 3.6.

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

Pump flowVelocity, load 1Velocity, load 2

Flow

and

velo

city

[-]

Time [-]

(a) The pump ow and both actu-ator velocities are constant.

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

Pressure dropOpening area, load 1Opening area, load 2

Time [-]

Pre

ssure

dro

pan

dop

enin

gar

ea[-]

(b) The pressure drop across the di-rectional valves will decrease whenthe opening areas are increased.

Figure 3.6 Flow sharing system characteristics. Both directional valveopening areas are increased without aecting the actuator velocities. The

pressure drop across both directional valves will decrease.

This system characteristic is dierent from most other valve controlledsystems. Instead of controlling the ow, the valves will serve as owdividers. One control approach is to open the valve section at the loadwith the highest ow demand to its maximum [34] [35]. Other activefunctions must always be opened in proportion to its ow request. Thisapproach will minimize the pressure drop across the directional valvesand thus save energy. This is further discussed in section 3.3.

Another control approach might be to use the valves to increase thedamping of the system. There is an optimal valve opening where the

27


damping is maximized. For example, when a function is oscillating thevalve opening could be reduced temporarily in order to dampen theoscillations. When no oscillations are present, a more energy ecientcontrol approach can be used. This is further discussed in section 4.3.1.

3.3 Energy Eciency

The energy eciency of ow control systems is similar to load sensingsystems. The pump pressure is adjusted according to the highest loadand high losses might occur when loads with dierent pressure demandsare operated simultaneously. However, instead of a prescribed pressuremargin, as in load sensing systems, the pressure drop between pump andload is given by the resistance in the hoses and in the valves. Further-more, it is also possible to lower the pressure drop across the directionalvalve by means of a more energy ecient control strategy.

In load sensing systems, the pump pressure margin is set to overcomethe losses in the pump hose, the compensator and the directional valve.These losses are system dependent and will change with internal andexternal conditions such as temperature, oil properties, hose length, etc.The pressure margin is set according to the worst case to ensure it ishigh enough at all operating points.

The pressure drop between pump and load can be divided into threedierent losses:

Losses between pump and valve There will be a pressure drop be-tween the pump and the valve. The magnitude will depend onthe internal and external properties mentioned above, but mostimportantly the ow rate. A simplied model is that the lossesincrease with the square of the ow rate.

Losses across the compensator There will be a pressure drop acrossthe compensator. High losses occur if the supply pressure is muchhigher than the load pressure. This is the case at partial loadingconditions. The smallest possible loss occurs when the compen-sator is fully opened. In that case, the required pressure dropincreases with the square of the ow rate.

Losses across the directional valve Typically, the compensatormakes sure that the pressure drop across the directional valveis constant. However, the smallest possible pressure drop occurs

28


if the valve is fully open. The pressure drop will then follow theow equation, similar to the compensator pressure drop.

In gure 3.7a, these three dierent losses are shown. If the pressuremargin is set perfectly, there would be no unnecessary losses at maxi-mum ow rate in load sensing systems. However, at lower ow rates,unnecessary losses will occur. In ow control systems, these losses willbe eliminated since the pump pressure is set by the resistance in thehose and the valve.

It is possible to further reduce the losses in ow control systems. Thisis done by opening the valve section with the highest ow demand to itsmaximum, as described in section 3.2.2, in which case the pressure dropacross the directional valve is minimized and additional energy savingsare possible, see gure 3.7b.

A ow control system without pressure compensators would increasethe eciency even further. In that case, the valve section at the highestload pressure might be opened completely. However, its functionalityrequires closed loop control and is therefore sensor dependent [21].

Pum

ppre

ssure

marg

in[-]

Flow [-]

unneces

saryloss

es

directional valve losses

hose lo

sses

com

pen

sato

rlo

sses

(a) The pump pressure marginis xed in load sensing systems.Therefore, unnecessary losses occurat lower ow rates.

