3.1 Step 1. Consistent asymptotic modelling
The fundamental concepts of the consistent asymptotic procedure are those of an averaging operation and fluctuation shape functions. In what follows, the above-mentioned concepts will be specified with respect to one-dimensional region \(\Omega \equiv (0,L_1 )\) defined in this contribution.
Let
f(
x) be a function defined in
\({\bar{\Omega }}\equiv [0,L_1 ]\), which is integrable and bounded in every cell
\(\Delta (x)\),
\(x\in \Omega _\Delta \).
The averaging operation of \(f(\cdot )\) is defined by
$$\begin{aligned} <f>(x)\equiv \frac{1}{\left| \Delta \right| }\;\mathop {\int }\limits _{\Delta (x)} {f(z)\mathrm{d}z,} \quad z\in \Delta (x),\quad x\in \Omega _\Delta . \end{aligned}$$
(8)
It can be seen that if
\(f(\cdot )\) is
\(\Delta \)-periodic, then
\(<f>\) is constant.
Denote by \(\partial _1^k \) the kth derivative of function defined in \(\Omega \). Let h(x) be a \(\uplambda \)-periodic, highly oscillating function defined in \({\bar{\Omega }}=[0,L_1 ]\), which is continuous together with derivatives \(\partial _1^k h,\; k=1,\ldots ,R-1,\) and has a continuous or piecewise continuous bounded derivative \(\partial _1^R h\). Function \(h(\cdot )\) will be called the fluctuation shape function of the Rth kind, \(h(\cdot )\in FS^{R}(\Omega ,\Delta )\), if it satisfies conditions: \(h\in O(\uplambda ^{R}),\, \partial _1^k h\in O(\uplambda ^{R-k}),\, k=1,2,\ldots ,R, \quad <\upmu h>=0\), where \(\upmu (x)\) is a shell mass density. Nonnegative integer R is assumed to be specified in every problem under consideration.
The first step of the combined modelling is based on
the consistent asymptotic averaging of lagrangian (
6). To this end, we shall restrict considerations to displacement fields
\(u_\upalpha =u_\upalpha (z,\upxi ,t)\),
\(w=w(z,\upxi ,t)\) defined in
\(\Delta (x)\times \Xi \times I\),
\(z\equiv z^{1}\in \Delta (x)\),
\(x\in \Omega _\Delta \),
\((\upxi ,t)\in \Xi \times I\). Then, we replace
\(u_\upalpha (z,\upxi ,t)\),
\(w(z,\upxi ,t)\) by families of displacements
\(u_{\upvarepsilon \upalpha } (z,\upxi ,t)\equiv u_\upalpha (z/\upvarepsilon ,\upxi ,t)\),
\(w_\upvarepsilon (z,\upxi ,t)\equiv w(z/\upvarepsilon ,\upxi ,t)\), where
\(\upvarepsilon =1/m,\, m=1,2,\ldots ,\) (
\(\upvarepsilon \)is a small parameter),
\(z\in \Delta _\upvarepsilon (x), \quad \Delta _\upvarepsilon \equiv (-\upvarepsilon \uplambda /2,\upvarepsilon \uplambda /2)\) (
scaled cell),
\(\Delta _\upvarepsilon (x)\equiv x+\Delta _\upvarepsilon ,\, x\in \Omega _{\Delta _\upvarepsilon } \) (
scaled cell with a centre at \(x\in \Omega _{\Delta _\upvarepsilon } )\).
We introduce
the consistent asymptotic decomposition of families of displacements
\(u_{\upvarepsilon \upalpha } (z,\upxi ,t)\),
\(w_\upvarepsilon (z,\upxi ,t)\),
\((z,\upxi ,t)\in \Delta _\upvarepsilon \times \Xi \times I\)$$\begin{aligned} \begin{array}{l} u_{\upvarepsilon \upalpha } (z,\upxi ,t)\equiv u_\upalpha (z/\upvarepsilon ,\upxi ,t)=u_\upalpha ^0 (z,\upxi ,t)+\upvarepsilon h_\upvarepsilon (z)U_\upalpha (z,\upxi ,t), \\ w_\upvarepsilon (z,\upxi ,t)\equiv w(z/\upvarepsilon ,\upxi ,t)=w^{0}(z,\upxi ,t)+\upvarepsilon ^{2}g_\upvarepsilon (z)W(z,\upxi ,t). \\ \end{array} \end{aligned}$$
(9)
Unknown functions
\(u_\upalpha ^0 ,U_\upalpha \) in (
9) are assumed to be continuous and bounded in
\(\Omega \) together with their first derivatives. Unknown functions
\(w^{0},W\) in (
9) are assumed to be continuous and bounded in
\(\Omega \) together with their derivatives up to the second order. Unknowns
\(u_\upalpha ^0 ,w^{0}\)and
\(U_\upalpha ^ ,W\) are called
macrodisplacements and
fluctuation amplitudes, respectively. They are independent of
\(\upvarepsilon \). This is the main difference between the asymptotic approach under consideration and approach which is used in the known homogenization theory, cf. Bensoussan et al. [
20], Jikov et al. [
21].
