Centroid, Area, and Moments of Inertia

Yong-Ming Li
January, 1997
1 Introduction
This design document was originally written for computation of area, centroid, and moments of
inertia of lamina (a thin plate of uniform density). It is expanded to compute physical properties
of wire and solid.
Area, centroid (or center of mass), and moments of inertia (a measure of an objects resistance to
changes in its rotation rate) are usually studied in engineering mechanics. For detailed discussion,
one may refer to [Hibbeler1991].
2 Set of points
Given n points (p
1
, p
2
, , p
n
) in space with associated mass m
1
, m
2
, , m
n
, then the position
vector of the center of mass is given by:
x
c
=
1
M
n

i=1
m
i
x
i
, y
c
=
1
M
n

i=1
m
i
y
i
, z
c
=
1
M
n

i=1
m
i
z
i
,
where M = m
1
+ m
2
+ + m
n
. If m
1
= m
2
= = m
n
= m (uniform mass), then
x
c
=
1
n
n

i=1
x
i
, y
c
=
1
n
n

i=1
y
i
, z
c
=
1
n
n

i=1
z
i
.
3 Wire
Assume that an open or closed thin wire can be mathematically described by a vector-valued
parametric curve r(t) = (x(t), y(t), z(t)) with t [a, b] and its density of mass is (x, y, z). Then,
Length: L =
_
b
a
_
x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt.
Mass: M =
_
b
a
(x, y, z)
_
x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt.
Moment about the y-z plane: M
yz
=
_
b
a
x(t)(x, y, z)
_
x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt.
Moment about the x-z plane: M
xz
=
_
b
a
y(t)(x, y, z)
_
x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt.
Moment about the x-y plane: M
xy
=
_
b
a
z(t)(x, y, z)
_
x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt.
1
Y.M. Li 2
The center of mass is given by
x
c
=
M
yz
M
, y
c
=
M
xz
M
, z
c
=
M
xy
M
.
In most engineering applications, the density of mass is constant so that can be taken out of the
integration sign. We may further assume that = 1 and hence discard it from our discussion. In
this case, the mass M has the same value as the length L though they have dierent units.
We now consider moments of inertia. To simplify the typing, we denote the small change of arc
length by
ds =
_
x

