Professional Documents
Culture Documents
531-570, 1997
1997 Elsevier Science Ltd
Printed in Great Britain. All fights reserved
PII:S0734-743X(97)00016-X
0734-743X/97 $17.00+0.00
Pergamon
Summary--Experimental results are provided from a series of tests on the uniaxial dynamic crushing of cylindrical
specimens of five species of wood selected for the density range they cover and tested up to impact velocities of
approximately 300 ms -1. An account of the macro-deformation and micro-deformation modes resulting from quasistatic and dynamic uniaxial compression is given. Measurements of the force pulses generated by the impact of the
wood specimens on the end of a Hopkinson bar load cell show that significant enhancements of the initial crushing
strengths of the specimens occur under dynamic loading conditions. The deformation mechanisms of wood are
localised under quasi-static compression and under dynamic loading conditions they become even more localised
and propagate through the material as crushing wave fronts which have some of the characteristics of shock waves.
A simple shock model based upon a rate-independent, rigid, perfectly-plastic, locking (r-p-p-l) idealisation of the
stress-strain curves for wood is proposed to provide a fast order understanding of the dynamic response. This model
is particularly successful in predicting the dynamic enhancement of the crushing strength of specimens loaded
across the grain as confirmed by comparisons between the experimental data and theoretical results. It is less
successful for those compressed along the grain. The source of the discrepancy is discussed and explanations are
provided for the fairly constant crushing stress enhancement factor observed at low to moderate impact velocities,
for the high impact velocity at which shock-type response is initiated and for the existence of clearly delineated
crush fronts which characterise these specimens. 1997 Elsevier Science Ltd.
Keywords: cellular materials, wood, impact energy absorption, crushing strength, microinertia, shock waves.
NOTATION
Ao
CL
Cs
hcell
Lo
M
At*
mw
Sr
t
u
v
Vc
Vcrit
vl
Vo
Vs
x
Xo
7
OL
El
/91
el
* Corresponding author.
531
532
Do
P~
17
I-Ilk
O"
O"A
O-comp
O'cr(O)
crD(0)
O"R
O"T
Crys
initial density
cell wall material density, po/p~=relative density
specific energy (energy per unit mass) of specimen
minimum value of 17 to produce full locking of the specimen
stress behind crushing wave front at time t
axial static crushing stress
applied stress in axial direction corresponding to uniaxial yield in the cell walls
crushing stress for grain orientation at 0 to the direction of loading
initial dynamic crush stress for grain orientation at 0 to the direction of loading
radial static crushing stress
tangential static crushing stress
yield stress of cell wall material
1. INTRODUCTION
There have been numerous studies of cellular structures and cellular materials such as metal ring
systems, honeycombs, polymeric foams and wood, some of which have been reviewed in the context of
dynamic compression by Reid e t al. [ 1]. Many aspects of the behaviour of cellular solids are summarised
well in the book by Gibson and Ashby [2]. One reason for this interest is the current and potential use of
these materials for packaging, as impact energy absorbers and their use as core materials in lightweight
sandwich structures. Gibson and Ashby devote a chapter of their book [2] to the selection of materials
for low speed impact applications. Wood in particular, whilst being one of the oldest constructional
materials, has also been used as a protective material for high velocity impact events for centuries (see
Johnson [3,4]) and is often used as an impact energy absorbing material in containment structures
surrounding systems which may disintegrate or, for example, as packaging material in transportation
flasks for nuclear fuel. The response of wood at high rates of loading is therefore of some interest. There
have been few systematic studies of the behaviour of wood under impact loading conditions as noted by
Johnson [3]. This paper presents the results of one such study in which specimens of wood were
subjected to uniaxial strain under quasi-static and impact loading conditions.
An introduction to the study of wood under dynamic uniaxial crushing conditions and when penetrated
by flat-ended cylindrical projectile was given by Reid et al. [5], who provided experimental data from a
preliminary study of Yellow Pine and American Oak. The sensitivity of the crushing stresses of these
woods to loading rate was noted. For example at an impact velocity of around 100 ms -1 the crushing
strength (i.e. the stress at which inelastic deformation is initiated) of the two woods increased by factors
of between 2 and 4 depending upon the density of the wood and the orientation of the grain with respect
to the direction of loading.
In the present paper, more extensive impact test data than those given in [1,5] are provided on the
significant changes in the crushing stresses of five species of wood, Balsa, Yellow Pine, Redwood,
American Oak and Ekki as the impact velocity is increased. The woods tested had densities ranging
approximately from 260-1200 kg m 3 and were subjected to uniaxial compression under impact loading
conditions at impact velocities in the range 30-300 ms 1 approximately. Loading both along the grain
and across the grain is considered and the deformation mechanisms corresponding to each are
contrasted.
The main emphasis in the paper is placed on predicting the initial crushing strengths of the various
specimens tested in terms of the impact velocity, the density of the wood and its material properties.
Simple theoretical models are developed to describe the effects of phenomena which are governed by
inertial effects active at the scale of the cellular structure, i.e. microinertial effects. From a materials
science perspective, the models are based on the premise that wood is rate-insensitive, the explanation
for the enhancement of the crushing stresses being sought in the influence of these inertial effects rather
than in viscoelastic or viscoplastic effects. In this sense a similar approach is followed to that proposed
by Jahsman [6] in a very interesting, earlier study of the dynamic behaviour of syntactic foams.
The predictions of a simple shock model are presented in which the enhancement of the crushing
stress is attributed primarily to the propagation of a plane surface of stress discontinuity through the
cellular material. Despite its simplicity, it is shown that the model agrees well with experimental data for
533
crushing across the grain but that, for a range of velocities up to a certain value which depends on the
density of the wood, it underestimates the enhancement of the initial crushing stress along the grain. The
reasons for this are discussed and another microinertial effect is suggested as the source of the crushing
stress enhancement for wood loaded in this direction.
To illustrate the range of phenomena involved when wood is crushed, the quasi-static behaviour of the
uniaxial crushing of woods is briefly described in the next section. This also provides a reference point
for the discussion of the dynamic test data which follows.
2. QUASI-STATIC COMPRESSION
Typical stress-strain curves for uniaxial compression tests on a variety of woods are given by Gibson
and Ashby [2] summarising their earlier work with Easterling and Harryson [7] and with Maiti [8]. A
detailed account is presented of the deformation mechanisms involved when wood is compressed in its
three principal directions namely axial (along the grain), radial (across the grain transverse to the growth
rings) and tangential (across the grain parallel to the growth rings). Elastic deformation occurs by
distortions of the cell structure which are distributed throughout the volume of the specimen. Inelastic
deformation occurs as the result of the collapse and gross deformation of the cells in a manner
qualitatively similar to that observed in macroscopic cellular systems as discussed by several authors and
summarised in [1]. Inelastic deformation in wood tends to be a localised phenomenon, especially for
specimens loaded along the grain. Following the attainment of the initial crushing stress, the deformation
progresses through the growth and multiplication of crushing zones at an applied load which is close to
the initial crushing load.
Fig. 1 shows typical load-compression curves for cylindrical specimens of the woods tested in this
study. The specimens were nominally 75 mm in diameter and 75 mm high and were contained in a thickwalled cylindrical steel vessel of 75 mm internal diameter to impose uniaxial strain conditions. The
moisture content of the specimens were measured in the conventional manner using a moisture meter.
For the Balsa, Pine, Redwood and Oak specimens the moisture content was within the range of 9 to 12%.
The Ekki specimens had a high moisture content in the range 15-24%. Ekki is normally used in a marine
environment. (Note that, in this paper, the only detailed data presented for specimens of Ekki relate to
those loaded across the grain.) Free specimens tend to split and fracture (see Fig. 2), although their
behaviour up to and just beyond initial crushing is very close to that shown in Fig. 1 since the Poisson's
ratios for wood tend to be relatively small [2,9]. For specimens loaded along the grain, there is a distinct
load drop after the initiation of crushing followed by a load plateau for all of the woods. For specimens
loaded across the grain, crushing continues essentially at or slightly above the crush-initiation load,
increasing at an increasing rate (i.e. the curve is convex towards the displacement axis) with displacement. There is no strain-softening. The rate of increase of the load is significantly higher for woods with
higher initial densities such as Oak and Ekki. In all cases, the deformation reaches a stage where the
crushing of the individual cells becomes exhausted and further deformation is the result of solid phase
compression of the cell wall material. This gives a much stiffer response as the material 'locks up'.
