← All notes & connection map

Numerical Integration

Created on: 2024-12-03
Tags: #numericalIntegration #trapezoidalRule #simpsonsRule #NetwonCoteRule #GaussQuadrature
Course: FEM


Definition

Methods to perform integration numerically specifically in computer program.

Trapezoidal Rule

  • Uniformly spaced discrete points and corresponding function values.
  • Linear approximation within each segment

\(\int f(x)\,dx \approx \left( \frac{d}{2} \right)f_{0} + df_{1} + df_{2}+ \dots + df_{n-1} + \left( \frac{d}{2} \right)f_{n}\)

Simpson’s Rule

  • Quadratic approximation within each segment

\(\int f(x)\,dx \approx \frac{d}{3}\,\left(f_{0} + f_{n} + 4\sum_{i=1,3,5,\dots}^{n-1} f_{i} + 2\sum_{i=2,4,6,\dots}^{n-2} f_{i} \right)\)

Newton-Cotes Formula

  • Value of $f(x)$ known at $n$ arbitrary points $x_{i}$.
  • Approximate $f(x)$ using a $(n-1)^{th}$ order polynomial $g(x)$ \(f(x) \approx g(x) = c_{0} + c_{1}x + c_{2}x^2 +\dots+c_{n-1}x^{n-1}\)
  • Solve the n equations in n unknowns and solve for $c_{0}, c_{1}, c_{2},\dots,c_{n-1}$.
  • Once coefficients are determined, the integral is \(\int_{-1}^1 f(x)\,dx \approx \int_{-1}^1 g(x)\,dx\)

\(\int_{-1}^1 g(x)\,dx = 2c_{0}\, +\, 0\, +\, \frac{2}{3}c_{2}\, +\,0\, +\, \frac{2}{5}c_{4}\, +\,\dots\)

Gauss Quadrature

  • Attempts to find the location of sampling points $x_{i}$ and the weight factors $W_{i}$ to reduce computational effort and improve the accuracy.
  • The function is approximated by a polynomial of degree p. \(f(x) \approx g(x) = c_{0} + c_{1}x + c_{2}x^2 +\dots+c_{p}x^{p}\) \(\int_{-1}^1 g(x)\,dx = 2c_{0}\, +\, 0\, +\, \frac{2}{3}c_{2}\, +\,0\, +\, \frac{2}{5}c_{4}\, +\,\dots\, + \frac{c_{p}}{p+1}[1-(-1)^{p+1}]\)

  • The general formula to express the integral is \(I \approx W_{1}f_{1} + W_{2}f_{2} + W_{3}f_{3} + \dots + W_{n}f_{n}\)
  • Putting the approximation of $f(x)$ in the above equation and comparing coefficients $c_{0}, c_{1}, c_{2},\dots,c_{p}$, we get p+1 equations for 2n unknowns.

  • Solving the above p+1 equations, we get the unknown locations as well as the weights.
  • For a given number of points, Gauss quadrature can integrate a much higher order polynomial $p = 2n-1$.
  • For n=1, p = 1 (Linear polynomial).
  • For n=2, p = 3 (Cubic polynomial).

gaussQuadrature


Linked from