The 15th Scandinavian International Conference on Fluid Power, SICFP’17, June 7-9, 2017, Linköping, Sweden

A Hardware-In-The-Loop (HIL) Testbed for Hydraulic Transformers Research

Sangyoon Lee and Perry Y. Li

Department of Mechanical Engineering, University of Minnesota, Minneapolis, MN USAE-mail: sylee@umn.edu, lixxx099@umn.edu

Abstract

A hardware-in-the-loop (HIL) testbed for testing and evaluating experimental hydraulic trans-formers has been developed. The testbed is capable of emulating a variety of loading con-ditions specified by the desired work cycle operation of various hydraulic machines, such ashumanoid robots or excavators. The HIL testbed is a useful research tool as it allows studyingthe dynamics of many desired loading conditions without having to obtain and reconfigure thesaid physical hardware for each case. This paper discusses the background, construction, andcontrol of the HIL testbed with the experimental results. Preliminary results demonstratingthe functionality of an experimental switch mode hydraulic transformers are also presented.

Keywords: Hydraulic Transformer, Switch mode Transformer, Hardware-In-The-Loop test-ing, Load Emulation

1 Introduction

A hydraulic transformer transforms an input pressure andflow combination into an output pressure and flow combina-tion in an energy conservative manner. It is a key componentin realizing a common pressure rail (CPR) system where acommon input pressure source feeds multiple services. Com-pared to valve based control, hydraulic transformers do notrely on throttling to transform pressures and allows energy re-generation. Thus they have the potential to improve hydraulicsystem efficiencies. Moreover, since pressure can be boostedas well as bucked, system pressure can be kept low to reducelosses.

Although much research has been done on hydraulic trans-former configured using a rotatable 3-ported port plate such asthe INNAS Hydraulic Transformer (IHT) in [1], a hydraulictransformer can also be configured simply as a combinationof a hydraulic pump and motor connected mechanically on acommon shaft. We refer to the latter as a Pump-motor or PMtransformer. In a PM transformer, hydraulic power is con-verted by varying the relative displacements of the pump andof the motor. If the hydraulic transformer is configured withaxial piston pump/motors, swash plate angles of the units willdetermine the pressure (or flow) transformation ratio.

In our previous work, we identified that 3 distinct configur-ations (PM-1, PM-2, PM-3) can be obtained by connectingthe ports of a PM transformer variously to the supply pres-sure line, the load, and the return line (Fig. 1). By switchingbetween these modes, a switch mode transformer has benefitsin compactness, efficiency and operating region [2, 3].

While methodologies to design a hydraulic transformer as acomponent in a hydraulic system are well studied [4, 5], con-

trol algorithms for hydraulic transformer had received less at-tention. Werndin and Palmberg [6, 7], Ahn and Ho [8] areamong the few who investigated control algorithms for a hy-draulic transformer to control actuators. In order to fill thisgap, our group designed a passivity based backstepping con-troller and experimentally demonstrated the control perform-ance for a trajectory tracking control [9] and a human poweramplifier control [10, 11]. These results were obtained foreach PM transformer configuration, and suggest using themore efficient transformers need not sacrifice control per-formance as compared to control valves.

A major challenge in experimentally validating the controlperformance and efficiency of transformers in various applic-ations and duty cycles is the difficulty of obtaining the neededhardware. It is often cost prohibitive to develop a machine andthe loads. Consequently, a controller could only be tested forlimited cases. Additionally, prototype transformers need to beduplicated to test a multi degree-of-freedom system.

A Hardware-In-the-Loop (HIL) testbed can provide a con-venient solution for the aforementioned challenges. In a HILsystem, the actual machine is replaced with a dynamic simu-lation executed in software, while the hydraulic component tobe tested are presented physically with the pressure and flowconditions as in the actual machine. This allows a hydrauliccomponent to be tested for a variety of operating scenarioswithout having to physically obtain the hardware.

