Ghostscript wrapper for D:\Digitalizacja\MTS85_t23z1_4_PDF_artyku³y\mts85_t23z1.pdf M E C H AN I KA TEORETYCZNA I STOSOWANA 1, 23 (1985) PH YSICAL CORRECTN ESS OF COSSERAT- TYPE MOD ELS OF H ON EYCOMB G RID PLATES TOM ASZ LE WI Ń SKI Politechnika W arszawska 1. Introduction Formulated via phenomenological considerations micropolar theory of elasticity (cf. [1]) can be applied for continuum description of dense, regular grids. This has been noted and applied by Woź niak et al. in numerous papers pertaining to lattice- type shells and plates, cf. [2]. Wozniak's approach is based on variational methods; an adequacy ofthe proposed differential model of a body with additional degrees of freedom is „ a priori" assumed. In the first- order approximation (see Sec. I ll in [2]) an in- plane plate motion is described by means of three independent functions approximating displacements and rotations of nodes. The governing equations of the theory have a similar form to those of the plane- stress theory of micropolar media. Therefore the Wozniak's approach is a heuristic one thus the recalled above procedure does not allow us to perform a physical correctness analysis of the model provided appropriate numerical tests are not carried out. In the case of simple layout grid plates (in which neighbourhoods of all nodes are con- gruent) Wozniak's algorithm leads to one set of effective constants describing elastic pro- perties of the structure. However in the case of complex layout grids, one can derive at least two sets of B and C tensors (cf. [3]). In the present paper an attempt is made to eluci- date questions concerning the mentioned difficulties in formulation of Cosserat- type mo- dels of complex geometry lattice plates. An attention will be focused on hexagonal grids belonging to the class of complex layout structures. I n order to have a new look at Wozniak's continuum models results of the work [4] (pertaining to differential models due to Rogula- Kunin's approach) are applied. It is easy to note that Cosserat- type equilibrium equations expressed in terms of dis- placements cannot be obtained by asymptotic method, e.g. by formal simplifications (neglect of terms of higher order) of equilibrium equations found in [4] by Rogula- Kunin's procedure; thus a simple correspondence between the latter and Woź niak's- type equa- tions is not valid. This fact is obvious since differential models derived in [4] (just contrary to theories outlined in [3]) do n ot satisfy stability conditions. In Sec. 5 a simple modification of the second- order differential approximation (obtai- ned in [4]) will be proposed. The aim of the procedures is to formulate a well- established 54 T. LEWIŃ SKI Cosserat- type equations, so called x- versions. I n Sec. 6 an attem pt is un dertaken to examine a range of applicability of the latter versions as well as of two varian ts resulting from Wozniak's concept, see [3]. 2. Basic assumptions A subject of our considerations is a plane- stress statical problem of a honeycom b grid com posed of bars whose axes constitute hexagons of sides being equal to /, cf. F ig. 1, I n order to make final results as clear as possible the bars are assumed to be prismatic (thus their heights h are constant while their depth is of un it dimension) an d m ade of an isotropic, elastic material whose properties are characterised by Youn g m odulus E and F ig. 1 P oisson's ratio v. External loads (subjected t o lattice join ts only) are assumed to yield plane- stress plate response hence the external forces are supposed to be subjected in- plane while moments should be n orm al to the mid- surface of the grid. A slenderness ratio of lattice rods is defined by, cf. [3] r] = t2lh2. A param eter Q defined as a quotient of opening's diam eter to a spacing between centres of neighbouring openings, see [3], reads 6 = f|/ 3»?- l)/ l/ 3»7 (2.1) As it has been pointed out in [3] EJjP = E/ I2r)3'2 where / den otes a m om en t of inertia of a constituent bar's cross section. Lattice rods are assumed t o be sufficiently slender so as to known m ethods of th e theory of structures could be applied. Analogously t o [3, 4] lattice nodes are divided in to two families of m ain an d in term ediate ones, see F ig. 1. D isplacements and rotation s of m ain nodes are approxim ated by con tin uous functions if(xP) an d
