Society of PetroleumEngineers

SPE 24583

Detection of BHA Lateral Resonances While Drilling With Surface

Longitudinal and Torsional Sensors
P.R. Paslay, Techaid Corp.; Yih-Min Jan,* ARC0 E&P Technology; and J.E.E. Kingman and
J.D. Macpherson,* EXLOG Inc.
*SPE Members

Copyright 1992, Society of Petroleum Engineers Inc.

This paper was prepared for presentation at the 67th Annual Technical Conference and Exhibition of the Society of Petroleum Engineers held in Washington. DC. October 4-7, 1992.

This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(@.Contents of the paper,
as'presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to pvblication review by Editorial Committees of the Society
of Petroleum Engineers. Permissionto copy is restricted to an abstract of not more than 300 words Illustrations may not be copied. The a b s t r p
of where and by whom the paper is presented. Write Librarian. SPE, P.O. Box 833836, Richardson, TX 750833836 U.S.A. Telex, 730989 SPEDAL.

Summary rotary speed or WOB (weight on bit). The ability to

An analytical model is presented in detail which pre-
dicts the oscillatory motions and forces at the top of a
drillstring due to forward whirl of the BHA (bottom hole
assembly) at a BHA lateral resonance. The geometry and
amplitude of the BHA mode shape are used in a scheme
which converts the lateral motions to sinusoidally varying
axial forces and tangential impulses acting on the BHA.
The entire string is then analyzed for wave propagation of
the axial forces and torsional impulses. In this manner,
the dynamic axial and torsional response at the top of the
drillstring can be used to predict BHA forward whirl at
each of the BHA lateral resonances. The model has
been incorporated in a computer program for practical
analyses.
Introduction
The direct observation while drilling (MWD) of dy-
namic downhole bit forces and motions is desirable but
the technology has not been developed to make it normal
practice at this time. In the meantime, valuable informa-
tion concerning the drilling process can be obtained from
measurements at the top of the string. In order to inter-
pret these measurements, it is necessary to postulate
possible motions of the lower end of the string and then
predict by analysis the resulting motion at the top of the
string. This paper addresses such a problem.
A procedure1 is described for predicting motions at
the top of the string owing to BHA forward synchronous
whirl. Detection of BHA forward whirl resonances (critical
speeds) can lead to their avoidance by variations in
References and.illustrationsat end of paper.
detect BHA critical speeds using surface monitoring is
of considerable practical importance. Since there are
so many critical speeds in the normal rotary speed
range, it is difficult to calculate all of them accurately
enough to insure they are all avoided. Several pub-
lished ~ o r k s ~are . ~related
.~ to this study.
Procedure
The following steps outline the procedure used to
predict motions that occur at the top of the drillstring as
a result of BHA forward synchronous whirl:
1) Determine the BHA lateral critical speeds and
mode shapes;
2) Determine the axial excitation at the BHA caused by
running at lateral critical speeds;
3) Determine the axial motions of the entire drillstring
resulting from the BHA axial excitations;
4) Determine the torsional excitation and motion of the
entire drillstring resulting from radial impacts.
1) Determine the BHA Lateral Critical Speeds and
Mode Shapes
The BHA is analyzed as a beam-column. This
means the influences of bending stiffness and axial
load on lateral displacements of the BHA are taken into
account. The top of the BHA is frequently taken as a
free end. Pin connections are assumed at the locations
of stabilizers along the BHA. Straightforward linear,
undamped, free vibration analysis (see Appendix A) is
used to find the natural frequencies and mode shapes.
The usual 'added mass" influence5on the BHA lateral
motions is included.

