Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 67 (2018) 81 – 86

11th CIRP Conference on Intelligent Computation in Manufacturing Engineering, CIRP ICME ‘17

State of the art and optimization of the energy flow in cooling systems of
motorized high-speed spindles in machine tools
Juliane Webera,*, Linart Shabia, Jürgen Webera
a
Institute of Fluid Power, TU Dresden, Helmholtzstraße 7a, 01069 Dresden, Germany

* Corresponding author. Tel.: +49-351-463-33320; fax: +49-351-463-32136. E-mail address: juliane.weber@tu-dresden.de

Abstract

In high-speed cutting thermo-elastic deformations of machine tool structures and motor spindles are main causes for manufacturing inaccuracies
because they lead to a TCP displacement. Therefore, cooling systems are essential for controlling thermo-elastic properties.
The paper describes different cooling system designs of motor spindles. Their thermal energy flow is analyzed by CFD simulation, and measures
improving the heat rejection are discussed. For future work, such models will be integrated into a virtual machine tool that enables the
estimation/control of the machines’ thermal behavior.
©2017
© 2017TheThe Authors.
Authors. Published
Published by Elsevier
by Elsevier B.V.
B.V. This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the scientific committee of the 11th CIRP Conference on Intelligent Computation in Manufacturing
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Engineering.
Peer-review under responsibility of the scientific committee of the 11th CIRP Conference on Intelligent Computation in Manufacturing Engineering

Keywords: Machine tool; Cooling system design; Motor spindle; CFD simulation; Heat transfer; Energy flow; Roughness.

1. Introduction can be controlled more efficiently in order to reduce thermo-

elastic deformations and the necessary energy demand.
In production technology, the demands on productivity,
accuracy and power efficiency have grown steadily in recent Nomenclature
years. Especially in high-speed cutting (HSC) thermo-elastic cp Specific heat capacity (J∙kg-1∙K-1)
deformations of the machine tool structure lead to a TCP-
݀௛ Hydraulic diameter (mm)
displacement and therefore, to dimensional and form errors of
D / ‫ܦ‬ௐ Mean curvature / mean diameter (mm)
the workpiece. Accordingly, they have a negative influence on
K Sand-grain roughness (µm)
manufacturing quality and productivity. To counteract this
problem, entire production areas were increasingly air- l Flow length (m)
conditioned or long warming-up phases were included before Nu Nusselt number (-)
starting the manufacturing process – only to guarantee P/ Q̇ Electrical power (kW) / heat flux (kW)
homogeneous boundary conditions. However, such p Pressure (bar)
conventional measures inevitably lead to increased energy Pr Prandtl number (-)
requirements. Therefore, within the Collaborative Research Re Reynolds number (-)
Center CRC/TR 96 the investigations focus on thermally ܴ௭ Peak-to-valley roughness (µm)
unsteady boundary conditions, where correction and T, ߴ Temperature (K, °C)
compensation measures influencing thermo-elastic properties v Flow velocity (m∙s-1)
of machine tools are developed, which ensure high ܸሶሶ Volume flow rate (l∙min-1)
manufacturing accuracy without an additional energy demand ߙ Heat transfer coefficient HTC (W∙m-2∙K-1)
[6]. In this respect, fluid cooling systems of machine tools play ߞሺௐሻ Darcy friction factor for straight (coiled) pipe (-)
a central role. With the optimization of their design and ߟ Dynamic viscosity (mPa∙s)
operation mode the heat fluxes within the machine structure

