Hybrid Testing Of Historic Materials
Miloš F. Drdácký, Associated Professor, DrSc., Ondřej Jiroušek, PhD.,
Zuzana Slížková, Dipl.Eng., Jaroslav Valach, PhD., Daniel Vavřík, PhD.
Institute of Theoretical and Applied Mechanics of the Czech Academy of Science, v.v.i.
190 00 Praha 9, Prosecká 76, Czech Republic
drdacky@itam.cas.cz, jirousek@itam.cas.cz, slizkova@itam.cas.cz, valach@itam.cas.cz,
vavrik@itam.cas.cz
ABSTRACT
This paper describes hybrid some approaches to testing the mechanical properties of historic materials, which are mostly
available in very small amounts and bulks, so that standard testing procedures and techniques are not appropriate. Three case
studies are presented. In all three, simple mechanical testing procedures are combined with special optical measurements and
tailored software tools. They involve measurements of mechanical characteristics of natural fibers (animal hair) and on small
samples of historic mortars, and measurements of physical characteristics of medieval sandstone before and after treatment
with ethylsilicate consolidants.
1.
Determining the mechanical characteristics of natural fibers
For research into the behavior of lime mortars with natural fibers, which are a quite frequently-used historic material in almost
all European countries, we need to know the mechanical characteristics of the natural fibers [1], [2]. They are usually very
short, of irregular cross section, with a quite high thickness-to-length ratio, typically ranging around 0.1mm : 45 mm. Therefore,
it is quite difficult to test them, and standard methods are not applicable. The rupture loads attain values of about 10 N.
A simple loading frame for testing animal fibers was
constructed from a laboratory balance, Fig.1. The
tested samples could be loaded with a force increment
of 0.02 N ± 0.005 N using calibrated steel balls 2 g in
weight. The ends of the test specimens - goat and
horse-hair fibers - were glued between two thin metal
sheets imitating grips. The lower fixture was firmly
clamped into the loading frame table using a watchmaker grip, and the upper grip hung on the arm of the
balance. After the weight of the upper grip was balanced, the frame was prepared for loading.
Before the test, two contrast marks were painted on
the fibers and, during the loading, changes in the
position of the contrast marks on the fibers were optically recorded using microscopy. Due to the small
dimensions and the surface characteristics of the
fibers, it was not possible to prepare sufficiently contrasting, regular and subtle strips. Therefore, it was not
possible to detect the edges or centers of the strips by
searching for their edges by means of a uniquely defined distribution function. Two characteristic spots
Figure 1 Arrangement for animal hair testing.
(marks) with a significant structure were selected on
the unloaded fiber. In most cases some paint nonhomogeneities served as such “points” – the contrast marks and their position were evaluated by special software written in
Matlab 6.1. A black and white 1.3 Mpixel CCD camera with macro-lenses was used to record the images. [3]
The measurement mark and a surrounding area of about 45 x 45 pixels were adopted as a mask in the first loading step. In the
next loading step the position with the least difference between the new record and the mask was looked for. The picture was
scanned in a small area of about 30 x 30 pixels around the coordinates of the mask in the previous phase. The grey level
values were normed with the median of grey levels of the analyzed image frame. This was necessary owing to low grey levels
outside the investigated fiber. When the mark was found, the mask and also the mark positions were updated and the procedure continued. It should be noted that, due to the quite large deformations of the fiber, the structure and dimensions of the
painted strips changed substantially.
The measurement accuracy of the described method reaches the level of one picture pixel. The measured length oscillated
around 900 pixels, which gives accuracy in the length of the measurement of the mark of about 0.11%. Because the measured
values oscillated within limits of one pixel around the actual value, the measured load-deformation diagrams can be considered reliable. The measured mechanical characteristics are shown in Table 1.
Fibers or fibrous
particles
Abbrev
goat hair
horse hair
Ko
Ku
TABLE 1. Measured properties of natural fibers
Tensile
Modulus of
Nominal
Equivalent
strength
elasticity
length
diameter (μm)
(MPa)
(GPa)
(mm)
50
30-100
110-230
6.2-7.7
50
50-140
110-200
3.9-5.1
For an evaluation of the mechanical characteristics it was necessary to determine the cross-section data of the tested fibers. For
this reason the fibers were embedded into resin after the test,
ground, polished and then the cross section picture was taken in
the microscope under 1000x magnification. From the photographs the cross sectional area was measured using a square
grid of 1 μm density, Table 2. The measured “deformation/force”
dependency could then be transformed into the “strain/stress”
function, from which the modulus of elasticity, elongation and
strength were determined.
