Priority Program 1253
|
|
- Deirdre Reynolds
- 5 years ago
- Views:
Transcription
1 Deutsce Forscungsgemeinscaft Priority Program 1253 Optimization wit Partial Differential Equations K. Deckelnick and M. Hinze Convergence and error analysis of a numerical metod for te identification of matrix parameters in elliptic PDEs April 2012 Preprint-Number SPP ttp://
2
3 Convergence and error analysis of a numerical metod for te identification of matrix parameters in elliptic PDEs Klaus Deckelnick & Micael Hinze Abstract: We analyze a numerical metod for solving te inverse problem of identifying te diffusion matrix in an elliptic PDE from distributed noisy measurements. We use a regularized least squares approac in wic te state equations are given by a finite element discretization of te elliptic PDE. Te unknown matrix parameters act as control variables and are andled wit te elp of variational discretization as introduced in [8]. For a suitable coupling of Tikonov regularization parameter, finite element grid size and noise level we are able to prove L 2 convergence of te discrete solutions to te unique norm minimal solution of te identification problem; corresponding convergence rates can be obtained provided tat a suitable projected source condition is fulfilled. Finally, we present a numerical experiment wic supports our teoretical findings. Matematics Subject Classification (2000): 49J20, 49K20, 35B37 Keywords: Parameter identification, elliptic optimal control problem, control constraints, H convergence, variational discretization, source condition. 1 Introduction In tis paper we are concerned wit a convergence analysis for a numerical metod tat identifies a diffusion matrix in te elliptic boundary value problem div (A y) = f in, y = 0 on (1.1) from distributed noisy measurements. Here, R n is a bounded, polyedral domain and f H 1 (). In addition we assume tat te diffusion matrix A(x) =(a ij (x)) n i,j=1 satisfies a ij L () and is uniformly elliptic. Te boundary value problem (1.1) ten as a unique weak solution y H0 1 () wic we denote by y = T (A, f). Our aim is to identify te unknown diffusion matrix from distributed noisy measurements (z (i) satisfying,f(i) ) L2 () H 1 (), 1 i N z (i) z (i), f (i) f (i) H 1. (1.2) Institut für Analysis und Numerik, Otto von Guericke Universität Magdeburg, Universitätsplatz 2, Magdeburg, Germany Scwerpunkt Optimierung und Approximation, Universität Hamburg, Bundesstraße 55, Hamburg, Germany. 1
4 Here, z (i) = T (A,f (i) ), 1 i N (1.3) for some symmetric diffusion matrix A wit a ξ 2 A (x)ξ ξ b ξ 2 for all ξ R n, a.e. in, (1.4) were 0 <a<b<. We sall employ a least squares approac in order to reconstruct te diffusion matrix, more precisely we consider te following optimization problem: (P ) min A M J(A) :=1 2 z (i) 2 + γ 2 A 2 subject to y (i) = T (A, f (i) ), 1 i N, were γ>0and we use te symbol for te L 2 -norm of scalar, vector- or matrix-valued functions. Te set M of admissible diffusion matrices will be defined in Section 2 below. One of te main difficulties in analyzing (P) is tat te mapping A y = T (A, f) is not weakly (sequentially) closed in L 2 (, R n,n ). Tis can be seen wit te elp of te following example taken from [16, Section 3]: let = (0, 1) R and a k L () be defined by m a, k a k (x) := x< m+ 1 2 k m =0,...,k 1. m+ b, 1 2 k x< m+1 k If we let y k = T (a k,f) it can be sown tat y k y = T (â, f) in L 2 (), were â ( 1 2 ( 1 a + 1 b )) 1. On te oter and, ak ain L 2 () wit a 1 2 (a+b), so tat y T (a, f), since â a. One possibility to overcome tis problem is to use a stronger norm in te Tikonov regularization (suc as H 1, see e.g. [15], [20]), but tis approac can be numerically cumbersome especially wen one takes ellipticity constraints into account. In [4] we were able to prove te existence of a solution to (P) by applying te concept of H convergence ([16]). Furtermore, we considered an approximation of (P) by discretizing (1.1) wit te elp of finite elements and establised te convergence of corresponding minimizers to a minimum of (P) in te case tat γ>0isfixedandz (i) = z (i), 1 i N. Te goal of te present work is to extend te convergence analysis to te case tat te regularization parameter γ tends to zero, wile we also take into account noisy measurements satisfying (1.2). Denoting by (P ) our approximation of (P) wit corresponding minimum A we sall prove in Section 3 tat A Ā in L 2 as te mes size and te noise level tend to zero provided tat γ is coupled to tese parameters in a suitable way. Here, Ā Mdenotes te norm-minimal (in te L 2 sense) diffusion matrix satisfying (1.3). Under a suitable projected source condition and appropriate smootness conditions on te data we ten sow in Section 4 tat an error bound of te form A Ā C olds. Section 5 presents a numerical test calculation wic supports te rates obtained. Let us briefly refer to related publications tat ave been concerned wit te identification of matrix valued parameters. In [1] and [9] a reconstructed matrix is obtained as te large time limit of a suitable dynamical system. A stability result, wic can also be used for te 2
5 convergence analysis of numerical metods, is derived in [11] by Hsaio and Sprekels for te reconstruction of matrices of te form A = p p. In [13], Kon and Lowe introduce a variational metod involving a functional wic is convex in te matrix A and te conductivities A y (i). Rannacer and Vexler prove in [17] error estimates for a matrix identification problem in wic a finite number of unknown parameters is estimated from finitely many pointwise observations. Te problem of identifying a scalar diffusion coefficient as been investigated muc more intensively compared to te matrix case. Identifiability results ave been obtained e.g. in [3], [18] and [19]. A survey of numerical metods for parameter estimation problems can be found in [14]. Error estimates for a least squares approac ave been obtained by Falk in [6] and more recently by Wang and Zou [20] taking into account Tikonov regularization. Te latter paper also contains a long list of furter references. 2 Notation and preliminary results Let us denote by S n te set of all symmetric n n matrices equipped wit te inner product A B =trace(ab). We introduce te subset K := {A S n a ξ 2 Aξ ξ b ξ 2 for all ξ R n }. Since K is a convex and closed subset of S n we may define te ortogonal projection P K : S n K, wic is caracterised by te relation (A P K (A)) (B P K (A)) 0 for all B K. (2.1) We define te admissible set for te optimization problem (P) by M := {A L (, R n,n ) A(x) K a.e. in } L (, S n ). According to [4, Teorem 2.2], (P) as a solution A M. For te convenience of te reader we include te necessary first order conditions satisfied by A. To begin, consider te mappings F i : M L 2 (), F i (A) :=T (A, f (i) ), 1 i N. It is not difficult to verify tat F i (A)H = w (i), H L (, S n ), were w (i) H0 1 () is te unique weak solution of te elliptic equation A w (i) vdx = H y (i) vdx for all v H0 1 () (2.2) and y (i) = T (A, f (i) ). Denoting by p(i) H0 1 (),i =1,...,N te solutions of te adjoint problems A p (i) vdx = (y (i) z (i) )vdx for all v H1 0 (), 3
