Applied Energy 210 (2018) 16–27

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Optimized design and characterization of motor-pump unit for energy- T

regenerative shock absorbers
⁎
Renato Galluzzi , Yijun Xu, Nicola Amati, Andrea Tonoli
Department of Mechanical and Aerospace Engineering (Mechatronics Laboratory), Politecnico di Torino, Corso Duca degli Abruzzi, 24, Turin, Italy

H I G H L I G H T S

• The energy dissipated in automotive suspensions can be recovered.

• The motor-pump unit is a key element for energy regeneration and damping control.
• The proposed design of this unit aims to maximize energy regeneration.
• The construction and testing of a prototype validate the research.

A R T I C L E I N F O A B S T R A C T

Keywords: The constant need to reduce emissions in the automotive sector has driven the electrification of powertrain and
Regenerative chassis. To comply with this trend and decrease the bound even further, the present paper proposes the use of
Automotive hydraulic regenerative shock absorbers for automotive suspension systems. The conversion of linear into angular
Shock absorber motion and the suitable control of an integrated electric machine allow to transform part of the vibrational
Electrohydrostatic actuation
energy into electricity. In these damping devices, the key element is the motor-pump unit that is interfaced onto
Energy harvesting
a conventional hydraulic cylinder architecture. Hence, the proposed research focuses on this component by
Efficiency
investigating different design aspects in all the domains of interest. The objective is to optimize the energy
conversion efficiency of the unit without affecting its damping control property. To give means of validation, a
motor-pump prototype is built and experimentally characterized through a dedicated test rig.

1. Introduction 2020, this value will be reduced to 95 g/km and by 2025, it will be
decreased even further (68–78 g/km) [1]. The awareness in CO2
Vehicle suspension systems play a fundamental role in filtering the emissions has pushed the electrification of the automobile as a means
vibrations induced by the road irregularities onto the automotive towards improved energy efficiency.
chassis. They are designed to meet requirements of road holding, ride As demonstrated by Zuo and Zhang [2], the average power dis-
comfort and handling performance. Conventional suspension systems sipated by the suspension is proportional to the tire stiffness, the vehicle
basically consist of an elastic member (spring), a damping element speed and the road roughness index defined by ISO [3]. For a typical
(shock absorber) and kinematic linkages. These systems are purely passenger car that travels at a speed of 32 km/h (the average speed of
passive: the damping task is achieved by converting the vibrational the New European Driving Cycle), the average dissipation on four
power into waste heat. Although simple and cost-effective, passive corners is 133 W on an ISO C-class road. According to the analysis
suspensions have a fixed response and cannot adapt to a variety of road presented in our previous works [4], the total recovery of the energy in
unevenness and vehicle dynamic conditions. To address this short- the aforementioned condition would lead to a CO2 emission reduction
coming, semi-active and active damping technologies have been de- slightly above 6 g/km for a D-class vehicle.
veloped in the last two decades. This background motivates the development of regenerative solu-
In addition, vehicle CO2 emissions have become a relevant aspect tions, i.e. devices able to vary their damping behavior while converting
during the past decade: regulations in this matter are more strict mainly part of the otherwise-dissipated power into electricity. These technol-
due to environmental concerns worldwide. In Europe, for instance, the ogies exploit the intrinsic reversibility of electric machines and a sui-
emission average to be achieved by all registered cars is 130 g/km. By table transmission system for their integration into the vehicle.

⁎
Corresponding author.
E-mail addresses: [email protected] (R. Galluzzi), [email protected] (Y. Xu), [email protected] (N. Amati), [email protected] (A. Tonoli).

http://dx.doi.org/10.1016/j.apenergy.2017.10.100
Received 4 August 2017; Received in revised form 3 October 2017; Accepted 29 October 2017
Available online 06 December 2017
0306-2619/ © 2017 Elsevier Ltd. All rights reserved.
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

Nomenclature lg gear axial length

lm active length
Ap piston cross section lp pocket depth
As slot cross section λp PM flux linkage amplitude
Aw wire cross section dynamic viscosity
αg1 outer gear aspect ratio Nc number of coils
αg2 inner gear aspect ratio Ng number of inner gear teeth
αt teeth aspect ratio Nt number of turns per coil
Az magnetic vector potential z component Ωg angular speed
cp linear damping coefficient p number of pole pairs
cref linear damping reference Pg harvested power
csc maximum rotary damping coefficient Pm,avg average power loss
Dp pocket diameter Qg input flow rate
dp minimum distance between ports q quadrature axis
Ds shaft diameter R phase resistance
d direct axis R1l outer gear lobe radius
ΔPg input pressure drop R1o outer gear outside radius
δg pulsation index R1r outer gear root radius
eg gear eccentricity R2i inner gear inside radius
ηe electrical efficiency R2o inner gear outside radius
ηg pump total efficiency R cyl,i cylinder inside radius
ηhm hydro-mechanical efficiency R cyl,o cylinder outside radius
ηt total conversion efficiency R ext external impedance
ηv volumetric efficiency Rp,i port inside radius
fa axial viscous drag function Rp,o port outside radius
Fp piston/damping force ρCu copper resistivity
fv chamber volume function ρcyl cylinder mass density
Fj emptying rate function of the jth chamber τ electrohydrostatic transmission ratio
ga axial clearance Ta axial viscous drag torque
gi inside diameter clearance Tm electromagnetic torque
go outside diameter clearance Tro radial viscous drag torque
gt inter-teeth clearance Θ0 initial temperature
γp port opening angle Θc winding temperature
Idc DC bus current θg pump shaft angular position
Jcyl cylinder moment of inertia Θp PM temperature
Kλ PM flux per unit length Vj volume of the jth chamber
K cp coil packing factor Vdc DC bus voltage
Ke back-EMF constant Vg volumetric displacement
Kt torque constant vp piston/damping speed
l cyl cylinder length
l et end-turn length

