You are on page 1of 8
Name: Answer k y Midterm Exam: Professor: Paul J. Atzberger Introduction to Numerical Analysis, 104B September 24, 2010 Scoring: Pot S508 Problema: 33-5" Problems: _ $5 Directions: Answer each question carefully and be sure to show all of your work. While you are permitted to use a calculator, please be sure to show intermediate steps in your calculations. If you have any questions please feel free to ask. Problem 1: Numerical Quadratures Consider approaches for estimating the value of the integral : Te i, f(a)dr. () Consider the following quadrature method for approximating the integral 1 e ynb-a, . (5) L= DZ) @) where cy =a, ci = 6. a) What is the degree of precision of this method? The degree of ptecision is the lavgest degree of a polynomial fur which the meted is exact, = Spay dx =(b-@)% = bag + Pha, = (balay % Polx) =o y z Fixjauyx, D= xs xx mai [bor]e = 2 (pea) 7 = Beh cia + Baye b = (bows bal) =( bo)" z = GP oxy = ay [herb 2 (Be) = BBog a 4 Behe yy <2 (boat beabh) 4 Jn gin\ Pil) = aX? This shows the degue of precisin js ones The quadvatime is euch for all poly nomiuls of degree ohe ov less, b) Consider the same method above, but used insteady in a composite quadra- ture scheme. In other words, we use for each k the nodes co = %,¢1 = Ck41 to estimate the integral on [zz, 241]. We sum the results to estimate the integral on the full interval [a,b]. The nodes are given by 2» = a+ kh, = (b-a)/n. What is the order of f accuracy of this composite quadrature method? Let Ty = 9" cat WAX. By the Lagvange interpola tien emavnder theo , ne have FOO =f, +a) 6 =e) (k-C)) Ix =e [ag ta,XJAX is £%5b)) (x- lo Mx~cy) dx Se Tae xpd + oth") = Be Pc) + BEAL e,) = SAL beta) + Capea] + 2G) - OF =ITntoth?) , sture enact for deg ove polye ae 24, = = Ee + Ol) = 6 erydy + he OI) = SPEM) Ax + ba oli?) = =LD+0l4) This shows +e Compost be Avadratire Ps second Orde accurnte, c) Consider the integral on’ the interval [—1,1] and non-composite quadra- tures. Suppose for the quadrature the nodes x, are chosen to be the roots of the n* degree Legendre Orthogonal Polynomial on [—1, 1]. NW He x I What is the degree of precision of the quadrature method? An-| Problem 2: Runge-Kutta Methods Consider the ODE Y — saye) @) (0) = w. O) For approximating solutions of the ODE consider the two-stage Runge-Kutta Method with the tablewx ofo 0 T|10 (5) 23 a) What is the order of accuracy of this Runge-Kutta Method? (Hint: Study the truncation error and use the chain rule derived from equation 3). Wie = wat Sn $4 S Fas) bf] c= [Pf ‘ 2 ae y helt 6) 3 Geka Flénth §)= Flt nth iy rh ltrwy hay +0(h*) a ky 5 OF : ve exact Sola hin Me Brant, + ol?) y (ens) =yltn) +y Maye Ly "(da dh? oll?) US) hy +he ObE, this au be cies a y Menai) = yen) + {bu yltal)h +f ABE o Re]t > +0th?) The treatin Rave! is qian by Conlh)= pete brea Whee whe PaO [HL hs BESTE ole lun ebb bE HE ikea 20E*>s h = o(h*) This Ghows the ov der of ACChw ay 1s tyre b) Consider f(t,y) = —Ay, A > 0 in equation 3. For what values of the time step h is the method stable? Werner = Wy + Blot) + ELH, ) Hh ue hells) Fov Flby)=~dy we have Witt = Oy + RA) + 2 (AG) = (1~ak) wy, + 20) Cy — rhe) (he ur) Wa Stability vequiees [[-Ah+ Rh le 4 I-Ah+¥h*= 4 > -AW+XhP 20 => 4yh(Ah~/) =o Dheo or hey I-\h +X W=-4 =) d-dht Me 0 Ph) =|~dh+ 2h, pllh) = P= i-padet This shows that Plh) hes the form vplex valad so not relevant -\+tA*h=o => h=d" eovided OSHE LA Fhe Runge ~ levtte font behaves stablely ay Problem 3: Multistep Methods Consider the ODE I 2 = fue) () 60) = 30 @ For approximating solutions of the ODE consider the two-step Multistep Method tons = te + 2 Sle) — Flora)] ®) wt = a9 ) wi = a. (10) a ‘What is the order of accuracy of this method? (ylbnad) = Plyltnd) - nt! tylbn ly (yldnl) + o¢h>) I [esd = ylin) + hy (bas ey ih) + ok) = yn) tht +2078 + olh®) The truncation evr Is giana by tno lb) = Br d— wert | with we =ylhe), keh Cralh) =a" Lylda rhe + Sere 4 oll) w+ SA — Rpt slid) = b' fy) sun) th(f-gted pte lee tye) 4 oli?) = 0h) - The wetHwd is second orden aaurates b) State the general criteria for a multistep method to be stable? Witt = Ami Wi Him Winey 01 + Ag Wat om thCbm ton thmitnt + bo frst] & Ge) Wo = M0) Whey 0+) Winey = otmeny Ret PEA) = dmg AL ans. Ls 0s = Ay Aap be the characteristic polynomial of (#) The mult step method is stable if fr all hots Ay of P, CPOW=0), ne have (3) the JE 4y 1) rH Prel=]o rb is 0 simple root, (c) Is the multistep method given by equation 8 stable? If so, show the criteria holds. PIN) = W-N=EO-DA => A= 1, Azo ave the rots, This callectren of vwts satisfies anditions (4), i) So the multistep woth given in eqnatun & is Stoble,

You might also like