Pum

ppre

ssure

marg

in[-]

Flow [-]

ecienc

y improv

ements

fully opened directional valve

hose lo

sses

com

pen

sato

rlo

sses

(b) The pump pressure margin isgiven by the system resistances inow control systems. Eciency im-provements are therefore possible.

Figure 3.7 Classication of the losses between pump and load. Threedierent losses occur; hose, compensator and directional valve losses. At

lower ow rates, unnecessary losses occur in load sensing systems. No

unnecessary losses occur in ow control systems.

As can be seen in gure 3.7, the two system layouts have the same e-ciency at maximum ow rate if the pump pressure margin is set perfectlyin the load sensing system. Flow control systems have higher eciencyfor smaller ow rates. However, it is important to consider the power

29


losses rather than the pressure losses. For low ow rates, the power losswill be small even for high pressure drops. Figure 3.8 shows the powersaving opportunities for ow control systems. The largest power savingsoccur in the medium ow rate area. If the directional valve is openedcompletely, even more power can be saved.

Pow

er[-]

Flow [-]

fully openeddirectional valve

power savings

Figure 3.8 Power savings in ow control systems compared to loadsensing systems. More power can be saved if the directional valve is com-

pletely opened. No power is saved at maximum ow rate.

Flow control systems have no unnecessary losses for the highest load.All losses that occur are necessary and limited by, for example, the diam-eter of the hoses and the maximum opening areas in the valve. However,ow control systems still have high losses at partial loading conditions.To increase eciency even further, individual metering valves or addi-tional hydraulic machines are required.

A ow control system with two hydraulic pumps has been studiedin [36] and [37]. The aim is to reduce the losses at partial loadingconditions without increasing the total installed displacement. This isachieved by connecting the two pumps when high ow rates are requiredby one load. Connecting several pumps at high ow rates is a commonsolution for more simple systems, for example, in excavators.

30

4Dynamic Analysis

The dynamic analyses in this thesis were made to show the fundamentaldierences between load sensing systems and ow control systems. Lin-ear models are used and dierent types of compensators are consideredin the analysis. The only dierence between the load sensing systemmodel and the ow control system model is the absence of the feed-back to the pump controller in the ow control system, see gures 4.1and 4.2. Nevertheless, there are fundamental dynamic dierences be-tween the two system layouts.

Qa AcVa,Pa

Vb,Pb

Qb

mL

U

Kcb

Qp

Vp,Pp

GpLS

Kca

Ppref

Xp

Figure 4.1 Dynamic load sensing system model.

31


Qa AcVa,Pa

Vb,Pb

Qb

mL

U

Kcb

Qp

Vp,Pp

GpFC

Kca

Qpref

Xp

Figure 4.2 Dynamic ow control system model.

4.1 Mathematical Model

A linear mathematical model is constructed to perform the dynamicanalyses. The derivation of the equations is shown in [38].

The pump controller can be described in two dierent ways. In loadsensing systems, the controller consists of a pressure controlled valvethat controls the displacement piston. If the pressure balance, Pp =Pp Pa, is disturbed, the valve is displaced and the pump setting is thenproportional to the integrated valve ow. Here, the pump is modelledas a pure inductance, see equation (4.1).

GpLS =Qp

Ppref Pp=

1

Lps(4.1)

The pump controller in ow control systems controls the displacement,and thereby the ow, directly instead of maintaining a certain pressuremargin above the highest load pressure. Such a pump controller has noexternal feedback from the system, similar to the load sensing feedback.Here, the transfer function describing the displacement controlled pumpdynamics is called GpFC , see equation (4.2).

GpFC =Qp

Qpref(4.2)

The continuity equation of the pump volume yields the transfer func-tion in equation (4.3).

Hs =Pp

Qp Qa =eVps

(4.3)

32

Dynamic Analysis

The model for the inlet orice in the directional valve will be dierentdepending on the compensator design. A non-compensated valve willhave a ow-pressure dependency according to equation (4.4). In thisanalysis, the valve is considered to be much faster than the rest of thesystem. The valve dynamics is therefore ignored. The dynamics ofpressure compensated valves have been studied in, for example, [39]and [40].