(t)
2
+ y

(t)
2
+ z

(t)
2
dt.
Since r(t) is a 3D space curve, the moments of inertia to three axes are
I
xx
=
_
L
[y
2
(t) + z
2
(t)]ds.
I
yy
=
_
L
[x
2
(t) + z
2
(t)]ds.
I
zz
=
_
L
[x
2
(t) + y
2
(t)]ds.
The products of inertia are
I
xy
= I
yx
=
_
L
xyds.
I
xz
= I
zx
=
_
L
xzds.
I
yz
= I
zy
=
_
L
yzds.
We can represent all inertias in a matrix called the inertia matrix as follows:
_
_
_
I
xx
I
xy
I
xz
I
yx
I
yy
I
yz
I
zx
I
zy
I
zz
_
_
_,
which will be used to determine the principal inertias and axes later.
The inertias we have discussed so far are dened on a coordinate system (O, x, y, z) where the
origin may not coincide with the centroid. Let (C, x, y, z) be a new coordinate system with C
being the centroid; x, y, and z parallel to the x, y, and z axes. Then, the moment of inertia to
the x-axis is
I
xx
=
_
L
[(y
c
+ y)
2
+ (z
c
+ z)
2
]ds = I
x x
+ L(y
2
c
+ z
2
c
) + 2y
c
_
L
yds + 2z
c
_
L
zds,
where
I
x x
=
_
L
[ y
2
+ z
2
]ds
is a moment of inertia to the axis x passing through the centroid. Noting that y and z passing
through the centroid, we have
_
L
yds =
_
L
zds = 0. Accordingly, I
xx
= I
x x
+ L(y
2
c
+ z
2
c
). We can
similarly derive I
y y
and I
z z
. The results are summarized as follows:
I
x x
= I
xx
L(y
2
c
+ z
2
c
).
I
y y
= I
yy
L(x
2
c
+ z
2
c
).
Y.M. Li 3
I
z z
= I
zz
L(x
2
c
+ y
2
c
).
The other products of inertia can be calculated in a similar manner and the results written as
follows
I
x y
= I
xy
Lx
c
y
c
.
I
x z
= I
xz
Lx
c
z
c
.
I
y z
= I
yz
Ly
c
z
c
.
We now consider the computation of principal moments of inertia and principal axes. It is readily
seen that
I
x x
+ I
y y
+ I
z z
= 2
_
L
r
2
ds,
where r is a distance to the centroid from an arbitrary point on the curve. It is invariant under
any transformation. Therefore, the sum of the moments of inertia is invariant with respect to a
coordinate system rotation. In terms of matrix notation, the sum of the moments of inertia is just
the sum of the elements on the principal diagonal of the inertia matrix and is known as the trace
of that matrix. So the trace of the inertia matrix is unchanged by a coordinate rotation, because
the trace of any square matrix is invariant under an orthogonal transformation.
Next the products of inertia are considered. A coordinate rotation of axes can result in a change
in the signs of the products of inertia. A 180
o
rotation about the x axis, for example, reverses the
signs of I
x y
and I
x z
, while the sign of I
y z
is unchanged. This occurs because the directions of the
positive y and z axes are reversed. On the other hand, a 90
o
rotation about the x axis reverses the
sign of I
y z
. It can be seen that the moments and products of inertia vary smoothly with changes in
the orientation of the coordinate system because the direction cosines vary smoothly. Therefore,
an orientation can always be found for which a given product of inertia is zero. In other words, the
inertia matrix is diagonal. The three mutually orthogonal coordinate axes are known as principal
axes in this case, and the corresponding moments of inertia are the principal moments of inertia.
The three planes formed by the principal axes are called principal planes.
If I is a principal moment of inertia, then I satises the cubic characteristic equation
_
_
_
I
xx
I I
xy
I
xz
I
yx
I
yy
I I
yz
I
zx
I
zy
I
zz
I
_
_
_ = 0.
This is a eignevalues and eignevvectors issue. The obtained eignevalue are the principal moments
of inertia and eignevvectors are the principal axes.
4 Generalization
In the previous section,we discussed how to compute physical properties for a wire. All the dis-
cussions can be generalized to planar or curved plates and solids. This is done by replacing L by
a continuous region D in integral and ds by the innitesimal mass element dm that may take one
of the following values:
dm =
V
dV ,
dm =
A
dA,
dm =
l
dl,
where
V
,
A
, and
l
are the volume density, area density and length density.
Y.M. Li 4
5 Lamina
5.1 Introduction
A lamina is a thin planar plate with uniform thickness. Mathematically, lamina is a two-dimensional
closed planar surface with mass and density. Since it is a planar surface, a closed boundary curve
is sucient for computing its physical properties. Accordingly, computations can be simpler and
faster. Therefore, we discuss its physical properties in a dierent way than that in the previous
sections though the end results should be the same if we still stick to the generic approach.
5.2 Denition
Denitions of an area, centroid, and moments of inertia for a lamina can be found in most text
books of engineering mechanics (e.g., [Hibbeler1991]). For completeness, we give a brief review of
these denitions in this section.
Let D be a region consisting of the boundary of the lamina and its interior; (x, y) be the density
of mass of the lamina. Then,
Area: A =
_ _
D
dxdy.
Mass: M =
_ _
D
(x, y)dxdy.
Moment about the x-axis: M
x
=
_ _
D
y(x, y)dxdy.
Moment about y-axis: M
y
=
_ _
D
x(x, y)dxdy.
Moment of inertia about the x-axis: I
xx
=
_ _
D
y
2
(x, y)dxdy.
Moment of inertia about the y-axis: I
yy
=
_ _
D
x
2
(x, y)dxdy.
Polar moment of inertia: I
O
= I
x
2
+y
2 =
_ _
D
(x
2
+ y
2
)(x, y)dxdy = I
xx
+ I
yy
.
Product of inertia: I
xy
=
_ _
D
xy(x, y)dxdy.
The center of mass (or centroid) is given by
x
c
=
M
y
M
, y
c
=
M
x
M
.
In most engineering applications, the density of mass of lamina is constant so that can be taken
out of the integration sign. We may further assume that = 1 by disregarding the constant factor.
Therefore, we have the following simpler formulas:
Mass: M =
_ _
D
dxdy.
Moment about the x-axis: M
x
=
_ _
D
ydxdy.
Moment about y-axis: M
y
=
_ _
D
xdxdy.
Moment of inertia about the x-axis: I
xx
=
_ _
D
y
2
dxdy.
Moment of inertia about the y-axis: I
yy
=
_ _
D
x
2
dxdy.
Polar moment of inertia: I
O
= I
xx
+ I
yy
.
Y.M. Li 5
Product of inertia: I
xy
=
_ _
D
xydxdy.
In engineering mechanics, we often want to know the moments of inertia about the axes passing
through the centroid. Let a user dened coordinate system be (O, x, y) and the centroid of a
lamina be C = (x
c
, y
c
). Assume the new coordinate system is (C, , ) with - and -axis parallel
to the x- and y-axis. Then, we have
I
xx
=
_ _
y
2
dxdy =
_ _
(y
c
+ )
2
dxdy = y
2
c
_ _
dxdy + 2y
c
_ _
dxdy +
_ _