By
\(h_\upvarepsilon (z)\equiv h(z/\upvarepsilon )\in FS^{1}(\Omega ,\Delta )\) and
\(g_\upvarepsilon (z)\equiv g(z/\upvarepsilon )\in FS^{2}(\Omega ,\Delta )\) in (
9) are denoted
\(\uplambda \)-periodic highly oscillating
fluctuation shape functions depending on
\(\upvarepsilon \). The
fluctuation shape functions are assumed to be known in every problem under consideration. They have to satisfy conditions:
\(h_\upvarepsilon \in O(\upvarepsilon \uplambda )\),
\(\uplambda \partial _1 h_\upvarepsilon \in O(\upvarepsilon \uplambda )\),
\(g_\upvarepsilon \in O((\upvarepsilon \uplambda )^{2})\),
\(\uplambda \partial _1 g_\upvarepsilon \in O((\upvarepsilon \uplambda )^{2}), \quad \uplambda ^{2}\partial _{11} g_\upvarepsilon \in O((\upvarepsilon \uplambda )^{2})\),
\(<\upmu h_\upvarepsilon>=<\upmu g_\upvarepsilon >0\). It has to be emphasized that
\(\partial _1 h_\upvarepsilon (z)\equiv \frac{1}{\upvarepsilon }\partial _1 h(z/\upvarepsilon ), \quad \partial _1 g_\upvarepsilon (z)\equiv \frac{1}{\upvarepsilon }\partial _1 g(z/\upvarepsilon ), \quad \partial _{11} g_\upvarepsilon (z)\equiv \frac{1}{\upvarepsilon ^{2}}\partial _{11} g(z/\upvarepsilon )\).
We substitute the right-hand sides of (
9) into (
6) and take into account that under limit passage
\(\upvarepsilon \rightarrow 0\), terms depending on
\(\upvarepsilon \) can be neglected and every continuous and bounded function of argument
\(z\in \Delta _\upvarepsilon (x)\) tends to function of argument
\(x\in \bar{{\Omega }}\). Moreover, if
\(\upvarepsilon \rightarrow 0\) then by means of a property of the mean value, cf. [
21], the obtained result tends weakly to the function being
the averaged form of starting lagrangian (
6)
under consistent asymptotic decomposition (
9). Then, applying the principle of stationary action we obtain
governing equations of the consistent asymptoticmodel for the unperiodic shells under consideration. These equations consist of partial differential equations for macrodisplacements
\(u_\upalpha ^0 , w^{0}\) coupled with linear algebraic equations for fluctuation amplitudes
\(U_\upalpha , W\). After eliminating fluctuation amplitudes from the governing equations by means of
$$\begin{aligned} \begin{array}{l} U_\upgamma = -(G^{-1})_{\upgamma \upeta } \;[<\partial _1 hD^{1\upeta \upmu \upvartheta }>\;\partial _\upvartheta u_\upmu ^0 +r^{-1}<\partial _1 hD^{1\upeta 11}>w^{0}],\; \\ W=-E^{-1}<\partial _{11} gB^{11\upgamma \updelta }>\partial _{\upgamma \updelta } w^{0} , \\ \end{array} \end{aligned}$$
(10)
where
\(G_{\upalpha \upgamma } =<D^{\upalpha 1\upgamma 1}(\partial _1 h)^{2}>\),
\(E=<B^{1111}(\partial _{11} g)^{2}>\), we arrive finally at
the asymptotic model equations expressed only in macrodisplacements \(u_\upalpha ^0 , w^{0}\)$$\begin{aligned} \begin{array}{l} D_h^{\upalpha \upbeta \upgamma \updelta } \partial _{\upbeta \updelta } u_\upgamma ^0 +r^{-1}D_h^{\upalpha \upbeta 11} \partial _\upbeta w^{0}-<\upmu>a^{\upalpha \upbeta }\ddot{u}_\upbeta ^0 =0, \\ B_g^{\upalpha \upbeta \upgamma \updelta } \partial _{\upalpha \upbeta \upgamma \updelta } w^{0}+r^{-1}D_h^{11\upgamma \updelta } \partial _\updelta u_\upgamma ^0 +r^{-2}D_h^{1111} w^{0}+<\upmu >\ddot{w}^{0}=0, \\ \end{array} \end{aligned}$$