Karnopp’s research [5] was the first to examine the feasibility of the suspension. Moreover, in a high-cycle task like vehicle damping,
using a permanent-magnet linear motor with variable resistors to sub- component wear and fatigue are critical aspects that have not been fully
stitute the conventional dampers. Gysen et al. [6] developed an active addressed.
electromagnetic suspension that employs a brushless tubular perma- Hydraulic shock absorbers with controllable damping and energy
nent-magnet actuator to control the roll and pitch of the vehicle. Linear harvesting features employ the electrohydrostatic actuation principle.
motors seem a straightforward choice for damping in most ground They use a hydraulic actuator directly interfaced to a motor-pump unit
vehicles due to their simple integration into the suspension layout and by means of a hydrostatic circuit to convert the linear motion of the
lack of a transmission system. However, their force density is limited for piston into rotation. The intrinsic lubrication of fluid-based solutions
the task and thus, they work inefficiently and add a substantial heft to overcomes the main tribology concerns of electromechanical systems.
the vehicle chassis. Moreover, since the fluid is used as a means for power transmission, the
In conventional suspension layouts, rotary electric motors require actuator offers better flexibility for its placement within the suspension.
complex systems to convert the linear motion between the wheel hub Many recent works have addressed the design and implementation
and the upper strut mount into an angular displacement. In this regard, of hydraulic regenerative shock absorbers. In previous works, it has
the state of the art exploits mainly mechanical or hydraulic working been studied for lead-lag motion damping in helicopter rotor blades
principles. [13]. Fang et al. [14] and Li et al. [15] assessed the damping and re-
The rack pinion mechanism can be used for linear-to-rotary con- generative capabilities of prototypes with external hydraulic rectifiers.
version. In recent works [7,8], it has been enhanced with clutch systems Zhang et al. [16] exploited the intrinsic fluid rectification of twin-tube
to allow unidirectional angular motion on an electric machine and shock absorbers to yield and validate a prototype. However, Levant
improve efficiency. Ball screw transmissions are also common for mo- Power Corporation has lead the development in this field with their
tion conversion: the literature offers prototypes with rotating screw [9] GenShock solution [17]. From the scientific point of view, they have
or rotating nut [10]. Furthermore, unidirectional rotary motion has provided useful guidelines to optimize the hydraulic pump for a re-
been also implemented [11,12]. Although promising, electro- generative damper [18].
mechanical technologies can be complex and difficult to integrate into The motor-pump unit is the core element in hydraulic regenerative

17
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

dampers, as the energy harvesting capability is heavily influenced by convert the kinetic energy of the suspension into electricity that can be
this component. Hydraulic, mechanical and electric subsystems must be stored in a battery.
integrated and optimized to maximize energy regeneration, while pre- In addition, the twin-tube architecture provides unidirectional mo-
serving the damping functionality. Nevertheless, the state of the art tion of the motor-pump group. The fluid flow is rectified with the
does not offer clear guidelines on the design of the motor-pump group aperture of the base check valve during rebound and the piston check
as a whole. Some of the previous research works deal with the design valve during compression. This working principle is already covered by
and validation of single subsystems. Others deal with system integra- previous research works [4,16].
tion at a proof-of-concept level and with the partial or total use of off- Fig. 1 also depicts the prototype implementation of the described
the-shelf components. system. The motor-pump group is fixed onto the shock absorber tube
To address these shortcomings, the present paper establishes a novel with a manifold. This approach yields a compact system and minimizes
system-level approach that investigates the role of different design as- the introduction of losses due to hydraulic lines. Furthermore, it pro-
pects on the performance of the motor-pump group. The unit is entirely vides modularity because the motor-pump group unit can be installed
designed to guarantee damping and energy harvesting features, while on any shock absorber tube. For this reason, this research deals with the
complying with important design constraints such as envelope and optimization of this unit, disregarding the selected shock absorber ar-
supply voltage. The study focuses on the subsystems of interest (hy- chitecture.
draulic pump, electric motor and power stage) to yield an optimized In general, the power conversion from the hydraulic domain of the
prototype. Subsequently, the device is numerically and experimentally pump to the electrical domain of the battery is inevitably affected by
characterized in different working points of the force-speed plane. different loss terms.
Unlike previous works, the study is focused on the partial and total The hydraulic power applied to the pump is initially reduced by a
efficiency terms calculated on each working point. The obtained effi- volumetric loss due to internal leakages. The volumetric efficiency is
ciency maps give means for validation. Furthermore, they prove to be given by
an adequate tool to identify the most critical aspects of the design. Vg Ωg
Although the design and characterization are based on a particular case ηv = ,
Qg (1)
study, the followed approach can be extended to any motor-pump unit
for hydraulic regenerative dampers. where Qg is the flow rate across the pump, Ωg its angular speed and Vg
The remainder of this research is organized as follows. In Section 2, its fixed volumetric displacement.
it gives an overview of the system and the efficiency terms that play a Subsequently, a hydro-mechanical loss term considers the pressure/
fundamental role in the performance of the unit. Then, relevant design torque drop due to hydraulic minor losses and mechanical friction af-
and construction aspects are analyzed for the different subsystems in fecting the rotating elements. The hydro-mechanical efficiency associated
Section 3. Finally, Section 4 deals with the experimental validation of a to these phenomena is
motor-pump prototype.
Tm
ηhm = .
Vg ΔPg (2)
2. System overview
Here, Tm is the electromagnetic torque of the motor and ΔPg the pump
pressure drop.
The present research focuses on the design and optimization of the
Finally, a third loss term considers the conversion from the me-
motor-pump unit employed in the regenerative shock absorber shown
chanical domain of the motor to the electrical domain of the battery,
in Fig. 1. The system exploits a conventional twin-tube shock absorber
i.e.
architecture connected to the hydraulic ports of a pump. During vehicle
cruise, the piston of the shock absorber oscillates inside the tube at a Vdc Idc
ηe = ,
speed vp due to the ground irregularities. This motion leads to an oil Tm Ωg (3)
flow rate Qg that drives the hydraulic pump and therefore, the linear
in which Vdc and Idc define the voltage and current of the power stage
movement is transformed into a rotary motion Ωg . Since the pump is
DC bus, respectively. This electrical efficiency can be attributed to motor
mechanically coupled to an electric machine, a suitable control strategy
and power-stage losses.
of the latter device allows to modify the damping characteristic of the
The product of the three aforementioned terms yields the total
shock absorber. Furthermore, the electric motor acts as a generator in a
conversion efficiency of the motor-pump unit:
portion of the damping quadrants. Hence, the system is also able to
Vdc Idc
ηt = ηv ηhm ηe = .
ΔPg Qg (4)

3. Design

The design of the motor-pump unit requires a system-level approach

to comply with different requirements in all the domains of interest. A
possible design method is outlined in Fig. 2. It considers three different
specification sets:

1. Damping requirements. The device must be able to work within a

defined force-speed envelope to guarantee its damping function-
ality.
2. Target envelope. The prototype must fit onto the shock absorber and
Fig. 1. Layout of a hydraulic regenerative shock absorber. Concept scheme (a): battery avoid any interference in static and dynamic conditions, such as
(1), power stage (2), electric motor (3), hydraulic pump (4), pressure-relief valves (5), suspension deflection and steering.
check valves (6), piston (7), gas accumulator (8), base (9). Prototype side view (b): 3. Power supply requirements. Assuming that the power stage of the
manifold (10), motor-pump unit (11), spring holder (12), rod (13), external tube (14), device is connected to a battery that is shared among different
anti-roll bar bracket (15), wheel hub bracket (16).