The essential differences between loading along the grain and transverse to the grain lie in the
differences in the collapse mechanisms induced in the cellular structure of the wood. The response to
compression along the grain is extremely complex and the initiation of crushing can involve failure of
the pyramidal end caps of the cells for low density woods such as Balsa [7] or the onset of various
buckling failures such as concertina progressive axial crushing of the cells (fold-microbuckling, see
[2,9]) or Euler-type buckling which can lead to kink-band formation especially in the more dense woods
[9]. Fig. 2 shows an example of the type of deformation that occurs in free (i.e. without any lateral
constraint) specimens of high density (in this case Ekki) loaded along the grain. Compression across the
grain (despite the influence of the rays in enhancing the crush stress by approximately 40% when the
wood is crushed in the radial as opposed to the tangential direction) is somewhat more straightforward. It
has been described in terms of plastic bending collapse of a two-dimensional hexagonal cellular array
which provides a good approximation to the structure of many woods [7]. Under quasi-static loading
conditions the deformation also tends to be more dispersed through the volume of the specimen than that
produced by loading along the grain.
The consequences of these observations on the structural form of the cell structure and the
mechanisms of failure are that simple relationships can be formulated between the initial crushing
534
500
(a)
450
400
:}50
(~~0
a k
:~00
[
/
/
/
~.oO
,-.
0 ~
Redwood
O'
Pine
.-""
200
1.50
- ...............
"
~/" -..
100
5O
0
10
~30
:50
Displacement
40
50
60
(nam)
500
45O
t
400
Ekk~
350
z
300
i 90 Pine
t
250
9 ( '( a k
/,
200
~
''
/
1o(
/
"
1 O0
50
/
/
./
i
-90
~ Balsa
0
0
10
'~0
:30
Displacenlent
10
50
60
(ram)
Fig. 1. Quasi-static uniaxial load-displacement curves for a variety of woods: (a) along (0); (b) across (90 ) the grain
stresses and the relative density of the wood, i.e. the average density of the w o o d sample, P0, divided by
the density of the cell wall material, Ps. These are expressed as follows [2]:
era
150 (Po/Ps) MN m ~2
(1)
(2)
535
Fig. 2. Compressedfree Ekki specimens: (a) kink bands in 0 specimen; (b) shear failure in 90 specimen.
where the suffices A, R and T refer to the three principal directions of the wood structure, viz. axial,
radial and tangential. Ps is virtually the same for all woods and has a value of 1500 kg m -3 [2]. The
numerical factors are derived empirically whilst the indices arise naturally from the particular collapse
mechanisms considered in the structural models. The differences in compressive strength in the radial
and tangential directions are relatively small when compared with differences between the strength
along the grain and that across the grain. In the tests reported, the specimens loaded across the grain
were cut with the grain perpendicular to their axis (i.e. the direction of loading) but were generally
neither fully radial nor fully tangential specimens. Therefore, no distinction is made herein between the
tangential and radial directions. A single crushing stress transverse to the grain defined by crcr(90) equal
to O"R has been used. Specimens cut in this way are referred to as 90 specimens herein. The axial
crushing stress is referred to as trot(0), the angle simply indicating the direction of loading relative to the
grain direction and such specimens are referred to as 0 specimens. Comparisons between Eqns (1) and
(2) and the measured initial crushing stresses are shown in Table 1. The low values for the 90 Ekki
specimens are due to the very high moisture content (~20%) of the specimens tested. The 'free'
specimen had a similar moisture content to the specimens tested dynamically.
As noted above, the inelastic deformation tends to be localised in wood. It develops by the growth of
local crush zones and crush bands, often initiated at the ends of the specimen adjacent to the loading
platens. Overall deformation of the laterally constrained specimen occurs by the propagation of crush
536
Table 1. Uniaxial quasi-static parameters for wood specimens tested under laterally constrained conditions (uniaxial compression);
75 mm diameter 75 mm high
Wood and grain
orientation
Balsa O'
Balsa 90:
Y. Pine O
Y. Pine 9 0
Redwood 0 '
Redwood 90'
A. Oak 0:
A. Oak 9 0
Ekki 90'*
Ekki 90"
Initial density,
P0 (kg m 3)
Locking
strain, c I
Specific locking
energy (kJ kg i)
277
264
383
396
367
409
725
695
1051
t 183
27.0
1.6
43.3
5.l
43.0
10.5
75.0
12.3
24
18.2
27.7
2.2
38.3
4.9
36.7
5.2
72.5
15.0
34.4
43.5
0.68
0.65
0.64
0.60
0.65
0.58
0.33
0.37
63.4
10
62.2
15.2
59.0
16.3
30.4
16.0
-3.7
0.13
* Free specimen.
fronts, these being more diffuse for 90 specimens but very distinct for 0 specimens. For 0 specimens
the deformation occurs in 'compressible kink bands' as shown in Fig. 3. These are similar to the kink
bands produced in the compressive failure of unidirectional, fibre-reinforced polymer composites
[10,11] but the considerable compressibility arising from cellular collapse makes them distinctive. The
angle between the normal to the kink band and the direction of loading increases with the density of the
wood, the angle being almost zero for Balsa (giving the appearance of a 'plane' crush front), 35 for Oak
and approximately 45 for Ekki. This latter feature has been briefly discussed by Reid e t al. [1].
A consequence of the localised nature of the deformation in wood is that, whilst it is conventional to
convert load-displacement curves into stress-strain curves by defining stress as load divided by crosssectional area and strain as change in specimen height divided by the undeformed height, it must always
be borne in mind that in the inelastic range the strain field is highly non-uniform throughout the
specimen. However, the rapid increase in the slope of the load-displacement curves occurs when the
crushed zones have grown to encompass most of the initial volume of the material. At this point one can
logically define a critical strain, the locking strain, e~, which reflects the transition from the cell wall
collapse mechanism to solid phase compression of the majority of the cell wall material. Using any of
the conventional definitions of strain before locking occurs produces a strain measure which varies
according to the size of the specimen and is therefore not representative of the material. Adopting the
procedure described above allows the load-displacement curves to be converted into stress-strain curves
to give an approximate continuum representation for the uniaxial deformation modes of the wood
specimens. Fig. 4 gives the stress-strain curves corresponding to the load-displacement curves shown in
Fig. 1. When discussing the uniaxial crushing of cellular materials, the key properties are the initial
crushing stress, Crcr, and the locking strain, el, [2]. These are listed for the five woods in Table I together
with the values given by Eqns (1) and (2). The graphical construction used to define these characteristic
parameters from the stress-strain curves is also shown in Fig. 4. It is noted that the crush stress for the
Ekki sample is significantly less than the value given by Eqn (2). This is because of the high moisture
content.
Maiti e t al. [8] discussed the shapes of the stress-strain curves for a variety of cellular solids
including wood. In particular they produced the following equation to estimate el for a wood of
density Po,
Cl-
(~(PO/P~).
(3)
In Ref. [8] a value a = 2 is suggested. Results from more recent tests [9] indicate that this is satisfactory
for low density woods such as Balsa but it cannot be used for moderate and high density woods. For Pine
and Redwood, c~= t.35 is more appropriate and for Oak c~ is approximately equal to t .3. For the samples
of Ekki tested, the transition to locking is more gradual. However, defining locking as shown in Fig. 4
results in a value for e~ of 0.13 and a value for c~ of 1.1. These results simply demonstrate that
characterising the inelastic properties of cellular materials in terms of the two parameters ~rcr and el,
whilst convenient, becomes imprecise when the relative density of the material increases. Jahsman [6]
and others have used more accurate representations for the stress-strain curves of their materials.
Fig. 3. Split 0 constrained specimens: (a) progressive 'plane' crush front in Balsa; (b) multiple kink bands in Oak.
537
538
"I (o)
/'
'90 L
,:
80
i;:
0-
6 H
40
"',
~,/
3o i /
1'
"*
o
.
......
./
....... - "
o: ~.t..,~
0 Balsa
. . . . .