Several HIL testbeds had been constructed by various re-search groups to overcome similar challenges in investigatinghydraulic system. For example, Zhang et al [12] developeda HIL testbed to develop a controller for a hydraulic earth-moving vehicle, addressing the challenge in reproducing thedigging cycles. Du et al [13] developed a hydrostatic dy-

Peer-reviewed Paper,Accepted for publication on 2017-04-18. 179 SICFP2017 7-9 June 2017Linköping, Sweden

PA

PB

D1 D2

PTQA

QB

PA

PB

D1 D2

PTQA

QB

PA

PB

D1 D2

QA

QB

PM-1 (Tank Shared) PM-2 (Output Shared) PM-3 (Input Shared)

Figure 1: Three distinct transformer configurations

namometer for testing hydraulic hybrid vehicles, capable ofemulating a variety of driving conditions and vehicle charac-teristics.

In this paper, a HIL testbed for hydraulic transformer is de-veloped to test the performances of hydraulic transformercontrolled systems. Each degree of freedom can be testedindividually without needing the physical actuator or inertialload. The same testbed can also be used to test the trans-former in different machine configurations and duty cycles bysimply reprogramming the dynamic simulation. This testbedhas a simple construction and can emulate both resistive andoverrunning loads.

The rest of the paper is organized as follows. Section 2 de-scribes the hydraulic transformer being studied in our group.Section 3 describes the HIL testbed construction and its con-trol strategies. Implementation results and concluding re-marks are given in Section 4 and 5.

2 Hydraulic Transformer in ResearchThe hydraulic transformer being studied in this paper is atraditional pump/motor type (PM transformer) configuredwith 3.15 cc/rev variable displacement micro-axial pistonpump/motors from Takako. Three distinct configurations(PM-1, PM-2, PM-3) can be obtained by connecting the portsof a PM transformer differently (Fig. 1).

For each of the configuration, the transformer shaft speed ωacting with a rotational inertia J to produce port flows out oftransformer at input port QA and output port QB are given byfollowing sets of equations:

PM-1:

Jω = (PA−PT )D1

2πu1 +(PT −PB)

D2

2πu2−btω−Tloss

QA =−ω · D1

2πu1−Qleak′

QB = ω · D2

2πu2−Qleak

(1)

PM-2:

Jω = (PA−PB)D1

2πu1 +(PT −PB)

D2

2πu2−btω−Tloss

QA =−ω · D1

2πu1−Qleak′

QB = ω ·(

D1

2πu1 +

D2

2πu2

)−Qleak

(2)

PM-3:

Jω = (PA−PT )D1

2πu1 +(PA−PB)

D2

2πu2−btω−Tloss

QA =−ω ·(

D1

2πu1 +

D2

2πu2

)+Qleak′

QB = ω · D2

2πu2−Qleak

(3)

where D1 and D2 are the maximum volumetric displacementsof the pump/motor units in m3/rev, u1 and u2 ∈ [−1,1] are thecontrol inputs which are the normalized displacements, bt isthe damping coefficient, Qleak′ , Qleak and Tloss are the lumpedvolumetric losses at the A and B ports and the mechanicalloss inside the transformer due to friction. These losses areconfiguration, pressure and shaft speed dependent.

Solenoid B

Solenoid A

u1 u2

PAQA

QBPB

Figure 2: Switch Mode Transformer

Adding two 3-way solenoid valves as in Fig. 2 allows allthree configurations to be realized. By switching betweenthe configurations, the total volumetric displacement of thepump/motors (hence overall size) can be reduced for a givenflow capability and shaft speed [2,3]. For example, a switchedmode PM transformer needs only be 13% larger than IHTwith equivalent flow capability as compare to 33% largerwhen compared to using just one configuration. In addition,by picking the configuration with the highest efficiency, theoverall transformer efficiency and operating region can be im-proved.

Fig. 3 shows the efficiency map of an early stage prototypetransformer with a constant 6.895 MPa (1000 psi) pressurerail for each configuration. The relatively low efficiency ofthis prototype is due primarily to the unusually high mechan-

Peer-reviewed Paper,Accepted for publication on 2017-04-18. 180 SICFP2017 7-9 June 2017Linköping, Sweden

PM−1

0.6

0.55

0.5

0.5 0.50.350.3

0

0.5

1

1.5

PM−2

0.35

0.4

0.45

0.5 0.55

0.6

load �ow rate (cc/s)

Tran

sfor

mat

ion

Ratio

PM−3

0.65

0.60.550.50.45

0.350.25

0 100 2000

0.5

1

1.5

Switched Mode

0.650.6

0.60.55

0.550.55

0.450.35 0.4

0.35

0.4

0 100 200

Figure 3: Efficiency contours of individual PM-1, PM-2, PM-3 transformer configurations and switched configurations

ical loss of the common shaft. Incorporating a bearing is ex-pected to significantly improve the efficiency. Nevertheless,this prototype can still demonstrate the implications of usinga hydraulic transformer for system control. One can also dis-cern that the switch mode transformer widens the operatingregion with high efficiency by picking the best configurationfor each operating point (Fig. 3, bottom right).