18
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

⎧Tm = τFp .
⎨ Ω = vp/ τ (6)
⎩ g

Furthermore, the damping coefficient imposed by the electric ma-

chine [10] is ideally expressed as

Kt K e
cp = ,
τ 2 (R + R ext ) (7)

where Kt is the torque constant of the electric motor, K e the back

electromotive force (EMF) constant, R is the phase resistance and R ext is
the impedance imposed by the power stage output. This formulation
implies that the power flows from motor to battery, as the latter ele-
ment acts as a power sink. Hence, the maximum attainable damping
c p,max is limited by the short-circuit response, i.e. R ext = 0 . Conversely,
the open-circuit response (R ext → ∞) will lead to a null damping coef-
ficient c p,min .
Since the linear-to-rotary conversion is not perfectly efficient, the
working envelope will be inevitably reshaped, as shown in Fig. 3
(right), because

1. the pump leakage paths will increase the effective damping speed,
thus lowering the slope of any attainable damping behavior, and
2. the hydro-mechanical losses on the motor and pump will add up to
the output force and increase the output damping behavior.

Note that the former phenomenon has a deeper impact at low speed,
while the latter has more presence as speed increases. Hence, it is ex-
pected that the maximum damping will be lower due to volumetric
Fig. 2. Design method for the motor-pump unit.
losses and the minimum damping will be non-null due to hydro-me-
chanical losses. It is worth observing that, although the coefficient c p,min
accessories, the design should account for intrinsic limitations like penalizes the conversion efficiency of the device, it also provides a
DC voltage, state of charge and charge-discharge rates to work re- passive damping action that is useful to maintain the vehicle stability in
liably. the case of system failure.

The first specification set is used to establish the design of the pump
in terms of gear geometry and clearances. The characteristics of the 3.1. Hydraulic pump
resulting device, together with the damping constraints, are used to
settle the electromagnetic design of the machine. The number of turns The present study considers the use of a gerotor pump. These fixed-
of the electrical machine windings, instead, is selected to meet the DC displacement devices offer important advantages with respect to other
supply voltage limitation. Once that motor and pump are fully de- hydraulic pump types [19,20], such as reduced tooth wear, low amount
signed, they are validated with a thermal analysis in the worst-case of components to machine and versatility for integration into any fluid
working condition. power system. Moreover, they are suitable for the low-mid pressure
A mechanical integration step aims to minimize the bulk and mass range (up to 150 bar).
of the final actuator, since both motor and pump designs must comply In gerotor pumps, the inner and outer gears rotate with pure relative
with the established geometric constraints. Thus, finding a configura- rolling motion. This type of motion ensures reduced wear and larger
tion that satisfies damping, DC voltage and mechanical packaging re- mechanical efficiency with respect to systems with sliding elements.
quirements usually requires multiple iterations. The outer gear presents an additional tooth with respect to the inner
The components of the power stage are selected according to the gear. Assuming that the pump shaft moves the inner gear at a speed Ωg ,
electric machine parameters and the supply requirements. Finally, a the relative rolling motion between gears takes place at rather low
suitable control strategy is implemented considering the damping en- speed, which also favors the mechanical efficiency:
velope and the physical limitations of the subsystems.
A correct design of the motor-pump group must favor the force-
speed envelope that the regenerative shock absorber should span. A
feasible choice is to constrain this working area as purely regenerative,
i.e. the battery can only absorb power from the system. Thus, the device
must be able to ideally satisfy the working region bounded by max-
imum damping c p,max , maximum force Fp,max and speed vp,max , as seen in
Fig. 3 (left).
Considering a fixed displacement Vg and a piston cross section Ap ,
the ideal transmission ratio of the system is given by

Vg
τ= ,
Ap (5)
Fig. 3. Force-speed maps of the regenerative shock absorber limited by damping, force
and hence, and speed values.

19
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

Ng ⎞ Ωg Note that the function fv depends on the geometric features of both

Ωg,rel = Ωg ⎛⎜1− ⎟ = , gears. In turn, The kinematic flow rate of the pump is expressed as
⎝ Ng + 1 ⎠ Ng +1 (8)
Ng
where Ng is the number of teeth of the inner gear. Qg (θg ) = Ωg ∑ Fj (θg ),
A gerotor pump is mainly characterized by its volumetric dis- j=0 (14)
placement Vg . In this design, Vg is limited by the maximum angular
speed that the motor-pump group can achieve. At low pressure, lea- where
kages are small and the maximum angular speed can be approximated dVj ⎞
as Fj (θg ) = −min ⎜⎛ ,0⎟
⎝ dθg ⎠ (15)
vp,max
Ωg,max ≅ . is a nonlinear function that denotes the volumetric emptying rate of
τ (9)
each chamber. The volumetric displacement is calculated as the ratio
This maximum angular speed constraint has to be pushed as high as between the flow rate and the angular speed:
possible to lower the required damping torque and hence, minimize the
envelope of the electric motor. Nevertheless, the design must also ac- Qg (θg )
Vg (θg ) = ,
count for the limitations in the filling capability of the gerotor unit at Ωg (16)
high speed [21], as well as the maximum supply voltage for the electric
and, in formal terms, the design target is its average value per revolu-
machine.
tion
After the target volumetric displacement has been defined, the
pump design must define the gear geometric features presented in Fig. 4 1 2π

and the following clearance dimensions between the gears and the
Vg =
2π
∫0 Vg (θg )dθg.
(17)
casing:
Another important aspect to consider is the pulsation index

• Inside diameter clearance. Difference between inner gear and shaft δg =

max(Vg (θg ))−min(Vg (θg ))
,
diameters: Vg (18)
gi = 2R2i−Ds, (10) which gives an indication of the flow ripple. The minimization of this
where Ds is the shaft diameter. phenomenon is desirable in shock absorbers because it could introduce

• Outside diameter clearance. Difference between pocket and outer gear unwanted vibrations into the suspension.
Fig. 5 illustrates the kinematic features for the designed pump. In
diameters:
this case, the selected profile yields a pulsation index of 4.36%. The
go = Dp−2R1o, (11) pulsation index is greatly influenced by the tooth profiles. The
where Dp is the pocket diameter. smoothness of the volumetric displacement function Vg (θg ) is de-
• Axial clearance. Difference between pocket depth and gear axial termined by the shape of the emptying function Fj (θg ) given by the
teeth geometry and the angular spacing between chambers established
length
by the number of teeth.
ga = l p−lg, (12) Additionally, previous analyses [18] demonstrate that the inter-
where l p is the pocket depth. teeth clearance gt and the teeth aspect ratio

3.1.1. Gear geometry

Notice that the pump preserves Ng + 1 fluid chambers sealed by the
teeth tips of both gears disregarding the angular position of its shaft
(θg ). The volume of any chamber is a generalized periodic function with
appropriate angular offset

2πj ⎞
Vj = fv ⎜⎛θg + ⎟ ∀ j ∈ [0,Ng ].
⎝ Ng + 1⎠ (13)

Fig. 5. Volumetric features of the designed pump. The volume and emptying function are
Fig. 4. Gear profile of the gerotor pump with Ng = 6 . Outer gear: outside radius R1o , lobe plotted for chamber 0 (solid) and the remaining chambers (dash-dot). The volumetric
radius R1l and root radius R1r . Inner gear: outside radius R2o and inside radius R2i . displacement (solid) is compared to its average value (dash-dot). Vertical-axis units are
Eccentricity eg , inter-teeth clearance gt , gear length lg and volume of the jth chamber Vj . normalized.

