Sie sind auf Seite 1von 7

Mathematical Geology, Vol. 24, No.

5, 1992

Least Squares Best-Fit Circles (with Applications to


Mohr's Diagram) 1

Richard J. Lisle 2

Procedures are outlined for the selection of a least squares best-fit circle to data points defined by
rectangular Cartesian coordinates. Equations are derived to allow fitting o f circles centered on the
x-axis as well as off-axis Mohr circles. These procedures are applicable to the estimation of second-
order tensors such as stress and strain by means of Mohr's diagram,

KEY WORDS: statistics, image analysis, strain, estimation, structural geology, tensors.

INTRODUCTION

The Mohr circle construction is widely used as a means of representing second-


order tensor quantities such as stress and strain as well as anisotropic physical
properties, e.g., magnetic susceptibility and thermal conductivity (Nye, 1972).
The construction is also used for the estimation of tensors from their components
measured in specific directions. For example in two-dimensional geological
strain analysis (Brace, 1961; Jaeger, 1962; Ramsay, 1967) measurements of
angular shear and longitudinal strain components made in a number of directions
in a planar slice through a deformed rock yield points (xi, Yi) on a Mohr diagram
which can be interpreted in terms of principal strains. The computation of the
principal axes of the finite strain ellipse and their directions is achieved by first
fitting a circle, the Mohr Circle, to these data. This note describes how a circle
can be objectively chosen in such cases by means of least-squares fitting.
In applications involving Mohr's diagram the procedures described below
offer an alternative to methods which find directly the best-fit ellipse (e.g.,
Kanagawa, 1990; Erslev and Ge, 1990). However, potential applications are
not limited to Mohr circles but include problems in the fields of remote sensing,
image analysis, surveying, archeology, and wherever a best-fit circle has to be
fitted to a set of co-planar points.

~Received 17 July 1991; accepted 6 November 1991.


2Department of Geology, University of Wales, Cardiff CF1 3YE, Wales.

455

0882-8121/92/0700-0455506.50/1© 1992InternationalAssociationfor MathematicalGeology


456 Lisle

T H E B E S T - F I T C I R C L E C E N T E R E D ON T H E X-AXIS

The problem in question is illustrated in Fig. 1 which shows a Mohr dia-


gram on which is plotted the N data points (xi, Yi where i = 1 to N) which are
to be used to estimate a symmetrical second-order tensor. For example, if we
are estimating the stress tensor, these xi, Yi data will consist of the normal and
shear stress components respectively measured on variably-oriented planes. The
estimation of the principal axes of the tensor requires drawing a circle which
passes through the data points. For symmetrical tensors like stress, the center
o f the Mohr circle is constrained to lie on the horizontal (x) axis of the diagram.
This implies that two Mohr points are sufficient for the construction of the circle.
In situations where more data are available, the problem is overdefined and leads
to the practical problem of choosing the circle which best fits the data.
Equations for the calculation of the best-fit circle are derived using the
standard least-squares approach (e.g., Neville and Kennedy, 1964, p. 186-187).
The equation of a circle of radius r and center at distance c along the x-axis
from the origin is:
(X -- C) 2 + y 2 = r 2 (1)

Using A = - 2 c and B -- w 2 = c 2 - r 2 (see Fig. 1), Eq. (1) can be re-written:


x 2 + Ax + y2 + B = 0 (2)

For a data point (xi, yi) which does not lie exactly on the circle, the fight hand
side of Eq. (2) will be non-zero and equal to Ei.
Considering all data points (xi, yi; i = 1 to N), the least squares best-fit
circle is the one which minimizes the sum of the squares of the error terms,
i.e.,
min Z (E~) = min Z (x~ + Ax~ + y2 + B)2 (3)

The minimum o f ~ (E~) is found by partial differentiation of Eq. (3) with respect
to circle parameters A and B and setting these partial derivatives to zero:

-5

' 1 /t3 ~ ' i ¢ " "\


X

Fig. 1. Fitting a circle with center on the x-axis to


data xi, Yi. The data and results are shown in Table 1.
Best-Fit Circles 457

d Y', ( E 2 ) / d A = Z (x~ + Ax 2 + xiy2~ + Bxi) = 0

d E ( E 2 ) / d B = • (x 2 + Axi + y2 + B) = 0
These equations simplify to give what are known as the normal equations:

