History-dependent contact models for viscoplastic materials

Abstract

We consider two mathematical models which describe the frictionless process of contact between a rate-type viscoplastic body and a foundation. The contact is modelled with normal compliance and memory term such that penetration is not restricted in the first problem, but is restricted with unilateral constraint in the second one.

For each problem, we derive a variational formulation in terms of displacements, which is in a form of a history-dependent variational equation and a history-dependent variational inequality. Then we prove the unique weak solvability of each model. Next, we prove the convergence of the weak solution of the first problem and the weak solution of the second problem, as the stiffness coefficient of the foundation converges to infinity.

Finally, we provide numerical simulations which illustrate this convergence result.

Authors

Mikael Barboteu
(Laboratoire de Mathématiques et Physique, Université de Perpignan)

Flavius Patrulescu
(Tiberiu Popoviciu Institute of Numerical Analysis, Romanian Academy)

Ahmad Ramadan
(Laboratoire de Mathématiques et Physique, Université de Perpignan)

Mircea Sofonea
(Laboratoire de Mathématiques et Physique, Université de Perpignan)

Keywords

viscoplastic material; frictionless contact; normal compliance; unilateral constraint; memory term; history-dependent variational inequality; weak solution; numerical simulations

Cite this paper as:

M. Barboteu, F. Pătrulescu, A. Ramadan, M. Sofonea, History-dependent contact models for viscoplastic materials, IMA J. Appl. Math., 79 (2014) no. 6, pp. 1180-1200.

PDF

About this paper

Publisher Name

Oxford University Press, Oxford

Print ISSN

0272-4960

Online ISSN

1464-3634

MR

3286321

ZBL

1307.74055

[1] P. Alart and A. Curnier, A mixed formulation for frictional contact problems prone to Newton like solution methods, Computer Methods in Applied Mechanics 26 and Engineering 92 (1991), 353–375.
[2] M. Barboteu, A. Matei, M. Sofonea, Analysis of quasistatic viscoplastic contact problems with normal compliance, Quarterly Journal of Mechanics and Applied Mathematics, 65 (2012), 555–579.
[3] N. Cristescu and I. Suliciu, Viscoplasticity, Martinus Nijhoff Publishers, Editura Tehnica, Bucharest, 1982.
[4] W. Han and M. Sofonea, Quasistatic Contact Problems in Viscoelasticity and Viscoplasticity, Studies in Advanced Mathematics 30, American Mathematical Society–International Press, 2002.
[5] I.R. Ionescu and M. Sofonea, Functional and Numerical Methods in Viscoplasticity, Oxford University Press, Oxford, 1993.
[6] H. B. Khenous, P. Laborde and Y. Renard, On the discretization of contact problems in elastodynamics, Lecture Notes in Applied Computational Mechanics 27 (2006), 31–38.
[7] H. B. Khenous, J. Pommier, Y. Renard, Hybrid discretization of the Signorini problem with Coulomb friction. Theoretical aspects and comparison of some numerical solvers, Applied Numerical Mathematics 56 (2006), 163–192.
[8] A. Klarbring, A. Mikeliˇc and M. Shillor, Frictional contact problems with normal compliance, Int. J. Engng. Sci. 26 (1988), 811–832.
[9] A. Klarbring, A. Mikeliˇc and M. Shillor, On friction problems with normal compliance, Nonlinear Analysis 13 (1989), 935–955.
[10] T. Laursen, Computational Contact and Impact Mechanics, Springer, Berlin, 2002.
[11] J. A. C. Martins and J. T. Oden, Existence and uniqueness results for dynamic contact problems with nonlinear normal and friction interface laws, Nonlinear Analysis : Theory, Methods and Applications 11 (1987), 407–428.
[12] J.T. Oden and J.A.C. Martins, Models and computational methods for dynamic friction phenomena, Computer Methods in Applied Mechanics and Engineering 52 (1985), 527–634.
[13] M. Shillor, M. Sofonea and J.J. Telega, Models and Analysis of Quasistatic Contact, Lecture Notes in Physics, 655, Springer, Berlin, 2004.
[14] M. Sofonea, C. Avramescu and A. Matei, A Fixed point result with applications in the study of viscoplastic frictionless contact problems, Communications on Pure and Applied Analysis 7 (2008), 645–658.
[15] M. Sofonea and A. Matei, History-dependent quasivariational inequalities arising in Contact Mechanics, European Journal of Applied Mathematics, 22 (2011), 471–491.
[16] M. Sofonea and F. P˘atrulescu, Analysis of a history-dependent frictionless contact problem, Mathematics and Mechanics of Solids, 18, no. 4 (2013), 409–430.
[17] P. Wriggers, Computational Contact Mechanics, Wiley, Chichester, 2002.

Paper (preprint) in HTML form

History-dependent Contact Models for Viscoplastic Materials

M. Barboteu 1, F. Pătrulescu 2, A. Ramadan 1 and M. Sofonea 1
1 Laboratoire de Mathématiques et Physique
Université de Perpignan, 52 Avenue de Paul Alduy, 66860 Perpignan, France
2 Tiberiu Popoviciu Institute of Numerical Analysis
P.O. Box 68-1, 400110 Cluj-Napoca, Romania
Abstract

We consider two quasistatic contact problems which describe the contact between a viscoplastic body and an obstacle, the so-called foundation. The contact is frictionless and is modelled with normal compliance and memory term of such a type that the penetration is not restricted in the first problem, but is restricted with unilateral constraint, in the second one. For each problem we derive a variational formulation, then we prove its unique solvability. Next, we prove the convergence of the weak solution of the first problem to the weak solution of the second problem, as the stiffness coefficient of the foundation converges to infinity. And, finally, we provide numerical simulations which illustrate the convergence result.

1 Introduction

Phenomena of contact between deformable bodies or between deformable and rigid bodies abound in industry and everyday life. Contact of braking pads with wheels, tires with roads, pistons with skirts are just a few simple examples. Common industrial processes such as metal forming, metal extrusion, involve contact evolutions and, for this reason, a considerable effort has been developed in their modelling, mathematical analysis and numerical solution. Owing to their inherent complexity, contact phenomena lead to nonlinear and nonsmooth mathematical problems.

The aim of this paper is to study two frictionless contact problems for rate-type viscoplastic materials, within the framework of the Mathematical Theory of Contact Mechanics. We model the behavior of the material with a constitutive law of the form

𝝈˙=ε(𝒖˙)+𝒢(𝝈,ε(𝒖)),\dot{\boldsymbol{\sigma}}=\mathcal{E}\varepsilon(\dot{\boldsymbol{u}})+\mathcal{G}(\boldsymbol{\sigma},\varepsilon(\boldsymbol{u})), (1.1)

where 𝒖\boldsymbol{u} denotes the displacement field, 𝝈\boldsymbol{\sigma} represents the stress and 𝜺(𝒖)\boldsymbol{\varepsilon}(\boldsymbol{u}) is the linearized strain tensor. Here \mathcal{E} is a fourth order tensor which describes the elastic properties of the material and 𝒢\mathcal{G} is a constitutive function which describes its viscoplastic behavior. In (1.1) and everywhere in this paper the dot above a variable represents derivative with respect to the time variable tt.

Various results, examples and mechanical interpretations in the study of viscoplastic materials of the form (1.1) can be found in [3, 5] and references therein. Displacement-traction boundary value problems with such materials were considered in [5], both in the dynamic and quasistatic case. Quasistatic frictionless and frictional contact problems for materials of the form (1.1) were studied in various papers, see 4 and 13 for a survey. There, various models of contact were stated and they variational analysis, including existence and uniqueness results, was provided. The numerical analysis of the corresponding models can be found in [4] and the references therein. In all the above papers the process of contact was studied in a finite interval of time.

In [14] a quasistatic frictionless contact problem for viscoplastic materials of the form (1.1) was considered. The process was assumed to be quasistatic and the contact was modelled by using the normal compliance condition with infinite penetration. The unique solvability of the solution was obtained by using a fixed point argument. In contrast, in [2] was considered a problem with normal compliance and finite penetration and was proved the unique solvability of the models using new arguments on history-dependent variational inequalities presented in [15]. The present paper represents a continuation of [2] and we consider the problem with normal compliance, finite penetration and memory term. The same contact condition for viscoelastic materials was used in [16]. Also, we state and prove the convergence of the solution of the problem with infinite penetration to the solution of the problem with finite penetration as the stiffness coefficient converges to infinity. And, finally, we provide numerical simulations which illustrate this convergence.

The rest of the paper is structured as follows. In Section 2 we introduce the notations we shall use as well as some preliminary material. In Section 3 we present the classical formulation of the two contact problems. In Section 4 we list the assumptions on the data and derive the variational formulation of the problems. Then we state and prove the unique weak solvability of each model. In Section 5 we state and prove a converge result, Theorem 5.1. Next, in Section 6 we present the numerical solution of the contact problem with normal compliance restricted by unilateral constraints and memory term. And, finally, in Section 7 some numerical simulations are presented including a numerical validation of the convergence result.

2 Notations and preliminaries

Everywhere in this paper we use the notation \mathbb{N} for the set of positive integers and +\mathbb{R}_{+}will represent the set of non negative real numbers, i.e. +=[0,+)\mathbb{R}_{+}=[0,+\infty). We denote by 𝕊d\mathbb{S}^{d} the space of second order symmetric tensors on d\mathbb{R}^{d} or, equivalently, the space of symmetric matrices of order dd. The inner product and norm on d\mathbb{R}^{d} and 𝕊d\mathbb{S}^{d} are defined by

𝒖𝒗=uivi,𝒗=(𝒗𝒗)12𝒖,𝒗d𝝈𝝉=σijτij,𝝉=(𝝉𝝉)12𝝈,𝝉𝕊d\begin{array}[]{lrl}\boldsymbol{u}\cdot\boldsymbol{v}=u_{i}v_{i},&\|\boldsymbol{v}\|=(\boldsymbol{v}\cdot\boldsymbol{v})^{\frac{1}{2}}&\forall\boldsymbol{u},\boldsymbol{v}\in\mathbb{R}^{d}\\ \boldsymbol{\sigma}\cdot\boldsymbol{\tau}=\sigma_{ij}\tau_{ij},&\|\boldsymbol{\tau}\|=(\boldsymbol{\tau}\cdot\boldsymbol{\tau})^{\frac{1}{2}}&\forall\boldsymbol{\sigma},\boldsymbol{\tau}\in\mathbb{S}^{d}\end{array}

Let Ω\Omega be a bounded domain Ωd(d=1,2,3)\Omega\subset\mathbb{R}^{d}(d=1,2,3) with a Lipschitz continuous boundary Γ\Gamma and let Γ1\Gamma_{1} be a measurable part of Γ\Gamma such that meas (Γ1)>0\left(\Gamma_{1}\right)>0. We use the notation 𝒙=(xi)\boldsymbol{x}=\left(x_{i}\right) for a typical point in ΩΓ\Omega\cup\Gamma and we denote by 𝝂=(νi)\boldsymbol{\nu}=\left(\nu_{i}\right) the outward unit normal at Γ\Gamma. Here and below the indices i,j,k,li,j,k,l run between 1 and dd and, unless stated otherwise, the summation convention over repeated indices is used. An index that follows a comma represents the partial derivative with respect to the corresponding component of the spatial variable, e.g. ui,j=ui/xju_{i,j}=\partial u_{i}/\partial x_{j}. We consider the spaces

V={𝒗H1(Ω)d:𝒗=𝟎 on Γ1},Q={𝝉=(τij)L2(Ω)d:τij=τji}V=\left\{\boldsymbol{v}\in H^{1}(\Omega)^{d}:\boldsymbol{v}=\mathbf{0}\text{ on }\Gamma_{1}\right\},\quad Q=\left\{\boldsymbol{\tau}=\left(\tau_{ij}\right)\in L^{2}(\Omega)^{d}:\tau_{ij}=\tau_{ji}\right\}

These are real Hilbert spaces endowed with the inner products

(𝒖,𝒗)V=Ω𝜺(𝒖)𝜺(𝒗)𝑑x,(𝝈,𝝉)Q=Ω𝝈𝝉𝑑x(\boldsymbol{u},\boldsymbol{v})_{V}=\int_{\Omega}\boldsymbol{\varepsilon}(\boldsymbol{u})\cdot\boldsymbol{\varepsilon}(\boldsymbol{v})dx,\quad(\boldsymbol{\sigma},\boldsymbol{\tau})_{Q}=\int_{\Omega}\boldsymbol{\sigma}\cdot\boldsymbol{\tau}dx

and the associated norms V\|\cdot\|_{V} and Q\|\cdot\|_{Q}, respectively. Here 𝜺\boldsymbol{\varepsilon} represents the deformation operator given by

𝜺(𝒗)=(εij(𝒗)),εij(𝒗)=12(vi,j+vj,i)𝒗H1(Ω)d.\boldsymbol{\varepsilon}(\boldsymbol{v})=\left(\varepsilon_{ij}(\boldsymbol{v})\right),\quad\varepsilon_{ij}(\boldsymbol{v})=\frac{1}{2}\left(v_{i,j}+v_{j,i}\right)\quad\forall\boldsymbol{v}\in H^{1}(\Omega)^{d}.

Completeness of the space ( V,VV,\|\cdot\|_{V} ) follows from the assumption meas (Γ1)>0\left(\Gamma_{1}\right)>0, which allows the use of Korn’s inequality.

For an element 𝒗V\boldsymbol{v}\in V we still write 𝒗\boldsymbol{v} for the trace of VV and we denote by vνv_{\nu} and 𝒗τ\boldsymbol{v}_{\tau} the normal and tangential components of 𝒗\boldsymbol{v} on Γ\Gamma given by vν=𝒗𝝂,𝒗τ=𝒗vν𝝂v_{\nu}=\boldsymbol{v}\cdot\boldsymbol{\nu},\boldsymbol{v}_{\tau}=\boldsymbol{v}-v_{\nu}\boldsymbol{\nu}. Let Γ3\Gamma_{3} be a measurable part of Γ\Gamma. Then, by the Sobolev trace theorem, there exists a positive constant c0c_{0} which depends on Ω,Γ1\Omega,\Gamma_{1} and Γ3\Gamma_{3} such that

𝒗L2(Γ3)dc0𝒗V𝒗V.\|\boldsymbol{v}\|_{L^{2}\left(\Gamma_{3}\right)^{d}}\leq c_{0}\|\boldsymbol{v}\|_{V}\quad\forall\boldsymbol{v}\in V. (2.1)

Also, for a regular stress function 𝝈\boldsymbol{\sigma} we use the notation σν\sigma_{\nu} and 𝝈τ\boldsymbol{\sigma}_{\tau} for the normal and the tangential traces, i.e. σν=(𝝈𝝂)𝝂\sigma_{\nu}=(\boldsymbol{\sigma}\boldsymbol{\nu})\cdot\boldsymbol{\nu} and 𝝈τ=𝝈𝝂σν𝝂\boldsymbol{\sigma}_{\tau}=\boldsymbol{\sigma}\boldsymbol{\nu}-\sigma_{\nu}\boldsymbol{\nu}. Moreover, we recall that the divergence operator is defined by the equality Div𝝈=(σij,j)\operatorname{Div}\boldsymbol{\sigma}=\left(\sigma_{ij,j}\right) and, finally, the following Green’s formula holds:

Ω𝝈𝜺(𝒗)𝑑x+ΩDiv𝝈𝒗dx=Γ𝝈𝝂𝒗𝑑a𝒗V\int_{\Omega}\boldsymbol{\sigma}\cdot\boldsymbol{\varepsilon}(\boldsymbol{v})dx+\int_{\Omega}\operatorname{Div}\boldsymbol{\sigma}\cdot\boldsymbol{v}dx=\int_{\Gamma}\boldsymbol{\sigma}\boldsymbol{\nu}\cdot\boldsymbol{v}da\quad\forall\boldsymbol{v}\in V (2.2)

For a normed space ( X,XX,\|\cdot\|_{X} ) we use the notation C(+;X)C\left(\mathbb{R}_{+};X\right) for the space of continuously functions defined on +\mathbb{R}_{+}with values on XX, and C1(+;X)C^{1}\left(\mathbb{R}_{+};X\right) for the space of continuous differentiable functions defined on +\mathbb{R}_{+}with values on XX. For a subset KXK\subset X we still use the symbols C(+;K)C\left(\mathbb{R}_{+};K\right) and C1(+;K)C^{1}\left(\mathbb{R}_{+};K\right) for the set of continuous and continuously differentiable functions defined on +\mathbb{R}_{+}with values on KK, respectively.