10 i~
0.0
iiiiililllJllJll[it~l~lltt[lllll,
0,1
0.2
0.3
0.4
Axial
0.5
0,6
ll,~ltll~
0.7
0.8
strain
t20
(b)
110
%2
100
+k
~'
%
-.~
90
0 Ekki
C]
i
-.
go
!
I
70
~90 Pine
60
90Oak
i/
50
/"
/'
~<
/'
/
4O
/'
f'
/
30
/"
:~
20
Redwood
................. " J
Ls"
...... / 9 0 B a l s a
10
0
0.0
//
./
0 1
0.2
0.3
0.4
Axial
0.5
0,6
0.7
0.8
strain
Fig. 4. Quasi-static uniaxial stress-strain curves derived from Fig. 1. The construction lines indicate how the crushing stress, trc~,
and the locking strain, el, are defined: (a) 0 specimens; (b) 90 specimens.
In this paper a rigid, perfectly-plastic, locking (r-p-p-l) material model is used for its simplicity to
give a first order understanding of the behaviour of wood subjected to impact against a rigid
target at velocities in excess of 30 ms -1. The limitations of this approach as well as the benefits are
discussed below.
539
(4)
Both 0 and 90 specimens were tested under free and laterally constrained conditions over a wide
range of impact velocities. The free specimens permitted high speed photography to be used to examine
and compare the global deformation processes; they also provided additional data on the initial crushing
stresses. Small samples cut from some of the constrained specimens were photographed through a
scanning electron microscope (SEM) to examine the nature of the deformation at the cell level. The
observations from these photographic studies will be described below.
540
(i)
=d
ll]mm
_LJ
[1~
llzo
I169
~UN
WO00
STRAIN{~AUG~ HCPKINSON
BARREL~ SAMPLE7
aELOWz
(hi
m
'.'I'lT31-'t
. . . . . ~^'-'
2
_U film
CAMERA
~
,I \lQr',n,F~,
I I]~IT~-
TOPlLOTL
SOLENOID
VALVE
\ \ ~ 2 XENON
~LIGHT! .
KDNTRON~ 700
--POWER SUPPtY/
l,JTIM)k~UNIT
"
'
I l~Gs~,,~R~L~'L~E
~
TRANSDUCER
AMPUFIER
CIRCUIT
EARTH
LINK
PC"TRAN.SENTI
RECORDER
DETAIL'A' {ENLARGED)
STRAINGAUGES.
Fig. 5. Pneumatic launch device for specimens and Hopkinson bar load cell used in impact tests: (a) photograph of system; (b)
schematic of whole system; (c) thick walled steel chamber to impose uniaxial strain conditions on specimens.
541
(c)
Fig. 5. (Continued)
significantly in the 90 specimens whereas the stress generally reduces significantly to virtually the
quasi-static plateau stress in the 0 specimens. Another feature evident in the 90 pulses is the existence
of a slight load drop after the initiation of crushing whereas under quasi-static loading conditions the
load increases monotonically for 90 specimens. A similar phenomenon was observed for syntactic
542
:300
(a)
27O
OOBalsa Vo=143
24O
....
OPlne
310
--
OOak
180
-G
150
0
_~
12.0
Vo=15:3
m/s
m's
Vo=172
m,'%
t,..,..
90
:%
60
"\,
',30
0
50
100
150
2,00
250
300
Time
260
350
400
450
__
90~Balsa
....
90Pine
600
(ps)
220
i
200
90Oak
Vo=12:3
Vo=143
m,,s
m/s
m/s
Vo=142
1~30
!\ t, i:."
160
i. . . . .
i ',:';
90 Ekki
Vo=129
m,'s
I
I
140
550
(b)
2,40
500
120
,,', %
',/
100
'~::~
I.I
80
60
i?
",
40
20
0
0
50
100
150 200
250
300
Time
350
400
450
500
550
600
650
(#s)
Fig. 6. Typical load-time pulses from uniaxial crushing tests: (a) 0 specimens; (b) 9 0 specimens; (c) 0 Redwood specimens; (d)
90 Redwood specimens.
foams by Jahsman [6]. This phenomenon is relevant to the accuracy of the shock theory predictions and
will be discussed further in Section 4.
From Fig. 7, the enhancement factor increases at an increasing rate with impact velocity for the 90
specimens. For the 0 specimens, Sr is almost constant, increasing only marginally up to a certain
velocity above which it also tends to rise more rapidly with increase in the impact velocity. These
wood
543
2,00
180
ORedwood
Vo=88 m/s
160
....
ORedwood
Vo=i54
nw"s
140
--ORedwood
Vo=213
rn/s
(o)
--\
/
i
120
\,
/
::
lO0
_~
! #"l ~"t/\
/I/
".
80
c
60
t:.~ \.. ~
~k,~_x,
.... ~- ....
4O
". ',
J "
X k .,
j)
20
/
0
0
50
100
150
200
250
300
Time
100
350
400
500
550
600
(/zs)
(d)
90
450
90Redwood
Vo=80
m/s
80
....
90Redwood
Vo=115
m/s
70
--
90Redwood
Vo=150
m/s
60
.-~
50
o
,.q
40
30
20
i0
~,
100
200
300
400
Time
F i g . 6.
500
600
700
800
(/xs)
(Continued)
observations appear to be independent of the test conditions whether free or laterally constrained, with or
without backing discs (see Fig. 7(c) and (d)), although some scatter occurs due to the variation of
specimen density within the same species, the inherent internal natural defects and due to slight
uncertainties in the impact velocity of constrained unbacked specimens due to deceleration of the
specimen prior to impact resulting from the presence of the air cushion between the specimen and the
load cell.
544
6.0
o' Oak
,-f,
O
r5
() a -
5.0
[]
0': R e d w o o d
45
0 Pin~
40
0 ~ Balsa*
3.5
~e$~
o ~ ~,v ~ . . ~ _ , ; ~
*f,
2.5
k~
20
1.5
[]
o
~, []
."
VO
~7
:3.0
[]
[]
[]
1.0
0.5
0.0
l0
80
120
1(30
Impact v e l o c i l v
:30
26
90 ~" E k k ]
24
90 Oak
22
-A
18
:240
(m
380
;32,0
360
s)
(b)
38
2O
:200
[] .q()~ R e d w o o d
:
V 90 Pine
[]
9(Y' Flals~,
16
14
-f.
1')
,-f~,
10
8
6
4
,)
0
0
40
80
i20
160
:200
240
280
320
:360
The sources of the enhancement of the crush loads are discussed in Section 4.
3.3.2. Description of macro-deformation and micro-deformation of the various woods. The dynamic
crushing of the specimens resulted in deformation mechanisms which were more localised than in the
samples tested statically. The main deformation was concentrated at the impact (proximal) end of the
545
6.0
r~
(o)
5.5
0 Constrained
backed
5.0
0 Constrained
unbacked
4.5
-x7 0 F r e e
backed
4.0
unbacked
0 Free
Oak
Oak
Oak
Oak
3.5
3.0
59
[] ,,~Zf Y
2.5
2.0
1.5
1.0
0.5
0.0
40
80
160
120
200
240
280
320
360
-A
90 Constrained
backed
12
:[]
90 Constrained
unbacked
11
-V
9 0 Free
Oak
10
~@
90 Free u n b a c k e d
backed
(d)
Oak
Oak
Oak
8
7
u]
(D
~A
v~
5
v
v~l
ve~
0
x7
[]
1
,
0
0
40
80
120
Impact
160
200
velocity
(m/s)
240
280
320
Fig. 7. (Continued)
specimens, particularly in the specimens without backing discs, as in bullet impact [ 13]. The progress of
the deformed region in free 0 specimens was by a mechanism reminiscent of projectile erosion, the
disintegration of the specimen taking place almost at the target surface, see Fig. 8(a). For the 90
specimens, the deformation was more like mushrooming, the deformation propagating into the
undeformed part of the specimen leaving the characteristic mushroomed profile, see Fig. 8(b). The free
546
(a)
(b)
Fig. 8. High speed films of free specimen impacts on rigid surface: (a) 0 Redwood specimen at 230 ms l; (b) 90 Pine specimen
at 108 ms 1 ; (c) 0 Balsa specimen at 166 ms 1 , with corresponding force pulse; (d) 90 Balsa specimen at 159 ms 1 with
corresponding force pulse. All times shown in microseconds.
c y l i n d e r s w i t h b a c k i n g discs o f t e n d i s i n t e g r a t e d in a n e x p l o s i v e m a n n e r w h e n i m p a c t e d at h i g h i m p a c t
velocities. F i g u r e s 8(c) a n d 8(d) s h o w the d e f o r m a t i o n o f t w o free B a l s a s p e c i m e n s s t r i k i n g the l o a d cell
at a b o u t 160 m s i t o g e t h e r w i t h t h e i r c o r r e s p o n d i n g force pulses.