Gva =Qa

Pp Pa = Kca (4.4)

A traditionally compensated valve will have no ow-pressure depen-dency since the pressure drop across the directional valve is constant,see equation (4.5).

Gva =Qa

Pp Pa = 0 (4.5)

A ow sharing pressure compensated valve will have a ow-pressuredependency, similar to a non-compensated valve, for the highest load.Lighter loads have no ow-pressure dependency, like traditional com-pensated valves. However, lighter loads will be disturbed by the highestload due to cross-coupling of the highest load pressure to all compen-sators [41].

Gva =Qa

Pp Pa = Kca , Pa = Pamax

Gva =Qa

Pp Pa = 0, Pa < Pamax (4.6)

Gvea =Qa

Pp Pamax= Kca , Pa < Pamax

A detailed investigation of valve models using dierent compensationtechniques can be found in [41] and paper [II].

A mass load with a gear ratio is considered to act on a cylinder. Thecontinuity equation for the cylinder chambers together with the forceequilibrium for the piston is shown in equations (4.7), (4.8) and (4.9).

Qa =Vae

sPa + AcsXp (4.7)

U2mLs2Xp + BpsXp = AcPa AcPb (4.8)

AcsXp Qb =Vbe

sPb (4.9)

33


It is also possible to describe a load which consists of a hydraulic motorby similar equations [II].

The outlet orice in the directional valve is considered to have a ow-pressure dependency according to equation (4.10).

Gvb =QbPb

= Kcb (4.10)

4.2 Pump Stability

Due to the absence of the load pressure feedback to the pump con-troller in ow control systems, there is a fundamental dynamic dier-ence between load sensing and ow control systems. To show this, themathematical model in section 4.1 can be simplied. A ow-pressuredependency at the inlet side of the valve is assumed and the outlet ori-ce is ignored. The simplications will not inuence the fundamentaldierences but is important to bear in mind when making other dynamicanalyses.

A transfer function from inlet ow to pressure in the cylinder canbe derived using equations (4.7) and (4.8). Ignoring the outlet oriceresults in a constant pressure on the piston rod side.

ZL

=PaQa

=U2m

Ls + Bp

Vae

U2mLs2 + Vae Bps + A

2c

(4.11)

4.2.1 Load Sensing Systems

The dynamic behaviour of load sensing systems can be described byequations (4.1), (4.3), (4.4) and (4.11). By reducing the block diagramin gure 4.3a, the open loop transfer function from desired pump press-ure margin, Ppref , to actual pressure dierence, Pp = Pp Pa, canbe derived according to equation (4.12). A complete investigation ofload sensing systems and their dynamic properties, including pump con-trollers, can be found in [42].

GpLSGo = GpLSHs

1 + Gva (ZL + Hs)(4.12)

By closing the control loop, the pump controller, GpLS , is a part of theloop gain, GpLSGo, as shown in gure 4.3b. To achieve a stable systemthe loop gain must be kept lower than unity when the phase crosses

34

Dynamic Analysis

+

GpLS+

Hs+

Gva

ZL

1Gva

Pp,ref Qp Pp Pp Qa

Pa

(a) Block diagram of a load sensing system derived from the transferfunctions (4.1) (pump controller), (4.3) (pump volume), (4.4) (inlet valve)and (4.11) (load).

+

GpLS Go Pp,ref Pp

(b) Rearranged block diagram with theloop gain GpLS Go.

Figure 4.3 Linear model of a load sensing system.

-180. On the other hand, it would be feasible to increase the gain ofthe pump and its controller to achieve a system that meets the responserequirements. To achieve a system, with desired response, the gain ofthe pump controller is increased, but at the same time the system isapproaching its stability limit. One should bear in mind that stabilityat one operational point will not guarantee stability at another, seegure 4.4.