a/3 and m" are dependent upon the loads subjected to intermediate nodes; their definitions are given in [3], [6]. 3.2. Strain energy as a positive definite function. Strain energy of the structure is postive definite provided, [3] fi > 0 , a > 0 , fi + X>Q, C > 0 , B 2 < CpL. (3,5) 4. Unstable quasicontinuum micropolartype equilibrium equations 4.1. Derivation of governing equations. Focus attention on Cosserattype equations (3.1). On noting that B ~ /, C ~ I2 one can make following remarks: a) two first equilibrium equations involve zeroorder (with respect to powers of /) terms of displacementtype and a firstorder term relevant to nodal rotations'; b) the last equation involves firstorder terms of displacementtype arid a secondorder term depending upon the rotations. The procedures put forward in [3] did not explain why: a) two first equations do not involve firstorder terms being dependent on displacements. Does it yield from approximations only or result from specific properties of the hexagonal grid? 56 T. LEWIŃ S KI b) the last equation does n ot involve secon d- order term s of displacem ent- type; also herein the same question arises. Answers to the above questions are supplied by quasicon tin uum con sideration s, [4]. C om pare consequent second- order equation s ((6.1) in [4]) with E qs. (3.1) of Cosserat- type theory. F irst of all it can be stated t h at m oduli A, / A an d a (being involved in bo t h compared sets of equations) have been identically defined (cf. ( 3. 8) 1 _ 3 in [3] an d (6.2) x_ 3 in[4]). Thus both approaches (based on Wozn iak's [3] an d R ogula- Kun in 's, [4], concepts) result in th e same definitions of elastic m oduli X, fi an d a, being depen den t u p o n slenderness ratio of bars only and th us being in depen den t of th e in tern ode spacing /. N ote t h at equations similar to th e Cosserat- type (3.1) can be derived from th e second- order Eqs. ((6.1) in [4]) provided in th e latters all th e term s involving derivatives u p to the second order are retain ed: 1 - 0, - 0, (4.1) [ ( ) ] [ = 0, where B° = /?/, C ° = yl2. (4.2) M oduli /3 an d y have been defined by E qs. ( 6. 2) 4, 6 in [4]. H en ce we obtain V ( , ^ T j^ rj+ f ( 3 V - T j^ rj+1 I" I 2 ' r, 1 EJ_ 1+ iyJ / ' Quantities p" and Y3 are equal to functions 'pa an d 'Y3 em ployed in [4]. I n a zero- order approxim ation we have j>« — pa | pa | ty _ e^ggY3, y 3 = y 3 | r]_~ *! Y3, (4 4) 2>r) + rj ff + 'ir) ' The derivation of Eqs. (4.1) (which will be called further quasicon tin uum micropolar- type equilibrium equations) violates accuracy principles form ulated in [4] where th e approxi- m ation procedure has been called consequent provided all th e term s proportion al to / ", p ^ s, (s is fixed) are being retained. I n th e n ext section an im proved accuracy analy- sis will be presented. Th e approxim ation s of govern in g equation s correspon d t o a certain form of density of strain energy of th e structure. Th us various approxim ation s of three equilibrium equations are reflected in the form of th e one scalar function which stan ds for the energy of the grid. This m eth od of error analysis is n o t new, th e idea was originated by Koitę r in th e paper [5] pertain in g to th e Kirchoff- Love shell th eory an d u p till n ow it is often applied to th e accuracy analysis of so called im proved theories describing plate and shell behaviour, see [6]. 4. 2. The micropolar- type approximation as a mode 1 of „moderate" rotations. The energy criterion, Strain energy of th e infinite hexagon al grid a m o u n t s to , P H YSI C AL C OR R E C TN E SS 57 E c = y P J £ K®«fa*wL «, /ff = 1, 2, 3 m, n where the h exagon 's area of a side / h as been den oted by P, P = l.5\ / J- 12\ wa, = «K, iv3 =<$ functions were given in [4]. On passing to fc- representation we arrive at (the proof, is om itted here) d 2 k = 4- n2- dk.d^ jP where th e dom ain of th e un it cell of the reciprocal lattice h as been denoted by P; a discrete F ourier tran sform of a discrete- argum ent function fm has been denoted by / ( k) . Certain approxim ations of of ^ ( k ) , cf. [4], result in differential models, particularly (as it will be shown further) — in C osserat- type m odels. I n troduce dimensionless variables \ F " = ua \ L , where [L ] = m. Let L = / / e. The density of strain energy of th e grid can be rearran ged to th e form (4.5) Let us define dim ensionless quan tities r an d 6 by m eans of the formulae T = / |k|, c o s© = fcj/ lkl, sin(9 = k 2 / \ k\ . Within th e fourth- order approxim ation (with respect to th e powers of £ a) the coefficients a a p can be expressed by m ean s of th e following equation s, cf. E qs. (6.1) in [4] T 2 F 3 T 2 T 2 1 - —£ O + a) + (A + i w- a ) c o s 2 6> - - jg- (/ « + a ) - — (A + ̂ - «) c o s 4 6> ,an £ i ) F 3 T 2 3 T 2 ( + ) - T - C O S4 0 + sin 4© + 2sin 2 6> c o s2 0) ] • a 33 = r2 8 + / T Ó . c o s0 ( c o s2 @- 3 si n2 0 ) L 58 T. LEWIŃ SKI «i8 = ~ TJ8COS2@- ^ ( 5 c o s 4 0 - 3 s i n4 0 - 6 c o s2 0 s i n 2 0 ) + - 2a sin 0+ - T - T2 sin 0| , T3 / 3 \ 1 - rB s'm20 + - jr- / 5sin20(cos26> + 3 si n 2 0 ) + ;' 2 a c o s@—~ r T 2 a c o s0 . 8 \ 4 / J The d m odulus has been defined by (6.2)5 in [4]. Let Xd m eans a wavelength of th e defor- m ation pattern in k direction; X d = 2^/ |k|. Thus the magnitude r is: r = 2nl/ X a = ~L —~ ~ 3.6275- — j / 3 A d / Id where b = / j/ 3 stands for th e spacing of m ain n odes. F r o m now on th e T quan tity will be supposed to be less th an one, T < 1; this yields X d > 3.6275 b. Th e param eters r and e are interrelated by means of the formula r/ e = 2TTL jX d . Let ę 0 and «0 stand for the absolute values of th e tran sform s ę an d m ax M" measured at the fixe9 node O of the grid.'The param eter
0 ~ — y>0 in t o (4.5) we h a ve
m= 0
Three first terms of this expansion correspond to th e approxim ation of e which yield a se-
cond- order model derived in [4], Sec. 6, E qs. (6.1). T h e presen ted derivation provides
a deeper insight into the assumptions (implicitly and tacitly assumed in [4]) which are
a basis of this model.
On neglecting all the term s except for th e two first ones th e first- order m odel, cf. [4],
Sec. 7, occurs.
The first term of the expansion is related to th e zero- order, asym ptotic or H orvay's
theory, see [4], Sec. 8.
P H YSI C AL C OR R EC TN ESS 59
1 A
Ad c) On substitutin g ć j>
0
~ W
o
in to (4.5) we obtain
e
co
m = 0
117/ |
By neglecting th e term s of higher order th an second (i.e. proportion al to r", p > 2) we
arrive at the expan sion which correspon ds to the micropolar- type approxim ation . Also
herein it can be poin ted out t h at the approach presented has revealed an d elucidated
assumptions which con stitute a basis of th e Cosserat- type models of fine hexagonal net-
works.
4.3. Stability. N ecessary an d sufficient stability condition s of Eqs. (4.1) (in the spirit of
Kunin, [7]) will be arrived at . Accordin g to this definition stability of equilibrium is satis-
fied provided an energy expressed in term s of th e wave vector components k
a
is positive
definite. Stability implies bo t h existence an d uniqueness of solutions.
The E qs. (4.1) are stable in the considered m ean in g when and only when the matrix
(2(i + X)y
2
+ (p + a)x
2
- (2B°xy + 2axi)Mx,y) -
B
0
(x
2
- y
2
)- 2ayi - (2B°x- y- 2u- x- i) C°(x
2
+y
2
) + 4a
is positive definite'for arbitrary x,yeR. I t can be shown (cf. [8]) th at the above condition
can be reduced t o t h e system of inequalities involving effective m oduli X, / x, a, B° an d C °
i t i > 0 Ar t > 0 A2 / a - ) -A > 0 A
(4.6)
> 221)
+ /
By virtue of the definitions (6.2) t _ 3 , [4], of X, fi an d a m oduli it can be stated th at a < p +
+ X. The last con dition ( 4.6) 3 reduces to th e form
C° > C ° = (B°)2IQ* + a). (4.7)
The m oduli X, ft an d « satisfy th e con dition s ( 4. 6) !_ 3 whereas th e inequality (4.7) is n ot
fulfilled for real grids. Therefore E qs. (4.1) obtain ed by formal (allthough justified in the
previous section) sim plification s of th e second- order Eqs. (6.1), [4], are unstable in th e
meaning of K un in .
4.4. Strong ellipticlty. Stron g ellipticity of a partial- differential equation system implies
(see [9]) the solution s featured by th e properties similar to those kn own from a classical
theory of well- established boun dary value elliptic problem s involving a one function to be
sought. If boun dary con dition s are admissible th e stron g ellipticity suffices for existence,
uniqueness and con tin uous depen den ce the solution u p o n th e boun dary conditions.
