The 16th International Scientific Conference

eLearning and Software for Education

Bucharest, April 23-24, 2020
10.12753/2066-026X-20-000

ANALYSIS OF TEMPERATURE-DEPENDENT NATURAL FREQUENCIES OF

SPINDLES WITH POSSIBLE APPLICATION IN MEDICINE

Andrei GHEORGHITA, PhD. , George CONSTANTIN

Polytechnic University of Bucharest, Splaiul Independentei 313, Bucharest, Romania
[email protected], [email protected]

Dragos Arotaritei
University of Medicine and Pharmacy “Grigore T. Popa”, Str. Universitatii 16, Iasi, Romania
[email protected]

Abstract: The calculus of natural frequencies plays an important role in modal analysis of spindle-
bearing systems, especially is micro-milling or micro-grinding applications in medicine (surgery and
dentistry). Each natural frequency is associated with spindle mass, and therefore material, on which the
heating generated in bearings causes axial and radial expansion. Because the radial expansion is
limited by the housing and bearings, the most important thermal deformation becomes the axial one.
The natural frequency changes caused by temperature evolution due to heating sources in a spindle is
analyzed. The heating interval is split in an equal number of intervals, each limit of interval
representing a simulation. A predictor is proposed in order to construct a function that shows the
dependency of natural frequencies of temperature. The continuous space in an interval is covered by
this predictor. The result of predictor is compared with a nonlinear universal approximator known for
its capability in prediction, a feedforward neural network. In this preliminary application, the spindle
has a predefined shape and the heating sources are placed in predefined locations of spindle - the
location of angular contact bearings that are the major sources of heat in the assembly. The spindle is
modeled as connected regular geometric bodies - hollow cylinders and hollow truncated cone. The
truncated cone is approximated by connected sections of hollow cylinders. The user can see in a
graphical interface the changes of natural frequencies with the temperature for didactic scope or a
possible optimization of the lifetime of the spindle-bearing assembly. The further development will
investigate the possibility of implementation online of simulation using finite elements method.

Keywords: natural frequencies, micro-milling, simulation, spindle

I. INTRODUCTION

The spindle bearings (rotor bearings) dynamics is studied extensively in the last years. The
main area of research encompasses the dynamic analysis based on forces and moments to which the
shaft is subjected, thermal analysis (temperature distributions or/and thermal expansions) and vibration
analysis ([1]-[4]). Thermo-mechanical approach is a thermal analysis with effects in mechanic
deformation. If axial and/or radial load is present, the effect is a combination of the two phenomena.
The spindle-bearing systems have angular-contact bearings and the effect of preload contributes to
deformations, the effect according to elasticity or rigidity of the ax. Dynamic behavior due to rotation
can contribute with moments of the balls to general displacement and deformation of the components
[1].
There is a considerable amount of papers related to spindle-bearings that has been published.
Thermo-mechanical studies consider the main source heat the bearings (the angular contact bearings)
([5]-[6]) and additional, for motorized spindles (the spindle is located on the rotor of the assembly),
the heat developed in stator, due to rotation of the rotor ([7]-[8]). In [9], an analysis is proposed
focused on change in spindle eigen frequency values in a transient time (until 8 hours) depending on
thermo-mechanical condition of bearings. Heat sources along with convective transfer are applied to
spindle bearings in [10]. The displacements are verified by experimental measurements on all three
directions (X, Y, and Z). An interesting analysis is made in [11]. The free vibration of square cross-
sectioned beams is studied by analytical and numerical methods depending on length of the beam and
cross-sectional area [11].