365
2 DETECTION OF BHA LATERAL RESONANCES WHILE DRILLING WITH SURFACE LONGITUDINAL AND TORSIONAL SENSORS
The natural frequencies in the rotary speed range are
the potential critical speeds of rotation. Results from
many actual strings have shown that there are many
natural frequencies in the normal rotary speed range,
typically ten or more. Figure 1 illustrates an example of
the mode shapes for a few selected natural frequencies
and their corresponding revolutions per minute (revlmin)
values for a specific BHA.
In addition, it is common to find that the BHA is being
run in a buckled condition. Running in a buckled state
will usually cause high cyclic bending stresses, leading to
accelerated fatigue, so this condition should be avoided.
Once the potential critical rotary speeds and their
associated mode shapes are known, it is possible to go
to the second step of this procedure.
2) Determine the Axiai Excitation at the BHA Caused
by Running at Lateral Critical Speeds
Since the accuracy of the calculated BHA lateral
critical speeds is limited, and there will normally be sev-
eral critical rotary speeds in the neighborhood (within 5%)
of a selected revlmin, it is prudent to calculate the ex-
pected response for several critical speeds nearest the
revlmin. For each of these critical speeds, the shortening
of the corresponding mode shape is found. Each BHA
mode shape is separated into upper (heavy weight) and
lower (drill collar) sections.
The maximum and minimum amplitudes of the mode
shape are found by establishing a stabilizer- to-hole
clearance. Then, at the maximum deflection point of the
mode, the minimum deflection is ((stabilizer diameter) -
(BHA diarneter))l2 while the maximum deflection is the
-
minimum deflection plus (hole diameter) (stabilizer
diameter). Figure 2 illustrates this calculation. It as-
sumes the stabilizer rests on one side of the hole and
moves through its possible extremes. Computer simula-
tions for single sections of BHA between two stabilizers
show that this is typically the case. Reference 6 confirms
this assumption.
When a BHA resonance occurs, the lateral deflections
cause a slight shortening of BHA sections, which pulls
the rest of the drillstring down (assuming the bit remains
on bottom) and excites axial resonances at once and
twice the revlmin. The variation of the amplitude of the
BHA resonance with each revolution then causes an
alternating force on the bottom of the drillstring. This
force induces longitudinalvibrations in the combined
BHA-drillstringsystem.
3) Determine the Axiai Motions of the Entire
Driilstring Resulting from the BHA Axial Excitation
The shortening of each section of the BHA induces
axial motions in the entire string. Normally, axial reso-
nance frequencies of the BHA alone are above the rev1
rnin range for the string. However, when the drill pipe is
SPE 24583
included with the BHA, the system usually has several
axial resonance frequencies in the running range. The
shortening is proportional to the integral of the square of
the slope, (Y')~,
of the mode shape. The slope varies with
time as
y' = A + Bsin (at)
where o is the rotary speed. The square of y' varies as
A2 + %B2 + 2ABsin (at) - %B2cos (2at)
Consequently, the shortening will excite axial resonances
at once and twice the rev/min. See Appendix B.
Appendix C describes the method of analysis used to
predict the response of the top of the drillstring to the
BHA shortening. The method is a classical structural
dynamics procedure which uses modal analysis and
assumes that the damping for the individual modes is
uncoupled from the other modes.
According to the boundary conditions set at the top of
the drillstring, the forces andlor the displacements may
be predicted. The amplitude of responses at once and
twice the revlmin are predicted. If a spectral analysis of
the measured motion andlor force at the top of the
drillstring is performed, the results at once and twice rev1
min may be compared. If the measured responses are
comparable in magnitude to the predicted responses (i.e.
maximum deflections are realized), the system is prob-
ably running at or very near a BHA lateral critical speed.
In this case, either the rotary speed or the WOB should
be altered to avoid the resonance and its associated
mechanical damage to the drillstring. If the computed
responses exceed the measured responses significantly,
a BHA resonance is not occurring. On the other hand, if
the measured responses exceed the computed re-
sponses significantly, this indicates that vibrations are
occurring which are not wholly due to a BHA resonance.
4) Determinethe Torsional Excitation and Motion of the
Entire Driilstring Resulting from Radial Impacts
Analysis of single BHA sections in forward synchro-
nous whirl showed that when the BHA was eccentric to
the hole axis, the center of the BHA section went through
the maximum possible radial excursions. Typically, the
BHA could not maintain contact with the hole wall while it
moved outward. This phenomenon resulted in a once
per revolution impact of the BHA against the hole wall.
The amount of the BHA momentum of this impact can be
estimated from the radial position variation determined in
Step 2 and the rotary speed.
By assuming the radial impact induces a tangential
impulse equal to the coefficient of friction times the radial
impulse, it is possible to estimate a tangential excitation
