Professional Documents
Culture Documents
CITATIONS READS
6 40
9 authors, including:
All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Markus Krger
letting you access and read them immediately. Retrieved on: 10 September 2016
Materials and Structures (2011) 44:10471062
DOI 10.1617/s11527-011-9734-9
RILEM 2011
repeating measurements, e.g. with a measurement the signals that are directly transmitted through the
interval of 15 min, the method allows for a quasi concrete. The actually needed distance depends on
continuous evaluation of hardening concrete right the type of concrete and has to be predefined before
after mixing. testing (see below). A slower mix (e.g. cellular
concrete) requires a longer side wall. For a rough
estimation of the needed distance between the
fixations the following formula could be used:
3 Testing devices
VPMMA
lmin d 2
Vmin
3.1 Principles of container geometry
at which lmin is the clear distance between the
The container has two parallel side walls made out of fixations, d the distance between the two side walls,
a transparent polymer material (polymethyl methac- vPMMA the pulse velocity of the compressional wave
rylate). A principle is shown in Fig. 1. The thickness of the side walls and vmin the minimum compres-
has to be at least 10 mm. In the middle of the side sional wave velocity of the concrete that is expected
wall some recesses allow for the insertion of nylon to be measured during the test.
wear caps in which the sensors (preferably Panamet- The fixation of the walls has to be demountable for
rics type V601-RB or V602-RB) are screwed. The calibration and cleaning. This ensures also that the
thickness of the plastic caps is app. 1 mm. Between container is reusable. The width of the U-shaped part
the sensors and the plastic caps some silicone oil has can be smaller for investigations of mortar and have
to be inserted to achieve adequate coupling. Between to be larger for concrete according to the maximum
the side walls the bottom part of the containerout aggregate size. The minimum width and height of the
of U-shaped closed cell EPDM foamis clamped in U-shaped part must ensure test specimen dimensions
which the fresh concrete can be filled. The foam has that exceed the sensor dimensions and sensor area. To
to be soft and water-impermeable and should be ensure proper sealing the width of the compressible
tested with respect to its acoustic wave propagation foam has to be in minimum the distance between the
properties before being used. The distance between side walls. The distance between the side walls has to
the fixations of the side walls is such that waves be at least three times the biggest aggregate size of
possibly propagating through the walls and the the mixture. An optimization of the container size is
fixations after signal excitation do not interfere with recommended to minimize the amount of wasted
material as well as to minimize the transit path of the
ultrasonic waves along with their attenuation. A
larger container can generally also be used to test
small grain size materials. However, every change in
the test setup and test container may influence the
results from the measurements. It has to be noted that
small distances between the side walls require high
sampling rates to assure required accuracy (see Sect.
6.2.3).
15 22 15
52
15 15 40 15 15 15 22 15
65
65
70
200
Fig. 2 Mortar container for materials with grain size of not larger than [ 4 mm (measures in mm) [1]
3.3 Container geometry for cementitious material 3.4 Container geometry for cementitious material
of larger than 4 mm grain size with low pulse velocity
Figure 3 represents the required dimensions of a The small and the big containers presented in the last
recommended container for materials with larger sections were recommended to be used for materials
grain sizes ([ larger than 4 mm but less than with pulse velocities of not less than app. 250 m/s. If
20 mm). The distance between the side walls is lower pulse velocities are expected, the distance
60 mm what results in a volume of about 450 cm3 for between the fixations has to be increased according to
the tested material. Eq. 2 and checked by preliminary tests.
than 20 mm (measures in
mm) [1]
20 28 80 28 20 15 60 15
130
130
136
400
1050 Materials and Structures (2011) 44:10471062
Fig. 4 Frequency transfer functions (sensitivity over frequency) of piezo electric sensors with different characteristics
Materials and Structures (2011) 44:10471062 1051
3.6 Description of the measurement device (see Sect. 5.3). An even shorter sampling interval is
recommended if the precision of the onset time
A DataAcQuisition (DAQ) system has to be used for determination (see Sect. 6.2.3) is not good enough as
logging data and simultaneously as a triggered function can be seen from velocity/time graphs. In these cases a
generator which enables to record the start of the digitizer with at least 5 MHz sampling frequency at 12
outgoing signal for example by using a positive slope bit amplitude resolution would be required.
with amplitude of 4 V as a trigger pulse (see Sect. 5.2).