Let XX be a real Hilbert space with inner product (,)X(\cdot,\cdot)_{X} and associated norm X\|\cdot\|_{X}. Assume given a set KXK\subset X, the operators A:KX,𝒮:C(+;X)C(+;X)A:K\rightarrow X,\mathcal{S}:C\left(\mathbb{R}_{+};X\right)\rightarrow C\left(\mathbb{R}_{+};X\right) and a function f:+Xf:\mathbb{R}_{+}\rightarrow X such that:

K is a closed, convex, nonempty subset of XK\text{ is a closed, convex, nonempty subset of }X\text{. } (2.3)
{ (a) There exists m>0 such that (Au1Au2,u1u2)Xmu1u2X2u1,u2K. (b) There exists L>0 such that Au1Au2XLu1u2Xu1,u2K.{ For every n there exists rn>0 such that 𝒮u1(t)𝒮u2(t)Xrn0tu1(s)u2(s)X𝑑su1,u2C(+;X),t[0,n]fC(+;X)\displaystyle\left\{\begin{array}[]{l}\text{ (a) There exists }m>0\text{ such that }\\ \left(Au_{1}-Au_{2},u_{1}-u_{2}\right)_{X}\geq m\left\|u_{1}-u_{2}\right\|_{X}^{2}\quad\forall u_{1},u_{2}\in K.\\ \text{ (b) There exists }L>0\text{ such that }\\ \left\|Au_{1}-Au_{2}\right\|_{X}\leq L\left\|u_{1}-u_{2}\right\|_{X}\quad\forall u_{1},u_{2}\in K.\\ \left\{\begin{array}[]{l}\text{ For every }n\in\mathbb{N}\text{ there exists }r_{n}>0\text{ such that }\\ \left\|\mathcal{S}u_{1}(t)-\mathcal{S}u_{2}(t)\right\|_{X}\leq r_{n}\int_{0}^{t}\left\|u_{1}(s)-u_{2}(s)\right\|_{X}ds\\ \forall u_{1},u_{2}\in C\left(\mathbb{R}_{+};X\right),\forall t\in[0,n]\\ f\in C\left(\mathbb{R}_{+};X\right)\end{array}\right.\end{array}\right. (2.4)

We proceed with the following existence and uniqueness result in the study of nonlinear equations involving monotone operators.

Theorem 2.1 Let XX be a Hilbert space and let A:XXA:X\rightarrow X be a strongly monotone Lipschitz continuous operator. Then, for each fXf\in X there exists a unique element uXu\in X such that Au=fAu=f.

The following result, proved in [15, will be used in Section 4 of this paper.
Theorem 2.2 Assume that (2.3)-(2.6) hold. Then there exists a unique function uC(+;K)u\in C\left(\mathbb{R}_{+};K\right) such that for all t+t\in\mathbb{R}_{+}, the inequality below holds:

(Au(t),vu(t))X+(𝒮u(t),v)X(𝒮u(t),u(t))X\displaystyle(Au(t),v-u(t))_{X}+(\mathcal{S}u(t),v)_{X}-(\mathcal{S}u(t),u(t))_{X} (2.7)
(f(t),vu(t))X\displaystyle\geq(f(t),v-u(t))_{X} vK\displaystyle\forall v\in K

We have the following consequence of Theorem 2.2.
Corollary 2.3 Let XX be a Hilbert space and assume that (2.3)-(2.6) hold. Then there exists a unique function uC(+;X)u\in C\left(\mathbb{R}_{+};X\right) such that

(Au(t),v)X+(𝒮u(t),v)X=(f(t),v)XvX,t+.(Au(t),v)_{X}+(\mathcal{S}u(t),v)_{X}=(f(t),v)_{X}\quad\forall v\in X,\quad\forall t\in\mathbb{R}_{+}. (2.8)

Following the terminology introduced in [15] we refer to (2.7) as a history-dependent quasivariational inequality. To avoid any confusion, we note that here and below the notation Au(t)Au(t) and 𝒮u(t)\mathcal{S}u(t) are short hand notation for A(u(t))A(u(t)) and (𝒮u)(t)(\mathcal{S}u)(t), i.e. Au(t)=A(u(t))Au(t)=A(u(t)) and 𝒮u(t)=(𝒮u)(t)\mathcal{S}u(t)=(\mathcal{S}u)(t), for all t+t\in\mathbb{R}_{+}.

3 The models

In this section we present the two problems which describe the frictionless contact process and present the assumption on the data. The physical setting is as follows. A viscoplastic body occupies a bounded domain Ωd(d=1,2,3)\Omega\subset\mathbb{R}^{d}(d=1,2,3) with a Lipschitz continuous boundary Γ\Gamma, divided into three measurable parts Γ1,Γ2\Gamma_{1},\Gamma_{2} and Γ3\Gamma_{3}, such that meas (Γ1)>0\left(\Gamma_{1}\right)>0. The body is subject to the action of body forces of density 𝒇0\boldsymbol{f}_{0}. We also assume that it is fixed on Γ1\Gamma_{1} and surface tractions of density 𝒇2\boldsymbol{f}_{2} act on Γ2\Gamma_{2}. On Γ3\Gamma_{3}, the body is in frictionless contact with a deformable obstacle, the so-called foundation. We assume that the problem is quasistatic, and we study the contact process in the interval of time +=[0,)\mathbb{R}_{+}=[0,\infty).

In the first problem, unlike in [2], the contact is modelled with normal compliance and memory term in such a way that the penetration is not limited. Under these conditions, the classical formulation of the problem is the following.

Problem 𝒫1\mathcal{P}_{1}. Find a displacement field 𝒖:Ω×+d\boldsymbol{u}:\Omega\times\mathbb{R}_{+}\rightarrow\mathbb{R}^{d} and a stress field 𝝈\boldsymbol{\sigma} : Ω×+𝕊d\Omega\times\mathbb{R}_{+}\rightarrow\mathbb{S}^{d} such that

𝝈˙=ε(𝒖˙)+𝒢(𝝈,𝜺(𝒖)) in Ω×(0,),Div𝝈+𝒇0=𝟎 in Ω×(0,),𝒖=𝟎 on Γ1×(0,),𝝈𝝂=𝒇2 on Γ2×(0,),σν=p(uν)+0tb(ts)uν+(s)𝑑s on Γ3×(0,),𝝈τ=𝟎 on Γ3×(0,),𝒖(0)=𝒖0,𝝈(0)=𝝈0 in Ω.\begin{array}[]{rll}\dot{\boldsymbol{\sigma}}=\mathcal{E}\varepsilon(\dot{\boldsymbol{u}})+\mathcal{G}(\boldsymbol{\sigma},\boldsymbol{\varepsilon}(\boldsymbol{u}))&\text{ in }&\Omega\times(0,\infty),\\ \operatorname{Div}\boldsymbol{\sigma}+\boldsymbol{f}_{0}=\mathbf{0}&\text{ in }&\Omega\times(0,\infty),\\ \boldsymbol{u}=\mathbf{0}&\text{ on }&\Gamma_{1}\times(0,\infty),\\ \boldsymbol{\sigma}\boldsymbol{\nu}=\boldsymbol{f}_{2}&\text{ on }&\Gamma_{2}\times(0,\infty),\\ -\sigma_{\nu}=p\left(u_{\nu}\right)+\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds&\text{ on }&\Gamma_{3}\times(0,\infty),\\ \boldsymbol{\sigma}_{\tau}=\mathbf{0}&\text{ on }&\Gamma_{3}\times(0,\infty),\\ \boldsymbol{u}(0)=\boldsymbol{u}_{0},\boldsymbol{\sigma}(0)=\boldsymbol{\sigma}_{0}&\text{ in }&\Omega.\end{array}

Here and below, in order to simplify the notation, we do not indicate explicitly the dependence of various functions on the variables 𝒙\boldsymbol{x} or tt. Equation (3.1) represents the viscoplastic constitutive law of the material introduced in Section 1 and equation (3.2) is the equilibrium equation. Conditions (3.3) and (3.4) are the displacement and traction boundary conditions, respectively, and condition (3.6) shows that the tangential stress on the contact surface, denoted 𝝈τ\boldsymbol{\sigma}_{\tau}, vanishes. We use it here since we assume that the contact process is frictionless. Finally, (3.7) represents the initial
conditions in which 𝒖0\boldsymbol{u}_{0} and 𝝈0\boldsymbol{\sigma}_{0} denote the initial displacement and the initial stress field, respectively.

The function pp is Lipschitz continuous, increasing and vanishes for a negative argument, i.e.

{ (a) p:+.(b) There exists Lp>0 such that |p(r1)p(r2)|Lp|r1r2|r1,r2. (c) (p(r1)p(r2))(r1r2)0r1,r2. (d) p(r)=0 for all r<0.\left\{\begin{array}[]{l}\text{ (a) }p:\mathbb{R}\rightarrow\mathbb{R}_{+}.\\ \text{(b) There exists }L_{p}>0\text{ such that }\\ \quad\left|p\left(r_{1}\right)-p\left(r_{2}\right)\right|\leq L_{p}\left|r_{1}-r_{2}\right|\quad\forall r_{1},r_{2}\in\mathbb{R}.\\ \text{ (c) }\left(p\left(r_{1}\right)-p\left(r_{2}\right)\right)\left(r_{1}-r_{2}\right)\geq 0\quad\forall r_{1},r_{2}\in\mathbb{R}.\\ \text{ (d) }p(r)=0\text{ for all }r<0.\end{array}\right.

In the second problem the contact is again modelled with normal compliance and memory term but in such a way that the penetration is limited and associated to a unilateral constraint. The classical formulation of the problem is the following.

Problem 𝒫2\mathcal{P}_{2}. Find a displacement field 𝒖:Ω×+d\boldsymbol{u}:\Omega\times\mathbb{R}_{+}\rightarrow\mathbb{R}^{d} and a stress field 𝝈\boldsymbol{\sigma} : Ω×+𝕊d\Omega\times\mathbb{R}_{+}\rightarrow\mathbb{S}^{d} such that

𝝈˙=𝜺(𝒖˙)+𝒢(𝝈,𝜺(𝒖))\displaystyle\dot{\boldsymbol{\sigma}}=\mathcal{E}\boldsymbol{\varepsilon}(\dot{\boldsymbol{u}})+\mathcal{G}(\boldsymbol{\sigma},\boldsymbol{\varepsilon}(\boldsymbol{u})) in Ω×(0,),\displaystyle\text{ in }\quad\Omega\times(0,\infty), (3.9)
Div𝝈+𝒇0=𝟎\displaystyle\operatorname{Div}\boldsymbol{\sigma}+\boldsymbol{f}_{0}=\mathbf{0} in Ω×(0,),\displaystyle\text{ in }\quad\Omega\times(0,\infty), (3.10)
𝒖=𝟎\displaystyle\boldsymbol{u}=\mathbf{0} on Γ1×(0,),\displaystyle\text{ on }\quad\Gamma_{1}\times(0,\infty), (3.11)
𝝈𝝂=𝒇2\displaystyle\boldsymbol{\sigma}\boldsymbol{\nu}=\boldsymbol{f}_{2} on Γ2×(0,),\displaystyle\text{ on }\quad\Gamma_{2}\times(0,\infty), (3.12)
uνg,σν+p(uν)+0tb(ts)uν+(s)𝑑s0\displaystyle u_{\nu}\leq g,\sigma_{\nu}+p\left(u_{\nu}\right)+\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds\leq 0 (3.13)
(uνg)(σν+p(uν)+0tb(ts)uν+(s)𝑑s)=0\displaystyle\left(u_{\nu}-g\right)\left(\sigma_{\nu}+p\left(u_{\nu}\right)+\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds\right)=0 on Γ3×(0,),\displaystyle\text{ on }\quad\Gamma_{3}\times(0,\infty), (3.14)
𝝈𝝉=𝟎\displaystyle\boldsymbol{\boldsymbol{\sigma}_{\tau}}=\mathbf{0} on Γ3×(0,),\displaystyle\text{ on }\quad\Gamma_{3}\times(0,\infty), (3.15)
𝒖(0)=𝒖0,𝝈(0)=𝝈0\displaystyle\boldsymbol{u}(0)=\boldsymbol{u}_{0},\boldsymbol{\sigma}(0)=\boldsymbol{\sigma}_{0} in Ω.\displaystyle\text{ in }\quad\Omega.

Here g0g\geq 0 is given and pp is a function which satisfies (3.8). Conditions (3.9)(3.12) and (3.14) - (3.15) have the same interpretation as in the contact problem 𝒫1\mathcal{P}_{1}.

We now present the new contact condition (3.13), condition (3.5) can be presented using similar arguments. It can be derived in the following way. First, we assume that the penetration is limited by the bound gg and, therefore, at each time moment t+t\in\mathbb{R}_{+}, the normal displacement satisfies the inequality

uν(t)g on Γ3.u_{\nu}(t)\leq g\text{ on }\Gamma_{3}. (3.16)

Next, we assume that the normal stress has an additive decomposition of the form

σν(t)=σνD(t)+σνR(t)+σνM(t) on Γ3\sigma_{\nu}(t)=\sigma_{\nu}^{D}(t)+\sigma_{\nu}^{R}(t)+\sigma_{\nu}^{M}(t)\quad\text{ on }\Gamma_{3} (3.17)

in which the functions σνD,σνR\sigma_{\nu}^{D},\sigma_{\nu}^{R} and σνM\sigma_{\nu}^{M} describe the deformability, the rigidity and the memory properties of the foundation, at each t+t\in\mathbb{R}_{+}. Also, we assume that the function σνD\sigma_{\nu}^{D} satisfies the normal compliance contact condition

σνD(t)=p(uν(t)) on Γ3.-\sigma_{\nu}^{D}(t)=p\left(u_{\nu}(t)\right)\quad\text{ on }\Gamma_{3}. (3.18)

Condition (3.18) combined with assumption (3.8) shows that when there is separation between the body and the obstacle (i.e. when uν<0u_{\nu}<0 ), then the reaction of the foundation vanishes (since σν=0\sigma_{\nu}=0 ); also, when there is penetration (i.e. when uνu_{\nu}\geq 0 ), then the reaction of the foundation is towards the body (since σν0\sigma_{\nu}\leq 0 ) and it is increasing with the penetration (since pp is an increasing function). Finally, we note that in this condition the penetration is not restricted and the normal stress is uniquely determined by the normal displacement.

Condition (3.18) was first introduced in [11, 12] in the study of dynamic contact problems with elastic and viscoelastic materials. The term normal compliance for this condition was first used in (8, 9. A first example of normal compliance function pp which satisfies condition (3.8) is

p(r)=cνr+p(r)=c_{\nu}r^{+} (3.19)

where r+=max{r,0}r^{+}=\max\{r,0\} and cνc_{\nu} is a positive constant. In this case condition (3.18) shows that the reaction of the foundation is proportional to the penetration and, therefore, (3.8), (3.18) model the contact with a linearly elastic foundation. A second example of normal compliance function pp which satisfies condition (3.8) is given by

p(r)={cνr+if rαcνα if r>αp(r)=\left\{\begin{array}[]{clr}c_{\nu}r^{+}&\text{if }&r\leq\alpha\\ c_{\nu}\alpha&\text{ if }&r>\alpha\end{array}\right.

where α\alpha is a positive coefficient related to the wear and hardness of the surface and, again, cν>0c_{\nu}>0. In this case the contact condition (3.18) means that when the penetration is too large, i.e. when it exceeds α\alpha, the obstacle backs off and offers no additional resistance to the penetration. We conclude that in this case the foundation has an elastic-perfectly plastic behavior.