In our prototype switch mode transformer, two Sun Hydraul-ics DMDA-MBN 3-way solenoid-operated directional spoolvalves are placed at both sides of the transformer as shown inFig. 2. At the default position, the transformer is configured in‘PM-1’ configuration, which has the shared line connected tothe return line. If solenoid valve B is triggered, then the trans-former is configured in ‘PM-2’ configuration, with the sharedoutput port. Lastly, if solenoid A is triggered, the transformeris in ‘PM-3’ configuration with the input port shared betweentwo units.

3 HIL Testbed for Hydraulic Transformer3.1 Components of HIL System Architecture

The hydraulic schematic of the HIL Testbed for hydraulictransformer is shown in Fig. 4. The system is powered bya main pressure compensated pump, which provides, via apressure reducing valve, the constant pressure PS to the sys-tem.

Switch mode transformer discussed in Section 2 is being stud-ied with this HIL Testbed. As mentioned, actuating the solen-oid valves switches the hydraulic transformer to be in one ofthe three configurations shown in Fig. 1. This setup can alsobe used to study other transformer designs, such as an IHT.

Pressure (P) and flow (Q) sensors are placed at various placesas shown in Fig. 4. The transformer shaft speed ω is meas-

LoadEmulatingValve

Solenoid B

Main Pump

Solenoid A

Switch Mode Transformer

QA

QB

PB

PT

PA1

PA2

u1 u2

uvPS

PS

PT

PB

Figure 4: HIL Circuit for testing Switch Mode Transformer

ured with an optical encoder. The transformer controller de-termines the displacement control inputs u1 and u2 that willdeliver the desired flow Qd

B to the actuator while maintainingthe desired shaft speed ωd [9, 10].

In the HIL testbed, a load emulating valve (LEV) is utilizedin place of the hydraulic actuator. A Moog 760 series servovalve (rated at 9.5 LPM), connected to the main pump priorto the pressure reducing valve and to tank, is used for thispurpose. One outlet port is connected to the transformer out-put while the other port is blocked. As will be described inSection 3.4, the servo valve command uv is used to create anappropriate pressure condition for the transformer that mim-ics the actual load condition.

The fact that the LEV is connected to a supply pressure PSis higher than the input port pressure of the transformer PAallows this HIL system to emulate both resistive and overrun-ning loads just using one main pump. This configuration isadvantageous as only one main pump is needed to power thewhole testbed as opposed to having separate pumps for emu-lating load and running a hydraulic transformer.

3.2 Data and Signal Flow in the HIL Testbed

TransformerController

Transformer HIL Load Emulator

HILController

x

Physical System

LoadFeedback

ω

u1 u2 PB

PSIM

uv

P

Simulated system

.. .

Inertia Pressure

QB

Figure 5: HIL Control Scheme

Fig. 5 shows the block diagram illustrating the data flowwithin the testbed. The output flow QB from the transformeris measured and is provided to simulate the pressure dynam-ics P, creating a simulated pressure PSIM in return. In case ofa hydraulic actuator shown in Fig. 6, the pressure dynamicsof the capside chamber which is to be (virtually) connected to

Peer-reviewed Paper,Accepted for publication on 2017-04-18. 181 SICFP2017 7-9 June 2017Linköping, Sweden

xP, A

QB

FL

Figure 6: Example cylinder being simulated

the transformer is given by:

PSIM =β

V0 +Ax(QB−Ax) (4)

where V0 is the volume in the capside chamber and hose whenthe actuator is at the position x = 0, and β is the bulk modulusof the fluid. Integrating PSIM yields a PSIM which is fed into aHIL controller to determine the LEV control uv to match theactual pressure PB with PSIM .

