Journal of Sound and Vibration 332 (2013) 145–153

Contents lists available at SciVerse ScienceDirect

Journal of Sound and Vibration

journal homepage: www.elsevier.com/locate/jsvi

Drill-string horizontal dynamics with uncertainty

on the frictional force
T.G. Ritto a,n, M.R. Escalante b,c, Rubens Sampaio d, M.B. Rosales e
a
Federal University of Rio de Janeiro, Department of Mechanical Engineering, Rio de Janeiro, Brazil
b
FRCU, Universidad Tecnológica Nacional, C. del Uruguay (ER), Argentina
c
FRCon, Universidad Tecnológica Nacional, Concordia (ER), Argentina
d
PUC-Rio, Department of Mechanical Engineering, Rio de Janeiro, Brazil
e
Universidad Nacional del Sur, Dpto. de Ingenierı́a, Bahı́a Blanca, Argentina

a r t i c l e i n f o abstract

Article history: This paper analyzes the dynamics of a horizontal drill-string. In this dynamics, the frictional
Received 12 February 2012 forces between the column and the borehole are relevant and uncertain. A stochastic model
Received in revised form is proposed for the frictional coefficient: a random field with exponential autocorrelation
9 August 2012
function. In the numerical analysis, the drill-string is modeled using the bar model (tension/
Accepted 10 August 2012
compression), and is discretized by means of the finite element method. An oscillatory force
Handling Editor: W. Lacarbonara
Available online 30 August 2012 (due to mud motor) is imposed on the system and there is a bit–rock interaction as the
column moves forward. We propose a new way to measure the efficiency of the process: an
output/input power ratio. The resultant random ratio is analyzed, and it turns out that it
presents a bimodal characteristic, an unexpected result.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction

A drill-string is a slender structure used to drill rocks in search of oil. It is composed of thin tubes (drill-pipes) and
thicker tubes (drill-collars) with a bit fixed at the end. In the beginning of oil exploration there were mainly vertical wells,
but nowadays directional (horizontal) drilling is very common [1,2].
There is a reasonable amount of papers in the literature that analyze the dynamics of a drill-string performing vertical
drilling. Each author uses a different approach to model the drill-string dynamics: one-mode approximation [3,4], beam model
together with the finite element method [5–7], Cosserat theory [8], discrete system with two degrees of freedom [9].
Concerning horizontal drilling, there are few papers treating the modeling of horizontal drill-string. Actually, to the best
of the authors knowledge, there is no previous work on the dynamics of horizontal drilling.
In the present paper, the part of the column composed by the drill-collars (called Bottom Hole Assembly—BHA) is
modeled using the bar model (tension/compression), and is discretized by means of the finite element method [10].
The mass of the bit is taken into account and an exponential dissipation model is proposed to model the bit–rock
interaction. Also at the bit, an oscillatory force (due to mud motor) is applied and, on the other side of the BHA, a constant
force is imposed to push the structure forward.

n
Corresponding author.
E-mail addresses: thiagoritto@gmail.com, tritto@mecanica.ufrj.br (T.G. Ritto), escalam@frcu.utn.edu.ar (M.R. Escalante),
rsampaio@puc-rio.br (R. Sampaio), mrosales@criba.edu.ar (M.B. Rosales).

0022-460X/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jsv.2012.08.007
146 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153

The goal of this paper is to analyze how uncertainties on the frictional forces propagate through the system. The
frictional forces are difficult to model because there are many uncertainties related to them (friction model, points of
contact, friction coefficient, etc.). A stochastic field with exponential correlation function is used to model the frictional
coefficient. To approximate the statistics of the response, the Monte Carlo method [11] is applied.
There are few papers treating the stochastic problem of the drill-string dynamics; and they are all related to vertical
wells. Among them, we may cite [7,12,13]. Ritto et al. [7] propose a probabilistic model for the bit–rock interaction model,
Spanos et al. [12] analyze stochastic lateral forces at the bit, and Kotsonis and Spanos [13] analyze a random weight-on-bit
using a simple two degrees of freedom drill-string model.
Another contribution of the present paper (besides the modeling of a horizontal drill-string and the analysis of
stochastic frictional forces) is to analyze the performance of the system in terms of output/input power. The relationship
between the output power and the input power is used to measure the efficiency of the system. The results are explored
for different values of a slenderness parameter (length of the BHA divided by its outer diameter), different excitation
frequencies, different levels of uncertainty of the frictional forces, etc.
This paper is organized as follows. The deterministic and stochastic models are presented in Sections 2 and 3, and the
numerical results are discussed in Section 4. Finally, the concluding remarks are done in Section 5.