The part σνR\sigma_{\nu}^{R} of the normal stress satisfies the Signorini condition in the form with a gap function, i.e.

σνR(t)0,σνR(t)(uν(t)g)=0 on Γ3\sigma_{\nu}^{R}(t)\leq 0,\quad\sigma_{\nu}^{R}(t)\left(u_{\nu}(t)-g\right)=0\quad\text{ on }\Gamma_{3} (3.20)

And, finally, the function σνM\sigma_{\nu}^{M} satisfies the memory condition

σνM(t)=0tb(ts)uν+(s)𝑑s on Γ3-\sigma_{\nu}^{M}(t)=\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds\quad\text{ on }\Gamma_{3} (3.21)

in which bb represents a given function, the so-called surface memory function. Contact conditions of the form (3.21) have a simple physical interpretation if there are no cycles
of contact and separation during the time interval of interest. For instance, assume in what follows that bb is a positive function. Moreover, assume that in the time interval [0,t][0,t] there is only penetration (i.e. uν(s)0u_{\nu}(s)\geq 0 for all s[0,t]s\in[0,t] ). Then (3.21) shows that the reaction of the foundation at tt is towards the body (since σν(t)0\sigma_{\nu}(t)\leq 0 ). Also, if in the time interval [0,t][0,t] there is separation (i.e. uν(s)<0u_{\nu}(s)<0 for all s[0,t]s\in[0,t] ) then there is no reaction at the moment tt (since σν(t)=0\sigma_{\nu}(t)=0 ).

Now, assume a situation in which uνu_{\nu} is positive in time interval [0,t0]\left[0,t_{0}\right] and negative on the time interval [t0,t]\left[t_{0},t\right]. Then, following (3.21) we have

σν(t)=0t0b(ts)uν+(s)𝑑s-\sigma_{\nu}(t)=\int_{0}^{t_{0}}b(t-s)u_{\nu}^{+}(s)ds

since the integral on the remaining interval [t0,t]\left[t_{0},t\right] vanishes. Assume, in addition, that the support of the function bb is included in the interval [0,δ][0,\delta] with δ>0\delta>0. Two possibilities arise. First, if tt0>δt-t_{0}>\delta it follows that b(ts)=0b(t-s)=0 for all s[0,t0]s\in\left[0,t_{0}\right] and (3.21) shows the normal stress σν(t)\sigma_{\nu}(t) vanishes. Second, if tt0δt-t_{0}\leq\delta (3.21) implies that σν(t)0\sigma_{\nu}(t)\leq 0 i.e. a residual pression exists at the moment tt on the body’s surface. We interpret this as a memory effect in which the foundation prevents the separation, moves towards the body and exerts a pression on a short interval of time of length δ\delta. Various other mechanical interpretation of the condition (3.21) could be obtained if bb is assumed to be a negative function.

We combine equalities (3.17), (3.18) and (3.21) to see that

σνR(t)=σν(t)+p(uν(t))+0tb(ts)uν+(s)𝑑s on Γ3\sigma_{\nu}^{R}(t)=\sigma_{\nu}(t)+p\left(u_{\nu}(t)\right)+\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds\quad\text{ on }\Gamma_{3} (3.22)

Then we substitute equality (3.22) in (3.20) and use inequality (3.16) to obtain the contact condition (3.13).

4 Existence and uniqueness results

In this section we list the assumptions on the data, derive the variational formulations of the problems 𝒫1\mathcal{P}_{1} and 𝒫2\mathcal{P}_{2} and then we state and prove their unique weak solvability. To this end we assume that the elasticity tensor \mathcal{E} and the constitutive function 𝒢\mathcal{G} satisfy the following conditions.

{ (a) =(ijkl):Ω×𝕊d𝕊d. (b) ijkl=klij=jiklL(Ω),1i,j,k,ld. (c) There exists m>0 such that 𝝉𝝉m𝝉2𝝉𝕊d, a.e. in Ω.\left\{\begin{array}[]{l}\text{ (a) }\mathcal{E}=\left(\mathcal{E}_{ijkl}\right):\Omega\times\mathbb{S}^{d}\rightarrow\mathbb{S}^{d}.\\ \text{ (b) }\mathcal{E}_{ijkl}=\mathcal{E}_{klij}=\mathcal{E}_{jikl}\in L^{\infty}(\Omega),1\leq i,j,k,l\leq d.\\ \text{ (c) There exists }m_{\mathcal{E}}>0\text{ such that }\\ \mathcal{E}\boldsymbol{\tau}\cdot\boldsymbol{\tau}\geq m_{\mathcal{E}}\|\boldsymbol{\tau}\|^{2}\forall\boldsymbol{\tau}\in\mathbb{S}^{d},\text{ a.e. in }\Omega.\end{array}\right.
{ (a) 𝒢:Ω×𝕊d×𝕊d𝕊d. (b) There exists L𝒢>0 such that 𝒢(𝒙,𝝈1,𝜺1)𝒢(𝒙,𝝈2,𝜺2)L𝒢(𝝈1𝝈2+𝜺1𝜺2)𝝈1,𝝈2,𝜺1,𝜺2𝕊d, a.e. 𝒙Ω. (c) The mapping 𝒙𝒢(𝒙,𝝈,𝜺) is measurable on Ω, for any 𝝈,𝜺𝕊d. (d) The mapping 𝒙𝒢(𝒙,𝟎,𝟎) belongs to Q.\left\{\begin{array}[]{l}\text{ (a) }\mathcal{G}:\Omega\times\mathbb{S}^{d}\times\mathbb{S}^{d}\rightarrow\mathbb{S}^{d}.\\ \text{ (b) There exists }L_{\mathcal{G}}>0\text{ such that }\\ \quad\left\|\mathcal{G}\left(\boldsymbol{x},\boldsymbol{\sigma}_{1},\boldsymbol{\varepsilon}_{1}\right)-\mathcal{G}\left(\boldsymbol{x},\boldsymbol{\sigma}_{2},\boldsymbol{\varepsilon}_{2}\right)\right\|\\ \quad\leq L_{\mathcal{G}}\left(\left\|\boldsymbol{\sigma}_{1}-\boldsymbol{\sigma}_{2}\right\|+\left\|\boldsymbol{\varepsilon}_{1}-\boldsymbol{\varepsilon}_{2}\right\|\right)\\ \quad\forall\boldsymbol{\sigma}_{1},\boldsymbol{\sigma}_{2},\boldsymbol{\varepsilon}_{1},\boldsymbol{\varepsilon}_{2}\in\mathbb{S}^{d},\text{ a.e. }\boldsymbol{x}\in\Omega.\\ \text{ (c) The mapping }\boldsymbol{x}\mapsto\mathcal{G}(\boldsymbol{x},\boldsymbol{\sigma},\boldsymbol{\varepsilon})\text{ is measurable on }\Omega,\\ \quad\text{ for any }\boldsymbol{\sigma},\boldsymbol{\varepsilon}\in\mathbb{S}^{d}.\\ \text{ (d) The mapping }\boldsymbol{x}\mapsto\mathcal{G}(\boldsymbol{x},\mathbf{0},\mathbf{0})\text{ belongs to }Q.\end{array}\right.

The surface memory function satisfies

bC(+;L(Γ3)).b\in C\left(\mathbb{R}_{+};L^{\infty}\left(\Gamma_{3}\right)\right). (4.3)

We also assume that the body forces and the surface tractions have the regularity

𝒇0C(+;L2(Ω)d),𝒇2C(+;L2(Γ2)d)\boldsymbol{f}_{0}\in C\left(\mathbb{R}_{+};L^{2}(\Omega)^{d}\right),\quad\boldsymbol{f}_{2}\in C\left(\mathbb{R}_{+};L^{2}\left(\Gamma_{2}\right)^{d}\right) (4.4)

In the study of Problem 𝒫1\mathcal{P}_{1} we assume that the initial data satisfy

𝒖0V,𝝈0Q\boldsymbol{u}_{0}\in V,\quad\boldsymbol{\sigma}_{0}\in Q (4.5)

and, finally, in the study of Problem 𝒫2\mathcal{P}_{2} we assume that

𝒖0U,𝝈0Q,\boldsymbol{u}_{0}\in U,\quad\boldsymbol{\sigma}_{0}\in Q, (4.6)

where UU denotes the set of admissible displacements defined by

U={𝒗V:vνg on Γ3}.U=\left\{\boldsymbol{v}\in V:v_{\nu}\leq g\text{ on }\Gamma_{3}\right\}. (4.7)

In the rest of the section we denote by cc a positive generic constant that may depend on time and whose value may change from line to line. Also, we use the symbol " \longrightarrow " to denote the weak convergence in the Hilbert space VV.

We turn now to the variational formulation of the problems 𝒫1\mathcal{P}_{1} and 𝒫2\mathcal{P}_{2}. To this end, we use Riesz’s representation Theorem to define the operator P:VVP:V\rightarrow V and the function 𝒇:+V\boldsymbol{f}:\mathbb{R}_{+}\rightarrow V by equalities

(P𝒖,𝒗)V\displaystyle(P\boldsymbol{u},\boldsymbol{v})_{V} =Γ3p(uν)vν𝑑a𝒖,𝒗V\displaystyle=\int_{\Gamma_{3}}p\left(u_{\nu}\right)v_{\nu}da\quad\forall\boldsymbol{u},\boldsymbol{v}\in V (4.8)
(𝒇(t),𝒗)V\displaystyle(\boldsymbol{f}(t),\boldsymbol{v})_{V} =Ω𝒇0(t)𝒗𝑑x+Γ2𝒇2(t)𝒗𝑑a𝒗V,t+\displaystyle=\int_{\Omega}\boldsymbol{f}_{0}(t)\cdot\boldsymbol{v}dx+\int_{\Gamma_{2}}\boldsymbol{f}_{2}(t)\cdot\boldsymbol{v}da\quad\forall\boldsymbol{v}\in V,t\in\mathbb{R}_{+} (4.9)

Assume in what follows that ( 𝒖,𝝈\boldsymbol{u},\boldsymbol{\sigma} ) are sufficiently regular functions which satisfy (3.1)-(3.7) and let t>0t>0 be given. We integrate equation (3.1) with the initial conditions (3.7) to obtain

𝝈(t)=𝜺(𝒖(t))+0t𝒢(𝝈(s),𝜺(𝒖(s)))𝑑s+𝝈0𝜺(𝒖0)\boldsymbol{\sigma}(t)=\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t))+\int_{0}^{t}\mathcal{G}(\boldsymbol{\sigma}(s),\boldsymbol{\varepsilon}(\boldsymbol{u}(s)))ds+\boldsymbol{\sigma}_{0}-\mathcal{E}\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{0}\right) (4.10)

Then we use the Green formula (2.2), the equilibrium equation (3.2), the boundary conditions (3.3)-(3.6) and notation (4.8)-(4.9) to see that

(𝝈(t),𝜺(𝒗))Q+(0tb(ts)uν+(s)𝑑s,vν)L2(Γ3)\displaystyle(\boldsymbol{\sigma}(t),\boldsymbol{\varepsilon}(\boldsymbol{v}))_{Q}+\left(\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds,v_{\nu}\right)_{L^{2}\left(\Gamma_{3}\right)}
+(P𝒖(t),𝒗)V=(𝒇(t),𝒗)V𝒗V\displaystyle+(P\boldsymbol{u}(t),\boldsymbol{v})_{V}=(\boldsymbol{f}(t),\boldsymbol{v})_{V}\quad\forall\boldsymbol{v}\in V (4.11)

We present the following existence and uniqueness result proved in 2 .
Lemma 4.1 Assume that (4.2) and (4.5) hold. Then, for each function 𝒖C(+;V)\boldsymbol{u}\in C\left(\mathbb{R}_{+};V\right) there exists a unique function 𝒮1𝒖C(+;Q)\mathcal{S}_{1}\boldsymbol{u}\in C\left(\mathbb{R}_{+};Q\right) such that

𝒮1𝒖(t)=0t𝒢(𝒮1𝒖(s)+ε(𝒖(s)),𝜺(𝒖(s)))𝑑s+𝝈0ε(𝒖0)t+.\mathcal{S}_{1}\boldsymbol{u}(t)=\int_{0}^{t}\mathcal{G}\left(\mathcal{S}_{1}\boldsymbol{u}(s)+\mathcal{E}\varepsilon(\boldsymbol{u}(s)),\boldsymbol{\varepsilon}(\boldsymbol{u}(s))\right)ds+\boldsymbol{\sigma}_{0}-\mathcal{E}\varepsilon\left(\boldsymbol{u}_{0}\right)\quad\forall t\in\mathbb{R}_{+}. (4.12)

Moreover, the operator 𝒮1:C(+;V)C(+;Q)\mathcal{S}_{1}:C\left(\mathbb{R}_{+};V\right)\rightarrow C\left(\mathbb{R}_{+};Q\right) satisfies the following property: for every nn\in\mathbb{N} there exists kn>0k_{n}>0 such that

𝒮1𝒖1(t)𝒮1𝒖2(t)Qkn0t𝒖1(s)𝒖2(s)V𝑑s\displaystyle\left\|\mathcal{S}_{1}\boldsymbol{u}_{1}(t)-\mathcal{S}_{1}\boldsymbol{u}_{2}(t)\right\|_{Q}\leq k_{n}\int_{0}^{t}\left\|\boldsymbol{u}_{1}(s)-\boldsymbol{u}_{2}(s)\right\|_{V}ds (4.13)
𝒖1,𝒖2C(+;V),t[0,n]\displaystyle\forall\boldsymbol{u}_{1},\boldsymbol{u}_{2}\in C\left(\mathbb{R}_{+};V\right),\forall t\in[0,n]

Using the operator 𝒮1:C(+;V)C(+;Q)\mathcal{S}_{1}:C\left(\mathbb{R}_{+};V\right)\rightarrow C\left(\mathbb{R}_{+};Q\right) defined in Lemma 4.1 we deduce that (4.10) and (4.11) are equivalent with

𝝈(t)=𝜺(𝒖(t))+𝒮1𝒖(t)(𝜺(𝒖(t)),𝜺(𝒗))Q+(𝒮1𝒖(t),𝜺(𝒗))Q+(0tb(ts)uν+(s)𝑑s,vν)L2(Γ3)+(P𝒖(t),𝒗)V=(𝒇(t),𝒗)V𝒗V.\begin{gathered}\boldsymbol{\sigma}(t)=\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t))+\mathcal{S}_{1}\boldsymbol{u}(t)\\ (\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t)),\boldsymbol{\varepsilon}(\boldsymbol{v}))_{Q}+\left(\mathcal{S}_{1}\boldsymbol{u}(t),\boldsymbol{\varepsilon}(\boldsymbol{v})\right)_{Q}+\left(\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds,v_{\nu}\right)_{L^{2}\left(\Gamma_{3}\right)}\\ +(P\boldsymbol{u}(t),\boldsymbol{v})_{V}=(\boldsymbol{f}(t),\boldsymbol{v})_{V}\forall\boldsymbol{v}\in V.\end{gathered}

We use again Riesz’s representation Theorem to define the operator 𝒮:C(+;V)C(+;V)\mathcal{S}:C\left(\mathbb{R}_{+};V\right)\rightarrow C\left(\mathbb{R}_{+};V\right) by equality

(𝒮𝒖(t),𝒗)V=(𝒮1𝒖(t),𝜺(𝒗))Q\displaystyle(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v})_{V}=\left(\mathcal{S}_{1}\boldsymbol{u}(t),\boldsymbol{\varepsilon}(\boldsymbol{v})\right)_{Q}
+(0tb(ts)uν+(s)𝑑s,vν)L2(Γ3)𝒖C(+;V),𝒗V\displaystyle+\left(\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds,v_{\nu}\right)_{L^{2}\left(\Gamma_{3}\right)}\forall\boldsymbol{u}\in C\left(\mathbb{R}_{+};V\right),\boldsymbol{v}\in V (4.14)

and we obtain the following variational formulation of the Problem 𝒫1\mathcal{P}_{1}.
Problem 𝒫1V\mathcal{P}_{1}^{V}. Find a displacement field 𝒖:+V\boldsymbol{u}:\mathbb{R}_{+}\rightarrow V and a stress field 𝝈:+Q\boldsymbol{\sigma}:\mathbb{R}_{+}\rightarrow Q, such that

𝝈(t)=𝜺(𝒖(t))+𝒮1𝒖(t)\displaystyle\boldsymbol{\sigma}(t)=\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t))+\mathcal{S}_{1}\boldsymbol{u}(t) (4.15)
(𝜺(𝒖(t)),𝜺(𝒗))Q+(𝒮𝒖(t),𝒗)V+(P𝒖(t),𝒗)V=(𝒇(t),𝒗)V𝒗V\displaystyle(\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t)),\boldsymbol{\varepsilon}(\boldsymbol{v}))_{Q}+(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v})_{V}+(P\boldsymbol{u}(t),\boldsymbol{v})_{V}=(\boldsymbol{f}(t),\boldsymbol{v})_{V}\forall\boldsymbol{v}\in V (4.16)

hold, for all t+t\in\mathbb{R}_{+}.
In the study of the problem 𝒫1V\mathcal{P}_{1}^{V} we have the following existence and uniqueness result.