The measured pressure PB in turn drives the simulated inertiadynamics, x:

mx =−bx+ PB(t)︸ ︷︷ ︸measured

A+FL (5)

where m is the mass of the cylinder and rod to be simulated, Ais the cap side area of the hydraulic actuator, b is the viscousfriction coefficient, and FL is the load force that would encap-sulate any external load including gravity and environmentforces. Integrating x yields the velocity x and the position xwhich are used in (4). Note that this simulated inertia dynam-ics take the measured pressure to account for any movementchange due to pressure dynamics associated with transformermode switch.

Finally, the transformer controller takes the feedback of theabove information to determine the transformer control in-puts u1 and u2 to achieve the desired flow for the (emulated)load and the desired torque for the (actual) transformer speedregulation.

The pressure dynamics P and the inertia dynamics x in (4)-(5)can be set to simulate any desired actuator type with specifiedloading conditions, generating appropriate motion and mech-anical load. If desired, the simulated system dynamics couldsimply be replaced by a duty cycle information providing thedesired pressure and flow traces.

3.3 Hydraulic Transformer Controller

The objective of the hydraulic transformer controller is toprovide the desired flow Qd

B for the given task and the desiredtorque T d to regulate the shaft speed [9, 10].

PM-1:[

u1u2

]=

[0 ω · D2

2π(PA−PT )

D12π (PT −PB)

D22π

]−1 [Qd

BT d

](6)

PM-2:[

u1u2

]=

[ω · D1

2π ω · D22π

(PA−PB)D12π (PT −PB)

D22π

]−1 [Qd

BT d

](7)

PM-3:[

u1u2

]=

[0 ω · D2

2π(PA−PT )

D12π (PA−PB)

D22π

]−1 [Qd

BT d

](8)

3.4 Load Emulating Valve (LEV) Controller

QB

uv

PS

PTPB

QvVB

Figure 7: Load Emulating Valve

The most important aspect of the HIL testbed is the load emu-lating valve (LEV) reproduced in Fig. 7, which will providethe desired loading condition through pressure. For uv > 0,the flow is released to the tank (as shown on the figure) andfor uv < 0, the flow is taken from the main pump to increasepressure. The flow Qv traveling across the valve from thetransformer output to be exposed to PT or PS is given by thesupply pressure or the tank pressure:

Qv =

{kvuv

√|PB−PT |, for uv > 0

kvuv√|PB−PS|, for uv < 0

(9)

(assuming PS >PB >PT ) where kv is the valve coefficient, anduv is the valve command to be designed. For the controllerdesign, consider the pressure dynamics within the hose thatlies between the hydraulic transformer and the LEV:

PB =βVB

(QB−Qv) (10)

where VB is the volume of the hose between the transformerand the LEV, QB is the flow out of the transformer that ismeasured, and Qv is the flow command to be designed. Definepressure error as e = PB − PSIM . For a constant (or slowlychanging) desired simulated pressure, e = PB. Defining thedesired valve port flow Qd

v as:

Qdv = QB +Kpe+KIeI (11)

where eI = e and Kp and KI are the proportional and integralgains. To show that the desired pressure PSIM can be achievedfor the HIL testbed, consider the Lyapunov function:

W =12

VB

βe2 +

12

KIe2I

W = e(QB−Qv)+ eKIeI

(12)

With Qv given by (11),

W =−Kpe2 (13)

which shows that for Kp > 0, KI > 0, e→ 0 and PB will trackPSIM , providing the simulated pressure.

In summary, the LEV controller is a PI controller with a feed-forward information coming from the measured flow out thetransformer.

Peer-reviewed Paper,Accepted for publication on 2017-04-18. 182 SICFP2017 7-9 June 2017Linköping, Sweden

4 Experimental Results4.1 Pressure Tracking Results

To validate that (11) is appropriate, the transformer displace-ments were held steady while only the desired pressure load-ing condition was varied. This type of experiment can be usedto analyze the steady-state operation for the hydraulic trans-former under various loading conditions.

17.3 17.35 17.4 17.45 17.5 17.55 17.6 17.65

4.1

4.2

4.3

4.4

4.5

Time [s]

Pres

sure

[MPa

]

PSIM

PB

Figure 8: Step response

Fig. 8 shows the step pressure response for a 0.345 MPa (50psi) step pressure command. Negligible overshoot with 10 msrise time is observed. Fig. 9 shows various responses of thesame step size.