4.2.2 Flow Control Systems

The dynamic behaviour of ow control systems can be described byequations (4.2), (4.3), (4.4) and (4.11). This results in almost the sameblock diagram as in gure 4.3. The only dierence is the absence ofthe feedback to the pump controller, see gure 4.5b. This results in afundamental dynamic dierence between load sensing systems and owcontrol systems. Since there is no closed loop for the pump controller,the stability issues described in section 4.2.1 are eliminated. The pumpand its controller can thereby be designed to meet the response require-

35


100 101 102 103 1046

5

4

3

2

1

0

1

100 101 102 103 104

270

180

90

log10

(Gp

LSG

o)

[-]

Phas

e[]

Frequency [rad/s]

mL

increasing

mL

increasing

Figure 4.4 Bode plot of the open loop gain in gure 4.3b, GpLSGo.

Table 4.1 Parameter values used in gure 4.4.

Parameter Value Unity

Ac 0.008 m2

Bp 10000 Ns/mVa 4103 m3Vp 5103 m3Kca 1109 m5/NsLp 5108 Pa s2/m3m

L[6000 12000 30000] kg

U 1 -e 1109 Pa

ments without considering system stability. This has been veried byexperiments in [22] and [34].

36

Dynamic Analysis

GpFC+

Hs+

Gva

ZL

Qpref Qp Pp Pp Qa

Pa

(a) Block diagram of a ow control system derived from the transferfunctions (4.2) (pump controller), (4.3) (pump volume), (4.4) (inletvalve) and (4.11) (load).

GpFC Go

Qpref Pp

(b) Rearranged block diagramwith no feedback present.

Figure 4.5 Linear model of a ow control system.

4.3 Damping

Hydraulic systems by themselves are normally poorly damped and needsome additional damping from the valves to prevent, or at least reduce,the tendency to oscillate. To obtain damping from a valve, the owshould increase when the pressure drop across the valve increases andvice versa. Andersson [43] gives an overview of the valves contributionto damping in mobile hydraulic systems. An overview of active oscil-lation damping of mobile machine structure is given by Rahmfeld andIvantysynova in [44].

Open-centre and constant pressure systems have a high damping asdescribed in section 2.1. Load sensing systems are poorly damped, es-pecially if pressure compensators are used. Valveless systems are ideallyundamped since no valves are present in those kinds of systems.

4.3.1 Active Control of the Inlet Orice

In this section, the damping contribution of the inlet orice in owcontrol systems is analysed. The cylinder friction and the outlet oriceare ignored to simplify the analysis, see gure 4.6. The inlet valve isassumed to have a ow-pressure dependency, which means that it could

37


be a non-compensated valve or a ow sharing valve at the highest loadaccording to equation (4.6).

Qa AcVa,Pa

mL

U

Qp

Vp,PpKca

Xp

Figure 4.6 Dynamic model of a ow control system with a mass load.The outlet orice and the cylinder friction have been ignored.

The system can be described by equations (4.2), (4.3), (4.4) and (4.11).An expression for the ow-pressure coecient of the inlet orice thatgives the highest damping has been derived in paper [II] according toequation (4.13).

Kcaopt =Ac

3/4i

Vp (i 1)eU2mL

(4.13)

where

i= 1 +

VpVa

(4.14)

The maximum damping in the system can be calculated using equa-tion (4.15).

hmax =1

2

(

i 1) (4.15)

Equation (4.15) shows that the maximum damping given by the inletorice only depends on the value of

i, which includes the pump hose

volume and the volume at the inlet side of the cylinder according toequation (4.14). To get a high damping contribution from the inletorice, the pump hose volume should be large compared to the volumeon the inlet side of the cylinder. However, this relationship will change

38

Dynamic Analysis

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Dam

pin

g[-]

Increased opening area

iincreasing

Figure 4.7 System damping as a function of the opening area of theinlet orice. A small value of

igives a low damping regardless of the

opening area. The damping will increase with higher values of i.

during the cylinder stroke. The damping as a function of the inlet oriceopening area for dierent values of

iis shown in gure 4.7.