547
p~tg4 ~ 'm 3
Yo~ 108 m '
I
i"
I/
r~
ao iJ I
~'~
jr
40
eo
leo
leo
a40
NO
4100
440
4410
~aO
(4~
eO0
(c)
'l'~me In jura
ii!!i!+i~i+i~
++!+iO~!:
+ii!!iiii!i!i!+i!i!!!~ii+iiii!?:!!!:i~
+~
m , d u t u t e d a c r o ~ the I r O n
( * t 90"
36
p 26S g4,'m 3
Vo I~,9 Ilx~
cl
2o
eo
llO
leo
140
'llum m/~
Fig+ 8+ (Continued)
(d)
548
II
~c
II
e,,
II
~9
II
7,
.fi
~9
"7.-~
(i)
549
r~
E
(b)
i)
e
Fig. 9. (Continued)
Post-impact examination of the constrained samples and the evidence from high speed films show that
for all specimens the deformation proceeded as an approximately plane wave front across which the cells
were crushed in a variety of mechanisms. This contrasts with the quasi-static behaviour (except for 0
Balsa specimens) where the deformation of a specimen is the result of the accumulation of local
deformation at sites distributed throughout the specimen which grow, multiply and eventually link up to
produce total crushing of the specimen.
Provided that the kinetic energy of the specimen was not so high as to cause all of the specimen to
deform into the locking regime, there was a clear distinction between the behaviour of the 0 and the 90
specimens. The 0 samples deformed by the propagation of a crushing front which was arrested before
reaching the distal end of the specimen (see the sequence of Pine specimens tested at increasing impact
velocity shown in Fig. 9(a) and the Oak and Balsa specimens in Fig. 9(b)). In contrast to the quasi-static
deformation mechanism, the front of the band of crushed cells progresses as an essentially planar surface
from the proximal end towards the distal end and the length of the crush band grew as the impact
velocity was increased, see Fig. 9(a). A band composed of crushed material appeared also at the distal
end in some of the specimens with backing discs as a result of the reflection of the compressive elastic
stress waves which precede the main crushing wavefront. The mean engineering compressive strain of
the crush band appears to be approximately the same as the uniaxial quasi-static locking strain.
However, the level of strain and the degree and nature of the deformation was not uniform throughout
the crushed zone as can be seen in Fig. 9(c). For the Pine specimen shown, the state of the material is
almost completely consolidated close to the impact surface (Fig. 9(c)(iii)) but is uncrushed just behind
the crush front (Fig. 9(c)(i)). At a certain impact velocity locking of the whole specimen is attained for
specimens with a backing disc.
In the 90 specimens, the leading edge of the region of crushed cells is not visible macroscopically as
in the 0 specimens, partly due to the recovery of the samples and partly because the deformation
progresses in multiple bands distributed according to the detailed structure of the wood as described
below. This leads to more non-uniform and distributed deformation patterns. Fig. 9(d) shows the
distinction between the deformation in 0 and 90 Balsa specimens. The nature of the bands of
550
(i) At the transition between crushed and uncrushed cells or crushing front,
(c)
(i)
551
(ii)
i
r~
(d)
Fig. 9. (Continued)
deformation in 90 specimens became visible only when observed through a scanning electron
microscope (SEM). The photomicrographs of the samples cut from the crushed parts of the 90
specimens, Fig. 10(a-c), show that the wood cells deform by bending modes similar to those observed in
square-packed ring systems referred to by Reid et al. [1] and described in detail by Stronge and Shim
[14]. As can be seen in Fig. 10(a), it would appear that the alternating band formation in the 90
specimens results from the propagation of shock waves in a heterogeneous medium having a repetitive
structure composed of microstructural layers of earlywood and latewood cells (the latter being thicker
and stronger) in which the latewood cell layers act as a rigid rear restraint for the earlywood cells. The
same effect can also be produced by the ray cells which also act essentially as rigid plates, concentrating
the deformation in the thinner-walled earlywood cells. This phenomenon is similar to that found in the
impact of metal ring and plate systems described by Reid and Reddy [15] and is probably responsible for
the more distributed deformation in the 90 specimens.
The photomicrographs of the clear deformation fronts in the 0 specimens show that the crushing front
is highly non-uniform (see Fig. 11) and that the micro-deformation is characterised by mechanisms
which involve kinking (strut-type buckling of the cells), progressive microbuckling (akin to concertina
buckling in metal tubes) and flow-microbuckling modes. Kinking (see Fig. 1 l(c) and (d)) is common in
man-made composites. Flow-microbuckling describes a mode of deformation associated with large fibre
rotations, see Fig. 1l(a) and (b). The clear delineation between crushed and uncrushed material evident
in the 0 specimens is reminiscent of the behaviour of two-dimensional hexagonally-close-packed metal
ring systems studied by Stronge and Shim [14]. The common behaviour is believed to be the
consequence of a degree of softening in the basic deformation characteristics of both this particular
cellular arrangement of rings and 0 wood specimens. Some of the dynamic response of systems with
softening have been exhibited by the heuristic spring-mass model produced by Shim, Tay and Stronge
[16] which will be discussed later.
552
(a)
(b)
(c)
Fig. 10. Photomicrographs of sectioned 90" specimens: (a) Redwood at 8 0 m s 1; (b) Pine at 6 0 m s 1 (c) Pine at 6 0 m s
magnification 25% of that in (b).
1,
553
t~
09
r~
"O
.=
-i
{o
li
kr.
tL
Fig. 11. Photomicrographs of crush front of 0 Balsa specimen without backing disc crushed at 239 ms-L: (d) has twice the
magnification of (a-c).
loading can be made for those constrained specimens which remained intact. This is achieved by relating
the specific energy, II (defined by dividing the kinetic energy of the projectile by the mass of the
specimen), to the average strain defined as the ratio of the compression and the initial length of the
specimen. The dynamic specific energy versus strain curves are compared with the quasi-static ones over
a range of global, average strains in Fig. 12. It should be noted that the dynamic average strains plotted
in these graphs are based upon measurements taken after the test when elastic recovery had occurred
554
.~
>"
160
150
140
130
120
110
100
90
80
70
60
~0
0 Balsa
0 Oak dyn
/%
data
data
- -
0 Oak static
.....
0 Balsa
/%
static
i%
/%
J'
i%
A
,,
50
,-~
dynamic
40
30 E
(a)
20
lO-Z
0--
0.0
05
O. 1
0.3
0.2
0.4
06
F I
07
0.8
Axial s t r a i n
150
140
_~ 90 Balsa
130
2~
dynamw
90 Oak dyn.
130
---90
llO
......
data
(b)
data
Oak static
90 B a l s a
static
100
90 -
ix
80 ~
~o
0
0
CZ
70 ~
/%
60 --50 ~
40 ~A
30 ~20 -
.1
1010
0.0
0 I
0.2
0.3
04
0.5
06
0,7
0.8
A~qal ~ t r a i n
Fig. 12. Experimental specific kinetic energy vs. average strain curves of dynamically crushed Balsa and Oak specimens: (a) 0
specimens and (b) 90 specimens.
This particularly distorts the 90 specimen results in Fig. 12(b). Notwithstanding this, the increase in
energy absorbing capacity under dynamic conditions is clear.
The specific locking energy ratio IIlk is defined as the ratio of the minimum kinetic energy per unit
mass required to produce the global locking strain (as indicated by the first appearance of locking at the
end of the force pulse) to the equivalent energy required under quasi-static conditions. The
555
YIlk(0)
FIlk(90)
1.49
1.30
1.48
1.90
3.59
3.50
2.60
1.64
experimentally determined ratios for the woods tested, are listed in Table 2. These global parameters
convey in an approximate way the influence of dynamic effects on the crushing strengths of the various
specimens.