2212-8271 © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the scientific committee of the 11th CIRP Conference on Intelligent Computation in Manufacturing Engineering
doi:10.1016/j.procir.2017.12.180
82 Juliane Weber et al. / Procedia CIRP 67 (2018) 81 – 86
ߣ Thermal conductivity (W∙m-1∙K-1)
ߩ Density (kg∙m-3)
Quasi-Series Parallel Single Helical Double Helical
Solid
. Domain
Fluid
Fig. 1. Typical designs of stator cooling sleeves and resulting flow paths in
the fluid domain.
2. State of the art
Previous studies focused on experimental investigation of
entire fluid cooling systems in modern machine tools [12, 13].
Here, and in accordance with [2], it was found, that the motor
spindle produces the largest heat amount because of losses in
x the built-in motor due to copper, iron and stray losses,
x the bearings depending on speed, preload and lubrication
x and the rotating elements due to viscosity shear of air.
Since the precision of machine tools considerably depends on
the relative motion between tool and workpiece, TCP displace-
ments resulting from thermal deformations in the main spindle
have to be avoided primarily. Therefore, an effective cooling
of the motor spindle with its heat losses is indispensable.
In [14, 15] the entire fluid systems and the basic structure of
a motor spindle with its main heat sources was already
introduced. Beside the hydraulic and the inner coolant supply
system for tool and process cooling, the cooling system is used
for cooling the stator and the bearings. The cooling system
itself is realized by a cooling liquid that directly flows through
the spindle shaft or a cooling sleeve [10]. To realize a good heat
transfer between the solid bodies, the stator is press-fitted into
the cooling sleeve, which consists of a thermally conductive
material. In addition, the sleeve may have ribs on its outer
radius improving the heat transfer into the fluid. Due to rib
structures, the design of the flow channels can vary
significantly, fig. 1 shows the most common ones. A similar
overview is already introduced by Gebert in [5], where beside
non-ripped outer contours helical and parallel channel
structures are mentioned. Nevertheless, the focus of his work is
on the enhancement of the spindle bearings’ life cycle in order
to reduce the TCP displacement. Further works deal with the
analysis and improvement of the performance of motor
spindles [3, 9]. Here, different spindle designs are introduced,
but the focus is on the investigation of the spindle bearings;
their design, lubrication, preload and resulting heat losses. In
[4] Chien and Jang analyze the heat transfer in double helical
cooling channels. They carry out a comparison between
experimental data and numerical simulations to determine the
temperature distribution in the fluid flow. Also in [7] helical
channel structures are examined in detail. The focus of this
work is the cooling efficiency depending on the adjustment of
the channel slope and width. A detailed analysis of the fluid
flow and heat transfer in helical tubes with different curvature
ratios is performed in [18]. From this analysis, it is found that
centrifugal forces due to channel curvature enhance the
pressure drop and heat transfer in comparison with straight
channels. Furthermore, Krishna experimentally investigates
single helical channels with different helix angles in order to
calculate the friction factor and to predict the pressure drop
depending on the flow characteristic (laminar / turbulent) [8].
The influence on the heat transfer is not discussed.
In addition to literature research, existing motor spindles
were analyzed, especially considering the design of their
cooling systems. Therefore, existing spindles at the Institute of
Fluid Power and further co-operating institutes within the
CRC/TR 96 (Chair of Machine Tools Development and
Adaptive Controls at Dresden University of Technology and
Laboratory for Machine Tools and Production Engineering of
RWTH Aachen University) were investigated. This turned out
that a meandering fluid flow according the quasi-series design
of the cooling sleeve (see fig. 1) is another common principle
in current motor spindle designs.
This paper is focusing on the comparison between the most
common cooling sleeve designs; single and double helical
channel structure as well as quasi-series structure. Therefore, a
3D numerical CFD analysis is carried out for all designs.
Furthermore, possibilities for the enhancement of the heat
transfer between solid and fluid are analyzed.
3. Modelling setup
3.1. General setup
The general CFD workflow was described in [14]. Three
major phases can be identified: preprocessing, solution and
post processing. Within preprocessing particular attention is
paid to the meshing quality. A grid independency study is
performed for the single and double helical channel structure
[14, 15]. Here it is shown that the fluid domain should have
minimum eight inflation layers with a first layer thickness of
10 µm. Furthermore, the solid bodies have to be considered
within simulation model due to effects of heat conduction
(conjugate heat transfer CHT). In addition, a study on the
parallel performance with different numbers of processors is
carried out in order to save additional computation time [15].
This is of special importance executing variation studies e. g.
in order to optimize geometry parameters.
Another step within preprocessing is the definition of boun-
dary conditions, explained in detail in chapter 3.2. Before
starting the solution process, the solver and physics are defined
according the problem definition. As already discussed in
[14, 15] the shear stress transport formulation (SST model) is
used. For temperature prediction within the fluid, the heat
transfer is taken into account by the energy transport equation.
Due to simplifications (see 3.2) this equation is transformed to
eq. (1). In addition to forced convection within the fluid, heat
conduction in the solid domains is taken into account (CHT).
Since the bodies are rigid, no inner energy sources exist and
steady state is considered, the conjugate heat transfer can be
calculated according the simplified equation (2).
ࢺ ή ሺߩࢁ݁ሻ ൌ ࢺ ή ሺߣࢺܶሻ ൅ ࣎ǣ ࢺࢁ (1)
Ͳ ൌ ࢺ ή ሺߣࢺܶሻ (2)
 Juliane Weber et al. / Procedia CIRP 67 (2018) 81 – 86 83