Theorem 4.2 Assume that (3.8) and (4.1)-(4.5) hold. Then, Problem 𝒫1V\mathcal{P}_{1}^{V} has a unique solution, which satisfies

𝒖C(+;V),𝝈C(+;Q)\boldsymbol{u}\in C\left(\mathbb{R}_{+};V\right),\quad\boldsymbol{\sigma}\in C\left(\mathbb{R}_{+};Q\right) (4.17)

Proof. We define the operator A:VVA:V\rightarrow V by equality

(A𝒗,𝒘)V=(ε(𝒗),ε(𝒘))Q+(P𝒗,𝒘)V𝒗,𝒘V.(A\boldsymbol{v},\boldsymbol{w})_{V}=(\mathcal{E}\varepsilon(\boldsymbol{v}),\varepsilon(\boldsymbol{w}))_{Q}+(P\boldsymbol{v},\boldsymbol{w})_{V}\quad\forall\boldsymbol{v},\boldsymbol{w}\in V. (4.18)

With this notation we consider the problem of finding a function 𝒖:+V\boldsymbol{u}:\mathbb{R}_{+}\rightarrow V such that, for all t+t\in\mathbb{R}_{+}, the following equality holds

(A𝒖(t),𝒗)V+(𝒮𝒖(t),𝒗)V=(𝒇(t),𝒗)V𝒗V.(A\boldsymbol{u}(t),\boldsymbol{v})_{V}+(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v})_{V}=(\boldsymbol{f}(t),\boldsymbol{v})_{V}\quad\forall\boldsymbol{v}\in V. (4.19)

To solve (4.19) we employ Corollary 2.3 with X=VX=V. We use (4.1), (3.8) and (2.1) to see that the operator AA verifies condition (2.4), i.e. it is strongly monotone and Lipschitz continuous. In addition, using (4.3) we note that the operator 𝒮\mathcal{S} satisfies condition (2.5) with

rn=knenkn+c2maxr[0,n]b(r)L(Γ3),r_{n}=k_{n}e^{nk_{n}}+c^{2}\max_{r\in[0,n]}\|b(r)\|_{L^{\infty}\left(\Gamma_{3}\right)}, (4.20)

for every nn\in\mathbb{N}.
Finally, using (4.4) and (4.9) we deduce that 𝒇\boldsymbol{f} has the regularity expressed in (2.6). It follows now from Corollary 2.3 that there exists a unique function 𝒖C(+;V)\boldsymbol{u}\in C\left(\mathbb{R}_{+};V\right) which solves the equality (4.19), for any t+t\in\mathbb{R}_{+}.

Based on the results above we deduce the existence of a unique function 𝒖C(+;V)\boldsymbol{u}\in C\left(\mathbb{R}_{+};V\right) which satisfies (4.16) for any t+t\in\mathbb{R}_{+}. Let 𝝈\boldsymbol{\sigma} be defined by (4.15). Then it follows that the couple ( 𝒖,𝝈\boldsymbol{u},\boldsymbol{\sigma} ) is the unique couple of functions with regularity (4.17) which satisfies (4.15)-(4.16).

Assume now that ( 𝒖,𝝈\boldsymbol{u},\boldsymbol{\sigma} ) are sufficiently regular functions which satisfy (3.9)-(3.15) and, again, let t>0t>0 be given. Then, using similar arguments as above we obtain the following variational formulation of Problem 𝒫2\mathcal{P}_{2}.
Problem 𝒫2V\mathcal{P}_{2}^{V}. Find a displacement field 𝒖:+V\boldsymbol{u}:\mathbb{R}_{+}\rightarrow V and a stress field 𝝈:+Q\boldsymbol{\sigma}:\mathbb{R}_{+}\rightarrow Q, such that

𝝈(t)=ε(𝒖(t))+𝒮1𝒖(t)\displaystyle\boldsymbol{\sigma}(t)=\mathcal{E}\varepsilon(\boldsymbol{u}(t))+\mathcal{S}_{1}\boldsymbol{u}(t) (4.21)
(ε(𝒖(t)),𝜺(𝒗)𝜺(𝒖(t)))Q+(𝒮𝒖(t),𝒗𝒖(t))V\displaystyle(\mathcal{E}\varepsilon(\boldsymbol{u}(t)),\boldsymbol{\varepsilon}(\boldsymbol{v})-\boldsymbol{\varepsilon}(\boldsymbol{u}(t)))_{Q}+(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v}-\boldsymbol{u}(t))_{V}
+(P𝒖(t),𝒗𝒖(t))V(𝒇(t),𝒗𝒖(t))V𝒗U\displaystyle+(P\boldsymbol{u}(t),\boldsymbol{v}-\boldsymbol{u}(t))_{V}\geq(\boldsymbol{f}(t),\boldsymbol{v}-\boldsymbol{u}(t))_{V}\forall\boldsymbol{v}\in U (4.22)

hold, for all t+t\in\mathbb{R}_{+}.
In the study of the problem 𝒫2V\mathcal{P}_{2}^{V} we have the following existence and uniqueness result.

Theorem 4.3 Assume that (3.8), (4.1) -(4.4) and (4.6) hold. Then, Problem 𝒫2V\mathcal{P}_{2}^{V} has a unique solution, which satisfies

𝒖C(+;U),𝝈C(+;Q).\boldsymbol{u}\in C\left(\mathbb{R}_{+};U\right),\quad\boldsymbol{\sigma}\in C\left(\mathbb{R}_{+};Q\right). (4.23)

Proof. We use the Theorem 2.2 with X=VX=V and K=UK=U and arguments similar to those used in the proof of Theorem 4.2.

5 A convergence result

Everywhere in this section we assume that the function pp satisfies condition (3.8) and let qq be a function which satisfies

{ (a) q:[g,+[+.(b) There exists Lq>0 such that |q(r1)q(r2)|Lq|r1r2|r1,r2g. (c) (q(r1)q(r2))(r1r2)>0r1,r2g,r1r2. (d) q(g)=0.\left\{\begin{array}[]{l}\text{ (a) }q:\left[g,+\infty\left[\rightarrow\mathbb{R}_{+}.\right.\right.\\ \text{(b) There exists }L_{q}>0\text{ such that }\\ \quad\left|q\left(r_{1}\right)-q\left(r_{2}\right)\right|\leq L_{q}\left|r_{1}-r_{2}\right|\quad\forall r_{1},r_{2}\geq g.\\ \text{ (c) }\left(q\left(r_{1}\right)-q\left(r_{2}\right)\right)\left(r_{1}-r_{2}\right)>0\quad\forall r_{1},r_{2}\geq g,r_{1}\neq r_{2}.\\ \text{ (d) }q(g)=0.\end{array}\right.

Let μ>0\mu>0 and consider the function pμp_{\mu} defined by

pμ(r)={p(r) if rg1μq(r)+p(g) if r>gp_{\mu}(r)=\left\{\begin{array}[]{cc}p(r)&\text{ if }\quad r\leq g\\ \frac{1}{\mu}q(r)+p(g)&\text{ if }\quad r>g\end{array}\right.

Using assumption (5.1) it follows that the function pμp_{\mu} satisfies condition (3.8), i.e.

{ (a) pμ:+.(b) There exists Lpμ>0 such that |pμ(r1)pμ(r2)|Lpμ|r1r2|r1,r2. (c) (pμ(r1)pμ(r2))(r1r2)0r1,r2. (d) pμ(r)=0 for all r<0.\left\{\begin{array}[]{l}\text{ (a) }p_{\mu}:\mathbb{R}\rightarrow\mathbb{R}_{+}.\\ \text{(b) There exists }L_{p_{\mu}}>0\text{ such that }\\ \quad\left|p_{\mu}\left(r_{1}\right)-p_{\mu}\left(r_{2}\right)\right|\leq L_{p_{\mu}}\left|r_{1}-r_{2}\right|\quad\forall r_{1},r_{2}\in\mathbb{R}.\\ \text{ (c) }\left(p_{\mu}\left(r_{1}\right)-p_{\mu}\left(r_{2}\right)\right)\left(r_{1}-r_{2}\right)\geq 0\quad\forall r_{1},r_{2}\in\mathbb{R}.\\ \text{ (d) }p_{\mu}(r)=0\text{ for all }r<0.\end{array}\right.

This allows us to consider the operator Pμ:VVP_{\mu}:V\rightarrow V defined by

(Pμ𝒖,𝒗)V=Γ3pμ(uν)vν𝑑a𝒖,𝒗V\left(P_{\mu}\boldsymbol{u},\boldsymbol{v}\right)_{V}=\int_{\Gamma_{3}}p_{\mu}\left(u_{\nu}\right)v_{\nu}da\quad\forall\boldsymbol{u},\boldsymbol{v}\in V (5.4)

and, moreover, we note that PμP_{\mu} is a monotone, Lipschitz continuous operator.
With these notation, we consider the following contact problem.

Problem 𝒫1μ\mathcal{P}_{1\mu}. Find a displacement field 𝒖μ:Ω×+d\boldsymbol{u}_{\mu}:\Omega\times\mathbb{R}_{+}\rightarrow\mathbb{R}^{d} and a stress field σμ:Ω×+𝕊d\sigma_{\mu}:\Omega\times\mathbb{R}_{+}\rightarrow\mathbb{S}^{d} such that

𝝈˙μ=ε(𝒖μ)+𝒢(𝝈μ,𝜺(𝒖μ))\displaystyle\dot{\boldsymbol{\sigma}}_{\mu}=\mathcal{E}\varepsilon\left(\boldsymbol{u}_{\mu}\right)+\mathcal{G}\left(\boldsymbol{\sigma}_{\mu},\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}\right)\right) in Ω×(0,),\displaystyle\text{ in }\quad\Omega\times(0,\infty), (5.5)
Div𝝈μ+𝒇0=𝟎\displaystyle\operatorname{Div}\boldsymbol{\sigma}_{\mu}+\boldsymbol{f}_{0}=\mathbf{0} in Ω×(0,),\displaystyle\text{ in }\quad\Omega\times(0,\infty), (5.6)
𝒖μ=𝟎\displaystyle\boldsymbol{u}_{\mu}=\mathbf{0} on Γ1×(0,),\displaystyle\text{ on }\quad\Gamma_{1}\times(0,\infty), (5.7)
𝝈μ𝝂=𝒇2\displaystyle\boldsymbol{\sigma}_{\mu}\boldsymbol{\nu}=\boldsymbol{f}_{2} on Γ2×(0,),\displaystyle\text{ on }\quad\Gamma_{2}\times(0,\infty), (5.8)
σμν=pμ(uμν)+0tb(ts)uμν+(s)𝑑s\displaystyle-\sigma_{\mu\nu}=p_{\mu}\left(u_{\mu\nu}\right)+\int_{0}^{t}b(t-s)u_{\mu\nu}^{+}(s)ds on Γ3×(0,),\displaystyle\text{ on }\quad\Gamma_{3}\times(0,\infty), (5.9)
𝝈μτ=𝟎\displaystyle\boldsymbol{\sigma}_{\mu\tau}=\mathbf{0} on Γ3×(0,),\displaystyle\text{ on }\quad\Gamma_{3}\times(0,\infty), (5.10)
𝒖μ(0)=𝒖0,𝝈μ(0)=𝝈0\displaystyle\boldsymbol{u}_{\mu}(0)=\boldsymbol{u}_{0},\boldsymbol{\sigma}_{\mu}(0)=\boldsymbol{\sigma}_{0} in Ω.\displaystyle\text{ in }\quad\Omega. (5.11)

The equations and boundary conditions in problem (5.5)-(5.11) have a similar interpretations as those in problem (3.9)-(3.15). The difference arises in the fact that here we replace the contact condition with normal compliance, unilateral constraint and memory term (3.13) with the contact condition with normal compliance and memory term (5.9). In this condition μ\mu represents a penalization parameter which may be interpreted as a deformability of the foundation, and then 1μ\frac{1}{\mu} is the surface stiffness coefficient. Indeed, when μ\mu is smaller the reaction force of the foundation to penetration is larger and so the same force will result in a smaller penetration, which means that the foundation is less deformable. When μ\mu is larger the reaction force of the foundation to penetration is smaller, and so the foundation is less stiff and more deformable.

Note that here and below uμνu_{\mu\nu} is the normal component of the displacement field 𝒖μ\boldsymbol{u}_{\mu} and σμν,𝝈μτ\sigma_{\mu\nu},\boldsymbol{\sigma}_{\mu\tau} represent the normal and tangential components of the stress tensor σμ\sigma_{\mu}, respectively.

Assume now that (4.1)-(4.4) and (4.5) hold. Using arguments similar as in Section 4 for contact problem 𝒫1\mathcal{P}_{1} we obtain the following variational formulation for Problem 𝒫1μ\mathcal{P}_{1\mu}.
Problem 𝒫1μV\mathcal{P}_{1\mu}^{V}. Find a displacement field 𝒖μ:+V\boldsymbol{u}_{\mu}:\mathbb{R}_{+}\rightarrow V and a stress field 𝝈μ:+Q\boldsymbol{\sigma}_{\mu}:\mathbb{R}_{+}\rightarrow Q, such that

𝝈μ(t)=𝜺(𝒖μ(t))+𝒮1𝒖μ(t)\displaystyle\boldsymbol{\sigma}_{\mu}(t)=\mathcal{E}\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right)+\mathcal{S}_{1}\boldsymbol{u}_{\mu}(t) (5.12)
(𝜺(𝒖μ(t)),𝜺(𝒗))Q+(𝒮𝒖μ(t),𝒗)V+(Pμ𝒖μ(t),𝒗)V=(𝒇(t),𝒗)V𝒗V\displaystyle\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right),\boldsymbol{\varepsilon}(\boldsymbol{v})\right)_{Q}+\left(\mathcal{S}\boldsymbol{u}_{\mu}(t),\boldsymbol{v}\right)_{V}+\left(P_{\mu}\boldsymbol{u}_{\mu}(t),\boldsymbol{v}\right)_{V}=(\boldsymbol{f}(t),\boldsymbol{v})_{V}\quad\forall\boldsymbol{v}\in V (5.13)

hold, for all t+t\in\mathbb{R}_{+}.
It follows from Theorem 4.2 that Problem 𝒫1μV\mathcal{P}_{1\mu}^{V} has a unique solution ( 𝒖μ,𝝈μ\boldsymbol{u}_{\mu},\boldsymbol{\sigma}_{\mu} ) which satisfies (4.17). Finally, it follows from Theorem 4.3 that Problem 𝒫2V\mathcal{P}_{2}^{V} has a
unique solution ( 𝒖,𝝈\boldsymbol{u},\boldsymbol{\sigma} ) which satisfies (4.23). The behavior of the solution ( 𝒖μ,𝝈μ\boldsymbol{u}_{\mu},\boldsymbol{\sigma}_{\mu} ) as μ0\mu\rightarrow 0 is given in the following result.