2
dxdy
= y
2
c
A + 2y
c
_ _
dd +
_ _

2
dd.
It is noted that the third term represents the moment of inertia about the centroidal axis . The
second term is zero since the - and -axis pass through the centroid, i.e.,
_ _
dd = 0. Thus,
we have
I
xx
= y
2
c
A + I

or I

= I
xx
y
2
c
A.
Similarly, we can drive:
I

= I
xy
x
c
y
c
A
I

= I
yy
x
2
c
A
I

2
+
2 = I

+ I

.
It should be pointed out that I

= 0 if a lamina is symmetric to either - or -axis.

5.3 Principal moments of inertia
Let I

, I

, I

2
+
2, and I

be moments of inertia dened in the coordinate system (C, , ), where

C is the centroid of lamina. We now want to compute moments of inertia about a new system
(C, u, v) which is obtained by rotating (C, , ) by . It is noted that
u = cos + y sin
v = cos x sin
Therefore, we can derive that
I
uu
=
_
v
2
dA = I

cos
2
+ I

sin
2
2I

sin cos .
I
vv
=
_
u
2
dA = I

sin
2
+ I

cos
2
+ 2I

sin cos .
I
uv
=
_
uvdA = I

sin cos I

sin cos + 2I

(cos
2
sin
2
).
I
u
2
+v
2 = I

2
+
2.
The above equations may be simplied by using the trigonometric identities
sin(2) = 2 sin cos and cos(2) = cos
2
sin
2
,
in which case
I
uu
=
I
xx
+ I
yy
2
+
I
xx
I
yy
2
cos(2) I
xy
sin(2) (1)
Y.M. Li 6
I
vv
=
I
xx
+ I
yy
2

I
xx
I
yy
2
cos(2) + I
xy
sin(2)
I
uv
=
I
xx
I
yy
2
sin(2) + I
xy
cos(2)
Dierent results in dierent moments of inertia. We now want to determine the
p
such that I
uu
and I
vv
are maximum and minimum. This particular set of axes is called the principal axes, and
the corresponding moments of inertia with respect to these axes are called the principal moments
of inertia. The
p
may be found by dierentiating, for example, I
uu
with respect to and setting
the result equal to zero. Thus,
dI
uu
d
= 2
I

I
2
sin 2 2I

cos 2 = 0.
Consequently, at =
p
, we have
tan 2
p
=
2I

I
(2)
and
I
max
, I
min
=
I

+ I

_
I

2
_
2
+ I
2

.
It is noted that equation (2) gives two solutions,
1
and
2
, which are 90
o
apart. To nd which one
is correct, we may substitute into equation (1) and check if I
uu
matches I
max
. If not, we should
set = + /2.
We now consider two principal axes. Let the x- and y-axis be dened respectively by two unit
vectors x = (x
1
, x
2
) and y = (y
1
, y
2
). Then, the unit vectors that dene the principal axes are
u = x cos
p
+y sin
p
and v = y cos
p
+x sin
p
.
Replacing in (1), we can obtain I
uu
and I
vv
which correspond to I
min
and I
max
.
5.4 Green Theorem
We start our discuss by considering the Greens theorem.
Theorem 5.1 (Green Theorem) Let C be a piecewise smooth simple closed curve and D be the
region consisting of C and its interior. If P(x, y) and Q(x, y) are functions that are continuous
and have continuous rst partial derivatives throughout an open region R containing D, then
_ _
D
_
Q
x