Furthermore, different timescale factors are specified on an x The physical properties of the solids (heating element,
equation-class basis in order to minimize computing time until cooling sleeve, housing) are assumed to be isotropic and
convergence. This is necessary to accelerate the energy temperature independent.
equation that normally converges much slower (especially
within solid regions) compared to momentum and continuity ఎ ್
ൌ ݁ ஺ା೅ (3)
equations within the fluid. To speed up CHT simulation the ሾ௠௉௔௦ሿ
energy equation is calculated with a time scale factor that is 50
times higher than the other ones. 1
After the solution process and writing out a result file, this 2
.
file is post processed in the last step (e. g. in CFD Post) to Q 3
visualize the field sizes and variables of interest.
4

3.2. Boundary conditions 5

6
The CHT model is build up according the developed test rig 7
that is depicted in fig. 2 and described in detail in [11, 14]. 8
Based on its modular setup, the cooling sleeve (2) can be
9
replaced easily to examine different designs. Moreover, the
heating element (3) with two centrally introduced heating 10

cartridges (4) substitutes the heat sources of the basic spindle M 11

configuration. This simplification allows a precise definition of Thermocouple mounting 12
the introduced heat known from the adjusted power ൌܳሶ at 1) Housing 4) Heating cartridge 7) Pressure sensor 10) Pump
the heating cartridges. In addition, influences of the 2) Cooling sleeve 5) End face isolation 8) Temperature sensor 11) Chiller
3) Heating element 6) Flow sensor 9) Pipe to reservoir 12) Throttle
environment are reduced by thermal insulations (5). The results
presented in this paper apply to the following modelling Fig. 2. CAD-model of the test rig and derived CHT model scope.
assumptions that are summarized in fig. 3:
x The fluid flow is considered steady state & incompressible. 1
x Gravity and buoyancy effects are neglected. 2
x The volumetric flow rate is ܸሶ ൌ ͳͶ݈Ȁ݉݅݊ and the fluid has
c 3
a temperature of ߴ௜ ൌ ʹͲι‫ ܥ‬at the velocity inlet.
x There is no heat transfer via front faces (adiabatic walls) F
because of the isolation (5). .
x A maximum heat flux of ܳሶ ൌ ʹ͹ͷ͸ܹ is introduced at the Q
2 13
3 41
14
1 15
5
1 12
0 11
inner walls of the heating element. 8 9
10
B
x Heat is dissipated from the outer surface of the housing to b 5 6 7
3 4
c
the ambient air (ߴ௢ ൌ ʹͲι‫ )ܥ‬by means of free convection 1 2

with ߙ ൌ ͷ ܹ Τ݉ଶ ‫ܭ‬. a Boundary

ry
y conditions

x The static pressure is assumed to be constant over the A A Ve

Velocity
ocity inlet:
B Pressure outlet:
,

pressure outlet and equals the environmental pressure. a Adiabatic wall

b Heat flux:
c Convection: ,
3.3. Cooling medium and material properties
1 Housing | 2 Cooling sleeve | 3 Heating element | F Fluid

First, the temperature dependency of the fluid and material Fig. 3. CFD model and boundary conditions (single helical cooling channel).
properties like dynamic viscosity ߟ, density ߩ, specific heat
Table 1. Fluid properties (dynamic viscosity ߟ, density ߩ, spec. heat capacity
capacity ܿ௣ and thermal conductivity ߣ is investigated. As
ܿ௣ , thermal conductivity ߣ) for different concentrations of AntifrogenN®.
shown in [11, 14, 15] only the dynamic viscosity ߟ has to be
conc. ߟ ߩ ܿ௣ ߣ
considered temperature dependent due to large changes
(34.8 % for pure water) within the considered temperature % A [-] b [K] [kg∙m-3] [J (kg∙K)-1] [W (m∙K)-1]
range ʹͲι‫ ܥ‬൑ ߴ ൑ ͷͲι‫ ܥ‬. Thus, the dynamic viscosity ߟ is 0 -6.568 1925.4 995.5 4.15 0.61
modelled in CFX expression language (CEL) considering the 20 -6.956 2207.5 1022.3 3.92 0.52
30 -7.163 2344.6 1034.8 3.75 0.48
Andrade equation (3) with the fluid specific parameters given
50 -7.725 2687.0 1059.8 3.36 0.41
in table 1 for different water AntifrogenN® mixtures. All other
quantities are considered with their mean values within this Table 2. Material specific properties of solid bodies (HE…Heating element,
temperature range. In table 1 the fluid and in table 2 the solid CS…Cooling sleeve, H…Housing).
properties are summarized. To summarize, the following
Body Material ߩ ܿ௣ ߣ
assumptions are made:
[kg∙m-3] [J (kg∙K)-1] [W (m∙K)-1]
x The specific heat capacity ܿ௣ , density ߩ and thermal HE AlMgSi1 2702 903 185
conductivityߣ are temperature independent, whereas CS CuZn39Pb3 8470 377 123
x the dynamic viscosity ߟ is temperature dependent. H X8CrNi 7854 500 15
84 Juliane Weber et al. / Procedia CIRP 67 (2018) 81 – 86