II. THERMAL ANALYSIS OF SPINDLE

It is known that thermal grow affects the spindle performances [12]. These performances are
related to lifetime, possibility of damages in ax, or generally speaking, a malfunction of the assembly.
The temperature distribution in the spindle has as effect a thermal deformation due to thermal
expansion. The expansion occurs in all coordinates but because the spindles are symmetric, the
calculus is made on axial and radial directions. The general equation is based on Hooks’ law:

  T (1)

where  is thermal strain,  is the coefficient of thermal expansion (depending of the material)
and T difference between final and initial temperature. There are clear that the heating sources
(bearings) transfer heat to all in components (spindles, fittings, spacers, housing, etc.) in a belt driven
spindle-bearing assembly. The main processes are conduction and convection [12]. Due to these
processes, the expansion of spindle occurs in radial and longitudinal direction. The length of the
spindle in this approach is much higher that the diameter, so the radial displacement is neglected. The
axial thermal growth for small machine tools is currently from few microns to hundreds of microns
[13].
After transitory stage, T = Tf –Tamb where Tf is the final temperature and Tamb is the ambient
temperature (usually Tamb = 25 C, or a slighter greater value, Tamb = 300 K). As result, depending of
number of sections of spindle, sections used in FEM (Finite Element Method), a temperature curve
throughout the spindle can be plotted. The spindle (and rotors in general) is considered symmetrical in
our approach, which is the most common case in rotating machinery. A simple model is shown in
figure 1-2. Each section Li is divided in a number of n similar sections, according to FEM [14].

Figure no. 1. A simplified spindle-bearing model with two angular-contact bearing and one cylindrical
roller bearing

Figure no. 2. The corresponding spindle-bearing model used in calculation of thermal analysis and
global stiffness matrix
From [3] and [15], a usually good approximation is given for one hollow cylinder when n10,
and as trade-off between precision and complexity, a value of n = 10, for each distinct segment of
spindle, i.e. Li, i=1,2, N (N=6 in the case of figure 1-2). The spindle-bearing is composed from
elementary geometrical objects as cylinders, hollow cylinders and truncated cone.
The total heat generated by angular contact rolling bearing is calculated as sum of heat
generated by all the Z balls (Z is the number of the balls), where the heat generated by a ball is given
by ([1], [16]):

H total  1.047 10 4 nM  1.047 10 4 n  ( M l  M s  M v ) (2)

where n [rpm] is the rotating speed (inner or outer) ring of the bearing, M is the total frictional torque
of the bearing as an additive process of the three torques (Ml - torque due to applied load, including
preload, Ms – torque due to spinning motion of the ball and Mv – torque due to viscous friction). The
formulas are complicated and the complete details are in ([1], [16]). The angular contact ball bearing is
considered along with grease lubrication and the viscosity–temperature characteristic is taken from
table [1]. The user has also an option to select oil lubrication. For cylindrical roller balls the method
used is that from ([1], [17]) (only friction on the contact line between cylindrical ball and the ring is
considered). The heat generated in contact angular bearing depend on contact angle and speed of
rotation which has many interdependencies in the system of equilibrium of forces and moments which
act on the ball bearings [1]. The system is usually solved by Newton-Raphson method, but complex
interdependencies among variables and because the convergence of method depend strongly of initial
conditions. In order to avoid these difficulties, a simplified method proposed in [18] is used in the
proposed educational tool.

Figure no. 3. Heat generated by bearing of type 7014 CE/HCP4A (number of balls Z= 25)

In order to solve the problem on conduction/convection a known method is proposed: thermal

networks [19]. The spindle is modelled by regular geometrical bodies: cylinder and hollow cylinder,
conic hollow and sphere. These elementary bodies have a calculated thermic resistance based on their
geometry and material, formulas given in [19] with details about notations in [20]. The values for
thermal contact resistance bearing-ball and inner ring spindle are taken, for steel from [21]. The
graphic for heat generated by a ball for an angular contact bearing of type 7014 CE/HCP4A
(geometrical dimensions and number of balls are given in specification of producer) is given in figure
3). For a cylindrical roller bearing N1012 KTNHA/SP, the heat generated by a single ball is much
smaller.
 After a transient time, the temperature is stabilized and the distribution of final temperature
along the shaft is calculated. The elementary elements used in modelling of the shaft (cylinders,
hollow cylinders, and truncated cone. For truncated cone form even it is possible to use ideas from
calculation of stiffness for variable cross section beam ([22]-[23]). Because the partition of the spindle
into elements can be used for modal analysis (calculation of natural frequencies), an approach that
approximate the truncated cone by a number of cylinders is used, similar to [24].