6 we ave for H L (, S n ) J (A)H = = (y (i) z (i) )w(i) dx + γ H y (i) p (i) dx + γ A Hdx = A Hdx = (2.3) A p (i) w (i) dx + γ A Hdx ( N y (i) p (i) + γa ) Hdx, were (p q) ij = 1 2 (p iq j + p j q i ),i,j =1,...,n for p, q R n.wenowave Lemma 2.1. Let γ>0 and A Mbeasolutionof(P).Ten ( ) 1 A(x) =P K y (i) (x) p (i) (x) a.e. in. γ Proof. Te optimality of A yields J (A)(B A) 0 for all B M, wic can be rewritten wit te elp of (2.3) as ( 1 y (i) p (i) A ) (B A ) dx 0, for all B M. γ A localization argument ten sows tat ( ) 1 y (i) (x) p (i) (x) A(x) (C A(x) ) 0 for all C K and a.a. x γ wic implies te result. Next, let T be a regular triangulation of wit mes size := max T T diam(t ). Let us denote by X := {v C 0 ( ) v is a linear polynomial on eac T T,v =0} H 1 0() te space of continuous, piecewise linear finite elements. For a given A Mand f H 1 (), te problem A y v dx = f,v for all v X as a unique solution y X wicwedenotebyy = T (A, f). Here,, is te duality between H 1 () and H0 1 (). Standard arguments yield te following bounds T (A, f) H 1 C f H 1, (2.4) T (A, f) T (A, f) H 1 C inf T (A, f) v H 1 (2.5) v X wit a constant C tat is independent of f H 1 (), A Mand te mes size. We are now in position to formulate our numerical metod wic is based on solving te following semi-discrete minimization problem: (P ) min A M J (A) :=1 2 z(i) 2 + γ 2 A 2 subject to y (i) = T (A, f (i) ), 1 i N. 4 (2.6)
7 Note tat following te idea of [8] te matrix parameters are not discretized, compare owever Remark 2.3 below. Lemma 2.2. Problem (P ) as a solution A M. Every solution A Mof (P ) satisfies ( ) 1 A(x) =P K y (i) γ (x) p(i) (x) a.e. in, (2.7) were y (i) = T (A, f (i) ) and p (i) A p (i) v dx = X are te solutions of te adjoint problems z(i) )v dx for all v X, 1 i N. (y (i) Proof. Let (A k ) k N Mbeaminimizing sequence for problem (P )sotatj (A k) inf A M J (A) as k. Since (A k) k N is bounded in L (, R n,n ) tere exists A L (, R n,n ) suc tat A k A in L (, R n,n ) for some subsequence. In addition one readily verifies tat A M. Te sequences y(i) k = T (A k,f (i) ), 1 i N are uniformly bounded in te finite-dimensional space X so tat we may assume tat y (i) k y (i) in H 1 (), 1 i N. Clearly y (i) = T (A,f(i) ) and terefore J (A ) = 1 2 z(i) 2 + γ 1 2 A 2 lim k 2 k lim inf J k (A k )= inf J (A). A M Te relation (2.7) is obtained exactly as in te continuous case. z (i) 2 + γ lim inf 2 A k k 2 Remark 2.3. Let us note tat in view of (2.7) A is piecewise constant on T so tat a discretization of te set M is not required. Variational discretization automatically yields solutions to (2.6) wic allow a finite-dimensional representation. In order to analyze te convergence of te above metod we sall make use of te concept of Hd-convergence. Tis concept was introduced in [5] in te context of finite volume discretizations of elliptic boundary value problems wit te aim of adapting H convergence results to te discrete setting. Te following teorem is a finite element version of te result obtained in [5]. Teorem 2.4. Let (A k ) k N be a sequence in M and (T k ) k N a sequence of triangulations wit lim k k =0. Ten tere exists a subsequence (A k ) k N and A Msuc tat for every f H 1 () T k (A k,f) T(A, f) in H 1 0 () and A k T k (A k,f) A T (A, f) in L 2 () n. We ten say tat te sequence (A k ) k N Hd converges to A and write A k Hd A. Proof. See [4, Teorem 3.1]. Please note tat te Hd limit in general will depend on te sequence (T k ) k N. 5
8 Corollary 2.5. Let (A k ) k N be a sequence in M and (T k ) k N asequenceoftriangulations Hd wit lim k k =0. Suppose tat A k A 0 and A k A1 in L (, R n,n ).Ten A 0 A 1 a.e. in and A 0 2 A 1 2 lim inf A k 2. k Proof. See [4, Corollary 3.1]. 3 Convergence To begin, note tat (1.3) and (1.4) imply tat te set M := {B M z (i) = T (B,f (i) ), 1 i N} (3.1) is not empty. Since M is a closed, convex subset of L 2 (, R n,n ), tere exists a uniquely determined Ā M suc tat Ā =min B L2 L2. (3.2) B M Te following lemma gives a sufficient criterion for a matrix function Ā to be norm-minimal. Lemma 3.1. Let Ā Mand z(i) H0 1() wit z(i) = T (Ā, f (i) ), 1 i N. Suppose tat tere exist ψ (i) H0 1 (), 1 i N suc tat ( N ) Ā(x) =P K z (i) (x) ψ (i) (x) a.e. in. Ten Ā L 2 =min B M B L 2. Proof. Let B M be arbitrary. Ten B 2 Ā 2 = B Ā 2 +2 In view of our assumption on Ā and (2.1) we ave so tat Ā (B Ā)dx. ( N z (i) (x) ψ (i) (x) Ā(x)) (B(x) Ā(x)) 0, a.e. in, B 2 Ā 2 2 = 2 wic finises te proof. ( z (i) ψ (i)) (B Ā) dx B z (i) ψ (i) dx 2 Ā z (i) ψ (i) dx =0, In order to formulate our first main result we introduce ρ (i) Note tat ρ 0as 0. := inf v X z (i) v H 1, 1 i N and ρ := max 1iN ρ(i). 6
9 Teorem 3.2. Let A Mbe a solution of (P ) and suppose tat γ 0, γ 0, ρ γ 0, as 0, 0. (3.3) Ten A Ā in L2 (, R n,n ) as 0, 0, wereā Mis as in (3.2). Proof. Given sequences ( k ) k N, ( k ) k N wit lim k k =0, lim k k =0,cooseγ k > 0 suc tat lim k γ k =0and k γk 0, ρ k γk 0ask. (3.4) Furtermore, let A k = A k k Mbe a solution of (P k k ). We deduce from Teorem 2.4 tat tere exists a subsequence (A k ) k N and A 0 M, A 1 L (, R n,n ) suc tat Corollary 2.5 implies tat A k Hd A 0, A k A 1 in L (, R n,n ). A 0 A 1 a.e. in, A 0 2 lim inf k A k 2. (3.5) For better readability we write again (A k ) k N instead of (A k ) k N. We first claim tat A 0 M. SinceA k is a minimum of (P k k )weinfertatj k k (A k ) J k (Ā), k so tat 1 2 k z (i) k 2 + γ k 2 A k 2 1 z (i) 2 k z (i) k 2 + γ k 2 Ā 2. Here, y (i) k = T k (A k,f (i) k ), z (i) k te elp of (1.2) k z (i) 2 2 (i) = T k (Ā, f k ), 1 i N. As a consequence we obtain wit 2 k z (i) k 2 +2 z (i) k z (i) 2 z (i) k z (i) k 2 +2γ k Ā 2 +2N 2 k. (3.6) In order to estimate te first term on te rigt and side we use (1.3), (1.2), (2.4), (2.5) and te definition of ρ and obtain for i =1,...,N z (i) k z (i) k L 2 z (i) k z (i) H 1 + z (i) z (i) k L 2 (3.7) (i) T k (Ā, f k f (i) ) H 1 + T (Ā, f (i) k ) T (Ā, f (i) ) H 1 + k C f (i) k f (i) H 1 + Cρ k + k C( k + ρ k ). Inserting (3.7) into (3.6) we infer tat y (i) k oter and we ave tat z (i) in L 2 () as k, 1 i N. Onte y (i) k = T k (A k,f (i) )+T k (f (i) k f (i) ) T(A 0,f (i) ) in H 1 0 () 7