Theorem 5.1 Assume that (3.8), (4.1)-(4.4), (4.6) and (5.1) hold. Then, the solution ( 𝒖μ,𝝈μ\boldsymbol{u}_{\mu},\boldsymbol{\sigma}_{\mu} ) of Problem 𝒫1μV\mathcal{P}_{1\mu}^{V} converges to the solution ( 𝒖,𝝈\boldsymbol{u},\boldsymbol{\sigma} ) of Problem 𝒫2V\mathcal{P}_{2}^{V}, that is

𝒖μ(t)𝒖(t)V+𝝈μ(t)𝝈(t)Q0\left\|\boldsymbol{u}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}+\left\|\boldsymbol{\sigma}_{\mu}(t)-\boldsymbol{\sigma}(t)\right\|_{Q}\rightarrow 0 (5.14)

as μ0\mu\rightarrow 0, for all t+t\in\mathbb{R}_{+}.

In addition to the mathematical interest in the result above, this result is important from the mechanical point of view, since it shows that the weak solution of the viscoplastic contact problem with normal compliance, memory term and finite penetration may be approached as closely as we wish by the solution of the viscoplastic contact problem with normal compliance, memory term and infinite penetration, with a sufficiently small deformability coefficient.

The proof of Theorem 5.1 is carried out in several steps.
Let μ>0\mu>0. In the first step we consider the auxiliary problem of finding a displacement field 𝒖~μ:+V\widetilde{\boldsymbol{u}}_{\mu}:\mathbb{R}_{+}\rightarrow V such that, for all t+t\in\mathbb{R}_{+},

(𝜺(𝒖~μ(t)),𝜺(𝒗))Q+(𝒮𝒖(t),𝒗)V+(Pμ𝒖~μ(t),𝒗)V\displaystyle\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}(\boldsymbol{v})\right)_{Q}+(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v})_{V}+\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\boldsymbol{v}\right)_{V} (5.15)
=(𝒇(t),𝒗)V𝒗V.\displaystyle=(\boldsymbol{f}(t),\boldsymbol{v})_{V}\forall\boldsymbol{v}\in V.

This problem is an intermediate problem between (5.13) and (4.22), since here 𝒮𝒖\mathcal{S}\boldsymbol{u} is known, taken from the problem with finite penetration 𝒫2V\mathcal{P}_{2}^{V}.

We have the following existence and uniqueness result.
Lemma 5.2 There exists a unique function 𝒖~μC(+;V)\widetilde{\boldsymbol{u}}_{\mu}\in C\left(\mathbb{R}_{+};V\right) which satisfies (5.15), for all t+t\in\mathbb{R}_{+}.

Proof. We define the operator Aμ:VVA_{\mu}:V\rightarrow V and the function 𝒇~:+V\widetilde{\boldsymbol{f}}:\mathbb{R}_{+}\rightarrow V by equalities

(Aμ𝒖,𝒗)V=(ε(𝒖),𝜺(𝒗))Q+(Pμ𝒖,𝒗)V𝒖,𝒗V,\displaystyle\left(A_{\mu}\boldsymbol{u},\boldsymbol{v}\right)_{V}=(\mathcal{E}\varepsilon(\boldsymbol{u}),\boldsymbol{\varepsilon}(\boldsymbol{v}))_{Q}+\left(P_{\mu}\boldsymbol{u},\boldsymbol{v}\right)_{V}\quad\forall\boldsymbol{u},\boldsymbol{v}\in V, (5.16)
(𝒇~(t),𝒗)V=(𝒇(t),𝒗)V(𝒮𝒖(t),𝒗)V𝒗V,t+\displaystyle(\tilde{\boldsymbol{f}}(t),\boldsymbol{v})_{V}=(\boldsymbol{f}(t),\boldsymbol{v})_{V}-(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v})_{V}\quad\forall\boldsymbol{v}\in V,t\in\mathbb{R}_{+} (5.17)

and note that (4.3), (4.4), (4.9), (4.12) and (4.14) yield

𝒇~C(+;V).\widetilde{\boldsymbol{f}}\in C\left(\mathbb{R}_{+};V\right). (5.18)

Let t+t\in\mathbb{R}_{+}. Based on (5.16)-(5.17), it is easy to see that the variational equation (5.15) is equivalent with the nonlinear equation

Aμ𝒖~μ(t)=𝒇~(t)A_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t)=\widetilde{\boldsymbol{f}}(t) (5.19)

Next, as in the proof of Theorem 4.2 by (4.1) and the properties of operator PμP_{\mu} it follows that AμA_{\mu} is a strongly monotone and Lipschitz continuous operator. And, Theorem 2.1 implies the existence of a unique solution 𝒖~μC(+;V)\widetilde{\boldsymbol{u}}_{\mu}\in C\left(\mathbb{R}_{+};V\right) for the nonlinear equation (5.19), which concludes the proof.

We proceed with the following convergence result.
Lemma 5.3 As μ0\mu\rightarrow 0,

𝒖~μ(t)𝒖(t) in V,\widetilde{\boldsymbol{u}}_{\mu}(t)\longrightarrow\boldsymbol{u}(t)\quad\text{ in }V,

for all t+t\in\mathbb{R}_{+}.
Proof. Let t+t\in\mathbb{R}_{+}. We take 𝒗=𝒖~μ(t)\boldsymbol{v}=\widetilde{\boldsymbol{u}}_{\mu}(t) in (5.15) to obtain

(𝜺(𝒖~μ(t)),𝜺(𝒖~μ(t)))Q+(𝒮𝒖(t),𝒖~μ(t))V\displaystyle\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}+\left(\mathcal{S}\boldsymbol{u}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V} (5.20)
+(Pμ𝒖~μ(t),𝒖~μ(t))V=(𝒇(t),𝒖~μ(t))V.\displaystyle\quad+\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}=\left(\boldsymbol{f}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}.

On the other hand, the properties (5.3) of the function pμp_{\mu} yield

(Pμ𝒖~μ(t),𝒖~μ(t))V0.\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\geq 0. (5.21)

We combine (5.20), (5.21) and use (4.1) (c) to obtain that

𝒖~μ(t)Vc(𝒇(t)V+𝒮𝒖(t)V).\left\|\widetilde{\boldsymbol{u}}_{\mu}(t)\right\|_{V}\leq c\left(\|\boldsymbol{f}(t)\|_{V}+\|\mathcal{S}\boldsymbol{u}(t)\|_{V}\right). (5.22)

This inequality shows that the sequence {𝒖~μ(t)}μV\left\{\widetilde{\boldsymbol{u}}_{\mu}(t)\right\}_{\mu}\subset V is bounded. Hence, there exists a subsequence of the sequence {𝒖~μ(t)}μ\left\{\widetilde{\boldsymbol{u}}_{\mu}(t)\right\}_{\mu}, still denoted {𝒖~μ(t)}μ\left\{\widetilde{\boldsymbol{u}}_{\mu}(t)\right\}_{\mu}, and an element 𝒖~(t)V\widetilde{\boldsymbol{u}}(t)\in V such that

𝒖~μ(t)𝒖~(t) in V, as μ0.\widetilde{\boldsymbol{u}}_{\mu}(t)\longrightarrow\widetilde{\boldsymbol{u}}(t)\text{ in }V,\quad\text{ as }\mu\rightarrow 0. (5.23)

It follows from (5.20) that

(Pμ𝒖~μ(t),𝒖~μ(t))V=(𝒇(t),𝒖~μ(t))V(𝜺(𝒖~μ(t)),𝜺(𝒖~μ(t)))Q(𝒮𝒖(t),𝒖~μ(t))V\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}=\left(\boldsymbol{f}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}-\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}-\left(\mathcal{S}\boldsymbol{u}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}

and, since {𝒖~μ(t)}μ\left\{\widetilde{\boldsymbol{u}}_{\mu}(t)\right\}_{\mu} is a bounded sequence in VV, we deduce that

(Pμ𝒖~μ(t),𝒖~μ(t))Vc.\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\leq c.

This implies that

Γ3pμ(u~μν(t))u~μν(t)𝑑ac\int_{\Gamma_{3}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\widetilde{u}_{\mu\nu}(t)da\leq c

and, since Γ3pμ(u~μν(t))g𝑑a0\int_{\Gamma_{3}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)gda\geq 0, it follows that

Γ3pμ(u~μν(t))(u~μν(t)g)𝑑ac\int_{\Gamma_{3}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)da\leq c (5.24)

We consider now the measurable subsets of Γ3\Gamma_{3} defined by

Γ31={𝒙Γ3:u~μν(t)(𝒙)g},Γ32={𝒙Γ3:u~μν(t)(𝒙)>g}.\Gamma_{31}=\left\{\boldsymbol{x}\in\Gamma_{3}:\widetilde{u}_{\mu\nu}(t)(\boldsymbol{x})\leq g\right\},\quad\Gamma_{32}=\left\{\boldsymbol{x}\in\Gamma_{3}:\widetilde{u}_{\mu\nu}(t)(\boldsymbol{x})>g\right\}. (5.25)

Clearly, both Γ31\Gamma_{31} and Γ32\Gamma_{32} depend on tt and μ\mu but, for simplicity, we do not indicate explicitly this dependence. We use (5.24) to write

Γ31pμ(u~μν(t))(u~μν(t)g)𝑑a+Γ32pμ(u~μν(t))(u~μν(t)g)𝑑ac\int_{\Gamma_{31}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)da+\int_{\Gamma_{32}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)da\leq c

and, since

Γ31pμ(u~μν(t))u~μν(t)𝑑a0\int_{\Gamma_{31}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\widetilde{u}_{\mu\nu}(t)da\geq 0

we obtain

Γ32pμ(u~μν(t))(u~μν(t)g)𝑑aΓ31pμ(u~μν(t))g𝑑a+c\int_{\Gamma_{32}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)da\leq\int_{\Gamma_{31}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)gda+c

Thus, taking into account that pμ(r)=p(r)p_{\mu}(r)=p(r) for rgr\leq g, by the monotonicity of the function pp we can write

Γ32pμ(u~μν(t))(u~μν(t)g)𝑑aΓ31p(u~μν(t))g𝑑a+cΓ3p(g)g𝑑a+c\int_{\Gamma_{32}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)da\leq\int_{\Gamma_{31}}p\left(\widetilde{u}_{\mu\nu}(t)\right)gda+c\leq\int_{\Gamma_{3}}p(g)gda+c

Therefore, we deduce that

Γ32pμ(u~μν(t))(u~μν(t)g)𝑑ac\int_{\Gamma_{32}}p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)da\leq c (5.26)

We use now the definitions (5.2) and (5.25) to see that, a.e. on Γ32\Gamma_{32}, we have

pμ(u~μν(t))=1μq(u~μν(t))+p(g),p(g)(u~μν(t)g)>0.p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)=\frac{1}{\mu}q\left(\widetilde{u}_{\mu\nu}(t)\right)+p(g),\quad p(g)\left(\widetilde{u}_{\mu\nu}(t)-g\right)>0.

Consequently, the inequality (5.26) yields

Γ32q(u~μν(t))(u~μν(t)g)𝑑acμ\int_{\Gamma_{32}}q\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)da\leq c\mu (5.27)

Next, we consider the function defined by

p~:+p~(r)={0 if rgq(r) if r>g\widetilde{p}:\mathbb{R}\rightarrow\mathbb{R}_{+}\quad\widetilde{p}(r)=\left\{\begin{array}[]{cll}0&\text{ if }&r\leq g\\ q(r)&\text{ if }&r>g\end{array}\right.

and we note that by (5.1) it follows that p~\widetilde{p} is a continuous increasing function and, moreover,

p~(r)=0 iff rg.\widetilde{p}(r)=0\quad\text{ iff }\quad r\leq g. (5.28)

We use (5.27), equality q(u~μν(t))=p~(u~μν(t))q\left(\widetilde{u}_{\mu\nu}(t)\right)=\widetilde{p}\left(\widetilde{u}_{\mu\nu}(t)\right) a.e. on Γ32\Gamma_{32} and (5.25) to deduce that

Γ3p~(u~μν(t))(u~μν(t)g)+cμ\int_{\Gamma_{3}}\widetilde{p}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(\widetilde{u}_{\mu\nu}(t)-g\right)^{+}\leq c\mu

where (u~μν(t)g)+\left(\widetilde{u}_{\mu\nu}(t)-g\right)^{+}denotes the positive part of u~μν(t)g\widetilde{u}_{\mu\nu}(t)-g. Therefore, passing to the limit as μ0\mu\rightarrow 0, by using (5.23) as well as compactness of the trace operator we find that

Γ3p~(u~ν(t))(u~ν(t)g)+𝑑a0\int_{\Gamma_{3}}\widetilde{p}\left(\widetilde{u}_{\nu}(t)\right)\left(\widetilde{u}_{\nu}(t)-g\right)^{+}da\leq 0

Since the integrand p~(u~ν(t))(u~ν(t)g)+\widetilde{p}\left(\widetilde{u}_{\nu}(t)\right)\left(\widetilde{u}_{\nu}(t)-g\right)^{+}is positive a.e. on Γ3\Gamma_{3}, the last inequality yields

p~(u~ν(t))(u~ν(t)g)+=0 a.e. on Γ3\widetilde{p}\left(\widetilde{u}_{\nu}(t)\right)\left(\widetilde{u}_{\nu}(t)-g\right)^{+}=0\text{ a.e. on }\Gamma_{3}

and, using (5.28) and definition (4.7) we conclude that

𝒖~(t)U.\widetilde{\boldsymbol{u}}(t)\in U. (5.29)

Next, we test in (5.15) with 𝒗𝒖~μ(t)\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t), where 𝒗U\boldsymbol{v}\in U, to obtain

(ε(𝒖~μ(t)),𝜺(𝒗)𝜺(𝒖~μ(t)))Q+(𝒮𝒖(t),𝒗𝒖~μ(t))V\displaystyle\left(\mathcal{E}\varepsilon\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}(\boldsymbol{v})-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}+\left(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V} (5.30)
+(Pμ𝒖~μ(t),𝒗𝒖~μ(t))V=(𝒇(t),𝒗𝒖~μ(t))V\displaystyle\quad+\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}=\left(\boldsymbol{f}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}

Since 𝒗U\boldsymbol{v}\in U we have pμ(vν)=p(vν)p_{\mu}\left(v_{\nu}\right)=p\left(v_{\nu}\right) a.e. on Γ3\Gamma_{3}. Taking into account this equality and the monotonicity of the function pμp_{\mu} we have

p(vν)(vνu~μν(t))pμ(u~μν(t))(vνu~μν(t)) a.e. on Γ3p\left(v_{\nu}\right)\left(v_{\nu}-\widetilde{u}_{\mu\nu}(t)\right)\geq p_{\mu}\left(\widetilde{u}_{\mu\nu}(t)\right)\left(v_{\nu}-\widetilde{u}_{\mu\nu}(t)\right)\text{ a.e. on }\Gamma_{3}

and, therefore, by using (5.4) we obtain

(P𝒗,𝒗𝒖~μ(t))V(Pμ𝒖~μ(t),𝒗𝒖~μ(t))V\left(P\boldsymbol{v},\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\geq\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V} (5.31)