20
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

R1l
αt =
R1r (19)

are relevant for the volumetric efficiency. The minimization of gt

shrinks the inter-teeth passage cross section, while the reduction of α t
increases the passage length. The final outcome is the diminution of
leakages in the tooth tip area.
The gear aspect ratios are another relevant design choice:

⎧ αg1 = R1o / lg .
⎨ α = R2o / lg (20)
⎩ g2
Their selection is driven by considerations on mechanical losses and
inertial contributions. Viscous drag can be split in two main torque
losses: radial and axial components. Radial drag affects only the outer
gear and originates from the thin fluid film between the pocket and the
gear outside diameter. Its contribution reported to the output shaft of Fig. 6. Hydraulic port geometry. Inside radius Rpi , outside radius Rpo , opening angle γp ,
the pump is approximated as minimum distance dp and gear eccentricity eg .

2
3
Ng ⎞ 4πμ f lg R1o
Tro ≅ ⎛⎜ ⎟ Ωg, attention is devoted to the number of cell layers within these gaps; at
⎝ Ng + 1 ⎠ go (21) least three cell layers are needed to accurately reproduce leakage flow
where is the fluid dynamic viscosity. Eq. (21) assumes a homogeneous [22]. The generated meshes must present a gradual cell density tran-
distribution of the gap between the outer gear and the pocket along sition [21]. The fluid domains are interfaced with mismatched grids,
their circumference. However, this simplification does not affect the which are useful to couple static and moving domains. Pumplinx im-
dependence with lg and R1o 3
. plements an efficient method that first identifies the common areas
Similarly, axial drag acts on the cross sections of both gears due to between mismatched grids and then treats them like generic interfaces.
the film between them and the casing. It yields a cumbersome analytic Therefore, the problem solution is fully conservative, fully implicit and
expression because it is predominantly influenced by the geometries of fast [23].
the gears and the casing (hydraulic ports and grooves). Nevertheless, it The software also needs the parameters of the employed fluid. Oil
can be stated that the axial viscous drag is a function of R1o 4 4
and R2o , properties are a determining factor for the efficiency characteristics of
since their cross section depends on the squared radius and the trans- the pump. Low-viscosity oils will induce more leakages and reduce
formation of linear variables into the rotary domain requires another viscous drag, while the opposite is true for high-viscosity fluids.
squared radius term. Nonetheless, viscosity is heavily affected by temperature.
To analyze the inertial contribution, the moment of inertia of a To understand the effect of clearances, different models are simu-
generic hollow cylinder is reported: lated with the (laminar) flow and cavitation modules. The models are
initialized with constant pressure drop values between the pump ports
πρcyl l cyl 4 4 (ΔPg ) and constant inner gear angular speeds (Ωg ). After the initial
Jcyl = (Rcyl,o −Rcyl,i ),
2 (22) transient is vanished, the average values of the gerotor torque (Tm ) and
the flow rate across the ports (Qg ) are extracted. The clearances, in-
where ρcyl is its mass density, l cyl its length, and R cyl,o and R cyl,i its out-
stead, are imposed in the geometry using the gerotor template mesher
side and inside radii, respectively. Hence, it can be deduced that both
4 in Pumplinx. Since the software does not consider fluid-body interac-
gear moments of inertia depend on R1o .
tions, the clearances remain constant as initially imposed at every time
All in all, the outlined scenario points towards the minimization of
step. Diameter clearances are set with the bodies perfectly centered
αg1 and αg2 within the established envelope limits.
with respect to their rotation axes. The axial clearance is imposed as
equally distributed among the front and the back sides of the gears. In
3.1.2. Hydraulic ports
practice, this is not always true because pressure unbalance in both
Twin-tube dampers provide an intrinsic fluid flow rectification.
axial and radial directions will displace the gears from their nominal
Nevertheless, a symmetric port architecture as the one in Fig. 6 sim-
position. Limitations in the representation of this phenomena will affect
plifies the applicability of the device for bidirectional operation, since it
results, especially those concerning the mechanical viscous losses.
yields the same hydraulic behavior in both directions.
A first set of simulations was devoted to assess the diameter clear-
Despite its machining complexity, the selected geometry ensures
ances gi and go . From the results, it was observed that the inside dia-
proper fluid filling and seal. The port depth is also crucial to guarantee
meter clearance has no effect on both the volumetric and hydro-me-
a correct fluid fill. According to typical pump design guidelines [20],
chanical efficiency terms. Thus, gi is selected to provide a tight fit
the port depth must be at least equal to the length of the rotors. The
without interference to minimize inner gear radial play, yet allowing its
resulting geometry can be replicated on the bottom of the gear pocket
axial motion on the shaft for pressure balancing. The outside diameter
as shadow groove features. This operation is recommended to reduce
clearance, instead, has a heavy influence on the hydro-mechanical
viscous drag.
losses, while volumetric ones are negligibly affected. This is expected,
as the radial viscous drag torque on the outer gear is inversely pro-
3.1.3. Clearances between gears and case portional to go (Eq. (21)). More importantly, the selection of go must
One of the most challenging steps in the design process is the de- consider the creation of a thin fluid film to lubricate and support the
finition of the clearances between the gears and the pump casing. The outer gear under pressure load.
accomplishment of this task was aided by Simerics Pumplinx. This Subsequently, a sensitivity analysis was carried out to understand
computational fluid dynamics package is able to obtain reliable simu- the optimal axial clearance of the gerotor unit. Models with different
lation results, yet keeping a relatively low computing overhead with axial clearance values were simulated at two different angular speeds.
respect to other software alternatives. The program requires the 3D For each speed, three damping coefficients spanning a portion of the
geometry of the fluid. Then, a mesh is created for each fluid domain: damping envelope were selected. Accordingly, the pressure drops were
ports, shadow ports, gerotor chambers and clearance gaps. Particular

21
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

80

70

60
efficiency [%]

50

40

30

20
0 50 100 150 200 250
axial clearance [%]
Fig. 7. Hydraulic pump efficiency with different normalized axial clearance values.
Minimum angular speed, minimum (square), mild (triangle) and maximum (circle)
damping coefficients. Results are interpolated with a piecewise-cubic function (solid).
The axial clearance axis is normalized with respect to the selected value.