Consider a correctly supported hexagon al grid plate. Solutions are un ique and always
exist as it clearly follows from th e theory of structures. This con tin uum theories ough t to
ensure (apart from specific cases which are n ot dealt with here) the solutions t o be unique
that holds good provided th e m oduli X, p, x, B° an d C ° satosfy the stron g ellipticity condi-
tion .
60 T. LEWIŃ SKI
The set of Eqs. (4.1) is strongly elliptic when an d only when th e m atrix
:
2
+ (fj. + ̂ 25.
One can require th e C(°K) m odulus to satisfy inequality (3.5) 3 resulting from positive de-
term ination of strain energy expressed in terms of strain com pon en ts y
a/ !
an d x
a
.
Cc°j > Cli, = (B°)
2
/
F
(5.10)
that decreases the upper boun d of «: x e [0, x,,) where ^ < «s . The definition of xd ex-
pressed in terms of r\ will not be reported here.
The upper bound of C°
xy
: C°
x)
< C°
0)
does n ot follow from physical consideration s but
from the condition of positivity of weighted coefficients in ( 5. 6) 2 . Th us th e modulus
C°
x)
can vary in the limits ,
Cp°.d. < C{% < C °0 ) or C? < Cgo > Q °0 ) . (5.11)
5.3. K- representation interpretation of the proposed „stabilisation procedure". Remarks on the range of
applicability of the micropolar- type «- models. T h e p r o p o se d m o d ific a t io n of i f 3 3 o p e r a t o r c an
be interpreted as an approxim ation of the function P~ x. $ 3 3 ( k ) defined by Eq. (3.5) in [4],
Accuracy analysis of this approxim ation is outlined below.
The function P ~ 1 0 3 3 ( k ) can be expressed by the form ula
^ ^ 3 k t J ) - 1] (5.12)
where the relations
$ = "4a, - 4 . 5 ^ ' |) = y
have been applied. The constants a an d y have been defined by E qs. ( 6. 2) 3 |f i in [4]. Com-
ponents of tj vectors are, see F ig. 1 in [4]
t , = ( - 0 . 5, - 1/ 3/ 2), t a = (0.5, - j/ 3/ 2), Jtm - d , 0) .
I t can be shown th at th e function $> 33(k) varies alm ost in depen den tly of the wave vector
direction provided r = / • |k| < 3, th us,
P H YSI C AL C OR R EC TN ESS 63
J ^ c o s( / j/ 3 (0, |k|) t , ) = 2- C O S( 1.5T) +
hence
= 4 a - 9 - y[ c o s( 1 . 5 r ) - l ],
The modification proposed in Sec. 5.1 correspon ds to the approxim ation
^ - ^ a s C k) = 4oc + y( J I ) /
2 |k|2 = 4a + y( K ) - r
2
where y( J t ) = C(°K)/ /
2, y(i> = y, Y(*i>y> if « s ( 0 , 1).
F ig. 3 displays variation s of the fu n c t io n /
9 a.OL- P
0 3 3 ( k )
and its approxim ation (Eq. 5.14)
1
(5.13)
(5.14)
(5.15)
(5.16)
The diagram s in F ig. 3, are foun d for r\ » rj = 50 (///i « 7) th us y/ a = 0.6275.
As it was possible t o suspect it can be stated t h at g( 1 ) curve yields the best approxima-
tion off. F or wave vectors satisfying the con dition r < 1.5 a relative error of approxima-
tion o f/ by g,
n
is less t h a n 10% an d less th an 2.5% provided x < 1. If an accuracy analysis
is confined to th e beh aviour of the function # 3 3 (k) on e can conclude th at unstable Cosserat-
- type m odel which em ploys t h e set of m oduli X, fi, a, B°, C°
u
approxim ates „ rotation al
waves" (i.e. two- variable con tin uous functions in terpolatin g rotation s of main nodes) of
lengths Ad = — > 2nl with some per cent errors only. C onsiderably worse result is yielded
from th e set of m oduli (X, fi, a, B°, C°
0)
). If r < 0.25 a relative error approxim ation of
/ b y g(o) is less t h a n 4.7%. T h e latter con dition m ean s t h at „ rotation al waves" of lengths
X
d
> 8nl X 23.14 / are adm itted with 5% error whereas th e wave pattern s of X
d
> 4 • n • I
are related t o 18% of error. Therefore accordin g to th e choice of the param eter x e (0, 1) the
„ rotation al wave p a t t er n s" of len gths X
d
> A(K) where Xw e (6.28 /, 25.14 /) are admissible.