To get the highest possible damping for a given value of i, the inlet

orice opening area has to be small. During certain points of opera-tion this might result in substantial power losses [II]. To avoid this itis possible to use the more energy ecient control strategy describedin section 3.2.2 while no oscillations are present. When damping is re-quired, the valve can temporarily be closed more to reach the peaks ingure 4.7. Finally, when the oscillations have died out, the energy e-cient control strategy can be applied again. This is possible to do in owcontrol systems without aecting the cylinder velocities if ow sharingpressure compensators are used.

Theoretically, a ow control system using traditional compensatorsobtains no damping from the inlet orice since the ow is independentof pressure changes, see equation (4.5). This is also true for lower loadsusing ow sharing compensators according to equation (4.6). One wayto obtain damping for such loads is to implement active damping, usingfor example a dynamic load pressure feedback.

A special case of this analysis is when the inlet orice opening areaapproaches innity. This is the case in valveless systems, which have no

39


orices at all. As can be seen in gure 4.7, the damping then approacheszero. Consequently, a valveless system is ideally undamped.

4.3.2 Design and Control of the Outlet Orice

In this section, the damping contribution of the outlet orice is analysed.This analysis is not limited to ow control systems, but is valid for allpump controller designs. The prerequisite is that the inlet ow can bemodelled as a perfect ow source, which is true if the inlet orice has noow-pressure dependency, see gure 4.8. This can be realized with, forexample, a traditional pressure compensator.

Pa

Qa

AcVa

Vb

Qb

Pbm

L

U

Kcb

Figure 4.8 Dynamic model of a ow controlled cylinder with a massload and an outlet orice. The pump controller can be of any design; it

does not aect the analysis.

The system can be described by equations (4.7)-(4.10). Similar to theanalysis in section 4.3.1, the viscous friction in the cylinder has beenignored to simplify the analysis. An expression for the ow-pressurecoecient of the outlet orice that gives the highest damping has beenderived in paper [III] according to equation (4.16).

Kcbopt = Ac

Vb

eU2mL (o 1)3/4

o(4.16)

where

o = 1 + 2VaVb

(4.17)

40

Dynamic Analysis

The maximum damping in the system can be calculated using equa-tion (4.18).

hmax =1

2

(o 1

)(4.18)

Equation (4.18) shows that the maximum damping of the system de-pends only on the value of o , which includes the volume at each side ofthe cylinder and the cylinder area ratio according to equation (4.17). Toget a high damping contribution from the outlet orice the volume onthe inlet side of the cylinder should be large compared to the volume onthe outlet side. However, this relationship will change during the cylin-der stroke. A high value of the cylinder area ratio increases the damping,which means that a symmetrical cylinder gives higher damping than anasymmetrical. The damping as a function of the outlet orice openingarea for dierent values of o is shown in gure 4.9.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Dam

pin

g[-]

Increased opening area

o increasing

Figure 4.9 System damping as a function of the opening area of theoutlet orice. Small values of

ogive a low damping regardless of the

opening area. The damping will increase with higher values of o.

A valve design that is suggested in paper [III] is to optimize the damp-ing when the piston is at its lower end position. While the piston movesupwards, the damping will increase. If a higher damping is required, itis possible to design the valve with a smaller orice area. The drawbackswith such a design are, however, that the damping will be slightly lowerat the pistons lower end position and that the losses across the outlet

41


orice will be higher. If lower losses are required it is possible to designthe valve with a larger opening area. However, this is at the expense ofa lower damping. There is no point in designing the valve with a toosmall orice area. The damping will then be low and the losses high.

