8.10. GAUSSIAN INTEGRATIO! So far the formulae derived for evaluation of (x) de, required the values of the function at equally spaced points of the interval. Gauss derived a formula which uses the same number of functional values but with different spacing and yields better accuracy. (Q) Gauss formula is expressed as [feed ae = wr, fx) +, fig) +. tw gfe = Ys rai | a 4 fet where w; and x; are called the weights and abscissae respectively. The abscissae and weights are symmetrical with respect to the middle point of the interval. There being 2n unknowns in (), 2n relations between them are necessary so that the formula is exact for all polynomials of degree not exceeding 2n — 1. Thus we consider Me) = 0g + Oy XH Oy Kb eer Hay gE +(2) ‘Then (1) gives, 1 {Go textepr.. fifa 2 2 = Beg 4 Foyt Foy ton . (8) Scanned with CamScanner206 NUMERICAL METHODS IN ENGINEERING AND SCIENCE Putting x =x; in (2), we get Hee) = Oy + ey, + Cyt? + gH 9 + vase +o py_ gH PM-? Substituting these values on the right hand side of (1), we obtain 1 J fO) de = wiley + 0,2, +ey 2 Heya, +. + Cyq_ 1 ,2"~1) 4 Wo(ey + Cy sty + Cy XQ? + 0g 2,9 + 2 3 + Woe + 04 25 +9 X42 + Cy 99 + + mat $544 %:28-2) an 1 Faq Hg 1) HOA (Cy +0 Hy + oy, + 05.294 as Hogg 4 2,2"-1) (Wy Wy + Wy Hr # Wy) + Cy (Wy { # yy + yy + sare + Wy) 2 2 2 2 Hey (Wye? + Wy)? + 5? + see + Wy 2) 4 Fe ee ta ls et But the equations (3) and (4) are'identical for all values of ¢;, hen ce comparing coefii- of; , we obtain 2n equations in 2n unknowns w, and x,(i = 1,2, cients Wy + Wy + wy +. y+ Wet, + wg, + (5) 2 2 2 Wy? + Wore? + wee? +. wr E+ wy.) 4 wees The solution of the above equations is extremely complicated. It.can however, be shown that x, are the zeros of the (n + 1)th Le 2 -gendre polynomial;. == (2) Gauss formula for n = 2 is : 1 Lf ds =v, fee, + wy fey) ‘Then the equations (5) become Wy + wes 2 Wwyr, + Wy %y=0 wyxy? + we x9) . wx, + wy x8 =0 Solving these equations, we obtain wy = w= 1,x,=—I1N3 and x, = 13. ‘Thus the Gauss formula for n = 2 is [foe == 1N3) + 113) 6). yhich gives the correct value of the integral of f(x) in the range (~ 1, 1) for any function upto third order. Equation (6) is also known as Gauss-Legendre formula. ‘The abscissae x, and the weights 1, in (1) are tabulated for different values of n. fe ieee table lists the abscissae and weights for values of n from 2 to 5. The follov —_— Scanned with CamScanner207 Tablo 8.1 Gauss Integration 1 Absclasae and Weights x wy = 0.67735 1.0000 : 0.67736 1.0000 a = 0.77460 0.66655 0.00000 0.88889 0.77460 0,55655 ooo 4 = 0.86114 0.34785 = 0.33998 0.65214 998 0.65214 0.86114 0.34785 5 = 0.90618 0.23693 = 0.63847 0.47863 0.00000 0.56889 0,63847 0.47863 0.90618 0.23693 (4) Gauss formula imposes a restriction on the limits of integr: In general, the limits of the integral i f formation : ¥ Example 8.13, Evaluate f, a 1 1+ x using Gauss formula forn=2 and n= 3. Sol, i) Gauss formula for n = 2 is =f she = wyfley) + wg) 1 where 4 fed" Tae 1 * te Tenor (i) Gauss formula for n = 3 is 1m w fly) + Wf) + 105 fle) where foe) = Ay 3in tho above Table 8. rs we obtain = 1.6833. 1 1 Oe ae Using the values of x, and 10, forgoing tons | a a 0555 tt 55 TConmar * 88° Tor * 555 T7746) Wy # Wy # 1,1 =~ INS, xy = INS. 3,3 aptpeis. Scanned with CamScanner -ation to be from — 1 to 1. » (4) de are changed to~ 1 to 1 by means of the trans- (7) (Anna, B.E., 2004)208 NUMERICAL METHODS IN ENGINEERING AND SCIENCE Obs. Actual value of J = 1.5708. 1 Example 8.14, Using three point Gaussian quadrature formula, evaluate [ ~ Lise (Madras, B. Sol. We first change the limits (0, 1) to - 1, 1 by (7) above, so that ., 2003) -)u+Z140)= 3 wed. 2 Lo dx 1 4du =[' 2 “Io tee Ja ded Laas es +3" = w, fly) + We flutg) + ws flus) where flu,) = Using the abscissae and weights crrempending ton =3 in the above table, we obtain a 1 I= 3007746 (0.5555) +350 9 8889) + 3407746 (0.5555) 6931. " = 0.4497 x’ 0.5555 + 3 (0.8889) + 0.2649 x'0.5555 = PROBLEMS 8.3 L 2. 3. 4. 5 Obtain an estimate of the number of sub-intervals that should be chosen so as to guarantee that the error committed in evaluating f dz/z by trapezoidal rule is less than 0.001. Fi ‘Use Romberg’s method to compute few ae ae \ Hence evaluate log, 2 correct to four decimal places. Apply Romberg’s method to evaluate f * tog x dx, given that 4.0 42° 44 46 48 5.0 5.2 x log,x + 1.3863 1.4361 1.4816 1.526 1.5686 1.6094 1.6486. i (Tirchirapalli, B.E., 2001) 5 : -[, Using Euler-Maclaurin formula, find the value of f sinx dz correct to five decimal places. Using Euler-Maclaurin formula, provo that ' oS # nnrDGn+D ' we {ese Apply Euler-Maclaurin formula, to evaluate —2,t,1, @) 700 * 402 * 404 *” 1 — (299)? © 7 1 om, 1 500 Gon? * wos? * Gos? * Scanned with CamScannerNUMERICAL DIFFERENTIATION AND INTEGRATION eee . A % Assuming that [y(x) dx = Aagy +a,y,) +h? (bye! +b,y,’), derive the quadrature formula, using the method of undetermined coefficients. Using Gaussian two-point formula compute [1 * eo abe ls 9. Evaluate f sin x dx by using Gauss-Legendre two point formula. (Bhopal, B.E., 2002) 10. Using three point Gaussian quadrature formula, evaluate : (i) (Madras, B.E., 2003 S) 7 4; a 15 @ [faseSae Git) (Me ae (Anna, B.E., 2003) oa 1 11. Using four point Gauss formula, compute Se dz correct to four decimal places. Scanned with CamScanner