25 30 35 40 452.5

3

3.5

4

4.5

5

Time [s]

Pre

ssur

e [M

Pa]

PSIM

PB

Figure 9: Various step responses

0 10 20 30 40 50

3.5

3.6

3.7

3.8

3.9

4

4.1

Time [s]

Pres

sure

[MPa

]

PSIM

PB

Figure 10: Chirp signal response

Figure 10 shows a response for the chirp signal and Fig. 10shows the same response zoomed around at select frequen-

14 14.5 15 15.5 16

3.6

3.8

4

3 Hz

PSIM

PB

23 23.5 24 24.5 25

3.6

3.8

4

Pres

sure

[MPa

]

5 Hz

48 48.5 49 49.5 50

3.6

3.8

4

Time [s]

10 Hz

Figure 11: Chirp signal zoomed in at select frequencies

cies. The reference chirp signal, an amplitude of 0.345 MPa(50 psi), started as 0.01Hz at t = 0 and approached 10 Hz att = 50. Phase lag of 36 degree was observed at the peak fre-quency, while phase lag is negligible around 3 Hz. Thus, itwill be safe to claim this testbed can be operated satisfactor-ily under 3 Hz load profile. Careful tuning of the parametersalong with better valve identification will lead to even betterperformance. As transformer itself has a limited bandwidthdue to the swashplate actuation, HIL testbed is more than cap-able of simulating the loads for the transformer.

80 100 120 140 1600

0.05

0.1

Traj

ecto

ry [m

]

80 100 120 140 1600.5

1

1.5

Pres

sure

[MPa

]

PSIM

PB

80 100 120 140 16040

60

80

Tran

sfor

er [r

ad/s

]

80 100 120 140 160

0123

Mod

e

80 100 120 140 160−1

0

1

Time [s]

D1,

D2

u1

u2

Figure 12: Simulated cylinder tracking result

Peer-reviewed Paper,Accepted for publication on 2017-04-18. 183 SICFP2017 7-9 June 2017Linköping, Sweden

4.2 Flow and Pressure Tracking Results

An operating condition where a LEV and transformer aresimultaneously controlled is shown in Fig. 12. For this im-plementation, a cylinder with vertical gravity load follow-ing a filtered trapezoidal trajectory was studied with a trans-former operating in 3 different modes. Controller determiningflow demand and the transformer displacement commands istaken from [9]. While the transformer mode switch is manu-ally triggered for this experiment, further development of thetransformer control will have the mode automatically determ-ined to maximize the operating efficiency.

In this operating mode, HIL provides a loading condition towhich trajectory tracking controller decides the transformercontrol inputs to be implemented. As described, the traject-ory is generated in simulation using the measured flow andpressure. It is observed that trajectory tracking is still satis-factory even as the operating mode switch causes some sharpchange of dynamics. In the meantime, HIL control is trackingthe simulated pressure PSIM .

5 ConclusionThis paper presents a development of a HIL testbed for hy-draulic transformer, with a focus on the control architecture.Using a single hydraulic power supply and a servo-valve, aHIL testbed is configured, providing a platform for furthertransformer control development.

A closed loop controller designed for the actual system for atrajectory tracking is easily implemented for the HIL testbed.Initial performance experiments indicates the ability of thistestbed to emulate various loading conditions.

AcknowledgmentThis work is performed within the Center for Compact andEfficient Fluid Power (CCEFP) supported by the NationalScience Foundation under grant EEC-05040834. Componentdonation from Takako Industries is gratefully acknowledged.Sangyoon Lee is supported by a 2016-2017 Doctoral Disser-tation Fellowship at the University of Minnesota.

