78 COMPUTER AIDED ENGINEERING DESIGN
The vector QP×QW can be computed as
QP×QW =[r(u + Δu)−r(u)]× [r(u + Δu)−r(u−Δu)] (3.28)
Using the first order Taylor series expansion and ignoring higher order terms in Δu, Eq. (3.28) can
be rewritten as
QP QW
r r rr rr
= +^1
2
2 + 2 =
2
2
2
2
2
2 2
2
× ⎡^3
⎣
⎢
⎤
⎦
⎥×
⎡
⎣
⎢
⎤
⎦
⎥ ×
⎡
⎣
⎢
⎤
⎦
⎥
d
du
u d
du
u
d
du
u
d
du
u
d
du
d
du
ΔΔ ΔΔ Δu (3.29)
AsΔs→ 0,
Δ
Δ
TT
r
s
d
ds
= ( )→ ′′s (3.25)
To determine the direction of T′(s) or r′′(s), consider r′(s) · r′ (s) = 1 differentiating which with
respect to s yields
r′ · r′′ + r′′ · r′ = 0 ⇒ 2 r′ · r′′ = 0 ⇒r′ · r′′ = 0
Thus,r′andr′′ are orthogonal to each other implying that r′′ is perpendicular to T. We may, therefore,
defineN (a unit normal vector) such that
κN
T
=
d
ds
(3.26)
whereκ =
d
ds
T
is the scaling factor to ensure that Nis a unit vector. Also note that
Nr r = /| | =
(/)
=
/(/)
=
(/)
′′ ′′
dds
d
ds
ddududs
d
du
du
ds
ddu
d
du
T
T
T
T
T
T
The binormal B can then be defined as
B=T×N (3.27)
The plane containing T and B is termed the rectifying plane while that containing T and N is referred
to as the osculating plane. To interpret the scalar κphysically, let P,Q and W be three points on
the curve in close vicinity with values r(u),r(u + Δu) and r(u−Δu), respectively, as shown in
Figure 3.9.
Figure 3.9 Determination of κκκκκ
W
P
r(u) Q r(u + Δu)
r(u – Δu)