INTEGRATION
Trapezoidal Rules
The trapezoidal rule improves upon the zeroth-order rule by replacing the constant with a linear representation of the function, f(s). This is accomplished by including another, higher-order term, in the polynomial 5↑. Since this is a “closed” integration with specific integration limits from A to B, the independent variable, s is redefined so that it is explicitly valid over the interval of integration. Using the definition of a derivative over a known interval
(8) f1(s) = f(A) + (f(B) − f(A))/(B − A)(s − A)
when this function is replaced into the integral approximation, eqn. 4↑, the integration yields
(9) F = (B − A)(f(B) + f(A))/(2)
with an error of
(10) E = − (f′′(s)(B − A)3)/(12)
where s ∈ (A, B). This is often referred to as the midpoint rule.
A modification or “correction” to the trapezoidal rule which can improve predictions is
While the truncation error and integrated value may be more accurate, this version of the trapezoidal rule requires that the derivatives of the points also be known or can be estimated.
Simpson’s Rules
Keeping another term (n = 2) in the polynomial approximation for fn in eqn. 5↑ results in a series of schemes known as Simpson’s rules. The basic Simpson rule yields an integrated equation
with a truncation error of
where s ∈ (A, B)
Higher-order Formulas
When equal spacing is available, these formulas can be combined to offer significant computational savings. The trapezoidal rule for example becomes
(14) F = (B − A)(f(s0) + n − 1⎲⎳i = 1f(si) + f(sn))/(2n)
for the entire length B − A with a truncation error of
and the Simpson basic (1/3) rule becomes
and a truncation error of
← Previous Page
← Previous Page