Fig. 9. Hydraulic pump efficiency map in the force-speed plane. Maximum damping
calculated following the ideal expression behavior cp,max (dashed) and damping behavior due to hydro-mechanical losses cp,min
τc p Ωp (dash-dot) linearized for low speeds. Axis units are normalized with respect to Fp,max and
ΔPg = vp,max .
Ap (23)

Figs. 7 and 8 show the results at both test speeds for different 3.1.4. Pump design verification
clearance gap lengths. The results show the total efficiency of the To verify the design, the identified pump configuration was simu-
gerotor pump, intended as lated at different working points. Results were interpolated to generate
ηg = ηv ηhm. (24) an efficiency map in the force-speed plane, as seen in Fig. 9. For this
purpose, the piston cross section is useful to convert hydraulic variables
At high speed, the test with the highest damping yields a pressure into the mechanical domain:
drop that goes well beyond the design limit ΔPg,max . Hence, this dataset
was discarded. ⎧ Fp = Ap ΔPg .
It is observed that the efficiency is a convex function of the axial ⎨ v = Qg / Ap (25)
⎩ p
clearance. Lower values favor volumetric over hydro-mechanical effi-
ciency. On the contrary, large gaps reduce viscous drag at the cost of The units in numerical and experimental plots have been normal-
increasing internal leakages. The obtained results are helpful to define ized to protect proprietary data. However, force and speed values can
an axial clearance value in simultaneous agreement with all the se- be associated to the typical damping characteristics in the automotive
lected working points. field [24]. The maximum force usually ranges from 1 to 4 kN, de-
pending on the size of the vehicle, while the speed reaches up to 2 m/s.
80 The behavior spans a relevant portion of the force-speed plane.
Furthermore, the maximum damping coefficient specification c p,max lies
70 inside the obtained map, while the damping behavior due to hydro-
mechanical losses can be linearized at low speed to obtain c p,min .
This result defines the design objective of the electric motor. The
60
maximum rotary damping provided by the motor must counteract
efficiency [%]

leakages and account for hydro-mechanical loss, i.e.

50
τ2
csc = (c p,max−c p,min ).
ηv (26)
40
The selected design constraint yields a motor able to accomplish the
30 damping task with the minimum footprint. Alternatively, it is possible
to optimize the machine to match the maximum efficiency point of the
pump map. However, this choice would potentially lead to a motor with
20
larger geometric envelope.

10 3.2. Electric machine

0 50 100 150 200 250
axial clearance [%] The electric machine was designed to comply with the damping
Fig. 8. Hydraulic pump efficiency with different normalized axial clearance values. specification in Eq. (26) when in short circuit. Thus, the motor does not
Maximum angular speed, minimum (square) and mild (triangle) damping coefficients. require active power to operate within the damping specifications.
Results are interpolated with a piecewise-cubic function (solid). The axial clearance axis In this case, a brushless permanent-magnet (PM) topology is ad-
is normalized with respect to the selected value.
vantageous to maximize the torque-to-mass ratio. Specifically, the

22
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

design targets an inner rotor, radial flux, surface-mounted PM motor K cp As

Aw =
due to its construction simplicity. A fractional-slot profile reduces Joule Nt (31)
losses and cogging torque [25]. The stator of the machine is devised
Substituting Eqs. (29) and (30) into (27) yields
with M270-35A steel and the rotor uses NdFeB PMs of grade N45SH and
soft iron. 3p2 Nc K cp Kλ2 As ⎛ l m2 ⎞
csc = ⎜ ⎟,
4ρCu ⎝ l m + l et ⎠ (32)
3.2.1. Electromagnetic design
The stator outside diameter was constrained to match the cross and thus, the constraint on csc can be used to determine the active
section of the gerotor unit (gears, o-ring seals and screws). Hanselman’s length of the machine l m .
analytical procedure [26] provides a starting point in the definition of
the machine 2D geometry by optimizing the PM flux paths and the 3.2.2. Winding number of turns
torque per unit length. Afterwards, COMSOL Multiphysics (AC/DC The number of turns per coil Nt is calculated from a constraint on the
Module) helps to refine the 2D design through magnetostatic finite- battery supply: the DC bus voltage is limited to 48 V (mild hybrid stan-
element analysis (FEA) models. The obtained FEA results for the de- dard) and the machine EMF should not exceed it for safety reasons, hence
signed motor are illustrated in Fig. 10.
Once the motor cross section is fully defined, its active length is Vdc = 3 pλ p Ωg,max , (33)
determined to match the target damping coefficient. The short-circuit
where λ p is a function of the number of turns per coil, as already de-
rotary damping coefficient of the machine is a particular case of Eq. (7)
monstrated in Eq. (29).
in which R ext = 0 and τ = 1. For a brushless PM motor
3p2 λ p2 3.2.3. Thermal analysis
csc = .
2R (27) A worst-case-scenario was simulated with Motor-CAD to validate
where p is the number of pole pairs of the machine and λ p is the PM flux the electric machine design. This software is based on lumped-para-
linkage amplitude. meter thermal network (LPTN) models which are the most used and
Assume that each motor phase is constituted by Nc coils in series, friendly solutions for a fast- and low-computation time thermal analysis
each one of Nt turns and active length l m , located inside slots of cross [27]. The problem formulation requires the machine and winding
section As . The PM flux linkage is calculated from the magnetic vector geometry, the power losses and a working duty cycle.
potential component Az orthogonal to As : For that purpose, the prototype mechanical layout is reproduced in
the program. Then, the machine losses are calculated when it works as a
Nc Nt plm generator. Joule losses are computed using the phase resistance. Iron
λp =
As ⎝
⎛ ∫A + Az dA− ∫ − Az dA⎞.
As ⎠ (28)
s and PM losses are obtained by implementing the discretized Bertotti
The superscript signs are used to denote the winding direction. equations [28] in a dedicated FEA model, where different currents and
Since the magnetic vector potential Az does not depend on the number speeds are applied to the machine. The resulting efficiency map of the
of turns per coil or active length, Eq. (28) is rewritten as machine is plotted in Fig. 11. The behavior at low angular speed and
high current is dominated by Joule losses, whilst iron and PM losses
λ p = Kλ Nt l m, (29) take place predominantly at high speed and low current.
where the PM flux per unit length constant Kλ depends on the estab- To assess the duty cycle of the motor-pump unit, a quarter car model
lished 2D geometry layout. can be built in Simulink with the parameters of the designed motor-
The phase resistance is computed as pump unit [4]. The model is set assuming motor short circuit to have
maximum dissipation in the stator windings. Furthermore, the vehicle
2Nc Nt ρCu
R= (l m + l et ),
Aw (30)
where ρCu is the copper resistivity and l et is the end-turn length. The
wire cross section Aw can be expressed in terms of the slot area and the
coil packing factor K cp :

Fig. 10. Electric machine magnetostatic FEA results. The color map denotes the magnetic
flux density norm, whereas the contours belong to the magnetic vector potential. The Fig. 11. Efficiency map of the electric machine working as a generator. Axis units are
letters and signs indicate the winding phase and direction. normalized with respect to Tm,max and Ωg,max .