9
6
7
6
5
3
2
1
J
/
/ /
so/ /
/ /
/
0.1 0
— — - ~
ł fl 2.5
—. —L^
ii
0 3.5
F ig. 3
64 T. LEWIŃ SKI
The above analysis is somewhat incomplete since our atten tion has been focused on the
function $ 3 3 ( k ) only. To make- up the considerations a complete analysis of approxima-
tions of all functions # a j 3 should be given.
M ention yet that the constant function h(r) = 4 supplies a better approxim ation of the
function / that g
ix)
functions provided ,
Y
w
/ y > 2. (5.17)
Therefore x parameter cannot be to sm all; if not the carried out modification of #>3 3 induces
the greater error th an a simple neglect of the term COV29? in ( 4. 1) 3. I t is easy to show that
lower boun ds of C °: C°,
d
. an d C ° do n ot satisfy (5.17), i.e. C°.
d
. > 2Cf1} = 2 • y • I
2
.
N evertheless it is m ore purposeful to retain the term C°V2q> in (4.1) 3 th an to neglect it
(and further to eliminate rotation s q> from Eqs, (4. l ) l f 2 , see [4], Sec. 6) in order to formulate
well- established stable theory. H owever the stability is achieved at the sacrifice of the ap-
proximation condition.
6. Comparison of Wozniak's and >c- raodels.
Further remarks on accuracy analysis
6.1. Governing equations. G overning equations (3.1) an d (4.1) have similar forms. As it
has been mentioned is Sec. 4 both systems of partial differential equations involve the same
moduli A, /« and a. Qualitative differences distinguish between two sets of m oduli B and C;
to this topic Sec. 6.4 will be devoted. Essential quan titative differences occur in functions
approximating effects of external loads. D efinitions (4.4) of/ ?" an d Y3 can n ot be rearranged
to the form (3.2), viz. Eqs. (3.2) do n ot involve fr bu t their derivatives only. The latter
fact can be treated as a shortcoming of the theory. Th e definitions of Y3 an d 'Y3 are also
different. I n the case of slender rods (Ijh > 7, r\ > 49, say) 'Y3 depends inconsiderably on
Y
3 and, if r\ -> oo we have 'Y3 -> Y3; on the other h an d in ^- models there is: Y3 - >• Y3-
*
—0.573 provided rj - » oo.
6.2. Existence and uniqueness of solutions. The existence and uniqueness of solutions are
ensured by:
a) stability (4.6) or strong ellipticity (4.8) conditions — in th e case of K- models; .
b) positive definiteness of strain energy expressed in terms of y and x tensors (3.5) —
in both approaches due to Wozniak's concept.
I t is worth stressing th at the latter condition is stronger t h an the former.
6.3. Boundary conditions. In Sec, 4 boun dary conditions of ^- versions have n ot been formu-
lated. H owever, since the governing equations of these models have similar form t o Woz-
niak's equations it seems reasonable t o subject th e solutions t o similar boun dary conditions,
cf. Eqs. (7.3) in [3].
6.4. Analysis of moduli B and C, I n the paper [3] (cf. Sec. 3 of th e present work) two ver-
sions of constitutive equations of Woź niak- type models resulting in different m oduli £
(J5V, B") an d C(CV, C A) have been proposed. C onsiderations based on th e quasicontinuum
PHYSICAL CORRECTNESS 65
approach (Sec. 4) yield the one definition (4.3) of B° and C°. In Sec. 5 the oneparameter
definition "(5.7) of the modulus C°x) has been derived.
Variations of moduli B(g), Q e [0.8, 1] (ratio Q being defined by (2.1)) are plotted in
Fig. 5 in the case of v = 0,3 (so that rj = r? + 3.12). The following inequalities hold true
B\Q) 0 and k t> 0, k 2 — 0
th e errors are equal t o th ose in duced by the I version : 2.2% for A,, > 8sr/ an d 8.5% if
la > Anl. I n th e case of k
x
= k
2
an approxim ation error of \ $
23
\ is equal to 16.5% provi-
ded X
d
> 8?r/ th us it is greater t h an in th e I version where this error is about 5%.
Therefore approxim ation errors of \ Anl correspon d t o 10% errors what is rather a small value if one
takes in to accoun t t h at a wavelength l
A
d min = Anl is relatively sh ort with respect to the
in tern ode distan ce. D isplacem en t waves of th e direction k = (k
lt
0) are related to the
errors of about 40%. Th us th e I I varian t is characterized by the lack of symmetry of
approxim ation s; direction s k
t
an d k
2
are n ot equivalent here.