10 recalling (2.4), (1.2) and te fact tat A k Hd A 0. As a result we infer tat z (i) = T (A 0,f (i) ), 1 i N and terefore A 0 M. Let us sow next tat A 0 = Ā. To see tis, let B M be arbitrary. Since J k k (A k ) J k k (B) we ave 1 2 k z (i) k 2 + γ k 2 A k 2 1 z (i) 2 k z (i) k 2 + γ k 2 B 2, (3.8) were again y (i) k = T k (A k,f (i) k ), wile z (i) k = T k (B,f (i) k ), 1 i N. Similarly as in (3.7) we obtain z (i) k z (i) k C( k + ρ k ) so tat (3.8) yields A k 2 C ( k 2 + ρ2 k ) + B 2. (3.9) γ k γ k Sending k we infer from (3.5) and (3.4) tat A 0 B for every B M, sotatwe deduce tat A 0 = Ā. Finally, A k Ā 2 = A k A 0 2 = A k 2 + A 0 2 2(A k,a 0 ) L 2 from wic we infer wit te elp of (3.9) (wit B = A 0 )andtefacttata k A1 in L (, R n,n ) lim sup A k Ā 2 2 A 0 2 2(A 1,A 0 ) L 2 0, k since A 0 A 1 a.e. in. In conclusion, A k Ā, k in L2 (, R n,n ). Since te limit is unique, te wole sequence converges to Ā and te teorem is proved. 4 Error bound In order to obtain an error estimate we require stronger conditions on te data of our problem. In wat follows we sall assume tat is a bounded, polygonal subset of R 2 and tat z (i) H 1 0() W 2,p (),f (i) L p (), 1 i N for some p>2. (4.1) Furtermore, we suppose tat tere exist ψ (i) H0 1() W 2,p (), 1 i N suc tat z (i) = T (Ā, f (i) ), 1 i N, were te matrix function Ā satisfies ( N ) Ā(x) =P K z (i) (x) ψ (i) (x) a.e. in. (4.2) In addition, we assume tat tere exists μ>2suctat T (Ā, f) W 2,q C f L q, f Lq (), 1 <q<μ. (4.3) Note tat Ā Mand satisfies Ā =min B in view of Lemma 3.1. Let us write Ā(x) = B M ( ) P K E(x),wereE(x) = N z(i) (x) ψ (i) (x). Since z (i), ψ (i) W 1,p (, R n )and 8
11 te embedding W 1,p () C 0 ( ) is continuous we ave tat E W 1,p (, S n ), wic we mayextendtoafunctione W 1,p (R n, S n ). Furtermore, te projection P K is Lipscitz continuous wit Lipscitz constant 1, so tat Rn ā kl (x + ) ā kl (x) p ( ( ) ( )) PK E(x + ) PK E(x) ek e l p p dx = p dx wic implies tat R n Rn E(x + ) E(x) p p dx C, ā kl W 1,p (), 1 k, l 2. (4.4) Let us next interpret (4.2) as a projected source condition (see e.g. [7], wic also contains furter references). To do so, we assume for a moment tat f (i) = f (i) and write te objective functional in (P) in te form J(A) = 1 2 F (A) Z 2 + γ 2 A 2, were F : M L 2 () N is given by F i (A) =T (A, f (i) )andz i = z (i), 1 i N. We claim: Lemma 4.1. Tere exists Θ L 2 () N suc tat Ā = P M ( F (Ā) Θ), were P M denotes te ortogonal projection onto M in L 2 (, S n ). Proof. Recalling (2.2) we see tat were w (i) H0 1 () are te solutions of Ā w (i) vdx = F (Ā)H =( w(i) ) 1iN, H L (, S n ), H z (i) vdx for all v H 1 0 (). (4.5) In view of (4.4) te functions θ (i) := (Ā ψ (i)) belong to L 2 (). Hence Θ = (θ (i) ) 1iN L 2 () N and we ave for H L (, S n ) F (Ā) Θ,H = ( Θ,F (Ā)H) L 2 = = Ā w (i) ψ (i) dx = = ( N z (i) ψ (i),h ). L 2 Here we ave used (4.5). As a consequence we may identify (Ā ψ (i)) w (i) dx H z (i) ψ (i) dx F (Ā) Θ= z (i) ψ (i) L 2 (, S n ) 9
12 and terefore we obtain for every B M ( F (Ā) Θ Ā, B Ā) L 2 ( N = z (i) ψ (i) ( N P K z (i) ψ (i))) ( (B N P K z (i) ψ (i))) dx 0, by (2.1), wic implies te result. Next, let us suppose tat te sequence of triangulations (T ) >0 is quasiuniform. We introduce teritzprojectionr : H0 1() X associated wit Ā by Ā R z v dx = Ā z v dx for all v X. In view of (4.3) and (4.4) we may apply [2, Teorem ] and [2, Teorem 8.5.3] togeter wit inequality (8.5.5) and deduce tat tere exists 0 > 0 so tat for all 0 < 0 R z W 1, C z W 1,, z W 1, () (4.6) z R z + (z R z) C 2 z H 2, z H 2 () H0 1 (). (4.7) In particular we obtain for all f L 2 () T (Ā, f) T (Ā, f) C2 f. (4.8) Teorem 4.2. Let te conditions (4.1) (4.3) be satisfied and A Mbe a solution of (P ). If < 0 and γ = ρ 2 for some suitable ρ>0, ten were y (i) A Ā C, z(i) + (y (i) z(i) ) C 2, 1 i N, = T (A,f(i) ), 1 i N. Proof. Clearly, γ 2 A Ā 2 = γ 2 A 2 γ(a, Ā) L 2 + γ 2 Ā 2 (4.9) = J (A ) 1 2 z(i) 2 + γ(ā A, Ā) L 2 γ 2 Ā 2. Since A is a solution of (P )weinfer J (A ) J (Ā) =γ 2 Ā z (i) z(i) 2, were z (i) (i) = T(Ā, f ). Furtermore, recalling (4.2) and (2.1) we ave N (Ā A, Ā) L 2 (Ā A ) z(i) ψ (i) dx = 10 (Ā A ) z(i) ψ (i) dx.
13 Inserting te above estimates into (4.9) we obtain γ 2 A Ā z (i) z(i) z(i) 2 + γ In order to estimate te first term we employ (2.4), (4.8) and (1.2) z (i) z(i) (i) T(Ā, f If we use tis bound in (4.10) we deduce (Ā A ) z(i) ψ (i) dx. f (i) ) + T (Ā, f (i) ) T (Ā, f (i) ) + z (i) z (i) C f (i) f (i) H 1 + C 2 f (i) + C + C 2. γ 2 A Ā z(i) 2 C 2 + C 4 + γ (4.10) S (i), (4.11) were S (i) = (Ā A ) z(i) ψ (i) dx = (Ā A ) z(i) (ψ (i) R ψ (i) )dx + (Ā A ) (z(i) y (i) ) R ψ (i) dx + (Ā A ) y(i) R ψ (i) dx S (i) 1 + S (i) 2 + S (i) 3. Using te embedding W 2,p () C 1 ( ) togeter wit (4.7) we ave Next, (4.6) implies S (i) 1 A Ā z(i) L (ψ (i) R ψ (i) ) (4.12) C ψ (i) H 2 z (i) W 2,p A Ā C A Ā. S (i) 2 A Ā (z(i) y (i) ) R ψ (i) L C ψ (i) W 1, A Ā (z(i) y (i) ) (4.13) C A Ā (z(i) y (i) ). Abbreviating z (i) = T (Ā, f (i) ) we deduce from (2.5), an inverse estimate, (1.2) and (4.8) tat (z (i) y (i) ) (z(i) z (i) ) + ( z(i) C z (i) H 2 + C 1 z (i) C + C 1( z (i) C + C 1 + C 1 z (i) Inserting tis bound into (4.13) we obtain y(i) ) y(i) ) z(i) + z (i) z (i) + z(i) y (i) y (i). ( S (i) 2 C A Ā z (i) y (i) ). (4.14) 11
14 Finally, te definition of R togeter wit te fact tat y (i) = T (A,f(i) ),z(i) = T (Ā, f (i) ) imply S (i) 3 = = = = = Ā y (i) Ā y (i) R ψ (i) dx A y(i) R ψ (i) dx ψ(i) dx f (i),r ψ (i) (Ā ψ (i)) y (i) dx (Ā ψ (i)) y (i) dx Ā z (i) R ψ (i) dx + f (i) f (i),r ψ (i) Ā z (i) (R ψ (i) ψ (i) )dx Ā z (i) ψ (i) dx + f (i) f (i),r ψ (i) (Ā ψ (i)) (z (i) y (i) )dx + Ā (z (i) R z (i) ) (ψ (i) R ψ (i) )dx + f (i) f (i),r ψ (i). Hence, we may estimate wit te elp of (4.7) and (1.2) S (i) 3 (Ā ψ (i)) z (i) y (i) + f (i) f (i) H 1 R ψ (i) H 1 + Ā L (z(i) R z (i) ) (ψ (i) R ψ (i) ) C ( z (i) z (i) + z(i) y (i) ) + C ψ (i) H 1 + C 2 z (i) H 2 ψ (i) H 2 C z (i) y (i) + C + C2. (4.15) Inserting (4.12), (4.14) and (4.15) into (4.11) we deduce tat γ 2 A Ā z(i) 2 (4.16) C 2 + C 4 + Cγ A Ā ( N +Cγ z (i) y (i) + Cγ + Cγ2 γ 4 A Ā 2 +(Cγ If we coose γ = ρ 2 wit ρ = 1 8C ) N +C ( γ 2 + γ γ + γ 2). ( A 2 Ā 2 C we finally obtain z(i) 2 γ + 4 γ γ since 2. Returning to (4.16) we infer N y(i) z(i) z (i) ) C 2 y (i) ) z(i) 2 C 4,sotat z(i) + z(i) z (i) C 2 + C 2, 1 i N (4.17) 12