to the BHA. The tangential impulse will excite the
drillstring torsionally. The details of how this torsional
excitation is computed and how the resulting motion at
the top of the drillstring is found are described in Appen-
SPE 24583 Paul Pashy, Yh-Mn Jan, John Kingman, and John hkqhemn
dix D. The summation of the responses of successive
impacts leads to the application of a convolution integral
to find the steady state response. The first three Fourier
components of the steady state response are found in
order to determine the torsional excitations at once,
twice, and three times the revlrnin.
The predicted once, twice and three times revlmin
torsional responses due to impacts associated with BHA
whirl are calculated for the top of the drillstring. These
responses can be compared to the measured torsional
spectral response at the top of the drillstring as a sepa-
rate evaluation of possible BHA resonance.
Field Results
Examples are presented which demonstrate both the
advantages and weaknesses of this attempt to indirectly
measure the dynamic behavior of the BHA. A vertical
well near Victoria, Texas was used for this example. The
BHA was 765 ft long, had an average weight per foot of
85 Iblft and there were three stabilizers. The drillstring
above the BHA was 4.5 inch, 16.6 Iblft drill pipe. The
swivel and hook weight was about 20,000 Ib. At the time
of these measurements, the measured depth was about
9400 ft, the weight on bit was 40,000 Ib and the rotary
speed was about 130 revlrnin.
The measurement system gave a real-time plot of the
spectrum of the response at the surface. Various calcu-
lated points were included on this plot. By comparing the
location of the calculated points to the spectrum it was
possible to estimate the severity of the measured dy-
namic response. Figures 3,4, and 5 show the axial force
and the torsional acceleration displays. The measure-
ments in Figures 3 and 4 were taken simultaneously.
Figure 3 shows the measured spectrum of axial force
at the top of the drillstring with two sets of plotted points.
The first five points are in the vicinity of the rotation rate
of 2.3 Hz (138 RPM). These points show that the mea-
sured axial force amplitudes are in the neighborhood of
the amplitudes predicted on the assumption of a forward
whirl resonance. The second set of five points near 4.6
Hz show that the calculated amplitudes are considerably
less than the corresponding measured amplitudes.
These results suggest that although a BHA resonance
may be occurring, there is vibration from some other
source in the twice revlmin region.
Calculated and measured values of torsional acceleration
are compared in Figure 4, using the same data set as
that used to generate Figure 3. The calculated values fall
well below the measured accelerations. This figure does
not show the well defined peaks at once and twice rev/
min that occur in Figure 3. These results suggest that
there is noise in the acceleration response that masks
the measurement of the BHA resonance.
Figure 5 is from a second data set, showing, like
3
Figure 3, the axial force at the top of the drillstring. The
character of this plot is similar to Figure 3, in that the calcu-
lated and measured results at once the revlmin shows a
reasonable pattern for BHA lateral resonance, while the
calculated response at twice the revlmin shows amplitudes
much smaller than the measured values. Since the second
data set is for the same well as the first set at a slightly
different depth, it is not surprising that a BHA resonance
may be occurring, and there is vibration from some other
source in the twice rotation rate region.
Conclusions
An analytical technique has been developed to predict
response at the top of the drillstring when the BHA
undergoes a forward synchronous whirl. Both axial and
torsional responses are predicted.
Experimental results confirm that responses similar to
those predicted occur in normal drilling operations.
Results to date suggest the analytical technique may be
combined with measured real time spectral analysis of
the motion at the top of the drillstring to help avoid
rotary speeds and WOB which excite forward whirl of
the BHA.
Based on analyses of many strings, it appears common
to run the BHA in a buckled condition. This practice
does not seem prudent to the authors. The addition of
stabilizers can correct this condition, while potentially
impacting the directional control of the well.
Many factors influence the response of the system.
Some of these, such as the mass of the swivel and
block, and the elasticity of the rotary table, need further
investigation.
Further studies have been conducted to improve the
current technique described in this paper7. In addition
to the factors described above (Conclusion 5), contin-
ued research is required in several areas. Specifically,
the analytical model should be extended to include
other BHA dynamic motions, e.g. backward whirl. Also,
simultaneously measuring surface and downhole
dynamic data will be required to further validate and
enhance the analytical technique described here.
Acknowledgements
The authors wish to express their appreciation to ARC0
Exploration and Production Technology and to EXLOG,
Inc. for permission to publish this paper, and for the com-
ments of many editors. Mr. Ronald B. Livesay of Hecate
Software, Inc. prepared the computer software.
References
1. P. R. Paslay and Y. M. Jan, 'Method of Determining