P
y
_
dxdy =
_
C
Pdx + Qdy
The sign of the right side integration is associated with the orientation of the curve. If we walk
along the curve and see the region D on our left side, then the integration is positive. This
requirement coincides with the denition of the ACIS about an orientation of a curve. Greens
theorem can also be applied to a region D which contains holes, provided we integrated over entire
boundary and always keep the region D to the left. This requirement again matches the denition
of the ACIS. Hence, Greens theorem will be used to compute properties of lamina which have
holes inside.
Y.M. Li 7
In CAD systems, curves are often given parametrically as r(t) = (x(t), y(t)) with t [t
0
, t
1
]. Let
P(x, y) = y and Q(x, y) = 0. From Green theorem, a formula of computing an area for a lamina
is given by
A =
_ _
D
dxdy =
_ _
D
_
Q
x

P
y
_
dxdy =
_
C
ydx =
_
t
1
t
0
y(t)x

(t)dt.
As an example, we compute an area of a circle x() = Rcos() and y() = Rsin():
A =
_
2
0
y()x

()d = R
2
_
2
0
(sin )
2
d = R
2
_

2

1
4
sin 2
_
2
0
= R
2
.
Similarly, we can derive
M
x
=
_ _
ydxdy =
1
2
_
y
2
dx =
1
2
_
t
1
t
0
y
2
(t)x

(t)dt (P =
1
2
y
2
, Q = 0)
M
y
=
_ _
xdxdy =
1
2
_
x
2
dy =
1
2
_
t
1
t
0
x
2
(t)y

(t)dt (P = 0, Q =
1
2
x
2
)
I
xx
=
_ _
y
2
dxdy =
1
3
_
y
3
dx =
1
3
_
t
1
t
0
y
3
(t)x

(t)dt (P =
1
3
y
3
, Q = 0)
I
xy
=
_ _
xydxdy =
1
2
_
xy
2
dx =
1
2
_
t
1
t
0
y
2
(t)x(t)x

(t)dt (P =
1
2
xy
2
, Q = 0)
I
yy
=
_ _
x
2
dxdy =
1
3
_
x
3
dy =
1
3
_
t
1
t
0
x
3
(t)y

(t)dt (P = 0, Q =
1
3
x
3
)
5.5 Gauss-Legendre integration
It is seen that computation of an area and moments is actually a requirement of computing the
integration of the type:
_
b
a
f(t)dt.
To compute the above integration numerically, the well-known method is the Gauss-Legendre
integration rule which has the following form:
_
b
a
f(t)dt
n

i=0
A
i
f(x
i
)
The above integration rule is exact if f(x) is a polynomial of degree not more than 2n +1. In our
case, f(t) is either polynomial or rational function. If it is former case, f(t) is a polynomial and
thus the integration rule is exact. If it is the latter case, f(t) can be approximated by a polynomial
function since f(t) is an analytic function. Therefore, we can usually obtain accurate results very
eciently using the Gauss-Legendre integration rule.
5.6 Applying Greens theorem to a composite curve
It is noted that Greens theorem do not require C to be a single closed curve. Instead, C could be
a piecewise closed curve. If C is made of piecewise curves C
i
(C
i
may be lines, arcs, and B-spline
curves), then the Greens formula becomes
_
C
P(x, y)dx + Q(x, y)dy =

_
C
i
P(x, y)dx + Q(x, y)dy.
Y.M. Li 8
As an extreme case where C
i
are all lines dened by points (x
i
, y
i
) and (x
i+1
, y
i+1
), then the Green
formula is given by
_
C
P(x, y)dx + Q(x, y)dy =

_
x
i+1
x
i
[P(x, y) + Q(x, y)y

(x)]dx.
We now consider area, moments, and moments of inertia. For the purpose of simplifying notation,
let any two adjacent points be denoted by (x
1
, y
1
) and (x
2
, y
2
). Then, a line joining the two points
is given by
y = y
1
+
y
2
y
1
x
2
x
1
(x x
1
) = y
1
+ (x x
1
).
Referring to previous sections, the area of lamina is given by
A =
_
ydx =

_
x
2
x
1
ydx =
1
2

(y
1
+ y
2
)(x
2
x
1
)
We note that (y
1
+ y
2
)(x
2
x
1
)/2 is the area of trapezoid. The moments of lamina are given by
M
x
=
1
2
_
y
2
dx =
1
2