Table 3. Energy flow (heat fluxes) in the fluid and solid domains across the sand-grain roughness ‫ ܭ‬required for the model and measured
interfaces (Heating Element, Cooling Sleeve, Housing, Fluid). surface roughness ܴ௭ is depicted in fig. 4.
Single helix Double helix Quasi-series
[W] [%] [W] [%] [W] [%] 4. Results
HE (3) → CS (2) 2756.0 100 2756.0 100 2756.0 100
CS (2) → H (1) 115.8 4.2 100.0 3.6 127.6 4.6
4.1. Comparison of different flow channel designs
CS (2) → F 2640.2 95.8 2656.7 96.4 2628.0 95.4
F ← CS (2), radial 1866.9 67.7 1796.6 65.2 1836.0 66.7
F ← CS (2), axial 773.3 28.1 860.1 31.2 792.0 28.7
Fig. 5 depicts the temperature development along the flow path
F ← H (1) 113.9 4.1 97.3 3.5 125.8 4.6 comparing the quasi-series, the single and double helical
F ← Viscous heating 15.7 0.6 17.3 0.6 11.2 0.4 cooling channel design. In the quasi-series channel higher
H (1) → Output 1.9 0.1 2.7 0.1 1.9 0.1 fluctuations occur, because the volume flow is first divided in
H (1) → F 113.9 4.1 97.3 3.5 125.8 4.6 two symmetrical flows within each channel segment and
afterwards combined in the passage to the next channel
segment. However, the solid temperatures in axial direction are
approximately equal in trend and differences comparing the
different designs. Due to reverse flow within the double helical
channel, the maximum temperature difference οܶ‫ ܪܦ‬ൌ ʹǤ͹‫ ܭ‬is
about 1 K smaller than the quasi-series design (due to double
pitch length, the maximum axial dimension of the sleeve is
already reached after half of the flow path, then the flow is
deflected between position 15 and 14). Fig. 6 shows the Nusselt
numbers ܰ‫ ݑ‬and the heat transfer coefficients ߙ of the
Fig. 4. Translation between measured surface roughness ܴ௭ and modeled different designs. However, comparing the different designs
sand-grain roughness ‫ܭ‬, following [1]. one can see that the heat transfer is almost equal. Nevertheless,
the single helix has the best heat transfer coefficients and the
quasi-series design the worst. Due to mutual impacts of the cold
inflow and the warmed backflow within the double helix, the
heat transfer strongly decreases in the backflow.
Fig. 7 depicts a vector plot of the resulting energy flow (heat
fluxes) within the fluid and solid domains. It is shown
exemplarily for the single helical design, because the directions
are quite the same and only the amounts are varying according
table 3. It is seen that about 96 % of the heat introduced into
the cooling sleeve is directly dissipated by the fluid and only
about 4% is conducted through the webs (ribs) to the housing.
Although these webs have the same dimensions, the conducted
Fig. 5. Temperature distribution ߴ in direction of the flow path at the cooling
sleeve (CS) and within the fluid (F).
heat amount varies about 1% between the double helix and the
quasi-series design. With regard to the housing, almost all heat
can be dissipated by the fluid at the inner housings wall (forced
convection), because the heat transfer coefficient at the
housings outer wall (free convection) is about magnitudes
lower. Focusing on the fluid, about two thirds of the heat is
dissipated in radial direction and about one third in the axial
one, whereas the surface ratio is 3 : 2 (radial : axial). This
difference is caused by increasing influence of the heat
conduction within the webs ((ribs). )