III. MODAL ANALYSIS

The structure spindle-bearings is susceptible to vibrations and in the case of resonance, this
phenomenon can cause the malfunction of entire system. As a result, it is important to conduct a
modal analysis of such type of spindles (rotors). The elastic bodies have natural frequencies,
depending of their number of freedom (body or multi-body systems). The resonance occurs when an
exciting frequency (frequency of rotation of spindle) is equal of one of natural frequency. The usually
way to avoid damages due to resonance phenomenon is the increase quickly the velocity over these
values [7].
The method used in this approach is that from [14] using assembly stiffness matrix, direct
method for calculation of global matrix of the spindle. The calculations are made synchronizing the
thermal partition from previous section with these used for calculation of the stiffness matrix (the
elements are the same. The temperature and stiffness of the element is considered uniform along the
element. The simple beam element is considered to have four degree of freedom [14]. The equation of
beam in matrix forms using the finite element method is [25]:

 M  w    K  w  { f } (3)

where M is the mass matrix, K is the stiffness matrix, and f the force vector. The eigenvalue
values  corresponding to natural frequencies are calculated using:

K   2 M  0, 2   (4)
D  I  0, DM 1
K (5)

where D is the known as dynamic matrix and I is the identity matrix. The real values of  are
sorted in ascending order meanwhile the complex values if they exists are rejected. The global
stiffness matrix assembly (GSMA) is obtained by direct stiffness method (DSM), which is practical,
the most common implementation of the finite element method (FEM).
The DSM construct large matrix is the division in elements is a high order of magnitude but
actually Matlab version deals with large matrix computation in a reasonable time for user in and
educational tool.

IV. THE EDUCATIONAL TOOL

The educational tool Sbfreq Analizer is constructed as a GUI using MATLAB support. The
main program permits to user two options: a simple predefined spindle as in figure 2 and a
construction of a new spindle with a table and dimensions. In the first case, a predefined thermic curve
is calculated and provided to user for the three bearings and rotation speed of the spindle can be set
between 0 rpm (static case) and 6000 rpm. The dimensions and showed in table along with
geometrical dimensions of the bearings. The material is considered bearing steel, density =7.810-9
[26].
The more complicated spindle is considered for a machine Schaublin milling center, with 3
NC axes, BT40 tool holder and a maximum speed of 9000 rpm in figure 4-5. The same rotor is
modeled in SolidWorks Premium 2012, in order to estimate the error in calculation of natural
frequencies as level of performance of the proposed method. The error in frequencies from same
condition was between 14.93% and 18.96%, but in the same direction, having a behavior of systematic
error. This is clearly due to some enough large approximations but having into account that the
proposed tool is only a didactic one, it is acceptable at this stage. The main idea is to see how
frequencies are moving as the frequency of rotation is increasing due to complex relation from heating
of bearings and as a result, the geometrical form suffer of thermal expansion.
The software tool is presented shortly in figure 6-7. In order to simplify the complexity of the
interface (GUI – Graphic User Interface) and to make it friendly, few options are passed to a single
edit field with number separated by the token “;” and thereafter parsed to obtain the values and
number of options, e.g. the plot for two frequencies modes 1 and 3 is set to be “1;3” and the speed of
rotor “500;1000;1500;2000;3000; 4000; 5000;6000”, see figure 7.