_
x
2
x
1
[y
1
+ (x x
1
)]
2
dx =
1
6

(x
2
x
1
)(y
2
1
+ y
1
y
2
+ y
2
2
)
M
y
=
1
2
_
x
2
dy =
1
2

_
y
2
y
1
[x
1
+
1

(y y
1
)]
2
dy =
1
6

(y
2
y
1
)(x
2
1
+ x
1
x
2
+ x
2
2
)
The moments of inertia are given by
I
xx
=
1
3
_
y
3
dx =
1
3

_
x
2
x
1
[y
1
+ (x x
1
)]
3
dx =
1
12

(x
2
x
1
)(y
3
1
+ y
2
1
y
2
+ y
1
y
2
2
+ y
3
2
)
I
yy
=
1
3
_
x
3
dy =
1
3

_
y
2
y
1
[x
1
+
1

(y y
1
)]
3
dy =
1
12

(y
2
y
1
)(x
3
1
+ x
2
1
x
2
+ x
1
x
2
2
+ x
3
2
)
I
xy
=
1
2
_
xy
2
dx =
1
2

_
x
2
x
1
x[y
1
+ (x x
1
)]
2
dx
=
1
6

_
x[y
1
+ (x x
1
)]
3

_
x
2
x
1
[y
1
+ (x x
1
)]
3
dx
_
=
1
6
_
x
2
y
3
2
x
1
y
3
1

1
4
(x
2
x
1
)(y
3
2
+ y
2
2
y
1
+ y
2
y
2
1
+ y
3
1
)
_
It is noted that if y
1
= y
2
, then
1

= . In this case, we represent the line parametrically as

y = y
1
and x = x
1
+ (x
2
x
1
)t, t [0, 1].
Thus,
I
xy
=
1
2
_
xy
2
dx =
1
4
y
2
1
(x
2
2
x
2
1
).
5.7 Ellipse: a special case
Let u and v denote the unit vectors of the major and minor axes of an ellipse, and a, b the length
of major and minor axes respectively. Then, we may write an ellipse parametrically as
u() = a cos and v() = b sin .
Y.M. Li 9
Let be the angle measured anti-clock-wisely from the x-axis to the major axis u. Then, the x
and y coordinates are given by
_
x()
y()
_
=
_
cos sin
sin cos
__
u()
v()
_
=
_
a cos cos b sin sin
a sin cos + b cos sin
_
.
In calculus, we have learned that

_
cos
2
d =
_
sin
2
d = ,

_
cos
3
d =
_
sin
3
d = 0,

_
sin
2
cos
2
d = /4,

_
cos
4
d =
_
sin
4
d = 3/4,

_
cos
3
sin d =
_
sin
3
cos d = 0.
Let a
s
= a sin , a
c
= a cos , b
s
= b sin , and b
c
= b cos . Then, x() = a
c
cos b
s
sin and
y() = a
s
cos + b
c
sin . Accordingly, we can derive
I
xx
=
1
3
_
y
3
()dx() =

4
(a
c
b
c
+ a
s
b
s
)(a
2
s
+ b
2
c
),
I
xy
=
1
2
_
y
2
()x()dx() =

4
(a
c
b
c
+ a
s
b
s
)(a
c
a
s
b
c
b
s
),
I
yy
=
1
3
_
x
3
()dy() =

4
(a
c
b
c
+ a
s
b
s
)(a
2
c
+ b
2
s
).
5.8 General organization
In applications, a lamina may be dened by a curve made of lines, arcs, and B-spline curves. For
ecient purpose, we may compute line/curve integrals of area and moments along lines and curves.
Due to the additive property of areas and moments, we can obtain the total area and moments of
lamina by simply add individuals together. To do so, it is however important to ensure that the
orientation of the curve is same.
It is easy to implement line integrals of area and moments. For a B-spline curve, the procedures
are summarized below:
1. Convert B-spline curve into Bezier curves.
2. With respect to each Bezier curve, compute signed area A
i
, moments (M
x
and M
y
) and
moments of inertia (I
xx
, I
xy
, I
yy
).
3. Compute

A
i
,

M
x
,

M
y
, etc..
References
[1] Hibbeler, R.C., Engineering Mechanics: Statics, Prentice Hall, 1991.