2. Deterministic model

This section describes the features of the model employed in the analysis. It should be noticed that this is a first approach to
the horizontal drilling dynamics, therefore the model is the simplest possible continuous model. A sketch of the system
analyzed is shown in Fig. 1. A constant force is imposed on the left side of the Bottom Hole Assembly (BHA), there is an
oscillatory force imposed at the bit, and there is an interaction force between the bit and the rock on the right side of the
structure. Besides these forces, gravity and the normal reaction associated to it (that, in the model, are balanced), and friction
(the main force in the process) are acting on the drill-string, as shown in the figure.
Only the axial vibrations of the BHA are considered and the equation of motion is given by
q2 uðx,tÞ q2 uðx,tÞ
rA 2
EA _
¼ f sta ðx,tÞ þ f har ðx,tÞ þ f bit ðuðx,tÞÞ _
þf fric ðuðx,tÞÞ €
þ f mass ðuðx,tÞÞ, (1)
qt qx2
where u is the axial displacement, r is the density of the column, A is the cross sectional area and E is the elasticity
modulus. The space and time variables are x 2 ½0,L, t 2 ½0,T, where L is the length of the structure and T is the duration of
the time analysis. The right-hand side of the equation presents the forces acting on the system, which are depicted in the
sequence. A constant force is imposed at x¼0
f sta ðx,tÞ ¼ F sta dðxÞ, (2)

where F sta is the force amplitude, and d is the Dirac-delta function. This force represents the effect of the drill-pipes on the
BHA; it is assumed constant since we are computing the displacement of about 0.1 m of penetration.
A harmonic force is imposed on the system because the driving source of a horizontal drilling is the mud motor, which
rotates about a given nominal rotational speed (in a steady operation). Therefore, the axial movement should be excited
about the same frequency. The harmonic force f har is given by
f har ðx,tÞ ¼ F o sinðof tÞdðxLÞ, (3)

where F o and of are the amplitude and frequency of the harmonic force; this force is applied at x¼ L. The force (per unit
length) related to the friction field f fric is given by
_
f fric ðuðx,tÞÞ ¼ mðxÞðrAÞg sgnðuðx,tÞÞ,
_ (4)

where m is the friction coefficient field, rA is the mass per unit length of the structure and g is the gravity. Note that the
friction depends on the sign of the speed, which makes it discontinuous. The force related to the bit mass is written as
€
f mass ðuðx,tÞÞ €
¼ mbit uðx,tÞ dðxLÞ, (5)

Fig. 1. Sketch of the system analyzed.

 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153 147

−200

−400

−600

Fbit [N]
c2 = 100
−800 c2 = 200
c2 = 400
−1000

−1200

−1400

−40 −20 0 20 40
bit speed [m/h]

Fig. 2. Bit–rock interaction model for c1 ¼ 1:4  103 N and different values of c2.

where mbit is the bit mass located at x ¼L. Finally, we propose a model for the bit–rock interaction force f bit . There is no
work in the literature (at the best of the authors’ knowledge) that considers the modeling of the bit–rock interaction of a
horizontal drilling. In this first attempt to construct a bit–rock interaction model we are assuming that there is a limit force
for the bit as the speed of the bit increases. We also assume that the bit force approaches this limit asymptotically (with
exponential decay); the model was inspired by the work of Wanhein et al. [14], which was developed in another context.
The exponential model is given by
(
_
f bit ðuðx,tÞÞ _
¼ ½c1 expðc2 uðx,tÞÞc _
1 dðxLÞ for uðL,tÞ 4 0,
(6)
f bit ðuðx,tÞÞ ¼ 0 for uðL,tÞ r 0,
_ _

where c1 and c2 are the two constants of the bit–rock interaction; this force is applied at x¼L. The constant c1 is chosen
such that the bit speed is close to a reasonable value ð  20 m=hÞ, and c2 defines the nonlinear shape of the function in the
range of analysis. Fig. 2 shows an example of the proposed bit–rock interaction model with c1 ¼ 1:3  103 N and different
values of c2. Initially there is no reaction force at the bit, but as long as the dynamics begins the column moves to the right
and f bit starts acting.
Using linear interpolation functions for each element that has one axial degree of freedom per node (see [10]
for details), and assembling the global matrices, the deterministic system of equations are written as
€ þ CuðtÞ
MuðtÞ _ þ KuðtÞ ¼ f sta þ f har ðtÞ þ f bit ðuðtÞÞ
_ _
þ f fric ðuðtÞÞ,

uð0Þ ¼ u0 , _
uð0Þ ¼ v0 , (7)

in which u is the displacement vector, M is the mass matrix and K is the stiffness matrix. The proportional damping matrix
C ¼ aM (a is a positive constant) is added a posteriori to the computational model. The initial conditions are given by u0
and v0 . The force vectors f sta , f har , f bit and f fric are related to the static, harmonic, bit–rock interaction and frictional forces,
respectively. The force related to the bit mass is included in the matrix M, since it depends linearly on the acceleration.