Figure 5 describes a typical test setup. For most
applications the test system consists of a frequency 4 Sample preparation
generator, a power amplifier, a measurement container
and a preamplifier as well as of an A/D converter and a Testing of the setting and hardening of different
computer. For the pulse excitation it is recommended materials requires reliable conditions with respect to:
to use hardware that could generate rectangular pulses Ambient temperature (20C recommended for
with amplitudes up to 800 V and a signal length down standard test procedure)
to 2.5 ls to excite broadband pulses. Excitation with Mixture composition and ingredients
half sine waves could be used as an alternative, but Mixture equipment and mixture procedure
results on smaller bandwidth of the excited pulses. In Filling the container and compaction of the
most cases it might be good to interconnect a pream- material
plifier between the receiving transducer and the A/D Storage of the container during the testing.
converter to improve the signal quality. In regard to
recording signals two parameters are crucial: the To keep variation in the testing procedure as low as
sampling interval s (ls) and the number of samples possible it is recommended to follow the standard EN
n. The measurement time is then equal to the total 196-1 [2] for most of the above mentioned aspects, i.e.
recorded time of each ultrasonic signal and is calcu- the arrangements for testing cementitious materials
lated by the product of the two. While the sampling are well defined. Special arrangements could be made
interval should not be larger than 1 ls (i.e. less data with respect to the individual testing objectives.
points) it is recommended to record signals with not However, such other arrangements should also be
less than 1 ms length including the pre-trigger time traceable, clearly specified and documented.
Pow
werr am
mplifier
Pre-a
amplifier
onta
Co aine
er
Fre
equ
uency gen
nera
ato
or
A//D ccon
nverterr
ata accqu
Da uisittion
n
Co
omp
pute
er
4.1 Mixing the material be used to allow self-compaction during the flowing
process and to achieve homogeneous filling.
A traceable mixing procedure is needed for all kinds The filling height of the container should be kept
of mixtures to be tested. If the mixture is prepared for within the following limits (see Fig. 6):
the ultrasonic testing only, a minimum amount of
For the small container: hmin = 40 mm; hmax =
0.5 l of material should be mixed. Mixing procedure
47.5 mm
as described in EN 196-1 is recommended if
For the big container: hmin = 80 mm; hmax =
applicable.
100 mm.
4.2 Filling the container and compaction
of the material 4.3 Protection of the test specimen against
evaporation and specimen storage
The container should be moulded immediately after
the preparation of the mix. For stiffer mortar as well The filling height should be about 5 mm less than the
as for concrete mixtures that have to be compacted it height of the container. To avoid evaporation and
is recommended to follow the procedures mentioned drying out the mix during the test period the
in EN 196-1. For most of the ultrasonic testing cases specimen should always be sealed with adhesive tape
the procedure should be the following: or equivalent material. Covering the specimen right at
the beginning of the test with a thin water layer
Fix the container to a jolting apparatus or other
(23 mm) might also be effective. Former tests have
reliable compaction equipment similar to that
shown that this does not affect the mixture compo-
described in EN 196-1 and its Appendix
sition (especially W/C ratio) neither the velocity
Fill the container with the first of two layers of
measurements significantly [1].
material
The container should be stored on top of an EPDM
Compact the first layer with the compaction
rubber foam similar to that of the U-shaped form of
apparatus
the container. If ambient temperature conditions are
Fill in the second layer of material
not guaranteed to be constant over the testing period
Compact the second layer with the compaction
additional temperature measurements are obligatory
apparatus.
for validation purposes.
For free-flowing or self compacting mixtures no
compaction is needed. However, for filling the
container a trough with an adequate length has to 5 Time signals and determination of the transit
time
0,5 15:10
0,0
-0,5
-1,0
0,9 0 2
0,6
0,3 19:10
0,0
-0,3
-0,6
-0,9
0 1 2 3
time in s
characterize the setting and hardening process (see together with the distance between the side walls the
Sects. 7, 8). basis for any pulse velocity calculations (see Sect. 6).