It should be noted that these energy ratios were produced for a ratio of backing disc mass to specimen
mass of 4.53 and one would expect that they are likely to vary with changes of the mass ratio.
4. SHOCK MODEL FOR DYNAMIC CRUSHING STRENGTH ENHANCEMENT
556
(a)
EL
,,
Vo
Lo
t~
(b)
/"
~cr, Vo
t=o
L
v I --
Vv/~ ch"o, pI
cr, Po
- - ~ .
',
,_uSu :i = L*SL
t6t
cr, Po
,
v.fv
8u~
,~
(i0
I front
/
/
I' Sho k
', .tsx I/
,4
cl, o,p~
i (ii0
cs6t
Fig. 13. Shockpropagation model for r-p-p-l material: (a) idealised stress-strain curve; (b) parameters defining the shock model.
inelastic stress-strain curves for wood namely a crushing strength, (rcr, and a locking strain, el. Clearly
from Fig. 4 this model will be more accurate for low density woods but it provides useful information
for other woods for which the two basic parameters can be defined using the simple construction shown
in Fig. 4. It results in analytical expressions for the dynamic initial crush stress which give a good
physical feel for the influence of the various parameters on the crushing process. It also exposes some of
the important differences between loading along and across the grain of the wood. The refinement of
using a more precise stress-strain curve will be pursued elsewhere.
4.2.1. Shock equations. Fig. 13(b)(i) shows a wooden cylinder of initial length Lo, cross-sectional area
Ao and density Po with an attached mass M striking a rigid target normally with an impact velocity Vo.
The wood is assumed to deform under uniaxial compressive strain conditions. When impact occurs a
D y n a m i c uniaxial c r u s h i n g o f w o o d
557
shock wave is initiated which travels along the cylinder. Because of the rigid nature of the initial
response, the stress in all the material ahead of the shock is instantaneously raised to ~rc, As it passes
through the material it brings it to rest, increases its density by compaction up to the locking strain el and
raises the stress to value ~r*. Fig. 13(b)(ii) shows the state of the system at time t when the shock has
advanced a distance x from the contact surface. The shaded region ahead of the shock front denotes the
material which is traversed by the shock in time 6t. It is shown in its crushed state in Fig. 13(b)(iii). The
equations governing the propagation of the shock are made up of kinematic equations and equations of
conservation of mass and momentum for material crossing the shock front.
If u is the displacement of the rigid mass at time t, then from Fig. 13(b) Xo, the undeformed length of
the crushed portion is given by
x0 = u + x,
(5)
(6a)
L = -(3 +/0,
(6b)
and so
where superposed dots denote differentiation with respect to time and /~ = d u / d t = v is the current
velocity of the rigid mass and the undeformed portion of the wood.
Assuming that no radial expansion occurs, conservation of mass gives,
(7)
where Po is the initial density and Pl is the density when fully crushed. Thus
7 = po _
Pl
(8)
u+x
By definition,
el .
X0 - - X
x0
(9)
u +x
Thus,
7 = 1-
e 1.
(10)
7u
1 - 7
1 - el
el
u.
(11)
The shock speed relative to the projectile, cs is equal to dx/dt. Thus from Eqn (11),
Cs --
7
1 -7
k =
"Y
v.
(12)
1 -7
This indicates that the shock speed reduces as the undeformed portion of the projectile is slowed
down. The mass 6m of the shaded element of length 6x = Cs6t is given by
6m =- plAocs6t -- poAocs6t
7
pOCsV
=O'cr + - -
558
Thus
Crcr _L
('7* ~
POV2
(13)
el
4.2.2. Initial crushing stress enhancement factor. Eqn (13) can be used to provide an estimate of the
level of stress generated in the crushed material at a given velocity v. The initial crushing stress is
provided by Eqn (13) by substituting cr~r from either Eqn (1) or Eqn (2) depending on the loading
direction and replacing v by the impact velocity Vo and el by Eqn (3). Thus for 0 specimens,
D(o ) =
Pov2
15O(po/ps) + (1
(14)
4po/p,))
pov~
70(pO/p,) 2 +
o@o/p,))
(1 -
(15)
The ratios of the dynamic stress to the quasi-static crushing stress, S,, can be defined as follows:
For 0 specimens:
p, vg
S,(O) = 1 + 150(1 - a(po/p,))"
(16)
For 90 specimens:
(psVo) 2
o~(po/p~))"
Sr(90 )
1 4 70po(1 -
(17)
4.2.3. Crushing force pulses. The shock model can also be used to calculate the variation in ~r* with
time. On the assumption that the material unloads rigidly, this function of time represents the stress pulse
exerted on the rigid target. The change in momentum of the end mass M and the undeformed portion of
the cylinder between time t and t+6t is
(18)
Thus
M* - PA u
] dv
= -Aocrcr,
(19)
C1
where M* = M +
poAo&).
Hence
dv
dv
--
dt
Integrating and using
- -
du
[M*
A0~cr
(poAou)lCl]"
2O'crC 1
ln(l
PL~,u~
(20)
Po
Hence
du
v = ~- =
vg +
Po
In 1
M'c1 J
(21)
~cr + P
%2+
In 1
(22)
559
In order to obtain tr* as a function of time, Eqn (21) must be integrated numerically to give u as a
function of time. By substituting the solution of Eqn (21) into Eqn (22), or* is obtained as a function of
time.
4.2.4. Critical impact velocity for full specimen locking. The position, x, of the compaction front from
the impact surface can be deduced from Eqns (3),(11) and (20). From Eqn (20)
u = M*(1-t~(po/Ps))
[p~o
1 - exp (P(V2--v2)-)]
\ 20"crel / J
(23)
x = p--s~ 1 - e x p
From Eqn (6)
L = L0 - u - x.
(25)
L=L0---
El
Thus
u = (Lo - L)el.
(26)
L=-
~ ( v ) 2+ 2Crcr ln(M+pAL~
poe1
-M~
-j.
(27)
The mass of the body being deformed is reducing and in the limit becomes zero. The primary crushing
of the wood is thus exhausted for some value of the impact velocity Vo. The limit of applicability of the
present simple analysis occurs when the material is fully crushed because the shock model for the r-p-p-l
material does not permit the reflection of the shock wave. Thus, the maximum value vent of impact
velocity Vo for which the theory can be used corresponds to L=0 and L = 0. It is given by
Vcrit = J2tYcrelv
P0 l n ( l + - ~ ) .
(28)
This is the impact velocity to produce full crushing of the wooden cylinder which has mass of
mw. If M=0 the material never fully locks because the shock cannot reach the free end of the
specimen.
From this simple model one can deduce IIlk, the ratio of the energy absorbed under dynamic loading
conditions at an impact velocity of vomitto that absorbed up to full locking under quasi-static conditions.
Thus
IIlk=
( M )
1+~
(mw)
In 1 + ~ - .
(29)
For M/mw = 4.53, 7rtk = 1.10. The greater benefits of the dynamic enhancement of the crushing
strength in absorbing the kinetic energy of the projectile occur for small values of the mass ratio, M/mw.
rqk > 1 because of the influence of inertia effects in producing higher stress levels than or= whilst the
material strain reaches el.
Because of the simplicity of the assumption that the material behaves in a rigid fashion when it locks,
the shock model does not account for the absorption of energy beyond locking. Thus the r-p-p-l model
can only be used to analyse the full history of deformation for velocities up to v~nt. The real material is
not perfectly rigid at strain el and the use of a finite stiffness would permit the final crush as well as the
maximum loads produced in the specimen for velocities in excess of v~it to be evaluated. The remaining
energy would be absorbed by crushing the wood beyond the locking strain. This limitation of the present
560
30
28I
Shock theory
(~)
36
24
E x p B a l s a 90
~,~
Exp. B a l s a 0
18
14
~
12
10 I
4[-
fi
0t-IIIIIIIII
0
80
ill
120
J , J , l , l l l
160
Impact
J,,,]
200
velocity
15
14
240
J l l , l l , , ,
280
5 ~
320
36C
(m/s)
(b)
r~
AA
AA ABa
4:
3
2
* ~
I1~1
40
ldk"
-:_ . . . . . . . . . .
_--
1 ~
"
-
~O'~L
0..,_..I~....~!~) OO.--~DO"~i)-
,*O
_000~....{] D ........
......