(11)
where
$$\begin{aligned} \begin{array}{l} D_h^{\upalpha \upbeta \upgamma \updelta } \equiv<D^{\upalpha \upbeta \upgamma \updelta }>-<D^{\upalpha \upbeta \upeta 1}\;\partial _1 h>(G^{-1})_{\upeta \upzeta }<\partial _1 hD^{1\upzeta \upgamma \updelta }>, \\ B_g^{\upalpha \upbeta \upgamma \updelta } \equiv<B^{\upalpha \upbeta \upgamma \updelta }>-<B^{\upalpha \upbeta 11}\partial _{11} g>E^{-1}\;<\partial _{11} gB^{11\upgamma \updelta }> . \\ \end{array} \end{aligned}$$
(12)
Since displacement fields
\(u_\upalpha (x,\upxi ,t),w(x,\upxi ,t)\) have to be uniquely defined in
\(\Omega \times \Xi \times \mathrm{I}\), we conclude that
\(u_\upalpha (x,\upxi ,t),w(x,\upxi ,t)\) have to take the form
$$\begin{aligned} \begin{array}{l} u_\upalpha (x,\upxi ,t)=u_\upalpha ^0 (x,\upxi ,t)+h(x)U_\upalpha (x,\upxi ,t),\; \\ w(x,\upxi ,t)=w^{0}(x,\upxi ,t)+g(x)W(x,\upxi ,t),\quad (x,\upxi ,t)\in \Omega \times \Xi \times I, \\ \end{array} \end{aligned}$$
(13)
with
\(U_\upalpha ,W\) given by (
10). Tensors
\(D_h^{\upalpha \upbeta \upgamma \updelta } \),
\(B_g^{\upalpha \upbeta \upgamma \updelta } \) given by (
12) are
tensors of effective elastic moduli for the considered composite uniperiodic shells.
In contrast to starting equations (
7) with discontinuous, highly oscillating and periodic coefficients, the asymptotic model equations (
11)
have coefficients constant but
independent of the microstructure size\(\uplambda \).
Hence, the above model is not able to describe the length-scale effect on the overall shell dynamics. That is why, the model derived in the first step of combined modelling is referred to as
the macroscopic model for the problem under consideration.
Unknown macrodisplacements \(u_\upalpha ^0 , w^{0}\) and fluctuation amplitudes\(U_\upalpha , W\) must be continuously bounded in \(\Omega \).
The resulting equations (
11) are uniquely determined by the postulated
a priori periodic
fluctuations shape functions,
\(h(x)\in FS^{1}(\Omega ,\Delta )\),
\(h\in O(\uplambda )\), and
\(g(x)\in FS^{2}(\Omega ,\Delta )\),
\(g\in O(\uplambda ^{2})\), representing oscillations inside a cell. These functions can be derived from the periodic discretization of the cell using, for example, the finite element method or obtained as exact or approximate solutions to certain periodic eigenvalue problems on the cell describing free periodic vibrations. If the fluctuation shape functions are not derived as solutions to periodic eigenvalue cell problems mentioned above, then
the effective moduli (
12)
of the shell are obtained without specification of the periodic cell problems. This situation is different from that occurring in the known asymptotic homogenization approach, cf., e.g. [
20], where
only solutions to the periodic cell problems make it possible to define the effective moduli of the structure under consideration.