In case of the inlet and outlet orices being decoupled, as in indi-vidual metering systems, it would be possible to optimize the dampingduring the cylinder stroke. While the piston is moving, the outlet oricecould be controlled in order to achieve the highest possible damping. Itwould also be possible to use a similar control approach as described insection 4.3.1. When no oscillations are present, the outlet orice couldbe fully opened, minimizing the losses. When damping is required, thecontroller could shift to optimize the damping and temporarily allowhigher losses.

42

5Experimental

Results

The energy eciency improvements described in section 3.3 have beenvalidated using a wheel loader application. Also, the theories concerningthe design and control of the outlet orice described in section 4.3.2 havebeen validated in a test rig.

5.1 Energy Eciency Improvements

5.1.1 Hardware Requirements

The hardware requirements in ow control systems are similar to loadsensing systems. To achieve the same system capacity, a pump size ofthe same magnitude is used. Only the pump controller needs to be dif-ferent. Instead of actively maintaining a certain pressure margin abovethe highest load pressure, the pump displacement is controlled directlyfrom the operators demand signals. This requires an electrically con-trolled displacement controller for the pump. However, the load sensinghose to the pump controller can be removed.

Flow control systems use the same type of valves as load sensing sys-tems. Flow sharing pressure compensators are favourable but work-arounds with traditional compensators also exist, see section 3.2.1. Insome valve designs, a traditional compensator placed upstream of thedirectional valve can be replaced with its ow sharing equivalent withouteven replacing the valve housing [23].

Sensors are not required to achieve the desired functionality in ow

43


control systems if pressure compensators are used. However, it wouldbe benecial to use sensors to detect if the cylinder end stops have beenreached. In that case, the valve could be closed and the pump owadjusted to avoid unnecessary energy losses.

5.1.2 A Demonstrator System

To verify the energy eciency improvements in the ow control concept,measurements where performed on a wheel loader application, see g-ure 5.1. The machine was equipped with a pump that can be operatedin both pressure and ow control modes and a valve prepared for usewith both traditional and ow sharing compensators, placed upstreamof the directional valve.

Figure 5.1 The machine used for experiments.

In gure 5.2c, the pump pressure margin for both the load sensing andthe ow control systems can be seen. The measurements agree with thetheoretical pressure margin shown in gure 3.7. The ow sent by thepump is similar in both systems, see gure 5.2a. It can also be observedin gure 5.2b that the pressure is more oscillative in the load sensingsystem. This is because the pump controller operates in a closed loopcontrol mode [42].

A short loading cycle has also been performed to compare load sensingand ow control. Only the working hydraulics have been taken into con-sideration, neither the steering nor the transmission. Figure 5.3a showsthe position of the actuators and gure 5.3b the energy consumption.The energy consumption was reduced by 14% for the ow control systemfor this particular application. This is the same order of magnitude asexperiments performed in [24] and [34].

44

Experimental Results

0 1 2 3 4 5 60

50

100

150

Flow

[l/m

in]

Time [s]

(a) Measured ow for both systems.The ow is increased from zero tomaximum.

0 1 2 3 4 5 60

5

10

15

20

25

30

35

40

45

50

Time [s]

Pre

ssure

[bar]

(b) Measured pump pressure mar-gin for both systems while the owis increased.

0 20 40 60 80 1000

5

10

15

20

25

30

Flow [l/min]

Pre

ssure

[bar]

Load sensing

Flow

contro

l

(c) Measured pump pressure mar-gin as a function of measured ow.Load sensing systems have a con-stant margin while ow control sys-tems have a margin given by the sys-tem resistances.

Figure 5.2 Experimental results showing the potential of reducing thepump pressure margin in ow control systems compared to load sensing

systems.

45


0 5 10 15 20 25 30 350

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Pos

itio

n[m

]

Time [s]

Lift

Tilt

(a) Measured positions of the actu-ators during the cycle.

0 5 10 15 20 25 30 350

20

40

60

80

100

120

140

160

Time [s]

Ener

gy[k

J]

Load sensing

Flow control

(b) Measured energy consumptionduring the cycle.