3. Stochastic model

The frictional force is hard to model, especially in the situation of horizontal drilling, where the points of contact and
the characteristics of the surface may vary. Acknowledging this fact, we propose to treat this force as a random force,
assuming a random frictional coefficient field. Thus, we define the random field fnðxÞ : x 2 ½0,Lg as a collection of real-
valued random variables from a probability space ðO,F ,PÞ, where O is the sample space, F is the s-algebra and P is the
probability measure.
It is assumed that n is a stationary truncated Gaussian random field on ½0,L with exponential autocorrelation function
 
9x2 x1 9
Rðx1 ,x2 Þ ¼ s2 exp  , (8)
b

where b is the correlation length, which measures the decay of the autocorrelation function. At this point we need to
justify this hypothesis. Choi et al. [15] have developed a discrete stochastic field to tackle the problem of micro-slip. They
derived the probability density function of the discrete friction coefficient field using the Maximum Entropy principle [16]
and ended up with a truncated Gaussian field with a given correlation length. However, the random generation of such a
148 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153

field is not trivial. We propose to use an exponential autocorrelation function and truncated Gaussian field that is very easy
to sample from rejection of a Gaussian field; if the random field generated has one or more values lower than zero or
greater than 1, this field is rejected. Actually, we need to do another approximation: the truncated Gaussian field is
expanded with Karhunen–Loe ve expansion using standard Gaussian random variables. As long as the probability of having
nðxÞ 2= ½0,1 is very low, this approximation will be good (which is the case of the present analysis).
The stochastic field n is then expanded using the Karhunen–Loe ve expansion (developed by Karhunen [17] and Loe ve
[18]; for some applications see [19,20]):
X
N pffiffiffiffiffi
nðx, xÞ ¼ m ðxÞ þ lk Z k ðxÞfk ðxÞ, (9)
k¼1

where m is the mean value of the frictional coefficient, lk and fk are the k-th eigenvalue and k-th eigenvector of the
autocorrelation function R, Zk are independent standard Gaussian random variables and N defines the precision of this
representation (as N increases the representation gets better). The random frictional force (per unit length) is written as
_
F fric ðuðx,tÞ, xÞ ¼ nðx, xÞðrAÞg sgnðuðx,tÞÞ:
_ (10)
And the stochastic system is written as
€ xÞ þ CUðt,
MUðt, _ xÞ þ KUðt, xÞ ¼ f sta þ f har ðtÞ þ f bit ðUðt,
_ xÞÞ þ Ffric ðUðt,
_ xÞ, xÞ,

uð0Þ ¼ u0 , _
uð0Þ ¼ v0 , (11)
where U is the random response and Ffric is the random vector related to the frictional random field. This is a nonlinear
system of equations because the bit–rock interaction is nonlinear and because the frictional force is discontinuous (if the
speed at a point changes sign, the friction force of this point also changes sign).

4. Numerical results

The data used in the simulation is given as follows. E¼210 GPa, r ¼ 7850 km=m3 , g ¼ 9:81 m=s2 , Di ¼ 0:10 m (inner
diameter), Do ¼ 0:15 m (outer diameter), L=Do ¼ 400, mbit ¼ 20 kg, a ¼ 0:2, c1 ¼ 1:4  103 N, c2 ¼ 400, m ¼ 0:1, s ¼ 0:1  m ,
b¼ 10 and of ¼ 100  2p=60 rad=s, t 2 ½0,10 s, Dt ¼ 0:0001 s, N ¼100, f sta ¼ 5500 N, F o ¼ 550 N.
The system is discretized with 100 finite elements (after analyzing convergence on the natural frequencies) and the
system of differential equations are integrated applying the Newmark integration scheme [10] assuming zero initial
condition. It should be noticed that the values of the parameters of the bit–rock interaction model, and of the mean value
of n are assumed to be known. These values should be the subject of an identification procedure, such as the one that was
done in [21], but they are such that the bit axial speed is coherent with the drilling process.
The above values are for the reference simulation. The response of the system is analyzed for different frequencies of
excitation of (100, 150 and 200 RPM), slenderness parameters L=Do (400, 600, 800), standard deviations s (0.1 and
0:2  m ), and correlation lengths b (10, 100).
Some deterministic results are presented before discussing the stochastic results. For the reference simulation, the first
five natural frequencies of the structures are {42, 86, 120, 174, 220} Hz and the first five damping ratios are about 2
percent. The displacement of the bit is shown in Fig. 3; remark that the column is free to move forward. Fig. 4 shows the bit
speed and its Fourier transform. It can be noted that besides the excitation frequency (100 RPM), some harmonics appear