"-~ ...............
"
rJ'i''''I'',ht'LLLIL,L,t,,,,],,J,I,~JJI,J,,
0
40
80
120
Impact
160
200
vetocitv
240
280
320
360
(m/s)
Fig. 14. Comparisons between the experimental and theoretical crushing stress enhancement factors, S, for (a) Balsa; (b) Pine; (c)
Redwood; (d) Oak; (e) Ekki. Dotted lines repesent the value of 2.3 predicted for 0 specimens.
shock model can be removed by constructing a shock wave model based upon a more realistic stressstrain curve as in [6,17,18] which allows reflected crushing waves to be generated. In this way the
increase in specific energy absorbed under dynamic conditions shown in Fig. 12 could be investigated
theoretically. This will be pursued elsewhere.
561
22
20
18
16
,'n
14 _
(e)
A Shock t h e o r y
Exp. Redwood 90
Exp. Redwood 0
:~
rj )
10
8
03
6
4
2
-OIIB~-03~-~'-'~----
0
40
80
120
160
Impact
200
240
280
320
360
velocity (m/s)
[4
13
(d)
Shock t h e o r y
12
11
Exp. Oak 90
10
Exp. Oak 0
9
7
6
5
4
3
h ..O~'a: . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
1
0 iLtll
0
{i)lLlOlrl~iL,,J
40
[30
. . . .
120
Impact
160
.....
200
240
380
f,r
. . . .
320
360
velocity (m/s)
562
S.R.
Reid
and
10
C.
Peng
(e)
A
Shock
Exp
theocv
Ekki
90 ~
7
/
/
/
/
/
:99
,A
t j
/
J
,-a
L I
30
L I
40
k i
L , I i i l , I
60
80
Impact
Fig.
. . . .
I , , i J J l i , J l l ] l l l l ] l L I I
100
veloeiLy
120
140
160
180
~00
(m/s)
14.(Continued)
generally the initial dynamic crushing stresses are greater than the predictions of the shock theory
although the form of their variation with increasing impact velocity is well represented by the theory. It
was noted in Section 3.3.2 that, in common with the results obtained by Jahsman [6] for syntactic foams,
the dynamic pulses for the 90 specimens developed a load drop. Thus one can define an initial crushing
stress (which has been compared with the shock theory in Fig. 14) and a plateau stress. Jahsman
compared his shock wave predictions with the plateau stresses rather than the initial crushing stresses
(termed the 'initial overshoot' in [6]). He argued that this procedure should be followed because the first
overshoot occurred at a time during which heterogeneity effects may be important. In his view, several
'zones of heterogeneity' should be exposed to the stress wave to ensure a statistically homogeneous
response. This led him to concentrate on the plateau stresses. It is certainly the case that plotting the
values of the dynamically enhanced plateau stresses brings the experimental data very close to the shock
theory predictions. For instance, examining the data for the 90 specimens shown in Fig. 6(b) and (d),
the differences between the initial peak stresses and the plateau stresses range from 12% to 37%
approximately. The data for Redwood shown in Fig. 6(d) shows also that the percentage difference
decreases as the impact velocity increases. The results in Fig. 14 suggest an almost constant value of 2
for the difference between the dynamic stress enhancement factors for the initial crushing stress and the
shock theory. The wood data therefore imply that the shock theory represents accurately the
enhancement of the plateau stress for the 90 specimens. What controls the value of the difference
between the initial and the plateau crush stresses remains to be clarified. It is most likely related to
inertial effects associated with the initiation of the predominant cell wall bending mode, a phenomenon
somewhat similar in character to the argument in Section 5.1 for the 0 specimens. However, such an
argument remains to be quantified.
For the 0 specimens, the theoretical results are about half of the measured values. However, the
convergence between the results for Oak and the shock theory as the impact velocity increases should be
noted. It is clear that the simple r-p-p-l shock model is not appropriate for dynamic loading along the
grain in the velocity range covered by the experiments. The enhancement of the initial crushing stress
and other distinctive features of the response of 0 specimens is discussed in the next section.
The load-time pulses obtained from Eqns (21) and (22) agree reasonably well with the experimental
pulses for the 90 specimens, see Fig. 15. The impact velocity to cause full crushing of the material
563
120
110
........
Shock theory
B a l s a 90
100
............
Shock theory
Balsa 0
~ 2
Balsa 0
Vo= 143 m / s
90
80
Z
(a)
70
Exp. 9 0 " B a l s a V o = 1 2 3 m / s
60
0
~_~
50
. . . . . .
}. . . . .
40
30
20
10
0
50
100
150
200
250
300
350
400
450
500
550
600
Time (#s)
400
(b)
360
..........
320
.............. S h o c k t h e o r y
Shock theory
Oak 90
Oak 0
Exp
280
Z
2<
240
"U
05
o
200
160
120
Exp. OOak V o = 1 4 2 I n / S
80
40
0
0
40
80
120
160
200
240
280
320
360
400
T i m e (/~s)
Fig. 15. Comparisons between measured force pulses and those predicted by shock theory: (a) Balsa; (b) Oak; (c) Ekki.
given by Eqn (28) and the experimental critical impact velocity (i.e. minimum impact velocity to
produce full locking strain in the specimen) for a mass ratio, r=4.53 are given in Table 3. For reasons
explained in Section 4.2.4, the deficiencies of the r-p-p-l model need to be removed to provide better
impact energy absorbing estimates for wood. The present theory provides a simple formula which could
assist in the sizing of a particular component in an application. Paradoxically, the predictions for the 0
564
500
(o)
450
400
350
Z
300
350
2,00
150
.....
Shock th Vo:::31.Srnj:
100
.......
5O
0
0
40
80
lEO
160
2,00
Time
a40
380
330
360
400
(/~s)
Balsa 0
190
Balsa 90"
Y. Pine 0~
Y. Pine 90"
Redwood 0
Redwood 90
A. Oak if'
A. Oak 90
80
170
80
154
80
157
70
148
52
136
44
122
57
specimens are better than for the 90 specimens. Amongst other things, this could be due to the fact that
the attainment of full locking is easier to identify in the specimens loaded along the grain.
565
increasing impact velocity. For 0 specimens the initial crushing stress increases more substantially
than the shock theory indicates, attaining and maintaining a value of Sr of between 2 and 3 over the
majority of the impact velocity range examined, the force pulses showing a rapid reduction down to
the quasi-static plateau level.
These features suggest that an alternative model for the prediction of dynamic crushing force pulses
for 0 specimens is required but this will not be pursued fully herein. Here an explanation is proposed for
the discrepancy between the shock theory and the measured initial crushing stress for the 0 specimens
and a number of other aspects of the behaviour of these specimens are briefly discussed.
5.1. Initial crush stress enhancement f o r 0 specimens
An explanation for the magnitude of the enhancement of initial crush stress for 0 specimens can be
produced by analogy with the behaviour of structural elements such as struts, axially compressed
circular tubes or honeycombs as discussed by Reid et al. [1].
The propagation of shock waves through the cellular structure of wood is governed by the inertia of
the cells, their crushing strength and the locking strain which reflects the steepening of the stress-strain
curve in a crude fashion. In this case the appropriate inertia (represented by the density of the material)
relates to the resistance to motion of the cells in the direction of the impact force. However, inertia (or,
more correctly, microinertia) has an additional role for the elongated wood cells which are loaded axially
when impact takes place in the direction along the grain. As indicated in Section 2 and as discussed in
[1], the cells undergo a variety of deformation modes which involve instabilities of various sorts, viz.
strut-like buckling and progressive collapse like circular tubes. These structural elements are
characterised by falling load-displacement curves, i.e. structural softening, after their initial collapse
load is attained. It is believed that the softening behaviour in the quasi-static load-compression curves
for wood loaded along the grain is the result of analogous behaviour in the cell structure as it crushes.
Recent studies of a class of softening structures, termed Type II structures, under dynamic loading
described in [21-26] have succeeded in explaining certain anomalies regarding their energy absorbing
capabilities. As explained theoretically by Su et al. [24], for structures made from rate-insensitive
materials, the lateral inertia of the strut-like elements opposes and delays the onset of the primary
deformation mode which is bending. The elements undergo an initial deformation phase which is
predominantly uniaxial compression, the force being governed by the uniaxial yield stress of the strut
material which results in a significant enhancement of the initial value of the crushing force under
impact which is of short duration. It is this result which provides a possible explanation for the level of
the enhancement factor for the initial crushing force in 0 specimens.