Figure 5.3 Experimental results showing the actuator positions andthe consumed energy in a short loading cycle. The ow control system

consumed 14% less energy during the cycle compared to the load sensing

system.

46

Experimental Results

5.2 Improved Damping

A test rig to validate the damping contribution by the outlet orice hasbeen constructed. It consists of a traditional pressure compensated valveon the inlet side, a cylinder with a mass load and a servo valve on theoutlet side, see gure 5.4. Dierent designs of the outlet orice can beachieved by controlling the opening area of the servo valve. A constantpressure pump supplies the system. Pressure sensors are attached onthe supply side and on both cylinder chambers. The cylinder and theservo valve are equipped with position sensors. External volumes aremounted on both sides of the piston. By using either one, it is possibleto manipulate the dead volumes on either side of the piston.

Figure 5.4 The experimental test stand. The pressure compensatedvalve can be seen at the lower right and one of the volumes to the left.

In the experiments, a step is made in the ow by opening the inletvalve. Oscillations in the cylinder velocity are then studied. The exper-imental results are presented in gure 5.5. In tests (a) and (b), thereis a large volume on the inlet side which means that a relatively highdamping can be expected. In test (a), the outlet orice area is dimen-sioned close to the maximized damping. As can be seen in gure 5.5a,there are almost no oscillations in the cylinder velocity. In test (b), theoutlet orice area is larger than in test (a) and the damping becomeslower, see gure 5.5b.

In test (c), there is a large volume on the outlet side of the cylinder,which means that the damping is expected to be low. The outlet oricearea is dimensioned close to the maximized damping. Nevertheless, thedamping is still low according to gure 5.5c. This is consistent with the

47


mathematical analysis according to equations (4.17) and (4.18).In test (d), the outlet orice area is so large that it can be equated

with having no outlet orice at all. Theoretically, the hydraulic systemwill not contribute any damping without an outlet orice as shown insection 4.3.2. This is almost the case in the measurements as can be seenin gure 5.5d. The damping that is still obtained is due to secondaryeects ignored in the mathematical analysis, such as friction and leakage.

0 0.5 1 1.5 2 2.5 3

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Veloc

ity

[-]

Time [s]

(a) A high damping is obtainedwhen there is a large volume on theinlet side of the cylinder and theoutlet orice is designed close to itsoptimum.

0 0.5 1 1.5 2 2.5 3

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Veloc

ity

[-]

Time [s]

(b) The damping becomes lowerwhen there is a large volume at theinlet side and the outlet orice areais too large.

0 0.5 1 1.5 2 2.5 3

0

0.5

1

1.5

2

2.5

3

3.5

Veloc

ity

[-]

Time [s]

(c) When there is a large volume atthe outlet side of the cylinder, thedamping becomes low even if theoutlet orice is designed close to itsoptimum.

0 0.5 1 1.5 2 2.5 3

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Veloc

ity

[-]

Time [s]

(d) Without an outlet orice, onlysecondary eects such as friction andleakage will contribute to the damp-ing.

Figure 5.5 Experimental results for dierent designs of the outlet ori-ce.

48

6Summary and

Conclusions

This thesis studies a system design where the pump displacement set-ting is controlled based on the operators command signals rather thanmaintaining a certain pressure margin above the highest load pressure.Conventional load sensing systems are state-of-the-art in industry todayand are therefore used as comparison base.

The fundamental dierence between ow control and load sensing isthat the load pressure feedback hose to the pump controller can beremoved. Instead of controlling the pump in a closed loop control mode,an open control mode can be used with no feedbacks present. Thismakes the system design process simpler since the pump can be designedto meet the response requirements without considering system stability.As long as the pump is stable as an isolated component, it will not causeany stability issues in the complete system. In load sensing systems onthe other hand, an apparently stable pump can cause instability in thecomplete system.