0.07

0.06

0.05
u (L,t) [m]

0.04

0.03

0.02

0.01

0
0 2 4 6 8 10
t [s]

Fig. 3. Displacement of the bit.

 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153 149

60

50
101
40

v (L, t) [m/h]
v (L, t) [m/h]

30 100

20

10−1
10

0
0 2 4 6 8 10 100 200 300 400 500 600
t [s] freq [RPM]

Fig. 4. Bit speed (a) and its Fourier transform (b).

0.35
0.3
0.25
pout(t)/pin(t)

0.2
0.15
0.1
0.05
0
6 6.5 7 7.5 8 8.5 9 9.5 10
t [s]

Fig. 5. Ratio between the input and output power.

in the frequency spectrum; this is due to the nonlinearity of the system. Furthermore, 4(a) shows that the system presents
some oscillations. There is a phase when the speed is close to zero, and a phase when the speed increases its magnitude.
This is close to stick-slip oscillations, but there is no stick (zero speed) phase.
We propose now a new way to analyze the efficiency of the horizontal drilling in the following way. We establish a
measure that indicates the ratio between output and input power; if this parameter increases, the efficiency of the process
also increases. The input and output power are given by
_
pin ðtÞ ¼ f sta uð0,tÞ _
þ f har ðtÞuðL,tÞ and _
pout ðtÞ ¼ f bit ðtÞuðL,tÞ: (12)

And the average power is given by

Z t1
1
p in ¼ _
ðf sta uð0,tÞ _
þ f har ðtÞuðL,tÞÞ dt,
t 1 t 0 t0

Z t1
1
p out ¼ _
f bit ðtÞuðL,tÞ dt: (13)
t 1 t 0 t0

Let us define the ratio between the output and input power in the time interval ðt 0 ,t 1 Þ by y ¼ p out =p in . Fig. 5 shows for
each instant the ratio pout ðtÞ=pin ðtÞ for the reference simulation. It can be noted that the efficiency of the system (energy
used for effective drilling) is lower than 25 percent. Of course, in a real operation, the efficiency may be lower due to other
losses in the process.
Now we are going to focus on the stochastic results. Fig. 6 shows a typical convergence curve for the variance of the
P s
ratio Y ¼ P out =P in , where conv ðns Þ ¼ 1=ns ni ¼ 2
1 Y i ; in which ns is the number of Monte Carlo simulations. This is a variable
of interest, since it measures the efficiency of the system under analysis. An acceptable convergence is achieved when
ns ¼500. The probability density function of the ratio Y is plotted for the different cases; see Figs. 7 and 8.
Fig. 7 shows that the performance of the system was similar for different excitation frequencies of and for different
correlation lengths b (Fig. 9 shows the autocorrelation function); the approximated probability density functions for these
150 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153

Standard Deviation

0.08

0.07

0.06

0.05

Conv
0.04

0.03

0.02

0.01

0 100 200 300 400 500

n

Fig. 6. Variance convergence of the random variable defined by the relation Y ¼ P out =P in .

12 14
100rpm b = 10
150rpm 12 b = 100
10 200rpm
10
8
8
fY
fY

6
6
4
4

2 2

0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
Y Y

Fig. 7. Probability density functions of Y (a) varying the rotational speed of and (b) varying the correlation length b.

14 25
L/D0 = 400 σ = 0.1μ
12 σ = 0.2μ
L/D0 = 600
20
L/D0 = 800
10

15
8
fY
fY

6 10

4
5
2

0 0
0 0.1 0.2 0.3 0.4 0 0.1 0.2 0.3 0.4
Y Y

Fig. 8. Probability density functions of Y (a) varying the slenderness parameter L=Do and (b) varying the uncertainty level of the random field s.
 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153 151

x 10−4
1.2
b = 10
b = 100
1

0.8

R
0.6

0.4

0.2

0
−30 −20 −10 0 10 20 30
Δx

Fig. 9. Autocorrelation function.