23
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

Table 1
Thermal simulation data.

Description Symbol Value Unit

Initial system temperature Θ0 50 °C

Final winding hotspot temperature Θc 137.7 °C
Winding temperature limit Θc,max 180 °C
Final PM temperature Θp 123.5 °C
PM temperature limit Θp,max 150 °C

speed and the road roughness index must be selected to reproduce a

worst-case scenario.
After the simulation, the average power dissipation and angular
speed of the machine are extracted from Simulink. Motor-CAD is in-
itialized with these data and the initial system temperature. The si-
mulation results after the initial transient show that the tested config-
uration is able to withstand the described working conditions because
the two most critical points, i.e. the winding hotspot and the PM stay
well below their temperature thresholds (see Table 1).

3.3. Power stage and control

In literature, many regenerative shock absorbers exploit a diode

rectifier connected to a passive load or a DC-DC converter that feeds the Fig. 12. Block diagram of the control strategy for the motor-pump unit.
battery [8,16]. Despite its simplicity, this solution intrinsically excludes
the possibility to work in active mode. Furthermore, in our case, the open circuit. In contrast, energy regeneration is possible through the
resistance of the windings is in the order of few milliohms due to the DC quadrature axis only if R q,ext > 0 , and the harvested power is ideally
bus voltage constraint. Hence, the use of a diode rectifier severely pe- given by
nalizes the conversion efficiency because its conduction losses are
3
comparable or larger than the motor Joule losses. This phenomenon Pg = R q,ext Iq2.
2 (36)
affects also the short-circuit damping, since this additional conduction
resistance adds up to the denominator of Eqs. (7) and (27). Note that the power term in Eq. (36) is not a power dissipation per
To provide full-active functionality, a power stage with a three- se, but rather a transfer from the motor phases to the DC bus. In prac-
phase full bridge is necessary. The employed power unit presents an tice, however, the variables in Eq. (36) are difficult to obtain and af-
inverter constituted by three MOSFET phase legs (IXYS FMM150- fected by the conduction and switching losses in the power stage.
0075X2F). These switches were carefully selected to comply with the Hence, the harvested power can be directly calculated with the DC bus
motor ratings and provide a low conduction resistance. voltage and current values, since these variables are easy to obtain and
For the control task schematized in Fig. 12, the power stage uses a free of any loss component:
32-bit floating-point digital signal processor (TI TMS320F28335). The
Pg = Vdc Idc. (37)
position feedback is accomplished with an array of three Hall-effect
latches (Allegro A1210) and a PM array fixed to the end of the rotor.
The DC bus and phase currents are measured using Hall-effect current 3.4. Mechanical packaging
probes (AMPLOC AMP25).
The adopted strategy exploits field-oriented control with a current The layout of the motor-pump prototype is depicted in Fig. 13. All in
loop constituted by a proportional-integral (PI) compensator and back- all, the design yielded a compact unit with a total length of 90 mm,
calculation anti-windup. The control routine runs synchronous to the maximum diameter of 88 mm and total mass of 3.5 kg.
phase switching frequency (40 kHz) and the PI compensator is tuned to Internally, the casing separates electric machine and pump gears,
mitigate the electromagnetic pole of the machine and yield a closed- but these two sections are oil-filled. This design choice minimizes
loop bandwidth of 500 Hz. Furthermore, the electric machine is con- friction losses because the seal element between shaft and case can be
trolled exclusively in the constant-torque region and hence, the direct- omitted. The rotor shaft transmits the torque from the pump gears to
axis current reference is always null. To assess the performance in the the electric machine through a rectangular key. Permanent magnets are
damping quadrant, the quadrature-axis current is assigned through a glued onto this shaft, which is supported by two ball-bearing units
reference damping coefficient cref . According to this method, the elec- axially preloaded by a wave spring. In addition, the electric machine
tromagnetic damping torque of the machine is proportional to the an- stator is locked between the casing and the motor cover using a spacer
gular speed Ωg : to ensure torque transmission.
Tm = τ 2cref Ωg. (34) The pump cover is integrated to a manifold that allows the in-
stallation of the motor-pump unit onto the outer tube of a shock ab-
In this control scheme, the direct and quadrature axes of the ma- sorber. Axial and radial o-rings on the pump and motor side, respec-
chine are loaded with the impedances synthesized by the power stage: tively, guarantee the fluid seal towards the exterior. A position sensor
array is fixed on the motor cover. The sensor and phase wirings are
⎧ R d,ext = −Vd/ Id ,
brought outside through four sealing cable glands.
⎩ R q,ext = −Vq / Iq
⎨ (35)

where V and I denote voltage and current values and subindexes d and q 4. Experimental validation
indicate direct or quadrature axis, respectively. Since current is ideally
null for d , the impedance R d,ext is infinite and thus, the direct axis is in The experimental characterization of the prototype aims to validate

24
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

Fig. 15. Test rig setup. Motor-pump prototype (1), power stage (2), driving motor-pump
unit (3), current probe (4), hydraulic lines (5), hand pump to fill the circuit (6), battery
array (7), data logging PC (8), driving motor switchboard (9), control PC (10), gas-loaded
accumulator (11).