In the first step of combined modelling, it is assumed that within the asymptotic model, solutions
\(u_\upalpha ^0 ,w^{0}\) to the problem under consideration are known. Hence, there are also known functions
$$\begin{aligned} \begin{array}{l} u_{0\upalpha } (x,\upxi ,t)=u_\upalpha ^0 (x,\upxi ,t)+h(x)U_\upalpha (x,\upxi ,t),\; \\ w_0 (x,\upxi ,t)=w^{0}(x,\upxi ,t)+g(x)W(x,\upxi ,t), \\ x\in \Omega ,\;\;(\upxi ,t)\in \Xi \times \mathrm{I}, \\ \end{array} \end{aligned}$$
(14)
where
\(U_\upalpha , W\) are given by means of (
10).
3.2 Step 2. Tolerance modelling
The second step of the combined modelling is based on
the tolerance modelling technique, cf [
7,
8].
The fundamental concepts of the tolerance modelling procedure under consideration are those of twotolerance relations between points and real numbers determined by tolerance parameters, slowly varying functions, tolerance-periodic functions, fluctuation shape functions and the averaging operation.
In what follows, some of the above-mentioned concepts and assumptions will be specified with respect to one-dimensional region \(\Omega \equiv (0,L_1 )\) defined in this contribution.
Let
F(
x) be a function defined in
\({\bar{\Omega }}=[0,L_1 ]\), which is continuous, bounded and differentiable in
\({\bar{\Omega }}\) together with their derivatives up to the
Rth order. Note that function
Fcan also depend on
\(\upxi \in {\bar{\Xi }}=[0,L_2 ]\) and time coordinate
t as parameters. Let
\(\updelta \equiv (\uplambda ,\updelta _0 ,\updelta _1 ,\ldots ,\updelta _R )\) be the set of tolerance parameters. The first of them is related to the distances between points in
\(\Omega \), the second one is related to the distances between values of function
\(F(\cdot )\) and the
kth one to the distances between values of the
kth derivative of
\(F(\cdot )\),
\(k=1,\ldots ,R\). A function
\(F(\cdot )\) is called
slowly varyingof the Rth kind with respect to cell
\(\Delta \) and tolerance parameters
\(\updelta \),
\(F\in SV_\updelta ^R (\Omega ,\Delta )\), if and only if the following two conditions are satisfied
$$\begin{aligned}&\displaystyle (\forall (x,y)\in \Omega ^{2})[(x\mathop \approx \limits ^\uplambda y)\Rightarrow F(x)\mathop \approx \limits ^{\updelta _0 } F(y)\;\;\hbox {and}\;\;\partial _1^k F(x)\mathop \approx \limits ^{\updelta _k } \partial _1^k F(y),\;\;\;k=1,2,\ldots ,R] , \end{aligned}$$
(15)
$$\begin{aligned}&\displaystyle (\forall x\in \Omega )[\uplambda \left| {\partial _1^k F(x)} \right| \mathop \approx \limits ^{\updelta _k } 0,\;\;k=1,2,\ldots ,R], \end{aligned}$$
(16)
where symbols “
\(\mathop \approx \limits ^\uplambda \)” and “
\(\mathop \approx \limits ^{\updelta _0 } \)”, “
\(\mathop \approx \limits ^{\updelta _k } \)” denote tolerance relations.
Roughly speaking, from (
15) and (
16) it follows that
slowly varying function\(F(\cdot )\) can be treated as constant on an arbitrary cell and that the products of derivatives of
slowly varying function in periodicity direction and
microstructure length parameter \(\uplambda \) are treated as negligibly small.
An integrable and bounded function f(x) defined in \({\bar{\Omega }}=[0,L_1]\), which can also depend on \({\upxi }\in {\bar{\Xi }}\) and time coordinate t as parameters, is called tolerance-periodic with respect to cell \(\Delta \) and tolerance parameters \(\updelta \equiv (\uplambda ,~\updelta _0 ), \)if for every \(x\in \Omega _\Delta \) there exist \(\Delta \)-periodic function \({\tilde{f}}(\cdot )\) such that \(f\left| {\Delta (x)\cap Dom f} \right. \) and \({\tilde{f}}\left| {\Delta (x)\cap Dom {\tilde{f}}} \right. \) are indiscernible in tolerance determined by \(\updelta \equiv (\uplambda ,\updelta _0 )\). Function \({\tilde{f}}\) is a \(\Delta \)-periodic approximation of f in \(\Delta (x)\). For function \(f(\cdot )\) being tolerance-periodic together with its derivatives up to the Rth order, we shall write \(f\in TP_\updelta ^R (\Omega ,\Delta )\), \(\updelta \equiv (\uplambda ,\updelta _0 ,\updelta _1 ,\ldots ,\updelta _R )\).