Then, using (5.31) and (5.30) we find that

(ε(𝒖~μ(t)),𝜺(𝒗)𝜺(𝒖~μ(t)))Q+(𝒮𝒖(t),𝒗𝒖~μ(t))V\displaystyle\left(\mathcal{E}\varepsilon\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}(\boldsymbol{v})-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}+\left(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V} (5.32)
+(P𝒗,𝒗𝒖~μ(t))V(𝒇(t),𝒗𝒖~μ(t))V𝒗U\displaystyle\quad+\left(P\boldsymbol{v},\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\geq\left(\boldsymbol{f}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\quad\forall\boldsymbol{v}\in U

We pass to the lower limit in (5.32) and use (5.23) to obtain

(ε(𝒖~(t)),ε(𝒗)ε(𝒖~(t)))Q+(𝒮𝒖(t),𝒗𝒖~(t))V\displaystyle(\mathcal{E}\varepsilon(\widetilde{\boldsymbol{u}}(t)),\varepsilon(\boldsymbol{v})-\varepsilon(\widetilde{\boldsymbol{u}}(t)))_{Q}+(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}(t))_{V} (5.33)
+(P𝒗,𝒗𝒖~(t))V(𝒇(t),𝒗𝒖~(t))V𝒗U\displaystyle\quad+(P\boldsymbol{v},\boldsymbol{v}-\widetilde{\boldsymbol{u}}(t))_{V}\geq(\boldsymbol{f}(t),\boldsymbol{v}-\widetilde{\boldsymbol{u}}(t))_{V}\quad\forall\boldsymbol{v}\in U

Next, we take 𝒗=𝒖~(t)\boldsymbol{v}=\widetilde{\boldsymbol{u}}(t) in (4.22) and 𝒗=𝒖(t)\boldsymbol{v}=\boldsymbol{u}(t) in (5.33). Then, adding the resulting inequalities we find that

(ε(𝒖~(t))ε(𝒖(t)),𝜺(𝒖~(t))𝜺(𝒖(t)))Q0(\mathcal{E}\varepsilon(\widetilde{\boldsymbol{u}}(t))-\mathcal{E}\varepsilon(\boldsymbol{u}(t)),\boldsymbol{\varepsilon}(\widetilde{\boldsymbol{u}}(t))-\boldsymbol{\varepsilon}(\boldsymbol{u}(t)))_{Q}\leq 0

This inequality combined with (4.1) implies that

𝒖~(t)=𝒖(t)\widetilde{\boldsymbol{u}}(t)=\boldsymbol{u}(t)

It follows from here that the whole sequence {𝒖~μ(t)}μ\left\{\widetilde{\boldsymbol{u}}_{\mu}(t)\right\}_{\mu} is weakly convergent to the element 𝒖(t)V\boldsymbol{u}(t)\in V, which concludes the proof.

We proceed with the following strong convergence result.

Lemma 5.4 As μ0\mu\rightarrow 0,

𝒖~μ(t)𝒖(t)V0\left\|\widetilde{\boldsymbol{u}}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}\rightarrow 0

for all t+t\in\mathbb{R}_{+}.
Proof. Let t+t\in\mathbb{R}_{+}. Using (4.1) we write

m𝒖~μ(t)𝒖(t)V2(𝜺(𝒖~μ(t))𝜺(𝒖(t)),𝜺(𝒖~μ(t))𝜺(𝒖(t)))Q\displaystyle m_{\mathcal{E}}\left\|\widetilde{\boldsymbol{u}}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}^{2}\leq\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)-\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t)),\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)-\boldsymbol{\varepsilon}(\boldsymbol{u}(t))\right)_{Q}
=(𝜺(𝒖(t)),𝜺(𝒖(t))𝜺(𝒖~μ(t)))Q(𝜺(𝒖~μ(t)),𝜺(𝒖(t))𝜺(𝒖~μ(t)))Q\displaystyle\quad=\left(\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t)),\boldsymbol{\varepsilon}(\boldsymbol{u}(t))-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}-\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}(\boldsymbol{u}(t))-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}

Next, we take 𝒗=𝒖(t)\boldsymbol{v}=\boldsymbol{u}(t) in (5.32) to obtain

(𝜺(𝒖~μ(t)),𝜺(𝒖(t))𝜺(𝒖~μ(t)))Q(𝒮𝒖(t),𝒖(t)𝒖~μ(t))V+(P𝒖(t),𝒖(t)𝒖~μ(t))V(𝒇(t),𝒖(t)𝒖~μ(t))V\begin{gathered}-\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}(\boldsymbol{u}(t))-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}\leq\left(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{u}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\\ +\left(P\boldsymbol{u}(t),\boldsymbol{u}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}-\left(\boldsymbol{f}(t),\boldsymbol{u}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\end{gathered}

and, therefore, combining the above inequalities we find that

m𝒖~μ(t)𝒖(t)V2(ε(𝒖(t)),𝜺(𝒖(t))𝜺(𝒖~μ(t)))Q\displaystyle m_{\mathcal{E}}\left\|\widetilde{\boldsymbol{u}}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}^{2}\leq\left(\mathcal{E}\varepsilon(\boldsymbol{u}(t)),\boldsymbol{\varepsilon}(\boldsymbol{u}(t))-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}
+(𝒮𝒖(t),𝒖(t)𝒖~μ(t))V+(P𝒖(t),𝒖(t)𝒖~μ(t))V(𝒇(t),𝒖(t)𝒖~μ(t))V.\displaystyle+\left(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{u}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}+\left(P\boldsymbol{u}(t),\boldsymbol{u}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}-\left(\boldsymbol{f}(t),\boldsymbol{u}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}.

We pass to the upper limit in this inequality and use Lemma 5.3 to conclude the proof of the lemma.

We are now in position to provide the proof of Theorem 5.1.
Proof. Let t+t\in\mathbb{R}_{+}and let nn\in\mathbb{N} be such that t[0,n]t\in[0,n]. Let also μ>0\mu>0. Then, testing with 𝒗=𝒖μ(t)𝒖~μ(t)\boldsymbol{v}=\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t) in (5.15) and (5.13), we have

(𝜺(𝒖~μ(t)),𝜺(𝒖μ(t))𝜺(𝒖~μ(t)))Q+(𝒮𝒖(t),𝒖μ(t)𝒖~μ(t))V+(Pμ𝒖~μ(t),𝒖μ(t)𝒖~μ(t))V=(𝒇(t),𝒖μ(t)𝒖~μ(t))V(𝜺(𝒖μ(t)),𝜺(𝒖μ(t))𝜺(𝒖~μ(t)))Q+(𝒮𝒖μ(t),𝒖μ(t)𝒖~μ(t))V+(Pμ𝒖μ(t),𝒖μ(t)𝒖~μ(t))V=(𝒇(t),𝒖μ(t)𝒖~μ(t))V\begin{array}[]{r}\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right)-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}+\left(\mathcal{S}\boldsymbol{u}(t),\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\\ +\left(P_{\mu}\widetilde{\boldsymbol{u}}_{\mu}(t),\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}=\left(\boldsymbol{f}(t),\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\\ \left(\mathcal{E}\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right),\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right)-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}+\left(\mathcal{S}\boldsymbol{u}_{\mu}(t),\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\\ +\left(P_{\mu}\boldsymbol{u}_{\mu}(t),\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}=\left(\boldsymbol{f}(t),\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\end{array}

We subtract the previous equalities and use the monotonicity of the operator PμP_{\mu} to deduce that

(𝜺(𝒖μ(t))𝜺(𝒖~μ(t)),𝜺(𝒖μ(t))𝜺(𝒖~μ(t)))Q(𝒮𝒖(t)𝒮𝒖μ(t),𝒖μ(t)𝒖~μ(t))V\begin{gathered}\left(\mathcal{E}\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right)-\mathcal{E}\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right),\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right)-\boldsymbol{\varepsilon}\left(\widetilde{\boldsymbol{u}}_{\mu}(t)\right)\right)_{Q}\\ \leq\left(\mathcal{S}\boldsymbol{u}(t)-\mathcal{S}\boldsymbol{u}_{\mu}(t),\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right)_{V}\end{gathered}

and, therefore,

𝒖μ(t)𝒖~μ(t)V1m𝒮𝒖(t)𝒮𝒖μ(t)V.\left\|\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right\|_{V}\leq\frac{1}{m_{\mathcal{E}}}\left\|\mathcal{S}\boldsymbol{u}(t)-\mathcal{S}\boldsymbol{u}_{\mu}(t)\right\|_{V}. (5.34)

We use (5.34) to find that

𝒖μ(t)𝒖~μ(t)Vrnm0t𝒖(s)𝒖μ(s)V𝑑s\left\|\boldsymbol{u}_{\mu}(t)-\widetilde{\boldsymbol{u}}_{\mu}(t)\right\|_{V}\leq\frac{r_{n}}{m_{\mathcal{E}}}\int_{0}^{t}\left\|\boldsymbol{u}(s)-\boldsymbol{u}_{\mu}(s)\right\|_{V}ds

where rnr_{n} is given by (4.20). It follows from here that

𝒖μ(t)𝒖(t)V𝒖~μ(t)𝒖(t)V+rnm0t𝒖μ(s)𝒖(s)V𝑑s\left\|\boldsymbol{u}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}\leq\left\|\widetilde{\boldsymbol{u}}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}+\frac{r_{n}}{m_{\mathcal{E}}}\int_{0}^{t}\left\|\boldsymbol{u}_{\mu}(s)-\boldsymbol{u}(s)\right\|_{V}ds

and, using a Gronwall argument, we obtain

𝒖μ(t)𝒖(t)V𝒖~μ(t)𝒖(t)V+rnm0ternm(ts)𝒖~μ(s)𝒖(s)V𝑑s\left\|\boldsymbol{u}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}\leq\left\|\widetilde{\boldsymbol{u}}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}+\frac{r_{n}}{m_{\mathcal{E}}}\int_{0}^{t}e^{\frac{r_{n}}{m_{\mathcal{E}}}(t-s)}\left\|\widetilde{\boldsymbol{u}}_{\mu}(s)-\boldsymbol{u}(s)\right\|_{V}ds (5.35)

Note that ernm(ts)ernmtenrnme^{\frac{rn}{m_{\mathcal{E}}}(t-s)}\leq e^{\frac{rn}{m_{\mathcal{E}}}t}\leq e^{\frac{nr_{n}}{m_{\mathcal{E}}}} for all s[0,t]s\in[0,t] and, therefore, (5.35) yields

𝒖μ(t)𝒖(t)V𝒖~μ(t)𝒖(t)V+rnmenrnm0t𝒖~μ(s)𝒖(s)V𝑑s\left\|\boldsymbol{u}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}\leq\left\|\widetilde{\boldsymbol{u}}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}+\frac{r_{n}}{m_{\mathcal{E}}}e^{\frac{nr_{n}}{m_{\mathcal{E}}}}\int_{0}^{t}\left\|\widetilde{\boldsymbol{u}}_{\mu}(s)-\boldsymbol{u}(s)\right\|_{V}ds (5.36)

On the other hand, by estimate (5.22), Lemma 5.4 and Lebesgue’s convergence Theorem it follows that

0t𝒖~μ(s)𝒖(s)V𝑑s0 as μ0\int_{0}^{t}\left\|\widetilde{\boldsymbol{u}}_{\mu}(s)-\boldsymbol{u}(s)\right\|_{V}ds\rightarrow 0\quad\text{ as }\quad\mu\rightarrow 0 (5.37)

We use now (5.36), (5.37) and Lemma 5.4 to see that

𝒖μ(t)𝒖(t)V0 as μ0.\left\|\boldsymbol{u}_{\mu}(t)-\boldsymbol{u}(t)\right\|_{V}\rightarrow 0\quad\text{ as }\quad\mu\rightarrow 0. (5.38)

Next, by (4.21) and (5.12) we obtain

𝝈μ(t)𝝈(t)Q𝜺(𝒖μ(t))𝜺(𝒖(t))Q+𝒮𝒖μ(t)𝒮𝒖(t)V\left\|\boldsymbol{\sigma}_{\mu}(t)-\boldsymbol{\sigma}(t)\right\|_{Q}\leq\left\|\mathcal{E}\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{\mu}(t)\right)-\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t))\right\|_{Q}+\left\|\mathcal{S}\boldsymbol{u}_{\mu}(t)-\mathcal{S}\boldsymbol{u}(t)\right\|_{V}

and, using (4.1), (4.13), (4.3) and (4.20) it follows that

𝝈μ(t)𝝈(t)Qc𝒖μ(t)𝒖(t))V+rn0t𝒖μ(s)𝒖(s)Vds\left.\left\|\boldsymbol{\sigma}_{\mu}(t)-\boldsymbol{\sigma}(t)\right\|_{Q}\leq c\|\boldsymbol{u}_{\mu}(t)-\boldsymbol{u}(t)\right)\left\|{}_{V}+r_{n}\int_{0}^{t}\right\|\boldsymbol{u}_{\mu}(s)-\boldsymbol{u}(s)\|_{V}ds

We use again the convergence (5.38) and Lebesque’s Theorem to find that

𝝈μ(t)𝝈(t)Q0 as μ0\left\|\boldsymbol{\sigma}_{\mu}(t)-\boldsymbol{\sigma}(t)\right\|_{Q}\rightarrow 0\quad\text{ as }\quad\mu\rightarrow 0 (5.39)

Theorem 5.1 is now a consequence of the convergences (5.38) and (5.39).

6 Numerical solutions

This section is devoted to the numerical solution of the contact problems presented in Section 3, including the numerical validation of the convergence result in Theorem 5.1. In order to avoid repetitions, we restrict ourselves to present details only on
the numerical approach of Problem 𝒫2\mathcal{P}_{2}, which is based on penalization and the augmented Lagrangean method. To this end we introduce a new variational formulation of Problem 𝒫2\mathcal{P}_{2}, more convenient for the numerical solution.

An adapted variational formulation. We consider the space Xν={vν|Γ3:𝒗V}X_{\nu}=\left\{v_{\left.\nu\right|_{\Gamma_{3}}}:\boldsymbol{v}\in\right.V\} with its usual norm and denote by XνX_{\nu}^{\prime} and ,Xν,Xν\langle\cdot,\cdot\rangle_{X_{\nu}^{\prime},X_{\nu}} the dual of XνX_{\nu} and the duality pairing mapping, respectively. We also consider the function φ:Xν(,+]\varphi:X_{\nu}\rightarrow(-\infty,+\infty] and the operators L,H:XνXνL,H:X_{\nu}\rightarrow X_{\nu}^{\prime} defined by

φ(uν)=Γ3I(uνg)𝑑a,uνXν\displaystyle\varphi\left(u_{\nu}\right)=\int_{\Gamma_{3}}I_{\mathbb{R}_{-}}\left(u_{\nu}-g\right)da,\quad\forall u_{\nu}\in X_{\nu}
Luν,vνXν,Xν=Γ3p(uν)vν𝑑auν,vνXν\displaystyle\left\langle Lu_{\nu},v_{\nu}\right\rangle_{X_{\nu}^{\prime},X_{\nu}}=\int_{\Gamma_{3}}p\left(u_{\nu}\right)v_{\nu}da\quad\forall u_{\nu},v_{\nu}\in X_{\nu}
Huν,vνXν,Xν=Γ3(0tb(ts)uν+(s)𝑑s)vν𝑑auν,vνXν\displaystyle\left\langle Hu_{\nu},v_{\nu}\right\rangle_{X_{\nu}^{\prime},X_{\nu}}=\int_{\Gamma_{3}}\left(\int_{0}^{t}b(t-s)u_{\nu}^{+}(s)ds\right)v_{\nu}da\quad\forall u_{\nu},v_{\nu}\in X_{\nu} (6.1)

where II_{\mathbb{R}_{-}}represents the indicator function of the set =(,0]\mathbb{R}_{-}=(-\infty,0].
We note that, for all t+t\in\mathbb{R}_{+}, condition (3.13) is equivalent to the subdifferential inclusion

σν(t)φ(uν(t))+Luν|Γ3(t)+Huν|Γ3(t) in Xν-\sigma_{\nu}(t)\in\partial\varphi\left(u_{\nu}(t)\right)+Lu_{\left.\nu\right|_{\Gamma_{3}}}(t)+Hu_{\left.\nu\right|_{\Gamma_{3}}}(t)\quad\text{ in }\quad X_{\nu}^{\prime} (6.2)

where φ\partial\varphi denotes the subdifferential of φ\varphi. This inclusion suggests to introduce a new unknown of the problem, the Lagrange multiplier, which represents the normal stress on the contact surface. Thus, proceeding in a standard way and using the inclusion (6.2) we obtain the following variational formulation of Problem 𝒫2\mathcal{P}_{2}, in terms of three unknown fields.