Fig. 6. Nusselt number ܰ‫ ݑ‬and heat transfer coefficient ߙ in direction of the 3

flow path for the forced convection within the fluid.
2

3.4. Surface Roughness

F
As known from [16, 17] defined surface structures (e. g.
protrusions or dimples) ensure homogeneous flow distributions axial
radial

avoiding recirculation zones to reduce the pressure drop and

1
they induce additional shear flow enhancing the heat transfer to
the fluid. To analyze the influence of wall conditions on the
heat transfer into the fluid, first, a rough estimation is made, Interfaces Heat
att fluxes
solid–solid Input Forced convection
where the wall roughness is varied within the numerical solid–fluid Conduction Free convection
simulation model. Therefore, the fluidic wall is modeled as
smooth (‫ ܭ‬ൌ Ͳߤ݉) and rough wall with sand-grain roughness Fig. 7. Energy flow in the fluid and solid domains of the single helical design,
interfaces and boundary conditions.
in the range of ʹͷߤ݉ ൑ ‫ ܭ‬൑ ͳͲͲͲߤ݉. The correlation between
 Juliane Weber et al. / Procedia CIRP 67 (2018) 81 – 86
Fig. 8. Resulting pressure drop ο‫ ݌‬depending on surface roughness ‫ ܭ‬and
cooling sleeve design.
Fig. 9. Nusselt number ܰ‫ ݑ‬and heat transfer coefficient ߙ depending on
surface roughness ‫ ܭ‬and cooling sleeve design.
Altogether, it can be confirmed that only small deviations
exist between the different cooling sleeve designs regarding the
heat transfer. Although the heat transfer coefficient in the
backflow of the double helix decreases to the smallest value of
all designs, more heat can be dissipated directly in the cooling
sleeve by the fluid. Nevertheless, the viscous heating has its
maximum in the double helix (due to the reflection that causes
large pressure drops) and its minimum in the quasi-series
design.
4.2. Influence of surface roughness
The pressure drop ߂‫ ݌‬in pipes due to viscous effects is
proportional to the flow length ݈ and can be characterized by
the Darcy-Weisbach equation (4). The flow coefficient ߞሺௐሻ , or
also called Darcy friction factor, is a dimensionless quantity.
Primarily its value depends on the flow regime characterized
by the Reynolds number ܴ݁. Since in heat transfer applications
almost turbulent flows are present in order to realize large heat
transfer coefficients, in further considerations, only turbulent
descriptions are taken into account. Here, especially for coiled
pipes the friction factor ߞௐ can be estimated by equation (5)
according to Mishra and Gupta. Herein ݀௛ is the hydraulic
diameter for a rectangular cross-section and D is the mean
curvature diameter of the coiled pipe. Furthermore, for
Reynolds numbers in the range of ͳͲସ ൑ ܴ݁ ൑ ͳͲ଺ equation
(6) of Konakov applies for straight pipes. Nevertheless, these
equations do not consider the pipe’s surface roughness ‫ܭ‬.
ఘ ௩మ
ο‫ ݌‬ൌ ߞሺௐሻ ή ή ή݈
ଶ ௗ೓
଴Ǥଷଵ଺ସ ௗ ଴Ǥହ
ߞௐ ൌ ൤ͳ ൅ ͲǤͲͻͷ ቀ ೓ ቁ ܴ݁ ଴Ǥଶହ ൨
ோ௘ బǤమఱ ஽
ଶ௕௛ ௞ ଶ
with ݀௛ ൌ and ‫ ܦ‬ൌ ‫ܦ‬ௐ ൤ͳ ൅ ቀ ቁ ൨
௕ା௛ గή஽ೈ
ߞ ൌ ሺͳǤͺ ή ݈݃ ܴ݁ െ ͳǤͷሻିଶ
85
ଵ ௄ ଶǤହଵ
ൌ െʹ ή ݈݃ ൬ ൅ ൰ (7)
ඥ఍ ଷǤ଻ଵήௗ೓ ோ௘ඥ఍
ଵ ௄ ଵǤଵଵ ଺Ǥଽ
ൌ െͳǤͺ ή ݈݃ ൤ቀ ቁ ൅ ൨ (8)
ඥ఍ ଷǤ଻ήௗ೓ ோ௘
ଵ ௄ ହǤ଻ସ
ൌ െʹ ή ݈݃ ቂ ൅ ቃ (9)
ඥ఍ ଷǤ଻ήௗ೓ ோ௘ బǤవ
ଵ ௗ
ൌ ʹ ή ݈݃ ቀ ೓ ቁ ൅ ͳǤͳͶ (10)
ඥ఍ ௄
Considering the surface roughness ‫ ܭ‬for calculating the
friction factor, available equations only consider the pipe
geometry to be straight. According to Colebrook et al the so-
called Colebrook equation (7) applies to turbulent flows in the
entire area of smooth and rough surfaces. Due to its implicit
nature it is usually solved numerically. In addition, there are
numerous approximations to calculate the friction factor
directly. These are for example the Haaland equation (8), the
Swamee-Jain equation (9) and eq. (10) for large Reynolds
numbers ܴ݁ ՜ λ according to Prandtl/v. Kármán. All these
three equations consider the surface roughness ‫ ܭ‬. Never-