Drillstring Bottom Hole Assembly Vibrations", US Patent
Pending.
DETECTIONOF BHA LATERAL RESONANCES WHILE DRILLING WITH SURFACE LONGITUDINAL AND TORSIONAL SENSORS
2. A. A. Besaisow, Y. M. Jan and F. J. Schuh, 'Detection
of Various Drilling Phenomena Utilizing High Frequency
Surface Measurements", SPE 14327,80th Annual Tech-
nical Conference and Exhibition, Las Vegas, Nevada,
September 22-25, 1985.
3. A. A. Besaisow and M. L. Payne, 'A Study of Excita-
tion Mechanisms and Resonances IndUcing Bottomhole
Assembly Vibrations", SPE Drilling Engineering, March,
1988, p~ 93-101.
4. A. A. Besaisow, F. W. Ng and D. A. Close, 'Application
of ADAMS (Advanced Drillstring Analysis and Measure-
ment Systems) and Improved Drilling Performance",
SPE 19998, 1990 IADCISPE Drilling Conference, Hous-
ton, Texas, February 27 - March 2,1990.
5. S. S. Chen, M. W. Wambsganss and J. A.
Jendtzejczyk, 'Added Mass and Damping of a Vibrating
Rod in Confined Viscous Fluids", Journal of Applied
Mechanics, June, 1976, pp 325-329.
6. J. D. Jansen, Whirl and Chaotic Motion of Stabilized
Drill Collars", SPE 20930, Europec 90, The Hague,
Netherlands, October 22-24, 1990.
7. P. R. Paslay, "A Simple Model for Drill Collar Whirl
Between Stabilizers", ARC0 Internal Report, June 1991.
-
Appendix A Determination of Natural Frequencies
and Mode Shapes
All systems analyzed in this work have equations of
motion with n degrees of freedom which may be written
in matrix form as
M~X+C.~+K*X=F (A-1
where
M = real, symmetric, positive definite, nxn mass matrix
C = real, symmetric, nxn damping matrix
K = real, symmetric, nxn spring matrix
X = nxl displacement vector. The real part of the ih
element of X is the displacement of the ihdegree of
freedom. The elements of X are time dependent.
F = nxl real, force vector. The value of the ih element of
F is the work-absorbing component of force acting on the
ih degree of freedom. The elements of F are time depen-
dent.
The natural frequencies and mode shapes for the
system described by Equation A-1 are found by solving
Equation A-1 after requiring C and F to vanish so that
where XO is an amplitude vector independent of time and
o is a circular frequency. Substituting Equation A-3 into
A-2 yields
SPE 24583
K*XO=&*M*XO (A-4)
which is a classical eigenvalue problem. The matrices K
and M can be constructed directly, once the system is
described. For cases here, M is diagonal with diagonal
elements being masses or moments of inertia. The K
matrix is related to the elastic properties of the system.
- For example, the stations along the BHA for lateral vibra-
tions each have a lateral displacement and a rotation so
that the solutions for the problems shown in Figure A-1
are all that are required to construct K.
The solution to Equation A-4 is a set of n eigenvalues
(natural frequencies) and for each eigenvalue a vector
solution (mode shape). The mode shapes are usually
stacked into an nxn modal matrix, a as
a = PO,, XO*, ...XOJ
Assume the mode shapes are ordered for increasing
natural frequencies and normalized with respect to the
mass matrix so that the orthogonality conditions become
aT*M-a=l (A-5)
aT* K * a = R 2 (A-6)
where Iis the nxn identity matrix and R2is a nxn diagonal
matrix whose ih diagonal term q 2 is the ih natural fre-
quency squared.
When the solution to the eigenvalue problem, Equa-
tion A-4, yields negative values for squares of any of the
frequencies, the system is buckled. The buckled configu-
ration is unstable, and the solution for the mode shapes is
not considered here.
One advantage of finding a and R for the system
defined by Equation A-1 is that normal coordinates,$, may
be introduced. The coordinate change
X=a-+ (A-7)
is introduced into Equation A-1, and the result is
premultiplied by aT. The orthogonality conditions, Equa-
.$.+ a T - C - a - $ + R 2 - $ = a T * F
tions A-5 and A-6, are then introduced to yield
(A-8)
in this equation, aT* F is known and R2is diagonal so that
if aT* C a were diagonal, the equations would uncouple.
In this work, as in many structural analyses, aT C a is
assumed to be diagonal with its IW diagonal element
equal to Q is the magnification factor at resonance,
taken to be the same for all vibration modes. Let the
elements of $ be $ (I), the elements of F be F(I) and the
elements of a be a (1,~)so that the IW uncoupled equation
.4).
is
(1) + 4
(4Q)(1) + q2 4) (1) = a (J,O (J) (A-9)
.
Equation A-9 may be solved for $ (1) . . $ (n) when F (J)
is specified as a function of time. Then the physical
coordinates of X may be found from Equation A-7.
An example is when
SPE 24583 Paul Paslay, Yih-Mn Jan, John IGngman,and John Macphersan
F (K) = FO sin ( o t), (A-10)
all other F (I) vanish, where FO is constant and o is a
forcing frequency. The steady state solution is
-
X (J) = (AS2 + AC2)% sin ( o t arctan (ASIAC))
where X (J) is the JTH component of X
= a (J, I) a (K, I) FO / ((oI2 - o?)~ + ( a
A (I) oI/Q)~)~
0 (I) = arctan ((a oI/ (Q ( q 2- d)))
AC = Z A (I) cos (0 (I))
AS = C A (I) sin (0 (1))
and the summations are taken over all modes.
A second example is for a system initially at rest. At
time equals zero, a unit impulse is applied to the X(K)
coordinate. For times greater than zero, the solution is
given by
X(J) = Z AR (I) e-k(i)' sin (al t)
where
The summation is taken over all modes.
-
Appendix B Shortening of BHA Mode Shapes
The normalized mode shape for the IhBHA mode