Problem 𝒫~2V\widetilde{\mathcal{P}}_{2}^{V}. Find a displacement field 𝒖:+V\boldsymbol{u}:\mathbb{R}_{+}\rightarrow V, a stress field 𝝈:+Q\boldsymbol{\sigma}:\mathbb{R}_{+}\rightarrow Q and a Lagrange multiplier λ:+Xν\lambda:\mathbb{R}_{+}\rightarrow X_{\nu}^{\prime} such that, for all t+t\in\mathbb{R}_{+},

𝝈(t)=𝜺(𝒖(t))+0t𝒢(𝝈(s),𝜺(𝒖(s)))𝑑s+𝝈0𝜺(𝒖0)\displaystyle\boldsymbol{\sigma}(t)=\mathcal{E}\boldsymbol{\varepsilon}(\boldsymbol{u}(t))+\int_{0}^{t}\mathcal{G}(\boldsymbol{\sigma}(s),\boldsymbol{\varepsilon}(\boldsymbol{u}(s)))ds+\boldsymbol{\sigma}_{0}-\mathcal{E}\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{0}\right) (6.3)
(𝝈(t),𝜺(𝒗))Qλ(t),vν|Γ3Xν,Xν=(𝒇(t),𝒗)V𝒗V\displaystyle(\boldsymbol{\sigma}(t),\boldsymbol{\varepsilon}(\boldsymbol{v}))_{Q}-\left\langle\lambda(t),v_{\left.\nu\right|_{\Gamma_{3}}}\right\rangle_{X_{\nu}^{\prime},X_{\nu}}=(\boldsymbol{f}(t),\boldsymbol{v})_{V}\quad\forall\boldsymbol{v}\in V (6.4)
λ(t)φ(uν(t))+Luν|Γ3(t)+Hu|Γ3(t)\displaystyle-\lambda(t)\in\partial\varphi\left(u_{\nu}(t)\right)+Lu_{\nu_{\left.\right|_{\Gamma_{3}}}}(t)+Hu_{\left.\right|_{\Gamma_{3}}}(t) (6.5)

Fully-discrete approximation. Let k>0k>0 be the time step size and define

tn=nk,0nN,t_{n}=nk,\quad 0\leq n\leq N,

where NN is a sufficiently large integer. Below, for a continuous function v(t)v(t) with values in a function space, we use the notation vj=v(tj)v_{j}=v\left(t_{j}\right), for 0jN0\leq j\leq N. Assume that Ω\Omega is a polyhedral domain. Moreover, let {𝒯h}\left\{\mathcal{T}^{h}\right\} be a regular family of triangular finite element partitions of Ω¯\bar{\Omega} that are compatible with the boundary decomposition
Γ=Γ¯1Γ¯2Γ¯3\Gamma=\bar{\Gamma}_{1}\cup\bar{\Gamma}_{2}\cup\bar{\Gamma}_{3}, i.e. if one side of an element T𝒯hT\in\mathcal{T}^{h} has more than one point on Γ\Gamma, then the side lies entirely in Γ¯1,Γ¯2\bar{\Gamma}_{1},\bar{\Gamma}_{2} or Γ¯3\bar{\Gamma}_{3}. The space VV is approximated by the finite dimensional space VhVV^{h}\subset V of continuous and piecewise affine functions, that is,

Vh={𝒗h[C(Ω¯)]d:𝒗h|T[P1(T)]dT𝒯h,\displaystyle V^{h}=\left\{\boldsymbol{v}^{h}\in[C(\bar{\Omega})]^{d}:\left.\boldsymbol{v}^{h}\right|_{T}\in\left[P_{1}(T)\right]^{d}\quad\forall T\in\mathcal{T}^{h},\right.
𝒗h=𝟎 at the nodes on Γ1},\displaystyle\left.\boldsymbol{v}^{h}=\mathbf{0}\text{ at the nodes on }\Gamma_{1}\right\}, (6.6)

where P1(T)P_{1}(T) represents the space of polynomials of degree less or equal to one in TT. The space QQ is approximated by the finite element space of piecewise constants, denoted QhQ^{h}. For any 𝝉Q\boldsymbol{\tau}\in Q, we denote by ΠQh𝝉\Pi_{Q^{h}}\boldsymbol{\tau} its finite element projection onto QhQ^{h}, that is

(ΠQh𝝉,𝝉h)Q=(𝝉,𝝉h)Q𝝉hQh.\left(\Pi_{Q^{h}}\boldsymbol{\tau},\boldsymbol{\tau}^{h}\right)_{Q}=\left(\boldsymbol{\tau},\boldsymbol{\tau}^{h}\right)_{Q}\quad\forall\boldsymbol{\tau}^{h}\in Q^{h}.

We also consider the discrete space YνhXνL2(Γ3)Y_{\nu}^{h}\subset X_{\nu}^{\prime}\cap L^{2}\left(\Gamma_{3}\right) related to the discretization of the Lagrange multiplier λ\lambda. see [6, 7] for considerations about the discretization step.

Let 𝒖0hVh\boldsymbol{u}_{0}^{h}\in V^{h} and 𝝈0hQh\boldsymbol{\sigma}_{0}^{h}\in Q^{h} be the finite element approximations of 𝒖0\boldsymbol{u}_{0} and 𝝈0\boldsymbol{\sigma}_{0}, respectively. Then, we consider the following fully discrete numerical approximation of Problem 𝒫~2V\widetilde{\mathcal{P}}_{2}^{V}.

Problem 𝒫Vhk\mathcal{P}_{V}^{hk}. Find a discrete displacement field 𝒖hk={𝒖nhk}n=1NVh\boldsymbol{u}^{hk}=\left\{\boldsymbol{u}_{n}^{hk}\right\}_{n=1}^{N}\subset V^{h}, a discrete stress field 𝝈hk={𝝈nhk}n=1NQh\boldsymbol{\sigma}^{hk}=\left\{\boldsymbol{\sigma}_{n}^{hk}\right\}_{n=1}^{N}\subset Q^{h} and a discrete Lagrange multiplier λhk={λnhk}n=1NYνh\lambda^{hk}=\left\{\lambda_{n}^{hk}\right\}_{n=1}^{N}\subset Y_{\nu}^{h} such that, for all n=1,,Nn=1,\ldots,N,

𝝈nhk=𝒫Qhε(𝒖nhk)+j=0n1k𝒫Qh𝒢jhk\displaystyle\boldsymbol{\sigma}_{n}^{hk}=\mathcal{P}_{Q^{h}}\mathcal{E}\varepsilon\left(\boldsymbol{u}_{n}^{hk}\right)+\sum_{j=0}^{n-1}k\mathcal{P}_{Q^{h}}\mathcal{G}_{j}^{hk} (6.7)
(𝝈nhk,𝜺(𝒗h))Qλnhk,vν|Γ3hXν,Xν=(𝒇n,𝒗h)V𝒗hVh\displaystyle\left(\boldsymbol{\sigma}_{n}^{hk},\boldsymbol{\varepsilon}\left(\boldsymbol{v}^{h}\right)\right)_{Q}-\left\langle\lambda_{n}^{hk},v_{\left.\nu\right|_{\Gamma_{3}}}^{h}\right\rangle_{X_{\nu}^{\prime},X_{\nu}}=\left(\boldsymbol{f}_{n},\boldsymbol{v}^{h}\right)_{V}\quad\forall\boldsymbol{v}^{h}\in V^{h} (6.8)
λnhkφ(uν|Γ3nnhk)+Luν|Γ3hkn+H~uν|Γ3hkn\displaystyle-\lambda_{n}^{hk}\in\partial\varphi\left(u_{\left.\nu\right|_{\Gamma_{3}}{}_{n}{}_{n}}^{hk}\right)+Lu_{\left.\nu\right|_{\Gamma_{3}}{}_{n}{}^{hk}}+\widetilde{H}u_{\left.\nu\right|_{\Gamma_{3}}{}_{n}{}^{hk}} (6.9)

Note that the sum in (6.7) corresponds to the approximation of the integral in (6.3) by using a rectangle method (Top-left corner approximation) for the time integration.

Furthermore, in (6.9) we propose to approximate the operator HH by using a trapezoidale rule for the time integral which appears in (6.1). The approximated operator H~\widetilde{H} is defined as follows:

H~uν=i=0nΔt2[b(tnti1)uνi1+b(tnti)uνi]uνXν.\widetilde{H}u_{\nu}=\sum_{i=0}^{n}\frac{\Delta t}{2}\left[b\left(t_{n}-t_{i-1}\right)u_{\nu i-1}+b\left(t_{n}-t_{i}\right)u_{\nu i}\right]\quad\forall u_{\nu}\in X_{\nu}. (6.10)

Here and below we use the short-hand notation 𝒢jhk=𝒢(𝝈jhk,𝜺(𝒖jhk))\mathcal{G}_{j}^{hk}=\mathcal{G}\left(\boldsymbol{\sigma}_{j}^{hk},\boldsymbol{\varepsilon}\left(\boldsymbol{u}_{j}^{hk}\right)\right).
Numerical method. In the rest of this subsection, to simplify the notation, we skip the dependence of various variables with respect to the discretization parameters nn, kk and hh, i.e., for example, we write 𝒖\boldsymbol{u} instead of 𝒖nhk\boldsymbol{u}_{n}^{hk}.

For the numerical treatment of the condition (6.9), we use the penalized method for the compliance contact combined with the augmented Lagrangean approach for the unilateral condition. To this end, we consider additional fictitious nodes for the Lagrange multiplier in the initial mesh. The construction of these nodes depends on the contact element used for the geometrical discretization of the interface Γ3\Gamma_{3}. In the case of the numerical example presented below, the discretization is based on "node-to-rigid" contact element, which is composed by one node of Γ3\Gamma_{3} and one Lagrange multiplier node. This contact interface discretization is characterized by a finite dimensional subspace HΓ3hYνhH_{\Gamma_{3}}^{h}\subset Y_{\nu}^{h}. Let Ntot hN_{\text{tot }}^{h} be the total number of nodes and denote by αi\alpha^{i} the basis functions used to define the space VhV^{h} for i=1,,Ntot hi=1,\ldots,N_{\text{tot }}^{h}. Moreover, let NΓ3hN_{\Gamma_{3}}^{h} represent the number of nodes on the interface Γ3\Gamma_{3} and let μi\mu^{i} be the shape functions of the finite element space HΓ3hH_{\Gamma_{3}}^{h}, for i=1,,NΓ3hi=1,\ldots,N_{\Gamma_{3}}^{h}, i.e.

HΓ3h={γhYνh:γh=i=1NΓ3hγiμi}H_{\Gamma_{3}}^{h}=\left\{\gamma^{h}\in Y_{\nu}^{h}:\gamma^{h}=\sum_{i=1}^{N_{\Gamma_{3}}^{h}}\gamma^{i}\mu^{i}\right\}

Usually, if a P1P_{1} finite element method is used for the displacement, then a P0P_{0} finite element method is considered for the multipliers. The expression of functions 𝒗hVh\boldsymbol{v}^{h}\in V^{h} and γhHΓ3h\gamma^{h}\in H_{\Gamma_{3}}^{h} is given by

𝒗h=i=1Ntot h𝒗iαi and γh=i=1NΓ3hγiμi,\boldsymbol{v}^{h}=\sum_{i=1}^{N_{\text{tot }}^{h}}\boldsymbol{v}^{i}\alpha^{i}\quad\text{ and }\quad\gamma^{h}=\sum_{i=1}^{N_{\Gamma_{3}}^{h}}\gamma^{i}\mu^{i}, (6.11)

where 𝒗i\boldsymbol{v}^{i} represents the value of the corresponding functions 𝒗h\boldsymbol{v}^{h} at the ii-th node of 𝒯h\mathcal{T}^{h}. Also, γi\gamma^{i} denotes the value of the function γh\gamma^{h} at the ii-th node of the contact element of the discretized contact interface. More details about this discretization step can be found in [1, 6, 7, 17].

It can be shown that the numerical approach of Problem 𝒫Vhk\mathcal{P}_{V}^{hk} is governed at each time step nn by a system of nonlinear equations of the form

𝐑(𝒖,𝝀)=𝐆~(𝒖)+(𝒖,𝝀)=𝟎,\mathbf{R}(\boldsymbol{u},\boldsymbol{\lambda})=\widetilde{\mathbf{G}}(\boldsymbol{u})+\mathcal{F}(\boldsymbol{u},\boldsymbol{\lambda})=\mathbf{0}, (6.12)

where the functions 𝐆~\widetilde{\mathbf{G}} and \mathcal{F} are defined below. Here the unknowns are the discrete displacement field 𝒖dNtot h\boldsymbol{u}\in\mathbb{R}^{d\cdot N_{\text{tot }}^{h}} and the Lagrange multiplier generalized vector 𝝀NΓ3h\boldsymbol{\lambda}\in\mathbb{R}^{N_{\Gamma_{3}}^{h}}, defined by

𝒖={𝒖i}i=1Ntoth,𝝀={λi}i=1NΓ3h,\boldsymbol{u}=\left\{\boldsymbol{u}^{i}\right\}_{i=1}^{N_{tot}^{h}},\boldsymbol{\lambda}=\left\{\lambda^{i}\right\}_{i=1}^{N_{\Gamma_{3}}^{h}}, (6.13)

where 𝒖i\boldsymbol{u}^{i} represents the value of the corresponding function 𝒖nhk\boldsymbol{u}_{n}^{hk} at the ii-th node of 𝒯h\mathcal{T}^{h}. Also, λi\lambda^{i} denotes the value of the corresponding function λnhk\lambda_{n}^{hk} at the ii-th node of the contact element of the discretized contact interface. The generalized elastic term 𝐆~(𝒖)dNtot h×NΓ3h\widetilde{\mathbf{G}}(\boldsymbol{u})\in\mathbb{R}^{d\cdot N_{\text{tot }}^{h}}\times\mathbb{R}^{N_{\Gamma_{3}}^{h}} is defined by 𝐆~(𝒖)=(𝐆(𝒖),𝟎NΓ3h)\widetilde{\mathbf{G}}(\boldsymbol{u})=\left(\mathbf{G}(\boldsymbol{u}),\mathbf{0}_{N_{\Gamma_{3}}^{h}}\right), where 𝟎NΓ3h\mathbf{0}_{N_{\Gamma_{3}}^{h}} is the zero element of NΓ3h,𝐆(𝒖)dNtot h\mathbb{R}^{N_{\Gamma_{3}}^{h}},\mathbf{G}(\boldsymbol{u})\in\mathbb{R}^{d\cdot N_{\text{tot }}^{h}} denotes the term given by