A Z (xp) + B Z (xi) = - Z (x2) - Z (xiy~)


A Z (Xi) "q- B = - Z (x 2) - Z (y~)
The solution to this system o f simultaneous equations gives expressions for A
and B, parameters for the circle. These can be manipulated to give the following
equations for the center (c) and radius (r) for best-fit circle:

r Ic2 +

Example. The above equations are employed to calculate the best circle
fitted to the data listed in Table 1 and represented graphically in Fig. 1. The
calculations are performed by means o f the BASIC program listed in Appendix

Table 1. Sample Data

Data in Fig. 1 Data in Fig. 2


(on-axis circle) (off-axis circle)

i XI YI XI YI

1 4.6 -1.9 4.6 3.3


2 6.0 -3.8 6.0 1.4
3 10.8 -1.7 10.8 3.5
4 10.7 -2.5 10.7 2.7
5 10.8 1.9 10.8 7.1
6 6.7 4.0 6.7 9.2
Best-fit circle

Center 7.6 0.0 7.6 5.2


Radius 3.9 3.9
Root-mean-square deviation as percentage of radius:
5.9% 5.9%
458 Lisle

A. The best-fit circle has center, c = 7.6 and radius, r = 3.9. The root-mean-
square deviation of the points away from this circle (expressed as a percentage
of the radius) is 5.9%.

B E S T - F I T OFF-AXIS C I R C L E
The need to consider circles with centers located away from the x-axis
arises in two situations. First, off-axis Mohr circles are employed to represent
asymmetrical second-order tensors quantities (De Paor and Means, 1984; re-
viewed by Means, 1990). Second, certain graphical techniques for finite strain
analysis (Ragan, 1983, p. 190-191; Lisle and Ragan, 1988) involve the con-
struction of the circle without a priori knowledge of the position of the origin
of the Mohr diagram; the center of the circle being used to fix the x-axis.
The equation of a circle of radius r and center at c, d (Fig. 2) is
(x - c) 2 + (y - d ) 2 = r 2
Using b = w 2 = c 2 + d 2 - r2 (Fig. 2), this becomes
x 2 - 2cx + y 2 _ 2yd + b = 0
For actual data points (xi, yi), error terms E i will appear on the right-hand side
which, as before, are to be subjected to least-squares minimalization. Partial
differentiations of the equation for ~ (E 2) with respect to circle variables c, d,
and b, respectively, and equating to zero yield:
d ~ (E2i)/dc = ~ (2x~c + 2xiYi d -- xib - x 3 - xiY 2)

=0
d Z (E2)/dd = Z (2xiyic + 2yZd - y~b - xZyi - y3)

=0
d Z (E2)/db = ~ ( - 2 x i c - 2yid + b + x 2 + y2)

=0

Y .10

I k

Fig. 2. Fitting an off-axis Mohr circle to data points


5 10 on the Mohr diagram. The data and results are shown
in Table 1.
Best-Fit Circles 459

and give the normal equations:


c • ~ (2x~) + d • ~ (2xiyi) + b • ~ ( - x i ) = ~ (x~ + xiy 2)

C • ~] (2xiYi) q'- d • E (2y 2) + b • ~ ( - Y i ) --- ~] (x2yi + y 3 )

c • Z(-xi) + d" ~(-yi) + b • ½N= Z - ~(xil 2 + y ~ )


The solution of this system of equations allows calculation of c, d, and b
K(DF - E 2) - L ( B F - E C ) + M ( B E - D C )
C =
A D F + 2 B C E - A E 2 - FB 2 - D C 2

-K(BF - CE) + L(AF - C 2) - M ( A E - B C )


d=
ADF + 2BCE- A E 2 - FB 2 - D C 2

K(BE - DC) - L(AE - BC) + M(AD - B 2)


b=
ADF + 2BCE - AE 2 - FB 2 - DC 2

where A = 2 E (x~), B = 2 E (xiYi), C = - Z (xi), D = 2 Z (y~), E =


-E ( Y i ) , F = 1N, K ~ (x~ + X i Y i2) , L = ~ (x2iyi + y 3 ) , a n d M = 1
(x 2 + y2). The radius of the best-fit circle is found from r = x/(c 2 + d 2 - b).
Example. The best-fit circle is found for the data points in Fig. 2 (listed
in Table 1). The equations above for the calculation of the best-fit circle are
incorporated in the computer program in Appendix B. The circle so found has
a center at c = 7.6 and d = 5.2 and a radius o f 3.9. The program also determines
the root-mean-square deviation of the points from the circle to be 5.9% of the
radius.