Figure no. 4 A simplified model for rotor, Schaublin milling center, with 3 NC axes, BT40 tool holder
and a maximum speed of 9000 rpm

Figure no. 5 SolidWorks model, Schaublin milling center, with 3 NC axes, BT40 tool holder and a
maximum speed of 9000 rpm
 Figure no. 6 SolidWorks model, Schaublin milling center, with 3 NC axes, BT40 tool holder

Figure no. 7 SolidWorks model, Schaublin milling center, with 3 NC axes, BT40 tool holder

The usually representation is made by 5 modes and a number of speeds of rotation, usually the
values 500;1000;1500;2000;3000; 4000; 5000 and 6000 rpm. It is necessary an interpolator for the rest
of speeds of rotation customized for the same geometry of the spindle, number of bearings and types,
location of it’s on the spindle and the material. The known nonlinear interdependencies between the
variable in thermo-mechanical equation and the difficulty to quantify the geometry of the spindle,
conduct to proposal to use of one tool, demonstrated mathematically as universal approximator,
feedforward neural networks [28]. The architecture an algorithm for learning of feedforward neural
networks (NN) are well known and the package in implemented in many statistical and simulation
software tools including Matlab/Simulink.
The interpolator/extrapolator made by the NN has n=1 input, m=10 outputs two hidden layers
each of them having 26 neurons. Depending on performance (MSE - mean square error of the
predicted signal at output versus the real signal) the number of neurons on each hidden layer can be
different. The number m can be different, the actually plot in figure 7 can be made for m=5 modes.
The input is speed of rotation and the NN is trained for 500;1000;1500;2000;3000; 4000; 5000 and
6000 rpm. The error of approximation MSE is  2.7810-5 for spindle from Fig. 2, and  1.1210-2 for
spindle from figure 4.
The development of interface is in progress. Actually, the most difficult work is to create a
module for design with smart dimension tool for design of spindle and location of the bearings is a
similar way that are used by CAD tools (like SolidWorks).
The electrical motors have many applications in medicine as arthroscopic shavers with
ceramic bearings, surgery and dentistry. In the actual stage it is very difficult to calculate exactly the
geometrical feature for rotor of one practical product, so we propose to use a “surrogate” as
preliminary approach, a micromachining tool presented in [27]. The axial vibration for a similar tool
line in [27], is practically constant for a rotation speed from 0 to 3×10 5 rpm. The balance equation for
temperature is very complicated and it is necessary a FEM (Finite Element Method) Tool to estimate
the temperature distribution (with cooling) along the axis. However, a preliminary supposition of
temperature value of 37.8 C conduct to an thermal expansion of  0.257 m, and an increasing of the
first natural frequencies from 1894 Hz (static case, ambient temperature T amb = 25 C) to 1896,3 Hz in
final evaluation (average Trotor = 37.8 C).

V. CONCLUSIONS

A didactic tool is proposed in order to estimate the natural frequency displacement depending
on temperature throughout the spindle as consequence of thermal expansion of geometry of the
spindle. The temperature distribution depends on the type of bearings, material, location of them and
the most important, speed of rotation and preload.
Due to complexity of problem for to calculate the contact angle (inner and outer) for angular
contact ball bearings under preload (axial force) and eventually radial force, the actual approximation
used a preload force negligible. The problem of solution using preload is under development in a
module of the tool.
The tool is useful especially for biomedical engineer from biomechanics section of study,
where electrical motor can play a role in prosthetic of surgical tools.