15 in view of te above relations betweeen γ, and. Finally, (4.7), an inverse estimate and (4.17) yield wic finises te proof. (y (i) z(i) ) (y (i) R z (i) ) + (R z (i) z (i) ) C 1 R z (i) + C z (i) H 2 C 1( z(i) + z (i) R z (i) ) + C C, 5 Numerical example Let := ( 1, 1) 2 R 2 and consider for N =1tedata(z,f) H 1 0 () H 1 () wit z(x) := (1 x 2 1)(1 x 2 2), f(x) := (Ā(x) z(x) ) for x =(x 1,x 2 ). Te diffusion matrix is given by Ā(x) :=P K ( z(x) z(x) ), were K = {A S 2 a ξ 2 Aξ ξ b ξ 2 for all ξ R 2 }.Byconstructionweavez = T (Ā, f), wile Lemma 3.1 implies tat Ā L 2 =min B M B L 2. It is not difficult to verify tat for a given matrix A S 2 we ave P K (A) =S t diag ( P [a,b] (λ 1 (A)),P [a,b] (λ 2 (A)) ) S, were λ 1 (A),λ 2 (A) denote te eigenvalues of A, S is an ortogonal matrix suc tat S t AS = diag ( λ 1 (A),λ 2 (A) ) and P [a,b] (κ) :=max(a, min(κ, b)). A calculation sows tat Ā(x) =ai 2 + P [a,b](η(x)) a η(x) were η(x) =4 ( x 2 1 (1 x2 2 )2 + x 2 2 (1 x2 1 )2). z(x) z(x), For te numerical verification of te estimates in Teorem 4.2 we coose a sequence of quasiuniform triangulations {T l } generated wit te POIMESH and REFINEMESH environment of MATLAB, were l =2 l,l N denotes te gridsize of T l. We consider te minimization problem (P )for = l, γ = γ l = l and take z = z l as te Lagrange interpolant of z, andf = f l as piecewise constant approximation to f on defined troug te function values in te barycenters of te triangles in T l. We note, tat since f is discontinuous (1.2) in general can not be satisfied wit 2 for tis coice of f. However, tis fact seems not to ave significant effects on te convergence istory of our numerical solutions presented in Tab. 1. Te constants a and b in te definition of K are cosen as a =0.5 andb = 10. Te 13
16 discrete problems (P ) are solved using te projected steepest descent metod wit Armijo step size rule (see e.g. [12]), wic we briefly describe for te convenience of te reader. In view of Remark 2.3 it is sufficient to iterate witin te class of matrices in M tat are piecewise constant on T.GivensucanA te new iterate is computed according to A + = A(τ) witτ =max r N {βr ; J l (A(β r )) J l (A) σ β r A(βr ) A 2 } were we ave abbreviated J l = J l l.furtermore,β (0, 1) and A(τ) T := P K (A T + τ ( y T p T γa T ) ), T T. Here, y = T (A, f l )andp X is te solution of te adjoint problem A v p dx = (y z l )v dx for all v X. In our calculations we cose as initial matrix A 0 := diag(1.01, 1.01)P L2 (A), were P L2 denotes te L 2 -projection onto te space of piecewise constant functions over te grid T. Te iteration on refinement level l was stopped if A + A(1) τ a + τ r A 0 A 0 (1) or te maximum number of l 100 iterations was reaced, were τ a =10 3 l and τ r = 10 2 l.furtermore,wecooseβ := 0.5 andσ := 10 4.For = l te numerical solution is denoted by Ā wit associated optimal state ȳ. Te numerical results are summarized in Tab. 1, were we display te refinement level l, te number of iterations, te value of γ l, te final L 2 -errors in te parameter, te final L 2 -errors in te states, bot togeter wit teir experimental order of convergence (EOC), and te convergence istory of te steepest descent algoritm. As predicted by Teorem 4.2, we observe quadratic convergence for te states. Te matrix parameters in te present example seem to converge faster tan predicted by te numerical analysis. Fig.1fromlefttorigtsowsȳ, te Lagrange interpolant of I z and ȳ I z for refinement level 5. As in te numerical experiments of [4] one observes tat te difference between ȳ and I z is comparatively large in regions were z (and tus I z) is small wic is in agreement wit classical results on te identifiability of scalar diffusion coefficients, see e.g. [18]. Acknowledgements Te presentation of te numerical results is partly based on a MATLAB code developed by Ronny Hoffmann in is diploma tesis [10] written under te supervision of te second autor. Te autors gratefully acknowledge te support of te DFG Priority Program 1253 entitled Optimization Wit Partial Differential Equations. 14
17 l it γ l Ā Ā EOC ȳ z EOC Ā A (1) τ a + τ r r e e e e e e e e e e e e e e e e e e e e e e e e-5 1/5: /5: 2.10 mean 1.71 mean 2.10 Table 1: Mes parameters, errors, experimental order of convergence, and convergence istory of te solution algoritm. Te table is supplemented wit te EOC between finest and coarsest level (1/5) and wit te mean value of te EOC over te refinement levels. Figure 1: Numerical solution, desired state, and error ȳ I z on refinement level 5. References [1] Alt, H.W., Hoffmann, K.H., Sprekels, J.: A numerical procedure to solve certain identification problems, Intern. Ser. Numer. Mat. 68, (1984). [2] Brenner, S. C., Scott, L. R.: Te Matematical Teory of Finite Element Metods. Tird edition. Texts in Applied Matematics, 15. Springer, New York, [3] Cicone, C., Gerlac, J.: A note on te identifiability of distributed parameters in elliptic equations, SIAM J. Mat. Anal. 18, (1987). [4] Deckelnick, K., Hinze, M.: Identification of matrix parameters in elliptic PDEs,toappear in Control & Cybernetics 40 (2011). [5] Eymard, R., Gallouët, T.: H convergence and numerical scemes for elliptic problems, Siam J. Numer. Anal. 41, (2003). [6] Falk, R.S.: Error estimates for te numerical identification of a variable coefficient, Mat. Comput. 40, (1983). 15
18 [7] Flemming, J., Hofmann, B.: Convergence rates in constrained Tikonov regularization: equivalence of projected source conditions and variational inequalities, Inverse Problems 27 (2011), doi:1088/ /27/8/ [8] Hinze, M.: A variational discretization concept in control constrained optimization: te linear-quadratic case, Comput. Optim. Appl. 30, (2005). [9] Hoffmann, K.H., Sprekels, J.: On te identification of coefficients of elliptic problems by asymptotic regularization. Numer. Funct. Anal. Optim. 7, (1984/85). [10] Hoffmann, R.: Entwicklung numeriscer Metoden zur Scätzung matrixwertiger verteilter Parameter bei elliptiscen Differentialgleicungen, Diploma tesis, TU Dresden, [11] Hsiao, G.C., Sprekels, J.: A stability result for distributed parameter identification in bilinear systems, Mat. Met. Appl. Sciences 10, (1988). [12] Kelley, C.T.: Iterative Metods for Optimization. SIAM, [13] Kon, R.V., Lowe, B.D.: A variational metod for parameter identification, RAIRO Modél. Mat. Anal. Numér. 22, (1988). [14] Kunisc, K.: Numerical metods for parameter estimation problems, Inverse problems in diffusion processes (Lake St. Wolfgang, 1994), , SIAM, Piladelpia, PA, [15] Kunisc, K., Sacs, E.W.: Reduced SQP metods for parameter identification problems, SIAM J. Numer. Anal. 29, (1992). [16] Murat, F., Tartar, L.: H-convergence, Topics in te matematical modelling of composite materials, page 21 43, Andrej Carkaev, Robert Kon Eds., [17] Rannacer, R., Vexler, B.: A priori estimates for te finite element discretization of elliptic parameter identification problems wit pointwise measurements, SIAM J. Cont. Optim. 44, (2005). [18] Ricter, G.R.: An inverse problem for te steady state diffusion equation, SIAM J. Appl. Mat. 41, (1981). [19] Vainikko, G., Kunisc, K.: Identifiability of te transmissivity coefficient in an elliptic boundary value problem, Z. Anal. Anwendungen 12, (1993). [20] Wang, L., Zou, J.: Error estimates of finite element metods for parameter identification problems in elliptic and parabolic systems, Discrete Contin. Dyn. Syst. Ser. B 14, (2010). 16