A P P E N D I X A. L I S T I N G OF A G W - B A S I C P R O G R A M T O F I N D
T H E BEST-FIT C I R C L E C E N T E R E D O N T H E X-AXIS

10 C O L O R 4,7,3: CLS
20 P R I N T " PROGRAM MOHR"
30 PRINT"==============~================================~=======================

40 P R I N T " F i n d s a l e a s t - s q u a r e s b e s t fit Mohr c i r c l e to a n u m b e r of p o i n t s x,y"


50 P R I N T " . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

60 DIM X ( 1 0 0 0 ) , Y ( 1 0 0 0 )
70 REM d a t a input
80 INPUT" N a m e o f input f i l e w i t h x,y d a t a " ; D $
90 O P E N D$ FOR INPUT AS ~I
100 C O U N T =I
110 IF E O F ( I } T H E N G O T O 150
120 INPUT4~ , X ( C O U N T ) , Y ( C O U N T )
130 C O U N T = C O U N T + I
140 G O T O 110
150 N=COUNT-I
160 F O R I=I TO N
170 T L = T L + X ( I ) ^ 2 + Y ( I ) ^ 2
180 T M = T M + X ( I )
190 T R = T R + X ( I ) ^ 3 + X ( I ) * Y ( I ) ~ 2
200 B R = B R + X ( I ) ^ 2
205 T Y = T Y + Y ( I )
460 Lisle

210 N E X T I
220 P R I N T
230 P R I N T
240 C E N T R E = . 5 * ( ( T L * T M - N * T R ) / ( T M ^ 2 - N * B R ) )
250 R A D I U S = S Q R ( C E N T R E ^ 2 + ( T R - 2 * C E N T R E * B R ) / T M )
260 P R I N T " C e n t r e of M o h r c i r c l e a t " ; C E N T R E ; " , 0 . 0 " ; " Radius = ";RADIUS
270 P R I N T " ...................... "
340 R E M r o o t m e a n s q u a r e v a r i a t i o n e x p r e s s e d as % of r a d i u s
350 F O R I=I T O N
360 D I S T C E N T = S Q R ( ( X ( I ) - C E N T R E ) ^ 2 + Y ( I ) ^ 2 )
390 T S V A R = T S V A R + ( D I S T C E N T - R A D I U S ) ^ 2
400 N E X T I
410 R M S V = S Q R ( T S V A R / N ) / R A D I U S * I O 0
420 P R I N T " R o o t M e a n S q u a r e D e v i a t i o n of p o i n t s f r o m c i r c l e " : P R I N T "
e x p r e s s e d as % of r a d i u s = " ; R M S V ; " %"

APPENDIX B. LISTING OF A GW-BASIC P R O G R A M TO FIND


THE BEST FIT OFF-AXIS CIRCLE

10 C O L O R 4,7,3: CLS
20 P R I N T " PROGRAM MOHROFF"
30 P R I N T " = = = . . . . . . . . . . . . . . . . . . . . ===== . . . . . . . = .... = ..... = . . . . . . . . . ==== . . . . . . = ....
========"
40 P R I N T " F i n d s a l e a s t - s q u a r e s b e s t fit M o h r c i r c l e t h r o u g h p o i n t s x , y "
50 P R I N T " T h e c i r c l e h a s r a d i u s r a n d a c e n t r e o f f the x - a x i s at x = c , y = d "
70 P R I N T " . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