The concepts of
fluctuation shape functions and
averaging operation have been explained in Sect.
3.1.
The tolerance modelling is based on two assumptions. The first assumption is called the tolerance averaging approximation. The second one is termed the micro-macro decomposition.
Let
\(f(\cdot )\) be an integrable periodic function defined in
\({\bar{\Omega }}=[0,L_1 ]\) and let
\(F(\cdot )\in SV_\updelta ^1 (\Omega ,\Delta )\),
\(G(\cdot )\in SV_\updelta ^2 (\Omega ,\Delta )\),
\(h(\cdot )\in FS^{1}(\Omega ,\Delta )\),
\(g(\cdot )\in FS^{2}(\Omega ,\Delta )\).
The tolerance averaging approximation has the form$$\begin{aligned} \begin{array}{l}<f \partial _1^R F>(x)=<f>\partial _1^R F(x)+O(\updelta ),\quad R=0,1,\quad \;\;\;\;\partial _1^0 F\equiv F, \\<f \partial _1^R G>(x)=<f>\partial _1^R G(x)+O(\updelta ),\quad R=0,1,2,\quad \;\partial _1^0 G\equiv G, \\ \end{array} \end{aligned}$$
(17)
and
$$\begin{aligned} \begin{array}{l}<f \partial _1 (hF)>(x)=<f \partial _1 h>F(x)+O(\updelta ),\quad \\<f \partial _1 (gG)>(x)=<f \partial _1 g>G(x)+O(\updelta ), \\<f \partial _1^2 (gG)>(x)=<f \partial _1^2 g>G(x)+O(\updelta ), \\ \end{array} \end{aligned}$$
(18)
In the course of modelling, terms
\(O(\updelta )\) in (
17) and (
18) are neglected. Let us observe that the slowly varying functions can be regarded as invariant under averaging.
Approximations given above are applied in the modelling problems discussed in this contribution. For details, the reader is referred to [
5‐
8].
The second fundamental assumption, called the micro-macro decomposition, states that the displacements fields occurring in the starting lagrangian under consideration can be decomposed into unknown averaged (macroscopic) displacements being slowly varyingfunctions in \(x\in \Omega \) and highly oscillating fluctuations represented by the known highly oscillating \(\uplambda \)-periodicfluctuation shape functions multiplied by unknownfluctuation amplitudes (microscopic variables) slowly varying in x.
In the second step of combined modelling, we introduce
the extra micro-macro decomposition superimposed on the known solutions
\(u_{0\upalpha } ,w_0 \) obtained within the macroscopic model.
$$\begin{aligned} \begin{array}{l} u_{c\upalpha } (x,\upxi ,t)=u_{0\upalpha } (x,\upxi ,t)+c(x)Q_\upalpha (x,\upxi ,t), \\ w_b (x,\upxi ,t)=w_0 (x,\upxi ,t)+b(x)V(x,\upxi ,t), \\ \end{array} \end{aligned}$$
(19)
where
fluctuation (
microscopic)
amplitudes\(Q_\upalpha ,V\) are
the new slowly varying unknowns, i.e.
\(Q_\upalpha \in SV_\updelta ^1 (\Omega ,\Delta )\),
\(V\in SV_\updelta ^2 (\Omega ,\Delta )\). Functions
\(c(x)\in FS^{1}(\Omega ,\Delta )\) and
\(b(x)\in FS^{2}(\Omega ,\Delta )\) are
the new periodic, continuous and highly oscillating
fluctuation shape functions which are assumed to be known in every problem under consideration. These functions have to satisfy conditions:
\(c\in O(\uplambda )\),
\(\uplambda \partial _1 c\in O(\uplambda )\),
\(b\in O(\uplambda ^{2})\),
\(\uplambda \partial _1 b\in O(\uplambda ^{2}), \uplambda ^{2}\partial _{11} b\in O(\uplambda ^{2})\),
\(<\upmu c>=<\upmu b>=0\), where
\(\upmu (x)\) is the shell mass density.