connected to four lead acid car batteries (12 V) in series. A PC is used to

monitor and control both the driving unit inverter (via RS232) and the
prototype power stage (via CAN). The test rig also includes different
Fig. 13. Isometric cut view of the motor-pump unit. Pump port plugs (1), pump cover sensors to characterize the prototype motor-pump unit. Two pressure
manifold (2), casing (3), gerotor gears (4), ball bearings [× 2 ] (5), stator (6), motor spacer sensors (GEFRAN TK) measure both branches of the hydraulic circuit.
(7), wave spring (8), position sensor (9), drain cap (10), cable gland [× 4 ] (11), motor The flow rate is measured with a turbine flow meter (KEM HM11).
cover (12), permanent magnet [× 10 ] (13), winding coil [× 6 ] (14), key (15), rotor shaft Battery voltage and current are also obtained with dedicated probes.
(16).
The angular speed of the prototype is monitored from one of the Hall
latches with a voltage probe. These signals are recorded in time using a
the proposed design and identify the main aspects that hamper its PC equipped with LMS SCADAS data acquisition hardware.
performance. This task is carried out with the system layout shown in The testing procedure is as follows:
Fig. 14. A driving motor-pump unit is connected to the prototype with
two hydraulic lines. The speed of the driving unit is controlled with a 1. The driving motor is controlled to work at a constant speed value.
dedicated inverter, while the prototype unit is electrically connected to 2. The power stage imposes a set of reference damping coefficients on
its power stage fed by a battery. To replicate the behavior of the pro- the prototype, starting from zero. This allows to cover different
totype on a shock absorber, one of the hydraulic lines is connected to a points of the characteristic.
gas-loaded accumulator. This allows to stabilize the low-pressure side 3. For each imposed damping coefficient, sensor measurements are
and avoid pump cavitation. During the tests, the driving unit imposes a acquired.
hydraulic flow rate, which in turn induces an angular speed on the 4. The driving motor speed is changed and the acquisition procedure is
prototype rotor. If the power stage simultaneously sets constant repeated (steps 1–3).
damping coefficients on the tested electric machine, it is possible to 5. In post-processing, the sensor data is interpolated and subsequently
attain different points in the force-speed plane and characterize the used to calculate the different efficiency contributions according to
prototype efficiency therein. Eqs. (1)–(4).
The complete test rig setup appears in Fig. 15. The driving system 6. The interpolated sensor data is used to calculate the different effi-
uses an inverter-controlled Kollmorgen AKM42G brushless PM motor ciency contributions according to Eqs. (1)–(4).
coupled to a Casappa PLP10 pump. The hydrostatic circuit is devised 7. The efficiency values can be plotted on a force-speed plane. Eq. (25)
with two G3/8 flexible pipelines and a FOX GR2 accumulator. The is useful in this regard to convert hydraulic variables into the linear
length, cross section and profile of the hydraulic circuit is carefully mechanical domain.
chosen to minimize unwanted pressure losses. Calculations in the
worst-case condition demonstrate that the used layout introduces a The characterization of the motor-pump unit is carried out in a
global loss (friction and minor components) below 5% of the measured single force-speed quadrant because the device intrinsically yields the
pressure drop. same hydraulic behavior in both directions.
The prototype unit is controlled by its dedicated power stage Fig. 16 exhibits the experimental map for each efficiency con-
tribution: volumetric, hydro-mechanical and electrical. The normalized
speed of the characterized area starts at 15%. This value is dictated by
the minimum measurable flow rate. By converse, the maximum speed
of 40% is related to the angular speed limit of the driving motor-pump
unit. The characterized area covers force values between 30.6% and
142.1%, which confirms the ability of the device to span different
damping characteristics. Furthermore, the maximum force of the pro-
Fig. 14. Test rig layout. Inverter (1), driving motor (2), driving pump (3), hydraulic lines totype is kept below 150% to limit the pressure drop and the motor
(4), prototype pump (5), prototype motor (6), gas-loaded accumulator (7), prototype current.
power stage (8), battery array (9). Experimental results show that the volumetric efficiency tends to

25
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

Fig. 16. Experimental volumetric, hydro-mechanical and electrical efficiency maps of the motor-pump prototype. Axis units are normalized with respect to Fp,max and vp,max .

Table 2
Statistic features of the experimental efficiency map: average (avg), standard deviation
(std), minimum (min), maximum (max) and optimal (opt). The optimal values are ob-
tained in the point where ηt = maxηt . All the values are percentages.

Feature Volumetric Hydro-mechanical Electrical Total

avg 85.6 44 78.3 28

std 6 17.4 7.5 8.4
min 69.7 0 50.1 0
max 98.5 80 100 41.7
opt 85.5 62.4 78.2 41.7

contains the average, standard deviation, minimum, maximum and

optimal values of the total conversion efficiency and its components for
the characterized envelope. It is observed that, for the tested region, the
volumetric and electrical terms keep a standard deviation below 8%
and relatively low difference between the maximum and minimum
values (below 50%). Conversely, hydro-mechanical efficiency has a
deviation of 17.4% and sweeps a wide range between 0% and 80%. The
steadiness of volumetric and electrical results is expected because the
characterized region does not cover very large damping coefficients.
Larger leakage and electrical losses will appear at damping values that
overcome those of the present experimental characterization. Never-
theless, if the control strategy prioritizes energy regeneration, the
Fig. 17. Experimental total conversion efficiency map of the motor-pump prototype. Axis damping request will be set predominantly near the maximum-effi-
units are normalized with respect to Fp,max and vp,max . ciency region of the map in Fig. 17. In this context, it is seen that the
maximum conversion efficiency assumes an optimal value of 41.7%.
Therein, the hydraulic losses correspond to 15% of the input power,
decrease as the damping value increases due to leakages. In contrast,
electrical losses are 12% and hydro-mechanical losses assume a larger
the hydro-mechanical efficiency is null-valued in the lower limit of the
portion of 32%. This fact indicates that the most critical aspect of the
map. This purely-dissipative behavior is attributed primarily to viscous
design regards the hydro-mechanical efficiency.
drag. As the damping increases, the drag plays a less-substantial role in
This statement is confirmed by the total efficiency map obtained
the total power. Conversely, electrical losses due to Joule effect dom-
from numerical analysis, as shown in Fig. 18. The maximum conversion
inate as the damping rises because the electric machine works closer to
efficiency in this case is 48.7%, against 41.7% obtained in the experi-
the short-circuit damping condition. In contrast, when the reference
ments. The difference can be mainly attributed to larger hydro-me-
damping is null, the power stage imposes null current, i.e. infinite im-
chanical losses. This phenomenon is expected because the pump gears
pedance, and the electrical efficiency tends to unit.
can move axially and radially within the clearance gaps, thus increasing
Consequently, the total conversion efficiency map in Fig. 17 depicts
viscous drag. As stated in Section 3, these phenomena are not re-
the superposition of all these contributions: its low-region behavior is
produced by the used simulation tools. At 40% of speed, the minimum
dominated by hydro-mechanical losses. Then, the total conversion ef-
force, which is dictated by these losses, reaches 34.8%. In experiments,
ficiency reaches a maximum and then drops again due to the influence
this force is larger (70%). Thus, the damping span of the device is
of leakages and electrical losses.
shifted to higher force values, as observed in Fig. 17, and the total
Table 2 proposes a quantitative analysis of the obtained data. It
conversion efficiency is penalized.

26
R. Galluzzi et al. Applied Energy 210 (2018) 16–27

contract 236/2016 between Politecnico di Torino and the funder.