As will be discussed in Sect. 5.3 it is recom- The outgoing time corresponds to the trigger time of
mended to do repetitive signal recordings in a time the piezoelectric transmitter sensor. A trigger pulse
window of not more than one second to reduce the (e.g. transistortransistor logic, TTL) is used to start
statistical error. The average signal usually has a the emission of the pulsethe emitting part of the
p
better signal-to-noise ratio by a factor of n; if n is ultrasound equipment has to be able to produce this
the number of repetitions. The number of repetitions TTL signal with an accuracy of better than 1 ls and
has to be stored along with the data as well. to record it. Any delay between the trigger time and
the actual emission time of the signal from the first
side wall needs to be subtracted according to Sect. 5.4
5.2 Measurement of the emission time with at least the same precision as the TTL pulse
time. Such a delay can occur due to the time
The transit time is the time difference between necessary to travel through the container wall and
outgoing and incoming signal. The transit time is the sensor case.
1054 Materials and Structures (2011) 44:10471062
Pn
5.3 Measurement of the arrival time n1 ttot;n
ttot s
n
The transmitted signal is recorded by a sensor 0
This results with ttot from Eq. 2 in
attached to the opposite side wall. This signal has Pn 0 Pn
to be converted from analogue to digital and to be n1 ttot;n t tot;n t c
t0 tot n1
s; 3
recorded by a transient recorder. The most crucial n n
part is then to extract as accurate as possible the onset
where t0 tot is the sum of the corrected total transit
time (i.e. the arrival time) of the signal in relation to
times. For simplification this is further on called t:
the trigger time. If the total recording time ttot is 1 ms
(see Sect. 3.6) a pre-trigger time interval of at least t t0 tot s: 4
15% (i.e. 150 ls) has to be set to establish for a
In most cases a repetition of three times (n = 3) is
reliable onset time detection. These first 15% of the
good enough, but it depends on the type of material
signals can be considered as noise. The onset of the
under investigation.
signals has to be discriminated from noise. There are
various possibilities to do this since there are,
threshold and energy based methods and auto-
6 Pulse velocity
regressive processes. An example how to use the
Akaike Information Criterion (AIC) [3] as a detector
6.1 Calculation of the pulse velocity
algorithm is given in the Appendix.
It is recommended to do repetitive measurements
When the average corrected total transit time t has
of ttot in a time window of not more than one second
been calculated the pulse velocity v of the compres-
to reduce the statistical error (see Sect. 5.4). The
sional wave yields analogue to Eq. 1
resulting ttot value has to be recorded together with
the estimated error that consists of a standard v L=t m/s; 5
deviation and a systematic error. Regarding the error
with L the clear distance of the container walls. Since
analysis the ones considered by the procedure
L is subjected to errors this effect has to be taken into
described in Sect. 5.4 (averaging of ttot) as well as
account for the error estimation of the total error.
the ones due to measurements of tc (see following
section) and L have to be included.
6.2 Calibration, accuracy and reproducibility
of the pulse velocity
5.4 Correction of the transit time 6.2.1 Calibration of the time/velocity measurement
When the total transit time ttot has been measured the As mentioned in Sect. 5.4 the time the signal needs to
real transit time has to be computed by subtracting travel through the sensor case and the walls of the
the time which is necessary to travel through the concrete/mortar container have to be measured prior
container wall and the sensor case tc (see Sect. 6.2.1; to investigations of setting and hardening. Usually,
0
calibration). So the real transit time ttot is this calibration time tc is in the range of several ls,
0
ttot ttot tc s 2 but it is necessary to subtract this time, because its
influence is increasing during the setting observation
tc has to be determined on the container material reciprocal to the reduction of the travel time. This
according to Sect. 6.2.1. Afterwards the mentioned calibration should be redone from time to time or
repetitive measurements of ttot in a time window of even prior to each measurement campaign. It is done
not more than one second to reduce the statistical by removing the U-shaped rubber foam part of the
error have to be done. If n (with n C 3) is the number container. Both side walls are pressed together and
of these measurements the influence of noise is the time tc is determined without any material in
p
reduced by this averaging by a factor of 1= n and between. If repetitive measures of tc show deviations
then ttot results in: an average value can be calculated similar to t.
