scieee Open visual document viewer

Some experimente with interval shooting methods for two point boundary value problems in ordinary differential equations

Valenga, María Raquel G. P.

Abstract

The purpose of this communication is to give some discussion, supported by numerical results of interval shooting techniques for 2 ptB.V.P. with a commentary on the relative advantages and disavantages of these methods.

Full text

Pub . Ma . UAB N° 22 No . 1980 Ac es VII JMH1 SOME EXPERIMENTE WITH INTERVALSHOOTINGMETHODS FORTWO POINT BOUNDARY VALÚE PROBLEMS IN ORDINARY DIFFERENTIAL EQUATIONS . Ma ía Raquel G .P . Valenga A ea de Ma emá ica Uni e sidade do Minho Abs ac The pu pose o his communica ion is o gi e some discussion, suppo ed by nume ical esul s o in e alshoo ing echniques o 2 p B .V .P . wi h  a commen a y on he ela i e ad an ages and disa an ages o hese me hods . 1 . The Aim o In e alAnalysis The aim o in e al analysis ís he simul aneous machine compu a ion o gua an eed bounds on he exac solu ion o a p oblem . Quan i ies a e ep esen ed by in e als con aining hei exac alue and echniques a e de eloped such ha h oughou compu a ion ound-o , un- ca ionand p opaga ion o ini ial e o a e all accoun ed o . 2 . In e al A i hme ic In he sequel we use he no a ion x I o an in e al wi h lowe and uppe bounds x_ , x such ha xI=  V x  , x i  ,  x 6 x  x_ .x  E  . The wid h o an in e al is deno ed by W(X I ) = x - x " A i hme icope a ions wi h in e als a e de ined in such a way ha he esul ing in e al will always con ain he exac esul o he ope a ion, Moo e (3) In a ious con ex s we shall ha e o compu e in e als which a e unc ions o in e als,Moo e (3) . 3 . Shoo ing Me hods o Linea P oblems We conside i s he Linea p oblem o second o de gi en by Y''=p(x)y'+q(x)Y+ (x),x E [a,b]  (3 .1) oge he wi hbounda y condi ions a ly(a) - a2y'  (a) =d- , bl Y(b)+b 2 y'  (b) =/ 1 3 (3 .2) whe e ai, bi, i=1,2,y'(a), y'(a),y(b),y'(b),o¿ .q3  may deno e in e als . We assume ha p (x) , q(x) and (x) a e con inuous on (a,b) and ha he homogeneous p oblemde i ed om (3 .1), (3 .2) has only he i ial solu ion y(x) =0 in which case (3 .1), (3 .2) has a unique solu ion . In heo y we can ind his by combining he solu ions o wo ini ial alue p oblems o sa is y he pai o bounda y condi ions . A s aigh , o wa d in e al ex ension o he classical shoo ing echnique was de eloped,Valenga (5) . The associa ed ini ial alue p oblems a e sol ed by in e al ini ial alue me hods,Moo e (3), Valenga (5), Hansen (2) . An es ima e o he o de o accu acy ob ained is ound Valenga (5) in e ms o he o de o he in e al ini ial alue me hods used . h=0 .1 max W'(Y I (x)) = Nhk Whe e N is a cons an , h he in eg a ion s ep and K he o de o he ini iál alue me hóds . Example 3 .1 . Y(2) =0 .Y(3)=0 . A ou h o de in e al ini ial alue me hod p oducedwi h w(y I )< 2 .0x10 -5 max y (x ) :z 0 . 55 The co esponding esul s wi h h=0,2 had a wid hwhich we e a mul iple e y close o 14 .6 o hose o h=0 .1 . This gi es easonable p oo o 0(h 4 ) accu acy .  ' In ce ain cí cuns ances, howe e he compu edin e al solu ionsmay la gely o e es ima e he e o in he solu ion . The me hod is uns able i he ini ial alue p oblems ha e solu ions ha inc ease apidly in absolu e alue while he solu ion o he bounda y alue p oblem emains almos s a ione y . Example in Valenga (5) . I he ini ial alue p oblems ha e oscilla o y solu ions a echnique o educe heso called "w apping e ec ",Moo e (3), mus be used . We no e howe e ha his is somewha unusual o well condi ioned bounda y alue p oblems . 4 . Shoo ing Me hods o NonLinea P oblems . We conside he nonlinea di e en ial equa ion o second o de y,'= (X,y,y,  (4 .1) oge he wi h linea bounda y condi ions (3 .2) We assume   con inuous and Lipschi zian onAaja,<x ~ b,y?+y'<~  Y  )0, I  Ó  C  M on R ; a ¡ - >, O,b i > 0,i=1,2, a l +a2 7 0, b 1 +b 2 7 0, a l +b l 70 . a y~ An i e a i e scheme which is essen ially an in e al ex ension o he 1 classical shoo ing echnique was de elopd Valenga (5) and condi ions o con e gence s a ed . Example 4 .1 max y(x )y0 .13 ,x=0 .5 . 5 . Conclusions y''=-1-0 .49y'2 , y(0)=0,y(1)=0 . Wi h s a ing in e als o wid h o o de 10-1 and h= 0 .1 using a ou h o de in e al ini ial alue me hod he algo i hm con e ges apidly p oducing a e h ee "s eps" an in e al solu ion o max w(y1)<3.5x10-5 In e alshoo ing me hods a e na u ally dependen on he pa icula i ies o in e al ini ial alue me hods . We we e able o compu e accu a eboundinj solu ions o some p oblems whe eas in o he cases e ypessímis ic bounds can be ob ained . One o he causes o he bad esul s p oduced in hese cases occu s in he classicalme hod as well . P ecau ions mus also be aken when he associa ed ini ial alue p oblems a e such ha he in luence o he "w apping e ec " is conside able . We no e ha mo e es ic ionsa e imposed wi h in e alshoo ing me hods han wi h he co esponding classical me hods bu his a e all may be he p ice o he knowledge o gua an eed e o bounding solu ions . BIBLIOGRAPHY : Fox, L, and Valenga M .R "Some expe imen s wi h In e al Me hods o 2 p B .V .P . in O .D.E ." o be published (2) Hansen, E . (ed) "Topics in In e al Analysis"Cla endon P ess . Moo e, R . E . "In e al Analysis" P en iceHall Nickel, K . "The applica ion o In e al Analysis o he Nume ical Solu ion o Di e en ial Equa ions" - In e ne Be ich des Ins . F . In o ma ik 69/9 . Uni e si ae Ka ls uhe Valenga, M . R . "In e al Me hods o O dina yDi e en ial Equa ions" D .Phil Thesis, Ox o d (1978) . A pa icipagáo da au o a nas VII Jo nadas Ma emá icas oi subsidiada pelo Ins i u o Nacional de In es igagáoCien i ica,Po ugal .