Nomenclature

Designation Denotation Unit

A Piston Area m2

b Cylinder viscous friction coeffi-cient

N/(m/s)

bt Transformer viscous frictioncoefficient

N·m·s

D1, D2 Volumetric displacements m3/reve Pressure error PaeI Integral of pressure error PaFL Load force NJ Transformer shaft inertia kg·m2

KI Integral control GainKp Proportional control gainkv LEV valve coefficientm Mass of cylinder kgPA, PB, PT Transformer port pressures PaPS Main pump supply pressure PaPSIM Simulated pressure PaQA, QB Transformer flows m3/sQd

B Desired flow of transformer m3/sQleak, Qleak′ Flow loss of transformer m3/sQv Flow across the valve m3/sQd

v Desired flow across the valve m3/sT d Desired torque for transformer N·mTloss Mechanical loss of transformer N·mu1, u2 Displacement ratio inputsuv Valve commandV0 Initial volume in cylinder cap-

sidem3

VB Fluid volume in hose m3

W Lyapunov energy Jx Piston displacement mβ Bulk modulus Pa∆Pv Pressure drop across the valve Paω Transformer shaft speed rad/s

References[1] P.A.J. Achten, Z. Fu, and G.E.M. Vael. Transforming fu-

ture hydraulics: a new design of a hydraulic transformer.In The Fifth Scandinavian International Conference onFluid Power SICFP ’97, page 287ev, 1997.

[2] Sangyoon Lee, Pieter Gagnon, and Perry Y Li. Systemconfiguration and control using hydraulic transformers.In Fluid Power Innovation and Research Conference,Chicago, IL, October. 14-16 2015.

[3] Pieter J Gagnon. Configuration and performance of hy-draulic transformer power distribution systems. Mas-ter’s thesis, University of Minnesota, 2016.

[4] Peter AJ Achten, Titus van den Brink, Johan van denOever, Jeroen Potma, Marc Schellekens, Georges Vael,Martijn van Walwijk, and BV Innas. Dedicated designof the hydraulic transformer. volume 3, pages 233–248,2002.

Peer-reviewed Paper,Accepted for publication on 2017-04-18. 184 SICFP2017 7-9 June 2017Linköping, Sweden

[5] PAJ Achten, TL Van den Brink, T Paardenkooper,T Platzer, HW Potma, MPA Schellekens, and GEMVael. Design and testing of an axial piston pumpbased on the floating cup principle. In The EighthScandinavian International Conference on Fluid PowerSICFP ’03, pages 805–820, May 2003.

[6] R. Werndin and J-O Palmberg. Controller design for ahydarulic transformer. In The Fifth International Con-ference on Fluid Power Transmission and Control ICFP’01, volume 5, pages 56–61, 2001.

[7] R. Werndin and J-O Palmberg. Hydraulic transformerin low-speed operation - a study of control strategies. InThe 5th International Symposium of Fluid Power, JFPS‘02, Nara, Japan, 2002.

[8] Triet Hung Ho and Kyoung Kwan Ahn. A study onthe position control of hydraulic cylinder driven by hy-draulic transformer using disturbance observer. In Inter-national Conference on Control, Automation and Sys-tems 2008, pages 2634–2639. IEEE, 2008.

[9] Sangyoon Lee and Perry Y Li. Passivity based back-stepping control for trajectory tracking using a hydraulictransformer. In ASME/BATH 2015 Symposium on FluidPower and Motion Control, pages V001T01A064–V001T01A064. American Society of Mechanical En-gineers, 2015.

[10] Sangyoon Lee and Perry Y Li. Passive control of a hy-draulic human power amplifier using a hydraulic trans-former. In ASME 2015 Dynamic Systems and Con-trol Conference, pages V002T27A004–V002T27A004.American Society of Mechanical Engineers, 2015.

[11] Sangyoon Lee, Fredrik Eskilsson, and Perry Y Li.Multi degree-of-freedom hydraulic human power amp-lifier with rendering of assistive dynamics. In ASME2016 Dynamic Systems and Control Conference, pageV001T07A002. American Society of Mechanical En-gineers, 2016.

[12] Rong Zhang, Don E Carter, and Andrew G Alleyne.Multivariable control of an earthmoving vehicle power-train experimentally validated in an emulated workingcycle. In ASME 2003 International Mechanical Engin-eering Congress and Exposition, pages 515–524. Amer-ican Society of Mechanical Engineers, 2003.

[13] Zhekang Du, Tan Cheng, Perry Y Li, Kai LoonCheong, and Thomas R Chase. Virtual vehiclecontrol concept for hydrostatic dynamometer control.Journal of Dynamic Systems, Measurement, and Con-trol, 139(2):021009, 2017.

Peer-reviewed Paper,Accepted for publication on 2017-04-18. 185 SICFP2017 7-9 June 2017Linköping, Sweden