theless, they do not consider the curvature of the pipe.
Fig. 8 depicts the resulting pressure drop ο‫ ݌‬in dependence
of surface roughness ‫ ܭ‬for each cooling sleeve design.
Additionally, a comparison between simulated and calculated
pressure drops (from the different friction factor relations) is
shown. First, you can see that for smooth surfaces with ‫ ܭ‬ൌ
Ͳߤ݉ the pressure drops are in the same range of 1.5 bar to
2.0 bar. Second, the equations of Haaland and Swamee-Jain
can predict the pressure drops in the helical structures quite
good until roughness of 150 µm. Third, with increasing surface
roughness, the deviations between the helical and the quasi-
series structure become larger. The major influence is the
design of the quasi-series structure, where the fluid flow is
divided in two symmetrical flows within each channel segment.
Thus, the velocity ‫ ݒ‬and the Reynolds numbers ܴ݁ are half of
that in the helical designs. Due to this division, a direct
comparison with the pressure drops calculated from the friction
factor equations is not possible in this case. Therefore, an
additional simulation with double of the flow rate ( ܸሶ ൌ
ʹͺ ݈ Τ݉݅݊) is carried out for the quasi-series design. Now one
can see better that the qualitative trends of all designs are
similar. Nevertheless, the pressure drop within the quasi-series
design is much higher (about a factor of 3.5 compared to the
helical designs and about 4 compared to the quasi-series design
with ܸሶ ൌ ͳͶ݈Ȁ݉݅݊ሻ . On the one hand, this behavior is
comprehensible due to the quadratic dependence between
friction factor and Reynolds number. On the other hand, the
equations cannot predict the simulated pressure drop values
due to neglecting the mixing zones between two channel
segments and the high turbulent effects occurring there.
The influence of the surface roughness ‫ ܭ‬on the heat
transfer regarding Nusselt number ܰ‫ ݑ‬and heat transfer
coefficient ߙ is depicted in fig. 9. With increasing surface
roughness, the Nusselt numbers ܰ‫ ݑ‬are increasing with a
degressive character. Here, a great potential for optimizing the
(4) heat transfer exist until roughness of about 100 µm to 200 µm,
then the heat transfer is rising only slightly. Furthermore, an
(5) increasing surface roughness has a greater influence on the heat
transfer in the double helical and the quasi-series design than
on that in the single helical one. Thus, until 50 µm the heat
transfer in the double helix is the best and in the quasi-series
(6) channel is worst. With increasing roughness, the balance is
86 Juliane Weber et al. / Procedia CIRP 67 (2018) 81 – 86
turning so that the quasi-series design has better heat transfer
than the single helical structure.
5. Summary
Thermo-elastic deformations negatively affect manufactu-
ring accuracy and production quality, especially in HSC
applications. Thus, precise machining requires a temperature
control, in particular of the main spindle. The focus of this
paper is the comparison of three common cooling sleeve
designs within a main spindle of a machine tool – a single and
double helix as well as a quasi-series design. Thereby a special
focus is on the heat transfer and the resulting pressure loss that
leads to viscous heating within the fluid itself. The method used
in this paper mainly consists of numerical simulation models.
Therefore, a validation with experimental results was carried
out for the single and double helix, explained in
[11, 13, 14, 15].
The results presented in this paper show that in case of a
smooth wall the single helix design has the best heat transfer
coefficients and the quasi-series design the worst. With respect
to the energy flow within the fluid and the solid bodies, only
small deviations exist between the different cooling sleeve
designs. The heating due to inner viscous friction of the fluid