Hamburger Beiträge zur Angewandten Mathematik
Hamburger Beiträge zur Angewandten Matematik Identification of matrix parameters in elliptic PDEs Klaus Deckelnick and Micael Hinze Nr. 2-5 February 2 Identification of matrix parameters in elliptic PDEs
More informationPriority Program 1253
Deutsce Forscungsgemeinscaft Priority Program 53 Optimization wit Partial Differential Equations K. Deckelnick, A. Günter and M. Hinze Finite element approximation of elliptic control problems wit constraints
More informationHamburger Beiträge zur Angewandten Mathematik
Hamburger Beiträge zur Angewandten Mathematik Numerical analysis of a control and state constrained elliptic control problem with piecewise constant control approximations Klaus Deckelnick and Michael
More informationPriority Program 1253
Deutsche Forschungsgemeinschaft Priority Program 1253 Optimization with Partial Differential Equations Klaus Deckelnick and Michael Hinze A note on the approximation of elliptic control problems with bang-bang
More informationHamburger Beiträge zur Angewandten Mathematik
Hamburger Beiträge zur Angewandten Matematik Variational metod for multi-parameter identification in elliptic partial differential equations Tran Nan Tam Quyen Nr. 2016-26 October 2016 Variational metod
More informationError analysis of a finite element method for the Willmore flow of graphs
Interfaces and Free Boundaries 8 6, 46 Error analysis of a finite element metod for te Willmore flow of graps KLAUS DECKELNICK Institut für Analysis und Numerik, Otto-von-Guericke-Universität Magdeburg,
More informationWeierstraß-Institut. im Forschungsverbund Berlin e.v. Preprint ISSN
Weierstraß-Institut für Angewandte Analysis und Stocastik im Forscungsverbund Berlin e.v. Preprint ISSN 0946 8633 Stability of infinite dimensional control problems wit pointwise state constraints Micael
More informationPreconditioning in H(div) and Applications
1 Preconditioning in H(div) and Applications Douglas N. Arnold 1, Ricard S. Falk 2 and Ragnar Winter 3 4 Abstract. Summarizing te work of [AFW97], we sow ow to construct preconditioners using domain decomposition
More informationError estimates for a semi-implicit fully discrete finite element scheme for the mean curvature flow of graphs
Interfaces and Free Boundaries 2, 2000 34 359 Error estimates for a semi-implicit fully discrete finite element sceme for te mean curvature flow of graps KLAUS DECKELNICK Scool of Matematical Sciences,
More informationNumerical Experiments Using MATLAB: Superconvergence of Nonconforming Finite Element Approximation for Second-Order Elliptic Problems
Applied Matematics, 06, 7, 74-8 ttp://wwwscirporg/journal/am ISSN Online: 5-7393 ISSN Print: 5-7385 Numerical Experiments Using MATLAB: Superconvergence of Nonconforming Finite Element Approximation for
More informationarxiv: v1 [math.na] 29 Jan 2019
Iterated total variation regularization wit finite element metods for reconstruction te source term in elliptic systems Micael Hinze a Tran Nan Tam Quyen b a Department of Matematics, University of Hamburg,
More informationOverlapping domain decomposition methods for elliptic quasi-variational inequalities related to impulse control problem with mixed boundary conditions
Proc. Indian Acad. Sci. (Mat. Sci.) Vol. 121, No. 4, November 2011, pp. 481 493. c Indian Academy of Sciences Overlapping domain decomposition metods for elliptic quasi-variational inequalities related
More informationDifferentiation in higher dimensions
Capter 2 Differentiation in iger dimensions 2.1 Te Total Derivative Recall tat if f : R R is a 1-variable function, and a R, we say tat f is differentiable at x = a if and only if te ratio f(a+) f(a) tends
More informationPoisson Equation in Sobolev Spaces
Poisson Equation in Sobolev Spaces OcMountain Dayligt Time. 6, 011 Today we discuss te Poisson equation in Sobolev spaces. It s existence, uniqueness, and regularity. Weak Solution. u = f in, u = g on
More informationThe Laplace equation, cylindrically or spherically symmetric case
Numerisce Metoden II, 7 4, und Übungen, 7 5 Course Notes, Summer Term 7 Some material and exercises Te Laplace equation, cylindrically or sperically symmetric case Electric and gravitational potential,
More informationCONVERGENCE OF AN IMPLICIT FINITE ELEMENT METHOD FOR THE LANDAU-LIFSHITZ-GILBERT EQUATION
CONVERGENCE OF AN IMPLICIT FINITE ELEMENT METHOD FOR THE LANDAU-LIFSHITZ-GILBERT EQUATION SÖREN BARTELS AND ANDREAS PROHL Abstract. Te Landau-Lifsitz-Gilbert equation describes dynamics of ferromagnetism,
More informationERROR BOUNDS FOR THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BRADLEY J. LUCIER*
EO BOUNDS FO THE METHODS OF GLIMM, GODUNOV AND LEVEQUE BADLEY J. LUCIE* Abstract. Te expected error in L ) attimet for Glimm s sceme wen applied to a scalar conservation law is bounded by + 2 ) ) /2 T
More informationA SPLITTING LEAST-SQUARES MIXED FINITE ELEMENT METHOD FOR ELLIPTIC OPTIMAL CONTROL PROBLEMS
INTERNATIONAL JOURNAL OF NUMERICAL ANALSIS AND MODELING Volume 3, Number 4, Pages 6 626 c 26 Institute for Scientific Computing and Information A SPLITTING LEAST-SQUARES MIED FINITE ELEMENT METHOD FOR
More informationarxiv: v1 [math.na] 28 Apr 2017
THE SCOTT-VOGELIUS FINITE ELEMENTS REVISITED JOHNNY GUZMÁN AND L RIDGWAY SCOTT arxiv:170500020v1 [matna] 28 Apr 2017 Abstract We prove tat te Scott-Vogelius finite elements are inf-sup stable on sape-regular
More informationA SHORT INTRODUCTION TO BANACH LATTICES AND
CHAPTER A SHORT INTRODUCTION TO BANACH LATTICES AND POSITIVE OPERATORS In tis capter we give a brief introduction to Banac lattices and positive operators. Most results of tis capter can be found, e.g.,
More informationquasi-solution of linear inverse problems in non-reflexive banach spaces
quasi-solution of linear inverse problems in non-reflexive banac spaces Cristian Clason Andrej Klassen June 26, 2018 Abstract We consider te metod of quasi-solutions (also referred to as Ivanov regularization)
More informationGradient Descent etc.