7. Concluding remarks
Th ree C osserat- type m odels for fine hexagonal grids have been analysed:
i) warian t I (due t o Woź n iak an d Klem m ) involving m oduli X, JJL , a, B an d C .
ii) warian t I I in which m oduli X, fi, a, B an d C are em ployed ...,- .,,4,\ ,.„ • •
68 T. LEWIŃ SKI
iii) *- models with m oduli X, fi, a, B° an d C°
K)
, « e [ 0 , x
s
).
I n order t o estimate ranges of applicability of the m en tion ed models approxim ation
errors of $ 3 p an d $>p3, ft — 1, 2, 3, induced by each approach , have been examined.
Ad i) I t has been shown th at approxim ation errors of \ 4nl. Th is value of errors results from the,
approxim ation of all of th e functions &
k3
, k = 1, 2, 3
Ad ii) The considerable approxim ation errors pertain to the functions u(k) have been given. N evertheless the proposed accuracy analysis allows us to formulate
the following rem arks.
1. I t is impossible to suspect which of the versions considered (i + iii) induces the
greatest errors of evaluating displacements u1 an d w2 whereas it is almost apparen t that
the (ii) version should yield th e best evaluation of n o d al rotation s (p.
2. The unstable varian t x — 1 (iii) can be employed for exam in in g local effects resulting
for instance from an influence of con cen trated n odal loads.
3. The version (iii) allows us to consider an effect of changes of the param eter x on
solutions of boun dary value problem s. C om putation s perform ed for several values of
H e (0, xs) yield results which are divisible in to two groups (a) an d (b). Results of (a) type
are stable with respect t o variations of « whereas th e (b)- results do n ot satisfy th e latter
condition. Apart from this, a zero- order H orvay's asym ptotic m odel (see [4]) involving
displacements u1, u2 only, can be employed. Results of (a) type vary inconsiderably thus
only these results are reliable. N evertheless results of (b) type are of great interest, because
despite the fact th at their values are evaluated incorrectly, valuable qualitative information
is obtained. I n cases of simple states of stress, to (a) group displacements belong while
rotation s belong to (b) group.
The remarks formulated above are confirmed in [10] in th e specific case of infinitely
long grid strip of hexagonal structure subjected to lon gitudin al forces. Specifically, the
rem ark 1 occured to be accurate; allthough approxim ation s of n odal rotation s are charged
with errors, the I I version seems to induce the smallest ones. N onetheless it should be
poin ted out here th at the Cosserat- type models provide th e description qualitatively correct,
particularly the altern ate vanishing variation of rotation s of m ain and interm ediate no-
des lieing along the lines perpendicular t o the strip's h orizon tal axis being successfully
predicted.
References
1. W. N OWACKI, T heory of unsymmetrical elasticity (in Polish), PWN , Warsaw 1981.
2. Cz. WoŻ NiAK, L attice- type shells and plates (in Polish), PWN , Warsaw 1970.
P H YSI C AL CORRECTN ESS 69
3. T. LEWIŃ SKI, T W O versions of W oiniak's continuum model of hexagonal- type grid plates, Mech. Teoret.
' Stos., 23, 3/4, 1984.
4. T. LEWIŃ SKI, Differential models of hexagonal- type grid plates, Mech. Teoret. Stos., 22, 3/ 4, 1984.
5. W. T. KOITER, A consistent first approximation in the general theory of thin elastic shells, Proc. IU TAM
Symp. Theory of thin shells, D elft 1959; N orth H olland P. C o., Amsterdam 1960, pp. 12 - 33.
6. W. PIETRASZKIEWICZ, Finite rotations and L agrangean description in the non- linear theory of shells, PWN
Warsaw- Poznań 1979.
7. I . A. K.UNIN, T heory of elastic media with microstructure (in Russian) N auka, Moscow 1975.
8. T. LEWIŃ SKI, Continuum models of lattice- type hexagonal plates, (in Polish), D octor's thesis, Technical
U niversity of Warsaw 1983.
9. M. T. VISIK, On strongly elliptic systems of differential equations (in Russian), M at. Sb., 29, (71), 3,1951.
10. T. LEWIŃ SKI, T he state of extension of honeycomb grid strip, Mech. Teoret. Stos., 23, 2,1985.
P e 3 w M e
AH AJI H 3