shape is given by a (J,I). The slopes between adjacent
stations may be found from the mode shape. The ampli-
tude of the mode shape is determined as described in the
text so that each slope is approximated as
actual slope = (A0 + Al sin (a t)) (mode dope),Al c A0 (El)
where o is the rotating speed (in radians per unit time).
Since the actual slopes are small, the shortening, S(I), is
given by
S (I) = 1/2 (actual slope)2 dz P-2)
where the integral is taken over the length of either the
top or bottom section of the BHA and z measures length
along the BHA axis. By combining Equations 8-1 and B-
2, the result is
S(1)=H(A0*+2-AO-Al*sin(o*t)+Al~-
sin2 (o t)) I (mode slope) dz (83)
The integral can be evaluated numerically when the
mode shape is found and the sin2 ( a t) term can be
expanded so that
S (I) = M (A02 + AI2/2 + 2 A 0 Al sin (a t) - -
(AI2/2) sin (2 o t)) I (mode shape) dz w)
Equation 8-4 shows there is a once per revolution short-
ening component of amplitude
A 0 Al (mode shape) dz
and a twice per revolution shortening component of ampli-
tude
- (AI2/4) I (mode shape) dz
These once per revolution and twice per revolution
shortenings excite the entire drillstring in longitudinal
vibrations. The determination of the responses at each
frequency is described in Appendix C.
-
Appendix C Longitudinal Excitation and response
due to BHA Shortening
The entire drillstring can be modeled as a series of
masses connected by springs to represent its longitudinal
dynamic characteristics. The natural frequencies and
mode shapes may be found as described in Appendix A,
and the BHA shortening may be found as described in
Appendix B. This appendix describes how the BHA
shortening can be converted to a force system, giving the
same response as the shortening excitation.
Figure C-1 shows identical systems subjected to two
different excitations. In the first, there is a variable length
element between adjacent masses whose length varies
sinusoidally in time. In the second excitation, there is a
set of opposing forces acting on the same adjacent
masses and varying sinusoidally in time. The amplitudes
of the forces are related to the shortening as indicated in
the figure. By writing the equations of motion for each
system and comparing them, it is clear that the steady
state solutions for each set of excitations is the same.
For the procedure described here, four steady state
solutions are found for each mode shape which is consid-
ered. The first two problems are for excitations at the
rotary speed. The first problem is for the upper part of the
BHA shortening and the second is for the bottom part.
The steady state response at the top of the drillstring is
found for each problem using the solution given in Appen-
dix A. The two responses are added, taking into account
phase, to find the predicted once per revolution response
at the top of the drillstring.
The second two problems are like the first two except
the shortening is for the twice per revolution component.
The result is the predicted twice per revolution response
at the top of the drillstring.
-
Appendix D The Torsional Response Problem
Torsional excitation is based on a once per revolution
radial impact of the BHA against the hole wall. The im-
pact for a given mode shape occurs at the location of the
largest displacement in that mode. The radial clearance
for the development of radial velocity for impact, DHS, is
DHS = (Hole Diameter) (Stabilizer Diameter)
For the ITH natural frequency, mi, the impact velocity, V (I),
is estimated to be
6 DETECTION OF BHA LATERAL RESONANCES WHILE DRILLING WITH SURFACE LONGITUDINAL AND TORSIONAL SENSORS
The radial momentum of the mode at impact is assumed
equal to the sum of the products of the masses and lateral
velocities along the BHA when the maximum velocity is
equal to V (I). Obviously, the modal station with the larg-
est lateral displacement has the velocity V (I) while the
remaining lateral velocities are proportional to the local
mode shape lateral displacements. The rotational mo-
mentum associated with the lateral mode shape is ne-
glected in this calculation. The resulting radial momentum
is designated as R (I). The change of momentum, RI (I),
which occurs at impact is found as
RI (1) = (1 + CR) R (I)
where CR is the coefficient of restitution. The torsional
impulse, TI (I) is determined from RI (I) as
I Maximum
deflection
SPE 24583
where p is the coefficient of friction and DCD is the drill
collar diameter.
The solution, given in Appendix A, for an impulsive
loading can be used for a torsional model of the entire
drillstring to find the torsional response at the top of the
drillstring. As the system is damped, the torsional re-
sponse becomes essentially zero in about 10 seconds.
Consequently, it is straightforward to convolve the re-
sponse for each revolution to find the periodic motion
caused by the repeated impulses. The first three Fourier
series components of the periodic motion are found to
determine the once, twice and three times rotary speed
responses.
Top of BHA
>
Figure 1:Selected BHA mode shapes
with natural frequencies and associated
RPM. Only five of the first seventeen
mode shapes are shown. The position
of maximum displacement in the BHA
is indicated by an arrow.
Drill 91
RPY
Figure 2: Stabilizer-to-hole clearance
Minimum
deflection
SPE 24583 Paul Pashy, YihMn Jan, John ICKlgman, and John Ma-