Flow control systems are more energy ecient compared to load sens-ing systems. This is because the pressure dierence between pump andload is given by the system resistance rather than a prescribed pumppressure margin. The two system layouts have the same eciency whenthe pump is saturated. However, in all other operational points, owcontrol systems have a higher eciency than load sensing systems. Thereare also potential energy savings tied to the absence of active control ofthe pump.

Pressure compensators are key components in ow control systems.

49


Traditional compensators control the absolute ow through the direc-tional valves. If the pump also controls the ow, the ow situation inthe system is over-determined and ow matching problems occur. Thiscan be solved by introducing sensors or a bleed-o valve. Another solu-tion is to use ow sharing pressure compensators. Instead of controllingthe ow, the valves will then serve as ow dividers, eliminating the owmatching problem. Both traditional and ow sharing compensators canbe realized by placing the compensators either upstream or downstreamof the directional valves. A drawback with ow sharing compensators isthat the highest load dynamically will disturb all lighter loads.

Damping is a desired property in uid power systems. Low dampingmakes the system oscillative, which has a negative impact on both theproductivity and the operator. Two dierent ways to obtain dampingin ow control systems are analysed in this thesis, either by the inletorice or by the outlet orice in the directional valve. If ow sharingcompensators are used, it is possible to control the inlet orice, with-out aecting the actuator velocity, with the objective to optimize thedamping. Active damping measures are required to obtain damping bythe inlet orice if traditional compensators are used. Some design rulesto obtain a high damping by the outlet orice are proposed. From anenergy eciency point of view, it is often better to obtain damping bythe outlet orice since a smaller pressure drop is required.

It is possible to combine ow control with other working hydraulicsystems. For example, ow control could be used as a complement topump controlled actuators. Some high power consumers could have onededicated pump while other, low power, consumers share one commonpump. In that case, the total installed displacement could be kept ata reasonable level while all pumps could be displacement controlled.Another possibility could be to use an electric motor in combinationwith a xed displacement pump, like in eha systems, but share it withseveral loads. This solution is more favourable in ow control systemsthan in load sensing systems because a lower bandwidth is required.

Experiments have been performed to show the capability of the owcontrol approach. For example, reduction of the superuous pump press-ure margin and energy saving potentials in a short loading cycle for awheel loader application have been demonstrated.

50

7Outlook

Today, both academia and industry devote a lot of eort to the area ofenergy ecient uid power systems. This will most likely continue inthe future. In the short term, ow control systems are a complement,or alternative, to load sensing systems in particular applications. Thechallenge is how, and to what extent, sensors should be used. It ispossible to design a ow control system without the need for sensors,but it might be desirable to use sensors in some operational cases. Oneexample is when a cylinder reaches its end stop. From an eciency pointof view, substantial power savings are possible if the pump controller hassuch information feedback from the consumers.

In the longer term, independent metering and valveless systems willprobably gain market shares. Those systems are more energy ecient,especially during partial loading conditions. Furthermore, they also havethe possibility to recuperate energy from the loads. This energy caneither be used to run the system pump as a motor or to be stored in, forexample, an accumulator. One interesting system layout in the futurecould be to use a xed displacement bidirectional pump powered by anelectric motor in ow control systems. This is a hybrid of eha and owcontrol.

Which hydraulic system to choose will always be a compromise be-tween, for example, eciency, dynamic characteristics, complexity andcost. It is possible that ow control systems will be the optimal solutionfor certain functions in some particular applications.

51


52

8Review of Papers

In this chapter, the three appended papers in this thesis are brieysummarized. Papers [I] and [II] analyse dierent aspects of ow controlsystems and compare the ndings with load sensing systems. Paper [III]is not limited to ow control systems; the conclusions are valid for severalsystem designs.

Paper I

Energy Ecient Load Adapting System Without Load Sens-

ing Design and Evaluation

This paper studies dierent pressure compensator techniques in owcontrol systems. The fundamental dierence between load sen

Hydrailic Systems Linkoping

Documents

flow control systems

fluid power systems

load sensing systems

linkping studies

pressure compensators

thesis possible

control approach

load pressure feedback