v̂( x ) = c0 + c1 x + c2 x2 + c3 x3
dv̂( x )
= c1 + 2c2 x + 3c3 x2
θ̂ ( x ) =
dx
Now, According to Finite Element methodology, we want to express this approximate function
finally in the following form.
v̂( x ) = v1 N1 ( x ) + θ1 N2 ( x ) + v2 N3 ( x ) + θ2 N4 ( x )
and,
dv̂( x ) dN ( x ) dN2 ( x ) dN3 ( x ) dN ( x )
θ̂ ( x ) = = v1 1 + θ1 + v2 + θ2 4
dx dx dx dx dx
Let’s see how we can do that.
At x = 0, N1 (0) = 1 and N2 (0) = N3 (0) = N4 (0) = 0
v̂(0) = v1 ⇒ c0 = v1
dN2 (0) dN1 (0) dN3 (0) dN4 (0)
Similarly, when dx = 1, dx = dx = dx =0
θ̂ (0) = θ1 ⇒ c1 = θ1
v̂(h) = v2 ⇒ v1 + θ1 h + c2 h2 + c3 h3 = v2
1
( v2 − v1 ) − θ1 h − c3 h3
c2 =
h2
dN4 (h) dN1 (h) dN2 (h) dN3 (h)
When dx = 1, dx = dx = dx =0
substituting for c2
( v2 − v1 ) − θ1 h − c3 h3
θ1 + 2h + 3c3 h2 = θ2
h2
(θ2 − θ1 ) − 2h (v2 − v1 ) + 2θ1 1
c3 = 2 2
= 3 (θ1 h + θ2 h − 2(v2 − v1 ))
3h − 2h h
This leads to,
(v2 − v1 ) − θ1 h − (θ1 h + θ2 h − 2(v2 − v1 ))
c2 =
h2
1
∴ c2 = 2 (3(v2 − v1 ) − 2θ1 h − θ2 h)
h
Substituting all constants,
1 1
v̂( x ) = v1 + θ1 x + (3(v2 − v1 ) − 2θ1 h − θ2 h) x2 + 3 (θ1 h + θ2 h − 2(v2 − v1 )) x3
h2 h
Rearranging terms,
x2 x3 x2 x3 x2 x3 x2 x3
v̂( x ) = v1 1−3 2 +2 3 + θ1 x−2 + 2 + v2 3 2 −2 3 + θ2 − + 2
h h h h h h h h
Comparing with
v̂( x ) = v1 N1 ( x ) + θ1 N2 ( x ) + v2 N3 ( x ) + θ2 N4 ( x )
x2 x3
N1 ( x ) = 1 − 3 2 + 2 3
h h
x2 x3
N2 ( x ) = x − 2 + 2
h h
2
x3
x
N3 ( x ) = 3 2 − 2 3
h h
x2 x3
N4 ( x ) = − + 2
h h
and corresponding derivatives,
2
x2
dN3 ( x ) x
= 6 2 −6 3
dx h h
x2
dN4 ( x ) x
= −2 + 3 2
dx h h
Now, Let’s check final condition, i.e.,
N1 ( x ) + N3 ( x ) = 1
x2 x3 x2 x3
1−3 + 2 + 3 − 2 =1
h2 h3 h2 h3
dN1 ( x ) dN3 ( x )
This also means dx + dx = 0 but,
x2 x2
dN2 ( x ) dN4 ( x ) x x
+ = 1−4 +3 2 + −2 + 3 2 6= 1
dx dx h h h h
What happened? Our conventional notion of shape function is violated? Well, the fact is, since
v̂( x ) and θ̂ ( x ) are not independent completely, their shape functions have some complex relation.
If you notice,
x2 x3 x2 x3
2
x3 x2 x3
x
v̂( x ) = v1 1 − 3 2 + 2 3 + θ1 x − 2 + 2 + v2 3 2 − 2 3 + θ2 − + 2
h h h h h h h h
x2 x3
2
x3 x2 x3 x2 x3
x
v̂( x ) = v1 1 − 3 2 + 2 3 + v2 3 2 − 2 3 + (θ1 − θ2 ) x − 2 + 2 + − 2
h h h h h h h h
x2 x3 x2 x3
x x
v̂( x ) = v1 1 − 3 2 + 2 3 + v2 3 2 − 2 3 + (θ1 − θ2 ) h 1−
h h h h h h
This last term is zero at x = 0 and at x = h, but has contributions for other locations. That means
v̂( x ) is not a simple linear interpolation between v1 and v2 but has some contribution from nodal
slopes also. Similarly, consider