1 Gradient Descent etc EE 13: Networked estimation and control Prof Kan) I DERIVATIVE Consider f : R R x fx) Te derivative is defined as d fx) = lim dx fx + ) fx) Te cain rule states tat if d d f gx) )
More information1. Introduction. We consider the model problem: seeking an unknown function u satisfying
A DISCONTINUOUS LEAST-SQUARES FINITE ELEMENT METHOD FOR SECOND ORDER ELLIPTIC EQUATIONS XIU YE AND SHANGYOU ZHANG Abstract In tis paper, a discontinuous least-squares (DLS) finite element metod is introduced
More informationMATH745 Fall MATH745 Fall
MATH745 Fall 5 MATH745 Fall 5 INTRODUCTION WELCOME TO MATH 745 TOPICS IN NUMERICAL ANALYSIS Instructor: Dr Bartosz Protas Department of Matematics & Statistics Email: bprotas@mcmasterca Office HH 36, Ext
More informationDownloaded 11/15/17 to Redistribution subject to SIAM license or copyright; see
SIAM J. NUMER. ANAL. Vol. 55, No. 6, pp. 2787 2810 c 2017 Society for Industrial and Applied Matematics EDGE ELEMENT METHOD FOR OPTIMAL CONTROL OF STATIONARY MAXWELL SYSTEM WITH GAUSS LAW IRWIN YOUSEPT
More informationParameter Fitted Scheme for Singularly Perturbed Delay Differential Equations
International Journal of Applied Science and Engineering 2013. 11, 4: 361-373 Parameter Fitted Sceme for Singularly Perturbed Delay Differential Equations Awoke Andargiea* and Y. N. Reddyb a b Department
More information3.4 Worksheet: Proof of the Chain Rule NAME
Mat 1170 3.4 Workseet: Proof of te Cain Rule NAME Te Cain Rule So far we are able to differentiate all types of functions. For example: polynomials, rational, root, and trigonometric functions. We are
More informationKey words. Optimal control, eikonal equation, Hamilton-Jacobi equation, approximate controllability
OPTIMAL CONTROL OF THE PROPAGATION OF A GRAPH IN INHOMOGENEOUS MEDIA KLAUS DECKELNICK, CHARLES M. ELLIOTT AND VANESSA STYLES Abstract. We study an optimal control problem for viscosity solutions of a Hamilton-Jacobi
More informationConvergence and Descent Properties for a Class of Multilevel Optimization Algorithms
Convergence and Descent Properties for a Class of Multilevel Optimization Algoritms Stepen G. Nas April 28, 2010 Abstract I present a multilevel optimization approac (termed MG/Opt) for te solution of
More informationOrder of Accuracy. ũ h u Ch p, (1)
Order of Accuracy 1 Terminology We consider a numerical approximation of an exact value u. Te approximation depends on a small parameter, wic can be for instance te grid size or time step in a numerical
More informationCopyright c 2008 Kevin Long
Lecture 4 Numerical solution of initial value problems Te metods you ve learned so far ave obtained closed-form solutions to initial value problems. A closedform solution is an explicit algebriac formula
More informationHamburger Beiträge zur Angewandten Mathematik
Hamburger Beiträge zur Angewandten Mathematik A finite element approximation to elliptic control problems in the presence of control and state constraints Klaus Deckelnick and Michael Hinze Nr. 2007-0
More informationarxiv: v1 [math.na] 20 Jul 2009
STABILITY OF LAGRANGE ELEMENTS FOR THE MIXED LAPLACIAN DOUGLAS N. ARNOLD AND MARIE E. ROGNES arxiv:0907.3438v1 [mat.na] 20 Jul 2009 Abstract. Te stability properties of simple element coices for te mixed
More information1. Introduction. Consider a semilinear parabolic equation in the form
A POSTERIORI ERROR ESTIMATION FOR PARABOLIC PROBLEMS USING ELLIPTIC RECONSTRUCTIONS. I: BACKWARD-EULER AND CRANK-NICOLSON METHODS NATALIA KOPTEVA AND TORSTEN LINSS Abstract. A semilinear second-order parabolic
More informationPolynomial Interpolation
Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximatinga function fx, wose values at a set of distinct points x, x, x,, x n are known, by a polynomial P x suc
More informationarxiv: v1 [math.na] 17 Jul 2014
Div First-Order System LL* FOSLL* for Second-Order Elliptic Partial Differential Equations Ziqiang Cai Rob Falgout Sun Zang arxiv:1407.4558v1 [mat.na] 17 Jul 2014 February 13, 2018 Abstract. Te first-order
More informationOn convexity of polynomial paths and generalized majorizations
On convexity of polynomial pats and generalized majorizations Marija Dodig Centro de Estruturas Lineares e Combinatórias, CELC, Universidade de Lisboa, Av. Prof. Gama Pinto 2, 1649-003 Lisboa, Portugal
More informationAnalysis of A Continuous Finite Element Method for H(curl, div)-elliptic Interface Problem
Analysis of A Continuous inite Element Metod for Hcurl, div)-elliptic Interface Problem Huoyuan Duan, Ping Lin, and Roger C. E. Tan Abstract In tis paper, we develop a continuous finite element metod for
More informationA Weak Galerkin Method with an Over-Relaxed Stabilization for Low Regularity Elliptic Problems
J Sci Comput (07 7:95 8 DOI 0.007/s095-06-096-4 A Weak Galerkin Metod wit an Over-Relaxed Stabilization for Low Regularity Elliptic Problems Lunji Song, Kaifang Liu San Zao Received: April 06 / Revised:
More informationDedicated to the 70th birthday of Professor Lin Qun
Journal of Computational Matematics, Vol.4, No.3, 6, 4 44. ACCELERATION METHODS OF NONLINEAR ITERATION FOR NONLINEAR PARABOLIC EQUATIONS Guang-wei Yuan Xu-deng Hang Laboratory of Computational Pysics,
More informationlecture 26: Richardson extrapolation
43 lecture 26: Ricardson extrapolation 35 Ricardson extrapolation, Romberg integration Trougout numerical analysis, one encounters procedures tat apply some simple approximation (eg, linear interpolation)
More informationA CLASS OF EVEN DEGREE SPLINES OBTAINED THROUGH A MINIMUM CONDITION
STUDIA UNIV. BABEŞ BOLYAI, MATHEMATICA, Volume XLVIII, Number 3, September 2003 A CLASS OF EVEN DEGREE SPLINES OBTAINED THROUGH A MINIMUM CONDITION GH. MICULA, E. SANTI, AND M. G. CIMORONI Dedicated to
More information1 The concept of limits (p.217 p.229, p.242 p.249, p.255 p.256) 1.1 Limits Consider the function determined by the formula 3. x since at this point
MA00 Capter 6 Calculus and Basic Linear Algebra I Limits, Continuity and Differentiability Te concept of its (p.7 p.9, p.4 p.49, p.55 p.56). Limits Consider te function determined by te formula f Note
More informationMath 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006
Mat 102 TEST CHAPTERS 3 & 4 Solutions & Comments Fall 2006 f(x+) f(x) 10 1. For f(x) = x 2 + 2x 5, find ))))))))) and simplify completely. NOTE: **f(x+) is NOT f(x)+! f(x+) f(x) (x+) 2 + 2(x+) 5 ( x 2
More informationch (for some fixed positive number c) reaching c
GSTF Journal of Matematics Statistics and Operations Researc (JMSOR) Vol. No. September 05 DOI 0.60/s4086-05-000-z Nonlinear Piecewise-defined Difference Equations wit Reciprocal and Cubic Terms Ramadan
More informationKey words. Sixth order problem, higher order partial differential equations, biharmonic problem, mixed finite elements, error estimates.