References

[1] EC Regulation No 443/2009. Setting emission performance standards for new

passenger cars as part of the Communitys integrated approach to reduce CO2
emissions from light-duty vehicles.
[2] Zuo L, Zhang PS. Energy harvesting, ride comfort, and road handling of regenerative
vehicle suspensions. J Vib Acoust 2013;135(1):011002.
[3] ISO 8608:1995, rev. 2011. Mechanical vibration – road surface profiles – reporting
of measured data.
[4] Galluzzi R, Tonoli A, Amati N, Curcuruto G, et al. Regenerative shock absorbers and
the role of the motion rectifier. SAE paper 2016-01-1552; 2016.
[5] Karnopp D. Permanent magnet linear motors used as variable mechanical dampers
for vehicle suspensions. Veh Syst Dynam 1989;18(4):187200.
[6] Gysen BLJ, van der Sande TPJ, Paulides JJH, Lomonova EA. Efficiency of a re-
generative direct-drive electromagnetic active suspension. IEEE Trans Veh Technol
2011;60(4):138493.
[7] Li Z, Zuo L, Kuang J, Luhrs G. Energy-harvesting shock absorber with a mechanical
motion rectifier. Smart Mater Struct 2013;22(2):025008.
[8] Zhang Z, Zhang X, Chen W, Rasim Y, Salman W, Pan H, et al. A high-efficiency
energy regenerative shock absorber using supercapacitors for renewable energy
applications in range extended electric vehicle. Appl Energ 2016;178:177188.
[9] Kawamoto Y, Suda Y, Inoue H, Kondo T. Electro-mechanical suspension system
considering energy consumption and vehicle manoeuvre. Veh Syst Dynam
2008;46(S1):10531063.
[10] Tonoli A, Amati N, Girardello Detoni J, Galluzzi R, Gasparin E. Modelling and va-
lidation of electromechanical shock absorbers. Veh Syst Dynam
2013;51(8):1186–99.
Fig. 18. Theoretical total conversion efficiency map of the motor-pump prototype. Axis [11] Yu C, Rakheja S, Wang W, Wang Q. Analysis of an on-off regenerative electro-
units are normalized with respect to Fp,max and vp,max . mechanical damper. Proc Inst Mech Eng Part D: J Autom Eng 2013;227(5):704722.
[12] Liu Y, Xu L, Zuo L. Design, modeling, lab and field tests of a mechanical-motion-
rectifier-based energy harvester using a ball-screw mechanism. IEEE/ASME Trans
The obtained results highlight the importance of the experimental Mech 2017;PP(99):1.
[13] Galluzzi R, Tonoli A, Amati N. Modeling, control and validation of electrohydro-
characterization of the prototype: hydro-mechanical efficiency is static shock absorbers. J Vib Acoust 2015;137:011017.
identified as the main responsible of performance reduction. This [14] Fang Z, Guo X, Xu L, Zhang H. Experimental study of damping and energy re-
phenomenon is more evident in the real prototype due to the limita- generation characteristics of a hydraulic electromagnetic shock absorber. Adv Mech
Eng 2013;5:943528.
tions of the used design tools. Hydro-mechanical efficiency can be [15] Li C, Zhu R, Liang M, Yang S. Integration of shock absorption and energy harvesting
further enhanced by introducing roller bearing elements on the peri- using a hydraulic rectifier. J Sound Vib 2014;333(17):390416.
meter of the pump outer gear, reducing the contact surface between the [16] Zhang Y, Chen H, Guo K, Zhang X, Eben Li S. Electro-hydraulic damper for energy
harvesting suspension: modeling, prototyping and experimental validation. Appl
gears and the case and improving radial and axial pressure balance on
Energ 2017;199:112.
both pump gears. [17] Giovanardi M, Tucker M, Wendell RJ, Anderson ZM, et al. Integrated active sus-
pension smart valve. Patent 0265169A1, USA; 2014.
[18] O’Shea C, Xia Y, Lowry S. Analysis and optimization of an electrohydraulic power
5. Conclusions
pack for use in a fully-active vehicle suspension through the use of computational
fluid dynamics. In: ASME/BATH 2013 symposium on fluid power and motion
The present paper described in detail a method to design a motor- control, Sarasota, FL, USA, 6–9 October 2013, paper no. 4402. ASME Digital
pump unit for hydraulic regenerative shock absorbers. A system-level Collection; 2013.
[19] Rundo M, Nervegna N. Lubrication pumps for internal combustion engines: a re-
approach was followed to guarantee the damping functionality and view. Int J Fluid Power 2015;16(2):5974.
maximize the energy regeneration capability of the system, while sa- [20] Nichols Portland. Gerotor selection and pump design. Report, USA.
tisfying design constraints such as geometric envelope and supply vol- [21] Altare G, Rundo M. CFD analysis of gerotor lubricating pumps at high speed: geo-
metric features influencing the filling capability. In: ASME/BATH 2015 symposium
tage. After a mechanical packaging step, a prototype was built and on fluid power and motion control, Chicago, IL, USA, 12–14 October 2015, paper
experimentally characterized. The obtained results were post-processed no. 9539. ASME Digital Collection; 2015.
in the form of efficiency maps. They put in evidence the role of the [22] Simerics Inc. Predicting pump leakage with computational fluid dynamics. White
paper, USA; 2014.
different components on the total conversion efficiency. The prototype [23] Ding H, Visser FC, Jiang Y, Furmanczyk M. Demonstration and validation of a 3D
was able to span a considerable portion of the force-speed plane and CFD simulation tool predicting pump performance and cavitation for industrial
yielded a maximum conversion efficiency of 41.7%. Furthermore the applications. J Fluids Eng 2011;133(1):011101.
[24] Dixon JC. The shock absorber handbook. 2nd ed. Warrendale, PA, USA: Society of
hydro-mechanical losses were identified as the most critical aspect of
Automotive Engineers; 1999.
the design. [25] Bianchi N, Dai Prè M, Alberti L, Fornasiero E. Theory and design of fractional-slot
PM machines. IEEE-IAS tutorial course notes. Italy: University of Padova; 2007.
[26] Hanselman DC. Brushless permanent magnet motor design. 2nd ed. Lebanon, OH,
Acknowledgments
USA: Magna Physics Publishing; 2006.
[27] Boglietti A, Cavagnino A, Staton D, Shanel M, Mueller M, Mejuto C. Evolution and
The authors would like to gratefully thank Piero Conti, Giordano modern approaches for thermal analysis of electrical machines. IEEE Trans Ind
Greco, Andrea Nepote and Fabio Cotto from Magneti Marelli for their Electr 2009;56(3):871–82.
[28] Hargreaves PA, Mecrow BC, Hall R. Calculation of iron loss in electrical generators
valuable support throughout this research activity. This work was funded using finite-element analysis. IEEE Trans Ind Appl 2012;48(5):14601466.
by Magneti Marelli Shock Absorbers in the context of the research

27