σ [M Pa]
200
160
120
40
ε [%]
0
0.1
negligible
negligible
TABLE 2 Fibre cross-section characteristics
2
Max / Min
Area (mm )
Tensile
diameter (:m)
strength
Cross section
(MPa)
Horse 4
0,005932
100 / 70
196.63
100.00
50.00
ε [%]
0.00
0.1
0.2
Horse 5
0,013194
130 / 130
111.27
0.3
σ [MPa]
0.2
150.00
0
2-29
4.5-32
Full profile
0
200.00
Moisture
(%)
Full profile
80
250.00
Elongation at
failure (%)
0.3
Figure 2 Stress-strain diagrams of horse hair (horse 4 upper)
and goat hair (goat 2 lower).
Goat 1
0,002948
70 / 60
Tube
Wall thickness
18 - 23
109.15
Goat 2
0,00393
100 / 68
230.27
Tube
Wall thickness
10 - 24
Two types of load/deformation or stress/strain diagrams were observed in the tests. In the first group, the deformability
changed during the loading and a “plateau” similar to plastic deformation of metals occurred, Fig.2. This was followed by a
short stiffening and rupture. The elastic phase of the load/deformation diagram can be described as bi-linear, with a high initial
modulus of elasticity (at stress up to about 25-30% of the strength value) subsequently changed into a branch with the
modulus of elasticity dropping to about one half of the initial value (up to stress of about 60 - 70% of the strength value). Then
σ [MPa]
σ [MPa]
the non-linear phase started. As shown in Figure 2, the two types of hair exhibited similar 100
100
characteristics. Due to this nonlinearity, the rate
of loading during the tests was adopted as 80
80
0.0065 N/s, and yielding was allowed under a
lower rate (corresponding to the usual rule for
60
settling increments between two subsequent 60
loading steps). The second group of fibers
40
failed without the plastic stage (mostly even 40
without the nonlinear stage) and usually at a
strength of about 50% of those in the first 20
20
eps
ε [ %]
group. Figure 3 shows a typical stress/strain
diagram for both types of hair. The behavior of
0
0
the fibers is influenced by the organization of
0
0.01
0.02
0
0.01
0.02
0.03
0.04
the collagen fibrils in the hair matrix and the
straightening of the hair in the course of loading.
Figure 3 Stress-strain diagrams of horse hair (horse 5 left) and goat hair
(goat 1 right).
Our optical method using a recurrently correlated mask, (similar to the image correlation method), performed very well in determining the mechanical characteristics of
short animal hair. Its accuracy and sensitivity provide credible experimental data on modulus of elasticity, ductility and elongation of the tested hair.
The method described here was also used with success for other non-standard measurements of material characteristics, for
measuring the mechanical properties of fresh bone samples. They have quite a rough, greasy surface. Therefore it is difficult
to create well-defined sharp marks on the surface, and the technique described above can be applied for compensating the
reduced legibility of such marks.
2.
Determining the deformation characteristics of mortars
In the second case, very flat small historic mortar samples had
to be tested for measurements of their deformability [4]. The
samples were cut from larger irregular plates, and their thickness was leveled with a thin polymer-cement layer. The
specimen was provided with resistance strain gages and
loaded in the Testatron electromechanical loading frame. The
loading force was measured by a load cell with a capacity of
100 kN, and the load was introduced with the cross head
velocity of 0.45 mm/min. The distance of the steel loading
plates was measured by means of four LVDT sensors
mounted in the corners, Fig.4, and one checking sensor for
measuring the cross-head movement.
Surface deformations of the specimens during the compression tests were further measured using a high-resolution CCD
camera and stored in bitmap format without compression. The
exact exposure time was recorded together with the image.
Figure 4 Test set-up – a specimen in the testing frame and
a camera with a circular lighting source.
Figure 5 Pixel calibration with well visible structure of
the tested mortar.
The deformation was then evaluated using the texture on the surface of the specimen as a natural marker. After several preprocessing techniques based on thresholding, four distinct markers
were chosen to calculate their time-dependent positions by identifying the center of gravity of each of the markers using the moments of inertia of the pixels weighted by their intensity values.
This enabled us to attain sub-pixel accuracy of the position measurement.