has its maximum in the double helix. Nevertheless, its share to
the total energy balance is only about 1 %, and therefore
negligible. For optimizing the heat transfer to the fluid
increasing surface roughness offer a great potential. For
example with roughness of 100 µm the heat transfer coefficient
is more than doubled. This effect of increasing heat transfer due
to increasing surface roughness is greatest within the double
helix and smallest in the single helix design.
Acknowledgements
The presented research activities are part of the project
“Thermo-energetic description of fluid systems” (Ref. No.
CRC/TR 96, A04). The authors would like to thank the German
Research Foundation (DFG) for financial support.
Funded by
References
[1] Adams T, Grant C, Watson H. A Simple Algorithm to Relate Measured
Surface Roughness to Equivalent Sand-grain Roughness. International
Journal of Mechanical Engineering and Mechatronics 2012; Volume 1:
66-71.
[2] Bossmanns B, Tu JF. A thermal model for high speed motorized
spindles. In: International Journal of Machine Tols and Manufacture
1999; Volume 39: 1345 – 1366.
[3] Butz F. Gestaltung der Loslagerung von Werkzeugmaschinenspindeln.
Dissertation. Aachen: Shaker Verlag; 2007.
[4] Chien CH, Jang JY. 3-D numerical and experimental analysis of a built-
in motorized high-speed spindle with helical water cooling channel.
Applied Thermal Engineering 2008; Volume 28: 2327-2336.
[5] Gebert K. Ein Beitrag zur thermischen Modellbildung von
schnelldrehenden Motorspindeln. Darmstädter Forschungsberichte für
Konstruktion und Fertigung. Aachen: Shaker Verlag; 1997.
[6] Großmann K, Ott G. Introduction. Thermo-energetic Design of Machine
Tools. A Systemic Approach to Solve the Conflict Between Power
Efficiency, Accuracy and Productivity Demonstrated at the Example of
Machining Production. Berlin: Springer Verlag; 2014: 1-11.
[7] Javelov I. Projektierung von Kühlsystemen für Elektrospindeln. Stanki i
Instrument 1983; Volume 54(4):25-26.
[8] Krishna B. Prediction of pressure drop in helical coil with single phase
flow of non-newtonian fluid. International Journal of Applied Research
in Mechanical Engineering 2012; Volume 2: 31-36.
[9] Tüllmann U. Das Verhalten axial verspannter, schnelldrehender
Schrägkugellager. Dissertation. Berichte aus der Produktionstechnik
1999, Volume 99(25).
[10] Walter R. Mit direkter Kühlung zu mehr Genauigkeit. Werkstatt und
Betrieb 2006; Volume 139(6): 129-130.
[11] Weber J, Weber J. Thermo-energetic analysis and simulation of the
fluidic cooling system of motorized high-speed spindles. 13th
Scandinavian International Conference on Fluid Power 2013; Linköping,
Schweden.
[12] Weber J, Weber J. Thermo-Energetic Analysis of the Fluidic Cooling
Systems in Tooling Machines. 9th International Fluid Power Conference
2014; Aachen.
[13] Weber J, Weber J. Thermo-energetic Modelling of Fluid Power Systems.
Thermo-energetic Design of Machine Tools. Heidelberg: Springer
Verlag; 2015: 49-60.
[14] Weber J, Shabi L, Weber J. Thermal Impact of Different Cooling
Sleeve’s Flow Geometries in Motorized High-speed Spindles of
Machine Tools. Proceedings of the ASME 2016 9th FPNI Ph.D
Symposium on Fluid Power 2016; Florianópolis, Brazil.
[15] Weber J, Shabi L, Weber J. Kühlstrukturen in Hauptspindeln von Werk-
zeugmaschinen. 34. CADFEM Ansys Conference 2016; Nürnberg.
[16] Su B. High-Resolution Temperature Measurement during Forced
Convective Heat Transfer at a Wall with a Dimple Structure.
Dissertation; Darmstadt: 2015.
[17] Turnow J. Heat transfer and vortex structures in corrugated plate heat
exchangers. Rostock: 2014.
[18] Yoo G, Choi H, Dong W. Fluid flow and heat transfer characteristics of
spiral coiled tube. Effects of Reynolds number and curvature ratio.
Journal of Central South University 2012; Volume 19: 471-476
.