1.6

1.4

1.2

1
u (L,t,ξ) [m]

0.8

0.6

0.4

0.2

0
0 2 4 6 8 10
t [s]

Fig. 10. Five Monte Carlo simulations of the bit displacement.

cases do not vary significantly. Therefore, for the frequencies analyzed, a variation in the load frequency does not change
much the magnitude of the displacement of the axial response; it should be remarked that the first natural frequency of
the system is far away from the excitation frequencies.
On the other hand, the probability density functions for different slenderness parameters L=Do and different levels of
uncertainty of the frictional coefficient s are quite different; see the shape of the probability density functions in Fig. 8.
Fig. 8(a) shows that, as the slenderness parameter increases, the efficiency is reduced; note that the values are shifted to
the left, indicating less efficiency. This result is expected, since there will be more column to resist the forward movement.
A less intuitive result happens when the level of uncertainty (s) is increased. Although the coefficient of variation of Y
ðsY =mY Þ increases from 36 percent to 50 percent, the resulting probability density function is more concentrated around
two values (0.04 and 0.25).
About the shape of the probability density function of Y, it should be remarked that this shape is quite unexpected.
We observe a bimodal distribution, indicating that there are two sets of behaviors. In this regard it makes no sense to
compute the average of the Monte Carlo simulations, because it would simply be meaningless. What makes sense is to
understand the behavior of the two sets of behaviors. A small set seems to be more efficient (y around 0.25) than the other
(y from 0.05 to 0.20); see Fig. 7, for instance. To understand this point, let us plot some Monte Carlo simulations of the bit
displacement; see Fig. 10. For some simulations the bit speed it kept small, and for others (fewer ones) the bit speed
increases a lot. Investigating these results, it was noted that there is a limit frictional force for which the axial speed of the
drill-string increases significantly. The system will accelerate differently depending on the system characteristics
152 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153

100

80

v (L,t,ξ) [m/h]
60

40

20

6 7 8 9 10
t [s]

Fig. 11. Four Monte Carlo simulations of the bit speed.

(stiffness, mass, etc.) and on the applied forces (constant force applied on the left, friction force applied along the column
and the bit–rock interaction force applied on the right). Fig. 11 shows four Monte Carlo simulations corresponding to one
set of simulations (the ones that do not have increasing speed). The stick-slip oscillations is observed for some simulations:
the speed reaches the value zero (stick) then it slips. Normally, in petroleum engineering, the term stick-slip is used for the
torsional vibration of a drill-string in vertical wells [22]. In that case there is only one contact point at the bit–rock
interaction. Nevertheless, in the present case of horizontal drilling, there is stick-slip in the axial direction due to the
frictional force, that is acting on all column extension. It is more appropriate to talk about a stick-slip field on ½0,L, this of
course is akin to a contact problem. To compute the stick-slip we simply regularize the frictional force function, so it is a
continuous function of time.

5. Conclusions

In this paper a dynamical model was proposed for the nonlinear dynamics of a horizontal drill-string. It should be
noticed that few papers treat this problem. In the present analysis, the drill-string is modeled using a bar model (tension/
compression), and is discretized by means of the finite element method. A constant and an oscillatory force is imposed on
the system, and there is bit–rock interaction. Friction between the column and the soil is also included in the model, and it
is modeled by a random field with an exponential correlation function. The Monte Carlo method is used to approximate
the propagation of uncertainty. Other sources of uncertainties, such as bit–rock interaction model, should be the subject of
futures works.
The efficiency of the system was measured by the ratio between the output power (obtained from the bit–rock
interaction) and the input power. The results were similar for different excitation frequencies and different correlation
lengths. But they were different for different slenderness parameters and different uncertainty levels of the random field.
Another interesting result is that the probability density function of the ratio between the output and input power is
bimodal. The bimodality indicates that there are two sets of behaviors: one presents considerably higher bit speed than the
other. We also observed stick-slip behavior in the axial direction; the usual analysis of stick-slip in drilling dynamics is
related to torsional vibrations in vertical wells.
This is an ongoing work, and there are many investigation to perform with this system. For instance, an identification
procedure for the parameters, a refinement of the stochastic modeling, a refinement of the bit–rock interaction model, the
inclusion of bending and torsional vibrations, and a robust optimization to maximize the performance of the system
considering the uncertainties.