Reference Text and Citations

[1] Harris, Tedric A., 2001. Rolling bearing analysis, Wiley, Michigan, 2001.
[2] Genta, Giancarlo, 2005. Vibration Dynamics and Control, Springer US, 2009.
[3] Tiwari, Rajiv, 2017. Rotor Systems: Analysis and Identification, CRC Press, 2017.
[4] Friswell, Michael; Penny, John; Garvey, Seamus; Lees, Arthur; Dynamics of Rotating Machines, Cambridge
University Press, 2010
[5] Jorgensen, Bert; Yung, Shin Dynamics of Machine Tool Spindle/Bearing Systems under Thermal Growth, Journal
of Tribology, 1997, Vol. 119,
https://asmedigitalcollection.asme.org/tribology/article-abstract/119/4/875/434910/Dynamics-of-Machine-Tool-
Spindle-Bearing-Systems?redirectedFrom=fulltext, pp. 875-882.
[6] Li, Hongqi; Integrated Dynamic Thermo-Mechanical Modeling of High Speed Spindles, Part 1: Model
Development, Journal of manufacturing Science and Engineering, 2004, vol. 126,
https://asmedigitalcollection.asme.org/manufacturingscience/article-abstract/126/1/148/462018/Integrated-
Dynamic-Thermo-Mechanical-Modeling-of?redirectedFrom=fulltext, pp. 148-158.
[7] Peng, Zhang; Xiaoan. Chen; Thermal–mechanical coupling model-based dynamical properties analysis of a
motorized spindle system, Proc IMechE Part B: J Engineering Manufacture, 2014,
https://journals.sagepub.com/doi/10.1177/0954405414558736, pp. 1–12
[8] Kuo, Liu; Fei, Chen; Tiejun. Zhu; Yuliang.Wu; Mingjia, Sun; Compensation for spindle’s axial thermal growth
based on temperature variation on vertical machine tools, Advances in Mechanical Engineering, 8/2016, vol. 8.
https://journals.sagepub.com/doi/full/10.1177/1687814016657733, pp. 1–10
[9] Kolar, Peter; Thomas Holkup; Prediction of Machine Tool Spindle's Dynamics Based on a Thermo-Mechanical
Model, Modern Machinery (MM) Science Journal, March 2010,
https://www.mmscience.eu/journal/issues/march-2010/articles/prediction-of-machine-tool-spindle-s-dynamics-
based-on-a-thermo-mechanical-model
[10] Uhlmann, Ekart; Hu, Jiangmin; Thermal modelling of a high speed motor spindle, Procedia CIRP 1, 2012 ,
https://www.sciencedirect.com/science/article/pii/S2212827112000571, pp. 313 – 318
[11] Avcar, Mehmet; Free Vibration Analysis of Beams Considering Different Geometric Characteristics and Boundary
Conditions, International Journal of Mechanics and Applications, 4(3), 2014,
https://www.researchgate.net/publication/269988705_Free_Vibration_Analysis_of_Beams_Considering_Different
_Geometric_Characteristics_and_Boundary_Conditions , pp. 94-100
[12] Than. Van-The; Huang, JinH; Nonlinear thermal effects on high-speeds pindle bearings subjected to preload,
Tribology International, 2016, 96, 2016,
https://www.sciencedirect.com/science/article/pii/S0301679X15005976, pp. 361–372
[13] Zivkovic, Aleksandar; Zeljkovic, Milan; Tabakovic, Slobodan; Milojevic, Zoran; Mathematical modeling and
experimental testing of high-speed spindle behavior, Int J. Adv Manuf Technol, 2015, 77,
https://link.springer.com/article/10.1007/s00170-014-6519-7; pp. 1071–1086
[14] Logan, Daryl; 2006. A First Course in the Finite Element Method, Fourth Edition, CL Engineering, 2006, pp. 647-
701
[15] Bergman, Theodore; Lavine, Adrienne; Incropera, Frank; Devitt, David; 2011. Fundamentals of Heat and Mass
Transfer, John Wiley & Sons, Inc., 2011, pp. 112-154
[16] Jin, Chao; Wu, Bo; Hu, Youmin; Heat generation modeling of ball bearing based on internal load distribution,