(𝐆(𝒖)𝒗)d×Ntoth=(𝝈nhk,𝜺(𝒗h))Q(𝒇,𝒗h)V𝒗={𝒗i}i=1Ntoth(\mathbf{G}(\boldsymbol{u})\cdot\boldsymbol{v})_{\mathbb{R}^{d\times N_{tot}^{h}}}=\left(\boldsymbol{\sigma}_{n}^{hk},\boldsymbol{\varepsilon}\left(\boldsymbol{v}^{h}\right)\right)_{Q}-\left(\boldsymbol{f},\boldsymbol{v}^{h}\right)_{V}\quad\forall\boldsymbol{v}=\left\{\boldsymbol{v}^{i}\right\}_{i=1}^{N_{tot}^{h}}

𝒗h\boldsymbol{v}^{h} is defined by (6.11) and 𝝈nhk\boldsymbol{\sigma}_{n}^{hk} is related to 𝒖nhk\boldsymbol{u}_{n}^{hk} by the discrete constitutive law (6.7). The contact operator (𝒖,𝝀)\mathcal{F}(\boldsymbol{u},\boldsymbol{\lambda}), which allows to deal with the contact condition (6.9) is defined by

((𝒖,𝝀)\displaystyle(\mathcal{F}(\boldsymbol{u},\boldsymbol{\lambda}) (𝒗,𝜸))dNtoth×NΓ3h=Γ3𝒖lνr(𝒖nhk,λnhk)𝒗hdΓ\displaystyle\cdot(\boldsymbol{v},\boldsymbol{\gamma}))_{\mathbb{R}^{d\cdot N_{tot}^{h}\times\mathbb{R}^{N_{\Gamma_{3}}^{h}}}}=\int_{\Gamma_{3}}\nabla_{\boldsymbol{u}}l_{\nu}^{r}\left(\boldsymbol{u}_{n}^{hk},\lambda_{n}^{hk}\right)\cdot\boldsymbol{v}^{h}d\Gamma (6.14)
+Γ3λlνr(𝒖nhk,λnhk)γh𝑑Γ+Γ3𝒖𝒫c([(uν)nhk]g)𝒗h𝑑Γ\displaystyle+\int_{\Gamma_{3}}\nabla_{\lambda}l_{\nu}^{r}\left(\boldsymbol{u}_{n}^{hk},\lambda_{n}^{hk}\right)\cdot\gamma^{h}d\Gamma+\int_{\Gamma_{3}}\nabla_{\boldsymbol{u}}\mathcal{P}_{c}\left(\left[\left(u_{\nu}\right)_{n}^{hk}\right]_{g}\right)\cdot\boldsymbol{v}^{h}d\Gamma
+Γ3𝒖𝒫H([(uν)nhk]g)𝒗h𝑑Γ\displaystyle+\int_{\Gamma_{3}}\nabla_{\boldsymbol{u}}\mathcal{P}_{H}\left(\left[\left(u_{\nu}\right)_{n}^{hk}\right]_{g}\right)\cdot\boldsymbol{v}^{h}d\Gamma
𝒖,𝒗\displaystyle\forall\boldsymbol{u},\boldsymbol{v}\in dNtoth,𝝀,𝜸NΓ3h,𝒖nhk,𝒗hVh,λnhk,γhHΓ3h\displaystyle\mathbb{R}^{d\cdot N_{tot}^{h}},\forall\boldsymbol{\lambda},\boldsymbol{\gamma}\in\mathbb{R}^{N_{\Gamma_{3}}^{h}},\boldsymbol{u}_{n}^{hk},\boldsymbol{v}^{h}\in V^{h},\forall\lambda_{n}^{hk},\gamma^{h}\in H_{\Gamma_{3}}^{h}

Here 𝒫c,𝒫H:\mathcal{P}_{c},\mathcal{P}_{H}:\mathbb{R}\rightarrow\mathbb{R} are derivable functions such that 𝒖𝒫c=p\nabla_{\boldsymbol{u}}\mathcal{P}_{c}=p on (,g](-\infty,g] and 𝒖𝒫H=H~,[]g:\nabla_{\boldsymbol{u}}\mathcal{P}_{H}=\widetilde{H},[\cdot]_{g}:\mathbb{R}\rightarrow\mathbb{R} is the function defined by

[s]g={s if sg0 if s>g[s]_{g}=\begin{cases}s&\text{ if }\quad s\leq g\\ 0&\text{ if }\quad s>g\end{cases}

and 𝒖\nabla_{\boldsymbol{u}} represents the gradient operator with respect the variable 𝒖\boldsymbol{u}; also, lνrl_{\nu}^{r} denotes the augmented Lagrangean functional given by

lνr(𝒖nhk,λnhk)=12r(λnhk)2+12r[(λnhk+r((uν)nhkg))+]2,l_{\nu}^{r}\left(\boldsymbol{u}_{n}^{hk},\lambda_{n}^{hk}\right)=-\frac{1}{2r}\left(\lambda_{n}^{hk}\right)^{2}+\frac{1}{2r}\left[\left(\lambda_{n}^{hk}+r\left(\left(u_{\nu}\right)_{n}^{hk}-g\right)\right)^{+}\right]^{2}, (6.15)

where rr is a positive penalty coefficient.
The solution of the nonlinear system (6.12) is based on a generalized Newton method, which permits to treat simultaneously the two unknowns 𝒖\boldsymbol{u} and 𝝀\boldsymbol{\lambda}. Nevertheless, to keep this paper in a reasonable length, we skip the presentation of the numerical algorithm and we pass in what follows to description of the numerical example. Details on this kind of algorithms can be found in [1, 10, 17.

7 Numerical simulations

Physical setting of the numerical example. For the numerical simulations we consider the physical setting depicted in Figure 1. There, Ω=[0,2]×[0,1]2\Omega=[0,2]\times[0,1]\subset\mathbb{R}^{2} with Γ1=({0}×[0.5,1])({2}×[0.5,1]),Γ2=([0,2]×{1})({0}×[0,0.5])({2}×[0,0.5]),Γ3=[0,2]×{0}\Gamma_{1}=(\{0\}\times[0.5,1])\cup(\{2\}\times[0.5,1]),\Gamma_{2}=([0,2]\times\{1\})\cup(\{0\}\times[0,0.5])\cup(\{2\}\times[0,0.5]),\Gamma_{3}=[0,2]\times\{0\}. The domain Ω\Omega represents the cross section of a three-dimensional deformable body subjected to the action of tractions in such a way that a plane stress hypothesis is assumed. On the part Γ1\Gamma_{1} the body is clamped and, therefore, the displacement field vanishes there. Vertical tractions act on the part [0,2]×{1}[0,2]\times\{1\} of the boundary Γ2\Gamma_{2} and the part ({0}×[0,0.5])({2}×[0,0.5])(\{0\}\times[0,0.5])\cup(\{2\}\times[0,0.5]) is traction free. No body forces are assumed to act on the body during the process. The body is in frictionless contact with an obstacle on the part Γ3=[0,2]×{0}\Gamma_{3}=[0,2]\times\{0\} of the boundary. For the discretization we use 7935 elastic finite elements and 129 contact elements.

The total number of degrees of freedom is equal to 8326 and we take a time step kk equal to 0.01s0.01s.

We model the material’s behavior with a constitutive law of the form (1.1) in which elasticity tensor \mathcal{E} satisfies

(𝝉)αβ=Eκ1κ2(τ11+τ22)δαβ+E1+κταβ,1α,β2(\mathcal{E}\boldsymbol{\tau})_{\alpha\beta}=\frac{E\kappa}{1-\kappa^{2}}\left(\tau_{11}+\tau_{22}\right)\delta_{\alpha\beta}+\frac{E}{1+\kappa}\tau_{\alpha\beta},\quad 1\leq\alpha,\beta\leq 2 (7.1)

where EE is the Young modulus, κ\kappa the Poisson ratio of the material and δαβ\delta_{\alpha\beta} denotes the Kronecker symbol.

Moreover, in order to facilitate the numerical implementation, we assume that 𝒢(𝝈,𝜺(𝒖))=𝒞𝜺(𝒖)\mathcal{G}(\boldsymbol{\sigma},\boldsymbol{\varepsilon}(\boldsymbol{u}))=\mathcal{C}\boldsymbol{\varepsilon}(\boldsymbol{u}), where the tensor 𝒞\mathcal{C} satisfies

(𝒞𝝉)αβ=γ1(τ11+τ22)δαβ+γ2ταβ,1α,β2(\mathcal{C}\boldsymbol{\tau})_{\alpha\beta}=\gamma_{1}\left(\tau_{11}+\tau_{22}\right)\delta_{\alpha\beta}+\gamma_{2}\tau_{\alpha\beta},\quad 1\leq\alpha,\beta\leq 2 (7.2)

For the computation below we use the following data:

t[0,T],T=1s,k=0.01s,N=100,\displaystyle t\in[0,T],\quad T=1s,\quad k=01s,\quad N=00,
E=10000N/m2,κ=0.3,γ1=1N/m2,γ2=2N/m2,\displaystyle E=0000N/m^{2},\quad\kappa=3,\quad\gamma_{1}=1N/m^{2},\quad\gamma_{2}=2N/m^{2},
𝒇0=(0,0)N/m2,\displaystyle\boldsymbol{f}_{0}=(0,0)N/m^{2},
𝒇2={(0,0)N/m on ({0}×[0,0.5])({2}×[0,0.5]),(2000,0)N/m on [0,2]×{1},\displaystyle\boldsymbol{f}_{2}=\left\{\begin{array}[]{rr}(0,0)N/m&\text{ on }(\{0\}\times[0,0.5])\cup(\{2\}\times[0,0.5]),\\ (-2000,0)N/m&\text{ on }[0,2]\times\{1\},\end{array}\right.
p(r)=cνr+,cν=200,g=0.05m,\displaystyle p(r)=c_{\nu}r^{+},\quad c_{\nu}=00,\quad g=05m,
b(r)=cν,q(r)=r+,pμ(r)=1μq(r)+p(g),1μ=200,\displaystyle b(r)=c_{\nu},\quad q(r)=r^{+},\quad p_{\mu}(r)=\frac{1}{\mu}q(r)+p(g),\quad\frac{1}{\mu}=00,

Numerical results. The main purpose of this part consists to present a numerical validation of the theoretical convergence result obtained in Theorem 5.1. Our results are presented in Figures 2-6 and are described in what follows.

First, the deformed configuration as well as the contact interface forces at t=1st=1\mathrm{~s} are plotted in Figure 2, which corresponds to the numerical solution of problem 𝒫2V\mathcal{P}_{2}^{V}.
In order to compare the deformed mesh related to Problem 𝒫2V\mathcal{P}_{2}^{V} with those obtained for the numerical solution of problem 𝒫1μV\mathcal{P}_{1\mu}^{V}, we plotted in Figures 3 and 4 , respectively, the deformed configurations for the numerical solution of problems 𝒫1μV\mathcal{P}_{1\mu}^{V} with memory term (in which the function b=cνb=c_{\nu} ) and without memory term ( b=0b=0 ), respectively. Then, in Figures 3 and 4, we note that the penetration of the contact nodes is no longer restricted by unilateral constraint and exceed the limit g. Moreover, the absence of the memory term leads to larger penetrations in the foundation.

In Figure 5 we present the evolution of the convergence of the discrete solution of the problem 𝒫1μV\mathcal{P}_{1\mu}^{V} to the discrete solution of the problem 𝒫2V\mathcal{P}_{2}^{V} as the deformability of the foundation μ\mu tends to zero. More precisely, we plot 4 deformed meshes and the associated contact forces for 4 values of 1/μ1/\mu which represents here the stiffness

of the foundation after the limit gg is reached. One can see that for 1/μ=101/\mu=10 all the contact nodes are in strong penetration, whereas for 1/μ=100001/\mu=10000 two-third of the nodes slightly exceed the limit g=0.05mg=0.05m and will come into a unilateral contact.

For the convergence result, we denote by 𝒖μhk\boldsymbol{u}_{\mu}^{hk} and 𝒖hk\boldsymbol{u}^{hk} the discrete solution of the contact problems 𝒫1μV\mathcal{P}_{1\mu}^{V} and 𝒫2V\mathcal{P}_{2}^{V}, respectively, for a given μ>0\mu>0. The numerical estimations of the difference

𝒖μhk𝒖hkV+𝝈μhk𝝈hkQ\left\|\boldsymbol{u}_{\mu}^{hk}-\boldsymbol{u}^{hk}\right\|_{V}+\left\|\boldsymbol{\sigma}_{\mu}^{hk}-\boldsymbol{\sigma}^{hk}\right\|_{Q}

at the time t=1st=1s, for various values of the coefficient μ\mu, are presented in Figure 6. It results from here that this difference converges to zero as 1/μ1/\mu tends towards infinity. We conclude that our results in Figure 6 represent a numerical validation of the theoretical convergence result obtained in Theorem 5.1.

References

[1] P. Alart and A. Curnier, A mixed formulation for frictional contact problems prone to Newton like solution methods, Computer Methods in Applied Mechanics
and Engineering 92 (1991), 353-375.
[2] M. Barboteu, A. Matei, M. Sofonea, Analysis of quasistatic viscoplastic contact problems with normal compliance, Quarterly Journal of Mechanics and Applied Mathematics, 65 (2012), 555-579.
[3] N. Cristescu and I. Suliciu, Viscoplasticity, Martinus Nijhoff Publishers, Editura Tehnica, Bucharest, 1982.
[4] W. Han and M. Sofonea, Quasistatic Contact Problems in Viscoelasticity and Viscoplasticity, Studies in Advanced Mathematics 30, American Mathematical Society-International Press, 2002.
[5] I.R. Ionescu and M. Sofonea, Functional and Numerical Methods in Viscoplasticity, Oxford University Press, Oxford, 1993.
[6] H. B. Khenous, P. Laborde and Y. Renard, On the discretization of contact problems in elastodynamics, Lecture Notes in Applied Computational Mechanics 27 (2006), 31-38.
[7] H. B. Khenous, J. Pommier, Y. Renard, Hybrid discretization of the Signorini problem with Coulomb friction. Theoretical aspects and comparison of some numerical solvers, Applied Numerical Mathematics 56 (2006), 163-192.
[8] A. Klarbring, A. Mikelič and M. Shillor, Frictional contact problems with normal compliance, Int. J. Engng. Sci. 26 (1988), 811-832.
[9] A. Klarbring, A. Mikelič and M. Shillor, On friction problems with normal compliance, Nonlinear Analysis 13 (1989), 935-955.
[10] T. Laursen, Computational Contact and Impact Mechanics, Springer, Berlin, 2002.
[11] J. A. C. Martins and J. T. Oden, Existence and uniqueness results for dynamic contact problems with nonlinear normal and friction interface laws, Nonlinear Analysis : Theory, Methods and Applications 11 (1987), 407-428.
[12] J.T. Oden and J.A.C. Martins, Models and computational methods for dynamic friction phenomena, Computer Methods in Applied Mechanics and Engineering 52 (1985), 527-634.
[13] M. Shillor, M. Sofonea and J.J. Telega, Models and Analysis of Quasistatic Contact, Lecture Notes in Physics, 655, Springer, Berlin, 2004.
[14] M. Sofonea, C. Avramescu and A. Matei, A Fixed point result with applications in the study of viscoplastic frictionless contact problems, Communications on Pure and Applied Analysis 7 (2008), 645-658.
[15] M. Sofonea and A. Matei, History-dependent quasivariational inequalities arising in Contact Mechanics, European Journal of Applied Mathematics, 22 (2011), 471-491.
[16] M. Sofonea and F. Pătrulescu, Analysis of a history-dependent frictionless contact problem, Mathematics and Mechanics of Solids, 18, no. 4 (2013), 409-430.
[17] P. Wriggers, Computational Contact Mechanics, Wiley, Chichester, 2002.

2014

Related Posts