Measurement of historical mortar
The optically measured modulus of elasticity was compared
to the measurements made by two different methods –
resistance strain gages glued on the surface of the speci12
men (marked in Table 3 as Modulus 2) and overall deformability (Modulus 1) measured by the change in distance
10
between the loading plates, (which was influenced by the
layer of leveling material, and the average value was 910
8
MPa). The results are presented in Table 3. In the first case,
the length of the measurement base was approximately
6
equal to the height of the specimen (Modulus 1), in the
second case it was equal to the strain-gage base 6 mm,
4
and in the optical measurements it oscillated between 7 mm
and 12 mm. Figure 6 presents numerically created stress2
strain diagrams constructed from the measured data. The
modulus of elasticity seems to be dependent on the meas0
0
0.1
0.2
0.3
0.4
0.5 urement base. Logically, the shorter the base, the higher
deformation [%]
the modulus. Another influence is due to the technique of
coupling the measured force during loading with the reFigure 6 Stress strain diagrams of mortars obtained by
corded images. This link is derived from time records in the
optical identification of natural markers
camera and in the data acquisition system. The optical
identification method finds the updated position of the natural marker based on the position established from the previous
image. The center of gravity of the marker is traced in the sequence of images. Only a small portion of the image in the vicinity
of the marker is used in calculating the center of gravity. This technique significantly speeds up the method. During the experiment, the images were taken manually approximately every 20 s, resulting in a small number of images and a great distance traveled by the natural markers in each step. The small number of images reduces the accuracy of the method. A
shorter time period between the images would overcome this problem.
14
stress [MPa]
sample 2 small dist
sample 2 large dist
sample 3
TABLE 3 Measured modulus of elasticity on a historic mortar
Specimen
Length
(mm)
Width
(mm)
Height
(mm)
Compression
strength (MPa)
Modulus 2
(MPa)
5.1
5.2
41.6
41.1
41.6
41.4
25.1
25.0
17.96
13.15
19750
5190
Ultimate
load range
(%)
15 – 30
14 -45
5.3
42.4
41.6
25.1
20.66
15840
14 – 27
Modulus 3
(MPa)
22330
Comment
Longer base
24550
The measured differences thus have various causes. Schueremans [5] shows that modulus of elasticity in compression calculated from the overall deformation measurements attains approximately one third of the value measured directly on the specimen by means of mounted LVDT sensors, when standard specimens are tested. In our experiments, the sample thickness
was very small compared to the size of the largest grains of sand in the mortar (about 8-10 mm), which were of the order of the
measurement base for precise methods. The resulting scatter of results is therefore not surprising. The non-homogeneity of
the material, together with the very short height of the specimen, results in an uneven non-uniform strain distribution across the
section which projects on the surface deformation. In such a case it is not possible to determine the characteristics of a
classical material corresponding to the “modulus of elasticity”, and the overall deformability measurement probably represents
the best “engineering” estimate. Nevertheless, the optical method provides the best opportunity to study the differences in
deformability of such a non-homogeneous material in relation to its structure, and helps to decide about homogenization
constraints. The optical results are in reasonable accord with the strain gage measurements.
3.
Measurements of physical characteristics of consolidated sandstone
The last example describes tests to detect changes in subtle material characteristics – thermal and moisture expansion of
sandstone – due to treatments using various products based on ethylsilicates. Again, only small samples were available, taken
from the original medieval structure of St. Vitus’ Cathedral in Prague. The task was made more difficult by the fact that the
tests had to be carried out within a period of just a few weeks [6]. The authors had to compare the performance of several
ethylsilicate products when used for consolidating green sandstone of the glauconitic type.
A set of tests was designed and
performed – from mechanical
(bending) tests on thin plates,
which aimed at determining
changes in mechanical characteristics in the depth profile,
through non-destructive ultrasonic tests and vapour permeability tests, to the tests of thermal and moisture expansion
characteristic that are described
in detail in this paper.
The task of detecting and measuring minute changes in dimensions, even smaller than the
surface and shape irregularities
of the specimen, meant that the
experiment had to be performed
in a single sequence, without
handling or repositioning the
specimen. The reason for this
Figure 7 Schematic drawing of the experimental setup, with details of the special rig where
requirement
becomes
clear
the sandstone plate specimen is placed, and a detail of a screen depicting the shape and
when we considers that the
intensity variations of the laser beam spot (the square size in the checker pattern in the
sandstone specimens consist of
screen frame is 1cm).
grit held in matrix, and the
individual grains of the grit
are typically 0.1mm in diameter. On the other hand, the
studied phenomenon - dilatation of the specimen - is in
the range of a few micrometers. Therefore the order of
difference between the two
values makes it impossible to
use anything other than a
relative measurement of the
length increment or decrement. Other approaches
would lead to the magnitude
of the error completely obscuring the signal, i.e., the
results.
Figure 8 Merged and time-synchronized plots of calculated position of the spot center on the
screen (solid line, crosses) and the average temperature of the specimen (dotted line, stars).
Similarity of the two plot shapes is an implication of linear relation between elongation translated in spot position and specimen temperature. In the upper inset of the figure is thermogram of the specimen at the beginning and at the end of the measurement.