We substitute the right-hand sides of (
19) into (
6). The resulting lagrangian is denoted by
\(L_{cb} \). Then, we average
\(L_{cb} \) over cell
\(\Delta \) using averaging formula (
8) and applying
the tolerance averaging approximation (
17), (
18). As a result, we obtain function
\(<L_{cb}>\) called
the tolerance averaging of starting lagrangian (
6)
in\(\Delta \)under superimposed decomposition (
19). Next, applying the principle of stationary action, under the extra approximation
\(1+\uplambda /r\approx 1\), we arrive at the system of Euler–Lagrange equations for
\(Q_\upalpha ,V\), which can be written in an explicit form as
$$\begin{aligned}&\underline{<D^{\upalpha 22\updelta }(c)^{2}>} \partial _{22} Q_\updelta -<D^{\upalpha 11\updelta }(\partial _1 c )^{2}>Q_\updelta -\underline{<\upmu (c)^{2}>} a^{\upalpha \upbeta }{\ddot{Q}}_\upbeta \nonumber \\&\quad =r^{-1}<D^{\upalpha 111}\partial _1 c w_0>+<D^{\upalpha \upbeta \upgamma 1}\partial _1 c \partial _\upbeta u_{0\upgamma } >, \end{aligned}$$
(20)
$$\begin{aligned}&\underline{<B^{2222}(b)^{2}>} \partial _{2222} V+[\underline{2<B^{1122}b\partial _{11} b>} -4\underline{<B^{1212}(\partial _1 b)^{2}> } ]\partial _{22} V \nonumber \\&\quad +<B^{1111}(\partial _{11} b)^{2}>V\underline{+<\upmu (b)^{2}> } \ddot{V}=-<B^{\upalpha \upbeta 11}\partial _{11} b \partial _{\upalpha \upbeta } w_0 >, \end{aligned}$$
(21)
Equations (
20) and (
21) together with
the micro-macro decomposition (
19) constitute
the superimposed microscopic model. Coefficients of the derived model equations are
constant, and some of them
depend on a cell size\(\uplambda \) (the underlined terms). The right-hand sides of (
20) and (
21) are known under assumption that
\(u_{0\upalpha } ,w_0 \) were determined in the first step of modelling. The basic unknowns
\(Q_\upalpha ,V\)of the model equations must be
the slowly varying functions in periodicity directions. This requirement can be verified only
a posteriori, and it determines the range of the physical applicability of the model. The boundary conditions for
\(Q_\upalpha ,V\) should be defined only on boundaries
\(\upxi =0\),
\(\upxi =L_2 \).
It can be shown that under assumption that fluctuation shape functions
h(
x),
g(
x) of macroscopic model coincide with fluctuation shape functions
c(
x),
b(
x) of microscopic model, we can obtain microscopic model equations (
20), (
21), in which
c(
x) and
b(
x) are replaced by
h(
x) and
g(
x), respectively, and in which the right-hand sides are equal to zero. Moreover, taking into account a symmetric form of tensor
\(D^{\upalpha \upbeta \upgamma \updelta }\) we arrive finally at three equations for unknown fluctuation amplitudes
\(Q_1 (x,\upxi ,t)\),
\(Q_2 (x,\upxi ,t)\) and
\(V(x,\upxi ,t)\), which are not conjugated with themselves
$$\begin{aligned}&\underline{<D^{1221}(h)^{2}>} \partial _{22} Q_1 -<D^{1111}(\partial _1 h )^{2}>Q_1 -\underline{<\upmu (h)^{2}>} {\ddot{Q}}_1 =0, \end{aligned}$$
(22)
$$\begin{aligned}&\underline{<D^{2222}(h)^{2}>} \partial _{22}Q_2 -<D^{2112}(\partial _1 h )^{2}>Q_2 -\underline{<\upmu (h)^{2}>}{\ddot{Q}}_2 =0, \end{aligned}$$
(23)
$$\begin{aligned}&\underline{<B^{2222}(g)^{2}>} \partial _{2222}V+[\underline{2<B^{1122}g\partial _{11} g>}-4\underline{<B^{1212}(\partial _1 g)^{2}> } ]\partial _{22} V \nonumber \\&\quad +<B^{1111}(\partial _{11} g)^{2}>V\underline{+<\upmu (g)^{2}> } {\ddot{V}}=0. \end{aligned}$$
(24)
Equations (
22)–(
24) are independent of the solutions
\(u_{0\upalpha } ,\;w_0 \)obtained in the framework of the macroscopic model. Hence, they
describe selected problems of the shell micro-dynamics (e.g. the free micro-vibrations, propagation of waves related to the micro-fluctuation amplitudes)
independently of the shell macro-dynamics. Moreover, micro-dynamic behaviour of the shell in the axial and circumferential directions can be analysed independently of its micro-dynamic behaviour in the direction normal to the shell midsurface.