According to the behaviour of Type II structures, under impact loading conditions one would expect
the initial peak stress for 0 specimens to reflect the yield strength of the wood cell material, trys. The
applied stress trcompcorresponding to uniaxial yield in the cell walls assuming a hexagonal cell geometry
(as assumed in the derivation of the form of Eqn (1) for the 0 crush strength) is given by ([2])
p0
O'comp = f f y s - - .
(30)
ps
Thus the enhancement factor for the dynamic strength relative to the quasi-static strength, Eqn (1), is
simply
O'cmp - - O'ys
O'cr (0 )
150'
(31)
where the units of trys are MN m -2. There is some debate about the value of trys [2], but Cave [27] gives
a value of 350 MN m -2, which implies a stress enhancement factor of 2.3 from Eqn (31) for the
dynamic compression of 0 specimens, regardless of the density of the wood. This value is shown as a
dotted line on Fig. 14 and provides a reasonable fit to the majority of the data for the 0 specimens. The
agreement is good throughout the impact velocity range for the low density woods (Balsa, Redwood and
Pine). For Oak the shock theory apparently takes over from this level of enhancement implying that the
stress enhancement mechanism has changed.
As noted in Section 4.3, it is also the case that the shock theory underestimates the experimental data
for the initial dynamic crushing stress for 90 specimens. One would also expect a microinertial effect
associated with the initiation of the cell crushing mode in the direction across the grain. This type of
566
phenomenon has been discussed for metal ring systems by Reid et al. [1,18]. Again the forces would
result from compressive yield along the cell walls and it is noted that an enhancement factor of 2.3 also
matches reasonably many of the 90 specimen data, especially at low impact velocities, which lends
some support to arguments based upon microinertia again inhibiting the onset of the bending
mechanism.
~ E ( = 35 109po/psNm -2)
po
--
3/~__ 109
V -ffo
-- 4830 m s - l,
where the expression for the Young's modulus along the grain shown in brackets has been taken from
Gibson and Ashby [2] and the value ps= 1500 kg m -3 has been used as before. This value for the elastic
wave speed has been confirmed by experimental wave speed measurements [9]. Disturbances will be
elastic for impact velocities up to that for which the stress generated is equal to the crushing stress. The
discussion in Section 5.1 indicates that this is equal to 350 (Po/ps) ~ m 2. Thus, the impact velocity at
which crushing is initiated in a 0 specimen, vc, is given by the bar wave equation
350 l 0 6 p = p 0 4 8 3 0 v c
p~
Thus
vc = 48ms 1.
Even though not many tests were performed at low velocities, this value compares well with the data
shown in Fig. 14, especially for Balsa wood where lower enhancement factors than 2.3 can be seen
below 40 ms -1. These lower values conform to predictions based on the elastic bar wave equation.
5.2.2. Role of the 'energy barrier' for 0 specimens. The presence of the softening region in the stressstrain curve for crushing along the grain is a distinctive feature which is ignored in the r-p-p-l model and
which does not lend itself to classical non-linear wave analysis since the implication is that the
governing equations are elliptic rather that hyperbolic and so do not permit wave-type solutions. Another
solution/modelling procedure is therefore required to make progress. Shim, Tay and Stronge [15] have
explored the behaviour of one-dimensional mass-spring systems in which the spring has a saddle
transition region bridging initial collapse and consolidation/compaction/locking behaviour. The sample
problems solved in [15] relate to an end mass being set in motion along the direction of the line of the
masses, compressing the springs between them. The distal end of the spring-mass system is fixed. The
motions of the individual masses and the deceleration of the end mass are deduced by solving the set of
differential equations of motion for the masses and the solutions have a number of interesting features.
As far as the transmission of deformation along the line of masses is concerned, the response of the
system is wave-like in that the masses move in sequence (see Fig. 16(a)), the disturbance transmission
diagrams resembling those for ring systems found in [ 18] and the influence of a precursor elastic wave in
triggering plastic deformation after reflection from the fixed end of the system is evident. However, the
most distinctive feature is the oscillatory nature of the deceleration trace of the end mass as exemplified
in Fig. 16(b). In all cases, the presence of the saddle transition in the spring characteristic leads to
sudden drops in the deceleration followed by rapid rises apparently as succeeding masses are brought
into motion.
This behaviour has the appearance of softening requiring that the deformation of a particular spring
has to exceed that required to take it through the saddle transition and into the locking phase. The extent
to which the deformation of the spring moves up the locking curve depends on the dynamics of the
spring-mass system. It would seem to be a necessary requirement for a model of such behaviour that the
shape of the spring characteristic, especially in the locking phase, be prescribed accurately. A rigidlocking approximation is inadequate. This contrasts with shock-type behaviour in which only the crush
567
~
~
]1t l ll i ili i /
1.0 ]
.............................................
!.g.... .......
~ ,
,In
117e
j7
,,
~3
,2
o:4!i !Milli
Iii/17,i//'i/
i I
o.~
t i i
a</.J, au
,
V]
. ~:
0.0
fo
"
1.0
t /iJ7
i,
:
" .'
'
2.0
,'
'
3.0
4.0
5.0
6.0
Time(s)
(a)
2.0
llr
1.5
0
1.0
L/LIy\i \j ~ / "-~,,~fWf--kj-~/
0.5
0.0
'
o.o
~b~
'
'
'
I . . . .
1.o
I . . . .
2.0
I . . . .
3.0
I . . . .
4.0
I . . . .
5.0
e.o
Time (s)
(yield) initiation point and the stress level to which the state jumps as the shock front passes are
significant. The detailed shape of the intervening stress-strain curve is irrelevant, provided only that it is
convex to the strain axis. This points to the need for an alternative approach to the r - p - p - l model when
considering solids with an element of softening in their behaviour. This will considered further
elsewhere. In this paper we consider two critical specimen velocities for which estimates can be made
without the need for a detailed softening-locking material model.
(i) S p e c i m e n v e l o c i t y a t w h i c h t h e c r u s h f r o n t c e a s e s to p r o p a g a t e
It is conjectured here that in order for deformations to be transmitted through the cellular structure of a
0 specimen the energy barrier represented by the saddle transition has to be surmounted. Using the
r - p - p - l approximation of the stress strain curves for wood, this energy can be estimated as crcrelAohw for
a layer of cells. If the kinetic energy remaining in the specimen is insufficient to crush the next layer of
568
cells then the deformation will cease. In order to test this conjecture we can use the information provided
in [7] for Balsa to estimate this critical velocity, vf, for one of the specimens used in the impact tests, the
details of which are given in Table 1.
Mass of specimen = poAohw ( m1 ~+)
-- 2777r 453
~ - 5.53
452
Crushing energy = ~rcrelAohcell ----277r ~ - - 635
10 -9 =
0.11 kg
1 0 - 6 0 . 6 8 ---- 1 8 . 5 J,
where a value of 635 #m has been used for the cell length, hcen [7]. Neglecting the mass of wood that has
been brought to rest and equating the kinetic energy of the specimen to the energy bamer represented by
the crushing energy for one layer of cells leads to v:=18.4 ms - l which is not unreasonable.
(ii) Impact velocity for transition to shock response
In a similar vein it can be conjectured that if the kinetic energy in the first layer of cells is alone
sufficient to cross the energy barrier then the deformation will proceed in a shock-like manner. If the
impact velocity corresponding to this condition is Vs, a shock will be produced if
'
p~
i.e.
Vs >
~/300 x 106 Cl
1500
~/~
103
ms -~
(32)
Here, in estimating the energy required to cross the softening barrier, the effect of the short duration
enhancement of the initial crushing strength represented by Eqn (31) is ignored. Using el equals 0.35 for
Oak and 0.65 for Balsa leads to values of V, of 265 and 360 ms -l, respectively, values which find
support in Fig. 14.
6. CONCLUSIONS
Even under the controlled conditions of quasi-static uniaxial compression, the inelastic deformation
of wood is a complex process governed by collapse and gross deformation of the cell structure.
The micro-mechanisms of collapse cover a wide range and are strongly influenced by the detailed
geometry and variability of the cell structure and, most importantly, by the direction of loading.