80 D I M X ( 1 0 0 0 ) , Y ( 1 0 0 0 )
90 R E M data input
100 I N P U T " N a m e of i n p u t f i l e w i t h x , y d a t a " ; D $
110 O P E N D$ F O R I N P U T AS ~I
120 C O U N T =I
130 IF E O F ( 1 ) T H E N G O T O 170
140 I N P U T ~ , X ( C O U N T ) , Y ( C O U N T )
150 C O U N T = C O U N T + I
160 G O T O 130
170 N = C O U N T - I
180 A = 0 : B = 0 : C = 0 : D = 0 : E = 0 : F = 0 : K = 0 : L = 0 : M = 0
190 F O R I=I T O N
200 A = A + 2 * X ( I ) ^ 2
210 B = B + 2 * X ( I ) * Y ( I )
220 C = C - X ( I )
225 D = D + 2 * Y ( I ) ^ 2
230 E = E - Y ( I )
240 F = F + . 5
250 K = K + X ( I ) ^ 3 + X ( I ) * Y ( I ) ^ 2
260 L = L + X ( I ) ^ 2 * Y ( I ) + Y ( I ) ^ 3
270 M = M - . 5 * ( X ( I ) ^ 2 + Y ( I ) ^ 2 )
280 N E X T I
290 R E M c o e f f s of e l l i p s e e q n ( a c , b c , c c )
300 D E T = A * D * F + 2 * B * C * E - A * E * E - F * B * B - D * C * C
310 C C = ( K * ( D * F - E * E ) - L * ( B * F - E * C ) + M * ( B * E - D * C ) ) / D E T
320 D C = ( - K * ( B * F - C * E ) + L * ( A * F - C * C ) - M * ( A * E - B * C ) ) / D E T
330 B C = ( K * ( B * E - D * C ) - L * ( A * E - B * C ) + M * ( A * D - B * B ) ) / D E T
335 R = S Q R ( C C * C C + D C * D C - B C )
340 P R I N T " . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

350 PRINT: P R I N T " B e s t fit M o h r C i r c l e "


360 P R I N T " .................... "
370 P R I N T " c =";CC; " d = " ; D C ; " r = ";R
380 R E M r o o t m e a n s q u a r e v a r i a t i o n e x p r e s s e d as % of r a d i u s
390 F O R I=I TO N
400 D I S T C E N T = S Q R ( ( X ( I ) - C C ) ^ 2 + ( Y ( I ) - D C ) ^ 2 )
410 T S V A R = T S V A R + ( D I S T C E N T - R ) ^2
420 N E X T I
430 R M S V = S Q R ( T S V A R / N ) / R * I 0 0
440 P R I N T " R o o t M e a n S q u a r e D e v i a t i o n of p o i n t s f r o m c i r c l e " : P R I N T " expr
e s s e d as % of r a d i u s = " ; R M S V ; " %"
Best-Fit Circles 461

ACKNOWLEDGMENTS
T h e a u t h o r is grateful to D e c l a n D e P a o r and an a n o n y m o u s r e v i e w e r f o r
t h e i r useful s u g g e s t i o n s f o r i m p r o v i n g t h e m a n u s c r i p t .

REFERENCES
Brace, W. F., 1961, Mohr Construction in the Analysis of Large Geologic Strain: Bull. Geol. Soc.
Am., v. 72, p. 1059-1080.
De Paor, D. G., and Means, W. D., 1984, Mohr Circles of the First and Second Kind and Their
Use to Represent Tensor Operations: J. Struct. Geol., v. 6, p. 693-701.
Erslev, E. A., and Ge, H., 1990, Least-Squares Center-to-Center and Mean Object Ellipse Fabric
Analysis: J. Struct. Geol., v. 12, p. 1047-1060.
Jaeger, J. C., 1962, Elasticity, Fracture and Flow: Methuen.
Kanagawa, K., 1990, Automated Strain Analysis Using an Image Analysis System: J. Struct. Geol.,
v. 12, p. 139-143.
Lisle, R. J., and Ragan, D. M., 1988, Strain Determination from Three Measured Stretches--A
Simple Mohr Circle Solution: J. Struct. Geol., v. 10, p. 905-906.
Means, W. D., 1990, Kinematics, Stress, Deformation and Material Behaviour: J. Struct. Geol.,
v. 12, p. 953-971.
Neville, A. M., and Kennedy, J. B., 1964, Basic Statistical Methods for Engineers and Scientists:
International.
Nye, J. F., 1972, Physical Properties of Crystals: Oxford.
Ragan, D. M., 1983, Structural Geology: An Introduction to Geometrical Techniques: Wiley, New
York.
Ramsay, J. G., 1967, Folding and Fracturing of Rocks: McGraw-Hill.

Das könnte Ihnen auch gefallen