Tribology International, 45/2012,
https://www.sciencedirect.com/science/article/pii/S0301679X11002386, pp. 8–15
[17] Junning, Li; Xue, Jiafan; Ma, Zhitao; Study on the Thermal Distribution Characteristics of High-Speed and Light-
Load Rolling Bearing Considering Skidding, Aplied science., 16/2018,
https://www.researchgate.net/publication/327565719_Study_on_the_Thermal_Distribution_Characteristics_of_Hig
h-Speed_and_Light-Load_Rolling_Bearing_Considering_Skidding; pp.607–611
[18] Antoine, Francois; Abba, Gabriel; Molinari, Alain; A New Proposal for Explicit Angle Calculation in Angular
Contact Ball Bearing, Journal of Mechanical Design, 128/2006,
https://www.researchgate.net/publication/245366290_A_New_Proposal_for_Explicit_Angle_Calculation_in_Angu
lar_Contact_Ball_Bearing, pp. 468-478
[19] Brecher, Christian; Shneor, Yair; Neus, Stephan; Bakarinow, Kolja; Fey, Marcel; Thermal Behavior of Externally
Driven Spindle: Experimental Study and Modelling, 7/2015, Engineering,
https://www.researchgate.net/publication/272832125_Thermal_Behavior_of_Externally_Driven_Spindle_Experim
ental_Study_and_Modelling , pp. 73-92
[20] Jorgensen, Bert; Shin, Yung; Dynamics of Machine Tool Spindle/Bearing Systems Under Thermal Growth,
Journal of Tribology, 119/1997
https://www.semanticscholar.org/paper/Dynamics-of-Machine-Tool-Spindle%2FBearing-Systems-Jorgensen-
Shin/0df01e22f4cef5bead3d6c318e5937db2a947327, pp. 875-882
[21] Nakajima, Katsuhiko; Thermal Contact Resistance Between Balls and Rings of a Bearing Under Axial, Radial, and
Combined Loads, Journal of Thermophysics and Heat Transfer, 9/1995,
https://arc.aiaa.org/doi/pdf/10.2514/3.632, pp. 88-95
[22] Friedman, Zack; Kosmatka, John.; Exact Stifness Matrix of a Nonuniform Beam - I. Extension, Torsion, and
Bending of A Bernoulli-Euler Beam, Computer & Structures, 42/1992,
https://www.sciencedirect.com/science/article/abs/pii/0045794992901794, pp. 671-682
[23] Friedman, Zack; Kosmatka, John; Exact Stifness Matrix of a Nonuniform Beam - II. Bending of a Timoshenko
Beam, Computer & Structures, 49/1993, https://www.sciencedirect.com/science/article/abs/pii/0045794992901794,
pp. 545-552
[24] Greenhill, Lyn; Bickford, Wayne; Nelson, Graburn; A Conical Beam Finite Element for Rotor Dynamics Analysis,
Journal of Vibration, Acoustics, Stress, and Reliability in Design, 107/1985,
https://asmedigitalcollection.asme.org/vibrationacoustics/article-abstract/107/4/421/439048/A-Conical-Beam-
Finite-Element-for-Rotor-Dynamics?redirectedFrom=fulltext , pp. 421-430
[25] Cao, Yuzhong; Altintas, Yusuf; A General Method for the Modeling of Spindle-Bearing Systems, Journal of
Mechanical Design, 126/2004,
https://asmedigitalcollection.asme.org/mechanicaldesign/article-abstract/126/6/1089/460082/A-General-Method-
for-the-Modeling-of-Spindle?redirectedFrom=fulltext, pp. 1089-1104
[26] http://www.nsk.com.br/upload/file/nsk_cat_e728g_10.pdf
[27] Li, Wei; Zhou, Zhi-Xiong ; Huang, Xiang-Ming ; He, Zhi-Jian; Du, Yong; Development of a High-Speed and
Precision Micro- Spindle for Micro-Cutting, International Journal of Precision and manufacturing, 15/2014,
https://www.researchgate.net/publication/278108423_Development_of_a_High-Speed_and_Precision_Micro-
Spindle_for_Micro-Cutting, pp. 2375-2383.
[28] Haykin, Symon; 2008. Neural Networks and Learning Machine, Pearson, 3rd edition, Ontario, Canada, 2008, pp
98-103