Acknowledgements

The authors are grateful to the financial support from the Brazilian agencies: CAPES, CNPq, and FAPERJ; and from the
Argentinean support: FRCU-UTN, FRCon-UTN, SGCyT-UNS, MINyT and CONICET.

References

[1] E. Allouche, S. Ariaratnam, J. Lueke, Horizontal directional drilling profile of an emerging industry, Journal of Construction Engineering and
Management ASCE 126 (2000) 68–76.
 T.G. Ritto et al. / Journal of Sound and Vibration 332 (2013) 145–153 153

[2] D.A. Willoughby, Horizontal Directional Drilling, Utility and Pipeline Applications, Civil Engineering Series, McGraw-Hill, New York, USA, 2005.
[3] A. Yigit, A. Christoforou, Coupled axial and transverse vibrations of oilwell drillstrings, Journal of Sound and Vibration 195 (1996) 617–627.
[4] A.P. Christoforou, A.S. Yigit, Dynamic modeling of rotating drillstrings with borehole interactions, Journal of Sound and Vibration 206 (1997) 243–260.
[5] Y.A. Khulief, H. AL-Naser, Finite element dynamic analysis of drillstrings, Finite Elements in Analysis and Design 41 (2005) 1270–1288.
[6] R. Sampaio, M.T. Piovan, G. Venero, Coupled axial/torsional vibrations of drill-strings by means of a nonlinear model, Mechanics Research
Communication 34 (2007) 497–502.
[7] T.G. Ritto, C. Soize, R. Sampaio, Nonlinear dynamics of a drill-string with uncertain model of the bit–rock interaction, International Journal of Non-
Linear Mechanics 44 (2009) 865–876.
[8] R.W. Tucker, C. Wang, An integrated model for drill-string dynamics, Journal of Sound and Vibration 224 (1999) 123–165.
[9] T. Richard, C. Germay, E. Detournay, A simplified model to explore the root cause of stick–slip vibrations in drilling systems with drag bits, Journal of
Sound and Vibration 305 (2007) 432–456.
[10] K.J. Bathe, Finite Element Procedures, Prentice-Hall, Inc., Upper Saddle River, New Jersey, 1996.
[11] R.Y. Rubinstein, Simulation and the Monte Carlo Method, Series in Probability and Statistics, 2nd edition, John Wiley and Sons, New Jersey, USA, 2007.
[12] P.D. Spanos, A.M. Chevallier, N.P. Politis, Nonlinear stochastic drill-string vibrations, Journal of Vibration and Acoustics 124 (4) (2002) 512–518.
[13] S.J. Kotsonis, P.D. Spanos, Chaotic and random whirling motion of drillstrings, Journal of Energy Resources Technology 119 (4) (1997) 217–222.
[14] T. Wanhein, N. Bay, A.S. Petersen, A theoretically determined model for friction in metal working processes, Wear 28 (1974) 251–258.
[15] B.-K. Choi, M.C. Sipperley, M.P. Mignolet, C. Soize, Discrete maximum entropy process modeling of uncertain properties: application to friction for
stick-slip and microslip response, EURODYN 2011, 8th International Conference on Structural Dynamics, Leuven, Belgium, 2011.
[16] E. Jaynes, Probability Theory: The Logic of Science, Cambridge University Press, Cambridge, UK, 2003.
[17] K. Karhunen, Uber lineare methoden in der wahrscheinlichkeitsrechnung (On linear methods in probability and statistics), Annales Academiae
Scientiarum Fennicae Seried A.I. Mathematica-Physica 37 (1947) 1–79.
[18] M. Loe ve, Probability Theory, Graduate Texts in Mathematics, 4th ed. Springer, USA, 1977.
[19] S. Bellizzi, R. Sampaio, POMs analysis obtained from Karhunen–Loe ve expansion for randomly vibrating systems, Journal of Sound and Vibration 297
(2006) 774–793.
[20] T. Ritto, F.S. Buezas, R. Sampaio, Proper orthogonal decomposition for model reduction of a vibroimpact system, Journal of the Brazilian Society
of Mechanical Sciences and Engineering, in press.
[21] T.G. Ritto, C. Soize, R. Sampaio, Probabilistic model identification of the bit–rock-interaction-model uncertainties in nonlinear dynamics of a drill-
string, Mechanics Research Communications 37 (6) (2010) 584–589.
[22] Y.A. Khulief, F.A. Al-Sulaiman, S. Bashmal, Vibration analysis of drillstrings with self excited stick-slip oscillations, Journal of Sound and Vibration 299
(2007) 540–558.