Microscopic model equations (
22)–(
24) also describe certain
time-boundary and space-boundary phenomena strictly related to the specific form of initial and boundary conditions imposed on unknown fluctuation amplitudes
\(Q_\upalpha ,V\). That is why, these equations are referred to as
the boundary layer equations, where the term “boundary” is related both to time and space.
Since equations (
22)–(
24) are not conjugated with themselves, the micro-dynamic behaviour of the shells in the axial, circumferential and normal directions can be investigated independently of each other.
3.3 Combined asymptotic-tolerance model
Summarizing results obtained in Sects.
3.1 and
3.2, we conclude that
the combined asymptotic-tolerance model of selected dynamic problemsfor the uniperiodic shellsunder consideration presented here following Tomczyk [
8] is represented by:
(a)Macroscopic model defined by Eq. (
11) for
\(u_\upalpha ^0 ,w^{0}\) with expressions (
10) for
\(U_\upalpha ,W\), formulated by means of
theconsistent asymptotic modelling and being independent of the microstructure length. Unknowns of this model must be continuous and bounded functions in
x. It is assumed that in the framework of this model, the solutions (
14) to the problem under consideration are known.
(b)Superimposed microscopic model equations (
20), (
21) derived by means of
the tolerance (non-asymptotic) modelling and having constant coefficients depending also on a cell size
\(\uplambda \) (underlined terms). Microscopic model equations (
20), (
21) are coupled with the macroscopic model equations (
11) by means of the known solutions (
14) obtained in the framework of the asymptotic model. Unknown fluctuation amplitudes
\(Q_\upalpha ,\;V\) of the tolerance model must be
slowly varying functions in
x.
(c)Decomposition
$$\begin{aligned} \begin{array}{l} u_\upalpha (x,\upxi ,t)=u_\upalpha ^0 (x,\upxi ,t)+h(x)U_\upalpha (x,\upxi ,t)+c(x)Q_\upalpha (x,\upxi ,t),\; \\ w(x,\upxi ,t)=w^{0}(x,\upxi ,t)+g(x)W(x,\upxi ,t)+b(x)V(x,\upxi ,t), \\ x\in \Omega ,\;\;(\upxi ,t)\in \Xi \times \mathrm{I}, \\ \end{array} \end{aligned}$$
(25)
where functions
\(u_\upalpha ^0 ,U_\upalpha ,w^{0},W\) have to be obtained in the first step of combined modelling, i.e. in the framework of
theconsistent asymptotic modelling.
Coefficients of all equations derived in the framework of combined modelling are constant in contrast to coefficients in starting Eq. (
7) which are discontinuous, highly oscillating and periodic in
x. Moreover, some of them
depend on a cell size\(\uplambda \). Thus,
the combined model can be applied to the analysis of many phenomena caused by the length-scale effect.
Under special conditions imposed on the fluctuation shape functions, we can obtain microscopic model equations (
22)–(
24), which are independent of the solutions obtained in the framework of the macroscopic model. It means that
an important advantage of the combined model is that it makes it possible to separate the macroscopic description of some special dynamic problems from the microscopic description of these problems.
For details, the reader is referred to Tomczyk [
8].
It should be noted that the combined asymptotic-tolerance model of dynamic problems for cylindrical shells with periodic structure in both circumferential and axial directions (biperiodic shells) proposed in Tomczyk and Litawska [
11] cannot be applied for analysis of dynamic problems for uniperiodic shells considered here. Model presented in [
11] is derived in the framework of the extended version of the tolerance modelling technique based on a new notion of
weakly slowly varying function, cf. Tomczyk and Woźniak [
12]. For this function, restrictive condition (
16) and approximations (
18) do not hold. Moreover, in the non-asymptotic-tolerance approach, the uniperiodic shells are not special cases of shells with two-directional periodic structure.
It should be also noted that the combined asymptotic-tolerance models for functionally graded cylindrical shells are presented by Tomczyk and Szczerba in [
17,
18]. Coefficients of governing equations of these models are not constant. They are smooth and slowly varying either in circumferential direction [
17] or in the axial one [
18].
Some applications of micro-dynamic Eqs. (
22)–(
24) will be shown in the next section.