Com~arisonof Measured Axial Force With Predicted at 138 RPM

.
-................................................................................................................::......................................................
-
J
- ........................................................
;
0
................................................. -..-
10-1
0 1 2 3 4 5 6
FREQUENCY (Hertz)

Figure 3

Comparison of Measured Angular Acceleration With Predicted at 138 RPM

............. .......................................
......................... ........
.. ......
.._
.
....... ....... .
..
.
...........................
....... .....
...........................+....
..

FREQUENCY (Hertz)

Figure 4
371
8 DETECTION OF BHA LATERAL RESONANCES WHILE DRILLING WITH SURFACE LONGITUDINAL AND TORSIONAL SENSORS SPE 24583

Comparison of Measured Axial Force With Predicted at 126 RPM

.........................+ ...........................+ ...........................!........................... +...........................!.........................
.............
.........................
.......................................................
; ...........................
...........................:>.. :...........................;
.....................................................
;
. ...........................::.........................
............................
.........................

2 3 4

FREQUENCY (Hertz)

Figure 5

I Displacement Problem
FRD I
~hbrteningUrment
FLD Length x Constant + A.sin(mt)

T - Axial Tension
FLD, FRD, FLR, FRR = Lateral Forces
MLD, MRD,MLR, MRR = Bending Moments
Figure GI: Two excitations which give the same motions
Figure A-1 :Beam column problems whose solution to the system.
determines K. For the displacement problem,
displacement at the right is unity and rotation is zero.
For the rotation problem, rotation at the right is unity and
displacement is zero.