From a solid mechanics point of view, the simple relationships for the crushing strengths in the
principal loading directions produced by Gibson and Ashby [2] and given by Eqns (1) and (2) have been
shown to provide useful estimates for these key mechanical properties and the dependence on the
relative density of the wood is both valid and useful. However, the prediction of the locking strain
given by Eqn (3) and originally due to Maiti et al. [8] has been shown to be less accurate. Indeed,
for higher density woods the concept of locking is questionable and a non-linear stress-strain curve is
more appropriate.
It has been demonstrated from the uniaxial impact tests that the initial crushing strengths of wood
samples can increase significantly, in some cases by an order of magnitude. And that this has
repercussions on the impact energy absorption capacity of wood. Notwithstanding the inaccuracies of a
material model which is based upon the crushing strength ~rcr and a locking strain el, it has been shown
that the level of enhancement of the crushing strength of wood across the grain is well-predicted by a
shock theory based upon a rate-independent, rigid, perfectly-plastic, locking material model in the
velocity range of 30-300 ms 1. The enhancement of the initial crushing stress for wood loaded along the
grain is apparently due to microinertial effects inhibiting the establishment of the characteristic cell
buckling mode for that loading direction, leading to an almost constant enhancement factor of
approximately 2.3. This has been predicted theoretically from a consideration of analogous dynamic
deformation mechanisms in Type II structures which delay the onset of the principle bending
deformation mode and generate forces which are the result of uniaxial compression of the deforming
element. A similar phenomenon may be responsible for the appearance of an initial peak in the
specimens loaded across the grain.
569
Considerations of the influence of local softening for loading along the grain lead to the conclusion
that, despite the similarities in the overall behaviour of wood specimens loaded along and across
the grain in terms of the progressive nature of the deformation and the influence of locking on that
response, a new 'wave' theory is required which recognises and incorporates the effect produced by the
presence of the saddle transition phase in the representation of the stress-strain curve, a continuum
theory with the characteristics exhibited by the spring-mass model due to Shim et al. [16]. This will
undoubtedly require the application of a regularised continuum theory involving higher order
displacement gradients introducing length scales characteristic of the cellular geometry or other nonlocal approaches. Elementary considerations of the energy barrier which various deformation processes
are required to surmount lead to both an explanation for the clear delineation between compressed
and undeformed wood in some 0 specimens and for the transition impact velocity for the onset of
a shock-type response.
The results obtained here are consistent with those obtained previously for syntactic foams [6] and
have been applied to other man-made cellular materials such as honeycombs (reminiscent of 0
specimens), the results of which will be reported elsewhere.
Acknowledgements-- The authors would like to express their gratitude for the support of British Nuclear Fuels plc who funded the
Ph.D. studies of the second author on which much of the present paper is based. Parts of this study were funded under EPSRC
research grant GR/K18832 for which the authors wish to express their thanks. Thanks are also due to the Department of Paper
Science (UMIST) for permitting the use of their electron scanning microscope for the examination of the structure of the deformed
specimens.
REFERENCES
1. Reid, S. R., Reddy T. Y. and Peng, C., Dynamic compression of cellular structures and materials, In: Structural
Crashworthiness and Failure, N. Jones and T. Wierzbicki, eds, pp. 295-339. Elsevier, Amsterdam (1993).
2. Gibson, L. J. and Ashby, M. E, Cellular Solids, Pergamon Press, Oxford (1988).
3. Johnson, W., Historical and present-day references concerning impact on wood, Int. J. Impact Engng., 1986, 4, 161-174.
4. Johnson, W., Mostly on oak targets and 19th century naval gunnery, Int. J. Impact Engng., 1986, 4, 175-183.
5. Reid, S. R., Peng, C. and Reddy, T. Y., Dynamic uniaxial crushing and penetration of wood, In: Mechanical Properties of
Materials at High Rates of Strain, J. Harding, ed., Inst. Phys. Conf. Series, No. 102, Bristol, pp. 535-542 (1989).
6. Jahsman, W. E., Static and dynamic material behaviour of syntactic foam, In: Mechanical Behaviour of Materials under
Dynamic Loads, U.S. Lindholm, ed., pp. 365-387, Springer-Verlag (1968).
7. Easterling, K. E., Harryson R., Gibson, L. J. and Ashby, M. E, On the mechanics of balsa and other woods, Proc. R. Soc. Lond.,
1982, A383, 31-41.
8. Maiti, S. K., Gibson, L. J. and Ashby, M. E, Deformation and energy absorption diagrams for cellular solids, Acta Metall.,
1984, 32, 1963-1975.
9. Peng, C., Crushing and indentation of wood under static and dynamic loading conditions, Ph.D. Thesis, University of
Manchester (UMIST) (1991).
10. Budiansky, B. and Fleck, N. A., Compressive failure of fibre composites, J. Mech. Phys. Solids, 1993, 41, 183-211.
11. Kyriakides, S., Arseculeratne, R., Perry, E. J. and Liechti, K. M., On the compressive failure of fiber-reinforced composites,
Int. J. SoL Struct., 1995, 32, 689-738.
12. Zukas, J. A., Nicholas, T. Swift, H. E Greszczuk, L. B. and Curran, D. R., Impact Dynamics, John Wiley and Sons, New York
(1982).
13. Taylor, G. I., The use of flat-ended projectiles for determining dynamic yield stress I: Theoretical considerations, Proc. R. Soc.
Lond., 1948, A194, 289-299.
14. Stronge, W. J. and Shim, V. P. W., Dynamic crushing of ductile cellular arrays, Int. J. Mech. Sci., 1987, 29, 381-406.
15. Reid, S. R. and Reddy, T. Y., Experimental investigation of inertia effects in one-dimensional metal ring systems subjected to
impact - I: Fixed-ended systems, Int. J. Impact Engng., 1983, 1, 85-106.
16. Shim, V. P. W., Tay, B. Y. and Stronge, W. J., Dynamic crushing of strain-softening cellular structures - a one-dimensional
analysis, Trans. ASME, J. Engng. Mater Technol., 1990, 112, 398-405.
17. Reid, S. R., Bell, W. W. and Barr, R. A., Structural plastic shock model for one-dimensional ring systems, Int. J. Impact
Engng., 1983, 1, 175-191.
18. Reid, S. R. and Bell, W. W., Response of one-dimensional ring systems to end impact, In: Mechanical Properties of Materials
at High Rates of Strain, J. Harding, ed., Inst. Phys. Conf. Series, No. 70, Bristol, pp. 471-478 (1984).
19. Lee, E. H. and Tupper, S. J., Analysis of plastic deformation in a steel cylinder striking a rigid target,, Trans. ASME, J. Appl.
Mech., 1954, 21, 63-70.
20. Zaretsky, E. and Ben-dor, G., Compressive stress-strain relations and shock Hugoniot curves of flexible foams, Trans. ASME, J.
Engng. Mater. Technol., 1995, 117, 278-284.
21. Calladine, C. R. and English, R. W., Strain-rate and inertia effects in the collapse of two types of energy absorbing structure,
Int. J. Mech. Sci., 1984, 26, 689-701.
570
22. Zhang, T. G. and Yu, T. X., A note on a 'velocity sensitive' energy absorbing structure, Int. J. Impact Engng., 1989, 8, 43-51.
23. Tam, L. L. and Calladine, C. R., inertia and strain-rate effects in a simple plate-structure under impact loading, Int. J. Impact
Engng., 1991, 11, 349-377.
24. Su, X. Y., Yu, T. X. and Reid, S. R., Inertia-sensitive impact energy-absorbing structures, Part 1: Effects of inertia and elasticity,
Int. J. Impact Engng., 1995, 16, 651-672.
25. Su, X. Y., Yu, T. X. and Reid, S. R., Inertia-sensitive impact energy-absorbing structures, Part 11: Effects of strain-rate, Int. J.
Impact Engng., 1995, 16, 673-689.
26. Karagiozova, D. and Jones, N., A note on the inertia and strain-rate effects in the Tam and Calladine model, Int. J. Impact
Engng., 1995, 16, 637-649.
27. Cave, I. D., Longitudinal Young's modulus of Pinus Radiata, Wood Sci. Technol., 1969, 3, 40-48.