Changes in the length of
small sandstone plates (15 x
40 x 3 mm) were measured
while they were cooling from
a preheated state to the ambient temperature of the laboratory, in the case of the
thermal dilatation study, or
while they were drying from
the saturated state to natural
moisture
content
corresponding to the temperature and relative humidity in the
laboratory. In both measurements, the specimens were
placed in the special rig and
optical system shown in Figure 7.
During the thermal dilation tests, the
changes in temperature were recorded
using the AGA Thermocamera. The optical system contained a mirror which
rotated in accordance with the
variations in the length of the specimen
and reflected a focused laser beam on
a screen. The movement of the laser
beam spot on the screen was recorded
in a digital camera with one pixel resolution (this was equal to 0.06 mm). Simultaneous recording of the position of
the light spot and the temperature field
in the specimen is vital for the measurement, as detailed knowledge of the
time evolution of the parameters substitutes for controlled conditions.
In the evaluation procedure, the data
from the image analysis must be connected with the data from the cooling
(or drying) records. This connection was
made with the use of specially developed software tools by means of time
synchronization (as shown in Figure 8).
Figure 9 Thermal expansion coefficients of sandstone for various treatment and
The position of the laser spot on the
distance from surface
screen was determined as the center of
gravity of the light points weighted by their intensity. The temperature change (cooling) was recorded in intervals of 10 seconds
over a period of ten minutes. From the thermograms, we calculated the average temperature of the specimen, (during cooling
a temperature gradient of several degrees was observed in the inclined specimen – clearly visible in the thermogram in Figure
8), and synchronized with the deformation measurements. The calculated values of the thermal expansion coefficient for various treatments of the sandstone, as a function of distance from the surface, are presented in Figure 9. It is seen that the
treatment has a very moderate influence on the thermal expansiveness. It seems that for some consolidants the thermal expansion coefficient is slightly higher than for virgin material (by 50% at most). However, a product was observed that may
cause the opposite behavior. A more detailed discussion is beyond the scope and intention of this paper.
The sensitivity and accuracy of the method presented here will be further tested on a larger series of samples of natural stone.
It is not possible to use a limited amount of very scarce historic material for this purpose.
4.
Conclusions
All three case studies presented here show that the intelligent application of relatively simple testing techniques combined with
sophisticated software tools is effective when dealing with complex problems for which standard procedures are not appropriate, and when limited time is available.
5.
Acknowledgement
The authors gratefully acknowledge support from ITAM institutional research plan grant AV0Z20710524, grant of the Czech
Ministry of Culture PK00-P04-OPP15, and grant of the Czech Grant Agency 103/06/1609. We are specially grateful Jaroslav
Lesák, MSc., for carrying out careful thermal measurements and evaluating the test data, and to Mgr. Petr Měchura from the
Prague Castle President’s Office for his fruitful collaboration.
6.
References
[1] Drdácký, M., Michoinová, D., Procházka, P.: Mortar mixtures with fibers for restoration of monuments (in Czech), Research
report of ITAM ARCCHIP, Prague, 146 p., 2002
[2] Drdácký, M., Michoinová, D.: Lime mortars with natural fibers, in „Brittle Matrix Composites 7“ Proceedings of the 7th Int.
Symposium (A.M. Brandt, V.C. Li, I.H. Marshall, eds.), pp.523-532, Woodhead Publishing Ltd./Zturek Research Sci.Inst., Cambridge and Warsaw, ISBN 1-85573-769-8, ISBN 83-917926-6-8, 2003
[3] Vavřík, D., Drdácký, M.: Experimental determination of stress-strain dependence for thin fibers, CD ROM "Engineering
Mechanics 2003", National Conf. with Int. Participation, ISBN 80-86246-18-3, ITAM CAS CZ, Prague, 2003
[4] Drdácký, M., Slížková, Z., Zeman, A.: Analysis and restoration of an exterior plaster floor of the 19th Century, In Proceedings “Heritage, Weathering and Conservation”, Fort, Alvarez de Buergo, Gomez-Heras & Vazquez-Calvo (eds.), Taylor &
Francis Group, London, ISBN 0-415-41272-2, pp.961-968, 2006
[5] Schueremans, L.: Probabilistic evaluation of structural unreinforced masonry, PhD Thesis, Katholieke Universiteit, Leuven,
December 2001
[6] Drdácký, M., Slížková, Z., Valach, J.: Influence of ethylsilicate consolidants on the behaviour and properties of glauconitic
sandstone used for construction of St. Vitus’ Cathedral in Prague (in Czech), Research report ITAM ARCCHIP, Prague, 39 pp,
2006