A MIXED FINITE ELEMENT METHOD FOR A SIXTH ORDER ELLIPTIC PROBLEM JÉRÔME DRONIOU, MUHAMMAD ILYAS, BISHNU P. LAMICHHANE, AND GLEN E. WHEELER Abstract. We consider a saddle point formulation for a sixt order
More informationA Hybrid Mixed Discontinuous Galerkin Finite Element Method for Convection-Diffusion Problems
A Hybrid Mixed Discontinuous Galerkin Finite Element Metod for Convection-Diffusion Problems Herbert Egger Joacim Scöberl We propose and analyse a new finite element metod for convection diffusion problems
More informationSuperconvergence of energy-conserving discontinuous Galerkin methods for. linear hyperbolic equations. Abstract
Superconvergence of energy-conserving discontinuous Galerkin metods for linear yperbolic equations Yong Liu, Ci-Wang Su and Mengping Zang 3 Abstract In tis paper, we study superconvergence properties of
More information5 Ordinary Differential Equations: Finite Difference Methods for Boundary Problems
5 Ordinary Differential Equations: Finite Difference Metods for Boundary Problems Read sections 10.1, 10.2, 10.4 Review questions 10.1 10.4, 10.8 10.9, 10.13 5.1 Introduction In te previous capters we
More informationVariational Localizations of the Dual Weighted Residual Estimator
Publised in Journal for Computational and Applied Matematics, pp. 192-208, 2015 Variational Localizations of te Dual Weigted Residual Estimator Tomas Ricter Tomas Wick Te dual weigted residual metod (DWR)
More informationarxiv: v1 [math.na] 9 Sep 2015
arxiv:509.02595v [mat.na] 9 Sep 205 An Expandable Local and Parallel Two-Grid Finite Element Sceme Yanren ou, GuangZi Du Abstract An expandable local and parallel two-grid finite element sceme based on
More informationarxiv: v1 [math.dg] 4 Feb 2015
CENTROID OF TRIANGLES ASSOCIATED WITH A CURVE arxiv:1502.01205v1 [mat.dg] 4 Feb 2015 Dong-Soo Kim and Dong Seo Kim Abstract. Arcimedes sowed tat te area between a parabola and any cord AB on te parabola
More informationTechnische Universität Graz
Tecnisce Universität Graz Boundary element metods for Diriclet boundary control problems Günter Of, Tan Pan Xuan, Olaf Steinbac Bericte aus dem Institut für Numerisce Matematik Berict 2009/2 Tecnisce
More informationMore on generalized inverses of partitioned matrices with Banachiewicz-Schur forms
More on generalized inverses of partitioned matrices wit anaciewicz-scur forms Yongge Tian a,, Yosio Takane b a Cina Economics and Management cademy, Central University of Finance and Economics, eijing,
More informationFourier Type Super Convergence Study on DDGIC and Symmetric DDG Methods
DOI 0.007/s095-07-048- Fourier Type Super Convergence Study on DDGIC and Symmetric DDG Metods Mengping Zang Jue Yan Received: 7 December 06 / Revised: 7 April 07 / Accepted: April 07 Springer Science+Business
More informationDifferent Approaches to a Posteriori Error Analysis of the Discontinuous Galerkin Method
WDS'10 Proceedings of Contributed Papers, Part I, 151 156, 2010. ISBN 978-80-7378-139-2 MATFYZPRESS Different Approaces to a Posteriori Error Analysis of te Discontinuous Galerkin Metod I. Šebestová Carles
More informationAn approximation method using approximate approximations
Applicable Analysis: An International Journal Vol. 00, No. 00, September 2005, 1 13 An approximation metod using approximate approximations FRANK MÜLLER and WERNER VARNHORN, University of Kassel, Germany,
More informationSUPERCONVERGENCE PROPERTIES FOR OPTIMAL CONTROL PROBLEMS DISCRETIZED BY PIECEWISE LINEAR AND DISCONTINUOUS FUNCTIONS
SUPERCONVERGENCE PROPERTIES FOR OPTIMAL CONTROL PROBLEMS DISCRETIZED BY PIECEWISE LINEAR AND DISCONTINUOUS FUNCTIONS A. RÖSCH AND R. SIMON Abstract. An optimal control problem for an elliptic equation
More informationA Mixed-Hybrid-Discontinuous Galerkin Finite Element Method for Convection-Diffusion Problems
A Mixed-Hybrid-Discontinuous Galerkin Finite Element Metod for Convection-Diffusion Problems Herbert Egger Joacim Scöberl We propose and analyse a new finite element metod for convection diffusion problems
More informationSOLUTIONS OF FOURTH-ORDER PARTIAL DIFFERENTIAL EQUATIONS IN A NOISE REMOVAL MODEL
Electronic Journal of Differential Equations, Vol. 7(7, No., pp.. ISSN: 7-669. URL: ttp://ejde.mat.txstate.edu or ttp://ejde.mat.unt.edu ftp ejde.mat.txstate.edu (login: ftp SOLUTIONS OF FOURTH-ORDER PARTIAL
More informationADAPTIVE MULTILEVEL INEXACT SQP METHODS FOR PDE CONSTRAINED OPTIMIZATION
ADAPTIVE MULTILEVEL INEXACT SQP METHODS FOR PDE CONSTRAINED OPTIMIZATION J CARSTEN ZIEMS AND STEFAN ULBRICH Abstract We present a class of inexact adaptive multilevel trust-region SQP-metods for te efficient
More informationMIXED DISCONTINUOUS GALERKIN APPROXIMATION OF THE MAXWELL OPERATOR. SIAM J. Numer. Anal., Vol. 42 (2004), pp
MIXED DISCONTINUOUS GALERIN APPROXIMATION OF THE MAXWELL OPERATOR PAUL HOUSTON, ILARIA PERUGIA, AND DOMINI SCHÖTZAU SIAM J. Numer. Anal., Vol. 4 (004), pp. 434 459 Abstract. We introduce and analyze a
More informationFinite Element Methods for Linear Elasticity
Finite Element Metods for Linear Elasticity Ricard S. Falk Department of Matematics - Hill Center Rutgers, Te State University of New Jersey 110 Frelinguysen Rd., Piscataway, NJ 08854-8019 falk@mat.rutgers.edu
More informationTHE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Math 225
THE IDEA OF DIFFERENTIABILITY FOR FUNCTIONS OF SEVERAL VARIABLES Mat 225 As we ave seen, te definition of derivative for a Mat 111 function g : R R and for acurveγ : R E n are te same, except for interpretation:
More informationMass Lumping for Constant Density Acoustics
Lumping 1 Mass Lumping for Constant Density Acoustics William W. Symes ABSTRACT Mass lumping provides an avenue for efficient time-stepping of time-dependent problems wit conforming finite element spatial
More informationMath 161 (33) - Final exam
Name: Id #: Mat 161 (33) - Final exam Fall Quarter 2015 Wednesday December 9, 2015-10:30am to 12:30am Instructions: Prob. Points Score possible 1 25 2 25 3 25 4 25 TOTAL 75 (BEST 3) Read eac problem carefully.
More informationNumerical analysis of a free piston problem
MATHEMATICAL COMMUNICATIONS 573 Mat. Commun., Vol. 15, No. 2, pp. 573-585 (2010) Numerical analysis of a free piston problem Boris Mua 1 and Zvonimir Tutek 1, 1 Department of Matematics, University of
More informationExam 1 Review Solutions
Exam Review Solutions Please also review te old quizzes, and be sure tat you understand te omework problems. General notes: () Always give an algebraic reason for your answer (graps are not sufficient),
More informationLecture 15. Interpolation II. 2 Piecewise polynomial interpolation Hermite splines
Lecture 5 Interpolation II Introduction In te previous lecture we focused primarily on polynomial interpolation of a set of n points. A difficulty we observed is tat wen n is large, our polynomial as to
More informationApproximation of the Viability Kernel
Approximation of te Viability Kernel Patrick Saint-Pierre CEREMADE, Université Paris-Daupine Place du Marécal de Lattre de Tassigny 75775 Paris cedex 16 26 october 1990 Abstract We study recursive inclusions
More informationarxiv: v1 [math.na] 27 Jan 2014
L 2 -ERROR ESTIMATES FOR FINITE ELEMENT APPROXIMATIONS OF BOUNDARY FLUXES MATS G. LARSON AND ANDRÉ MASSING arxiv:1401.6994v1 [mat.na] 27 Jan 2014 Abstract. We prove quasi-optimal a priori error estimates
More informationNumerical Differentiation
Numerical Differentiation Finite Difference Formulas for te first derivative (Using Taylor Expansion tecnique) (section 8.3.) Suppose tat f() = g() is a function of te variable, and tat as 0 te function
More informationarxiv: v1 [math.na] 12 Mar 2018
ON PRESSURE ESTIMATES FOR THE NAVIER-STOKES EQUATIONS J A FIORDILINO arxiv:180304366v1 [matna 12 Mar 2018 Abstract Tis paper presents a simple, general tecnique to prove finite element metod (FEM) pressure
More informationBlanca Bujanda, Juan Carlos Jorge NEW EFFICIENT TIME INTEGRATORS FOR NON-LINEAR PARABOLIC PROBLEMS
Opuscula Matematica Vol. 26 No. 3 26 Blanca Bujanda, Juan Carlos Jorge NEW EFFICIENT TIME INTEGRATORS FOR NON-LINEAR PARABOLIC PROBLEMS Abstract. In tis work a new numerical metod is constructed for time-integrating
More informationKey words. Finite element method; convection-diffusion-reaction; nonnegativity; boundedness
PRESERVING NONNEGATIVITY OF AN AFFINE FINITE ELEMENT APPROXIMATION FOR A CONVECTION-DIFFUSION-REACTION PROBLEM JAVIER RUIZ-RAMÍREZ Abstract. An affine finite element sceme approximation of a time dependent
More informationAN ANALYSIS OF THE EMBEDDED DISCONTINUOUS GALERKIN METHOD FOR SECOND ORDER ELLIPTIC PROBLEMS
AN ANALYSIS OF THE EMBEDDED DISCONTINUOUS GALERKIN METHOD FOR SECOND ORDER ELLIPTIC PROBLEMS BERNARDO COCKBURN, JOHNNY GUZMÁN, SEE-CHEW SOON, AND HENRYK K. STOLARSKI Abstract. Te embedded discontinuous
More informationQuasiperiodic phenomena in the Van der Pol - Mathieu equation
Quasiperiodic penomena in te Van der Pol - Matieu equation F. Veerman and F. Verulst Department of Matematics, Utrect University P.O. Box 80.010, 3508 TA Utrect Te Neterlands April 8, 009 Abstract Te Van
More informationLinearized Primal-Dual Methods for Linear Inverse Problems with Total Variation Regularization and Finite Element Discretization
Linearized Primal-Dual Metods for Linear Inverse Problems wit Total Variation Regularization and Finite Element Discretization WENYI TIAN XIAOMING YUAN September 2, 26 Abstract. Linear inverse problems
More informationNumerische Mathematik
Numer. Mat. DOI 10.1007/s0011-011-0386-z Numerisce Matematik Numerical analysis of an inverse problem for te eikonal equation Klaus Deckelnick Carles M. Elliott Vanessa Styles Received: 9 July 010 / Revised:
More informationarxiv: v2 [math.na] 5 Jul 2017
Trace Finite Element Metods for PDEs on Surfaces Maxim A. Olsanskii and Arnold Reusken arxiv:1612.00054v2 [mat.na] 5 Jul 2017 Abstract In tis paper we consider a class of unfitted finite element metods
More information3 Parabolic Differential Equations
3 Parabolic Differential Equations 3.1 Classical solutions We consider existence and uniqueness results for initial-boundary value problems for te linear eat equation in Q := Ω (, T ), were Ω is a bounded
More informationPOLYNOMIAL AND SPLINE ESTIMATORS OF THE DISTRIBUTION FUNCTION WITH PRESCRIBED ACCURACY
APPLICATIONES MATHEMATICAE 36, (29), pp. 2 Zbigniew Ciesielski (Sopot) Ryszard Zieliński (Warszawa) POLYNOMIAL AND SPLINE ESTIMATORS OF THE DISTRIBUTION FUNCTION WITH PRESCRIBED ACCURACY Abstract. Dvoretzky
More informationA posteriori error analysis for time-dependent Ginzburg-Landau type equations
A posteriori error analysis for time-dependent Ginzburg-Landau type equations Sören Bartels Department of Matematics, University of Maryland, College Park, MD 74, USA Abstract. Tis work presents an a posteriori
More informationFunction Composition and Chain Rules
Function Composition and s James K. Peterson Department of Biological Sciences and Department of Matematical Sciences Clemson University Marc 8, 2017 Outline 1 Function Composition and Continuity 2 Function
More informationCrouzeix-Velte Decompositions and the Stokes Problem
Crouzeix-Velte Decompositions and te Stokes Problem PD Tesis Strauber Györgyi Eötvös Loránd University of Sciences, Insitute of Matematics, Matematical Doctoral Scool Director of te Doctoral Scool: Dr.
More informationLEAST-SQUARES FINITE ELEMENT APPROXIMATIONS TO SOLUTIONS OF INTERFACE PROBLEMS
SIAM J. NUMER. ANAL. c 998 Society for Industrial Applied Matematics Vol. 35, No., pp. 393 405, February 998 00 LEAST-SQUARES FINITE ELEMENT APPROXIMATIONS TO SOLUTIONS OF INTERFACE PROBLEMS YANZHAO CAO
More informationSolving Continuous Linear Least-Squares Problems by Iterated Projection
Solving Continuous Linear Least-Squares Problems by Iterated Projection by Ral Juengling Department o Computer Science, Portland State University PO Box 75 Portland, OR 977 USA Email: juenglin@cs.pdx.edu
More informationChapter 5 FINITE DIFFERENCE METHOD (FDM)
MEE7 Computer Modeling Tecniques in Engineering Capter 5 FINITE DIFFERENCE METHOD (FDM) 5. Introduction to FDM Te finite difference tecniques are based upon approximations wic permit replacing differential
More informationPolynomial Interpolation
Capter 4 Polynomial Interpolation In tis capter, we consider te important problem of approximating a function f(x, wose values at a set of distinct points x, x, x 2,,x n are known, by a polynomial P (x
More informationERROR ESTIMATES FOR A FULLY DISCRETIZED SCHEME TO A CAHN-HILLIARD PHASE-FIELD MODEL FOR TWO-PHASE INCOMPRESSIBLE FLOWS
ERROR ESTIMATES FOR A FULLY DISCRETIZED SCHEME TO A CAHN-HILLIARD PHASE-FIELD MODEL FOR TWO-PHASE INCOMPRESSIBLE FLOWS YONGYONG CAI, AND JIE SHEN Abstract. We carry out in tis paper a rigorous error analysis
More informationApplications of the van Trees inequality to non-parametric estimation.
Brno-06, Lecture 2, 16.05.06 D/Stat/Brno-06/2.tex www.mast.queensu.ca/ blevit/ Applications of te van Trees inequality to non-parametric estimation. Regular non-parametric problems. As an example of suc
More informationThe Convergence of a Central-Difference Discretization of Rudin-Osher-Fatemi Model for Image Denoising
Te Convergence of a Central-Difference Discretization of Rudin-Oser-Fatemi Model for Image Denoising Ming-Jun Lai 1, Bradley Lucier 2, and Jingyue Wang 3 1 University of Georgia, Atens GA 30602, USA mjlai@mat.uga.edu
More informationPart VIII, Chapter 39. Fluctuation-based stabilization Model problem
Part VIII, Capter 39 Fluctuation-based stabilization Tis capter presents a unified analysis of recent stabilization tecniques for te standard Galerkin approximation of first-order PDEs using H 1 - conforming
More informationConsider a function f we ll specify which assumptions we need to make about it in a minute. Let us reformulate the integral. 1 f(x) dx.
Capter 2 Integrals as sums and derivatives as differences We now switc to te simplest metods for integrating or differentiating a function from its function samples. A careful study of Taylor expansions
More informationSymmetry Labeling of Molecular Energies
Capter 7. Symmetry Labeling of Molecular Energies Notes: Most of te material presented in tis capter is taken from Bunker and Jensen 1998, Cap. 6, and Bunker and Jensen 2005, Cap. 7. 7.1 Hamiltonian Symmetry
More informationFlavius Guiaş. X(t + h) = X(t) + F (X(s)) ds.
Numerical solvers for large systems of ordinary differential equations based on te stocastic direct simulation metod improved by te and Runge Kutta principles Flavius Guiaş Abstract We present a numerical
More informationCONVERGENCE ANALYSIS OF FINITE ELEMENT SOLUTION OF ONE-DIMENSIONAL SINGULARLY PERTURBED DIFFERENTIAL EQUATIONS ON EQUIDISTRIBUTING MESHES
INTERNATIONAL JOURNAL OF NUMERICAL ANALYSIS AND MODELING Volume, Number, Pages 57 74 c 5 Institute for Scientific Computing and Information CONVERGENCE ANALYSIS OF FINITE ELEMENT SOLUTION OF ONE-DIMENSIONAL
More informationRobust approximation error estimates and multigrid solvers for isogeometric multi-patch discretizations
www.oeaw.ac.at Robust approximation error estimates and multigrid solvers for isogeometric multi-patc discretizations S. Takacs RICAM-Report 2017-32 www.ricam.oeaw.ac.at Robust approximation error estimates
More information1. Questions (a) through (e) refer to the graph of the function f given below. (A) 0 (B) 1 (C) 2 (D) 4 (E) does not exist
Mat 1120 Calculus Test 2. October 18, 2001 Your name Te multiple coice problems count 4 points eac. In te multiple coice section, circle te correct coice (or coices). You must sow your work on te oter
More informationA Finite Element Primer
A Finite Element Primer David J. Silvester Scool of Matematics, University of Mancester d.silvester@mancester.ac.uk. Version.3 updated 4 October Contents A Model Diffusion Problem.................... x.
More information