Materials and Structures (2011) 44:10471062 1055
6.2.2 Calibration of the sensors 6.3 Presentation of the pulse velocity over setting
time
The calibration of the sensors used can be done
according to international standards. We recommend For the presentation of the pulse velocities over
the phase-to-phase calibration [4] because it is easy to setting times a graph showing the evolution of the
be implemented and can be done with less time and velocities calculated as described in Sect. 6.1 and
little expenses. based on the onset time data measured as described in
Sect. 5 can be plotted. Usually the velocity is a
6.2.3 Accuracy and reproducibility of the velocity function of the setting time as well as of the material
calculations tested velocity plots of different materials can be
combined into one graph. An example for material
It also has to be ensured that the distance L is not mixes with different additives added to a standard
changed throughout the measurements. Usually, the CEM 42.5 N is given in Fig. 8.
measurement of L isin addition to tsubjected to The curves can be represented in different ways
errors as well. If that is the case, error propagation including the single data points or just showing the
algorithms considering L and t have to be used to data curve. It is valid to smooth the plots by using
determine the overall accuracy of v. The error interpolation algorithms as long as the curvature is
calculated in this way represents the reproducibility not significantly changed. In this case it is mandatory
of the velocity determination. to mention the smoothing of the plots. This could in
If in particular during the initial phase of the particular be helpful for the higher velocity values
measurements the calculated velocity values are not where the accuracy is usually lower (see Sect. 6.2.3).
changing and/or are abnormally high, it has to be Generally, the determined errors should also be
checked, whether the side wall distance is too small represented in the graph.
resulting in waves travelling around the container
instead of through it (see Sect. 3.1).
The transit time gets significantly shorter during 7 Transmitted energy ratio
the setting process. The determination of onset times
is therefore more and more difficult the lower the 7.1 Calculation of the energy ratio
t value becomes. A sampling interval (see Sect. 3.6)
short enough to enable for accurate onset picking is The transmitted ultrasonic energy E is determined by
recommended. numerical integration of the squared amplitude values
following the onset time.
E Rx2i with i ionset . . .N 6
3500 accelerator
xi is the amplitude at sample point i of the signal with
no additives
N sampling points.
3000
This value is then divided by the reference energy
pulse velocity [m/s]
air entrainer
2500
Eref to eliminate the energy loss due to divergence
2000 (beam spreading) and reflection at the contact inter-
1500
faces (transducercouplant, couplantplastic wear
retarder cap, plastic wear capsample). The reference energy
1000
is measured during the calibration of the set-up. The
500 reflection at the plasticsample interface changes in
0 time as the mortar or concrete sets so it cannot be
0 500 1000 1500 2000 completely eliminated.
age of concrete [min] To allow an unambiguous determination of the
energy ratio, the following specific settings for the
Fig. 8 Example of velocity development over setting time of
concrete (age of concrete) depending on the type of additives transducer RB-602 were obtained from comprehen-
added to a standard mix (curves are smoothed) sive examinations and are preferred to be:
1056 Materials and Structures (2011) 44:10471062
Rectangular pulse excitation: Excitation pulse 7.3 Representation of the energy over setting
length: 2.5 ls (mortar samples), 5 ls (concrete time
samples); Excitation pulse amplitude: at least
400 V (mortar samples), 750 V (concrete samples) Analogously to the velocity measurement, a graph
Recording length: 1.638 ms (0.164 ms pretrigger, showing the change of the energy calculated as
1.474 ms recorded signal). described in Sect. 7.1 can be made (see Fig. 9). The
energy is plotted on a logarithmic scale in correspon-
dence with the wave attenuation which is commonly
7.2 Calibration, accuracy and reproducibility
expressed in dB. Generally, smoothing of the energy
of the energy ratio
curves is not required to produce clear graphs.
7.2.1 Calibration of the energy measurement
7.4 Criteria to determine initial and final setting
The reference energy Eref is measured with the
The initial and final setting times according to the
same set-up and settings as during actual measure-
penetration resistance test (ASTM C403) [5] can also
ments, but with the container filled with water. The
be indicated by threshold values of the ultrasonic
average value of at least six measurements of Eref is
wave energy [6].
calculated.
For mortar mixtures:
tinitial set h tE=Eref 0:02 h 7a
7.2.2 Accuracy and reproducibility of the energy
calculations tfinal set h tE=Eref 0:13 h 7b
For concrete mixtures:
Shrinkage of the tested specimen can cause the
ultrasonic transducers to decouple from the sample. tinitial set h tE=Eref 0:02 h 8a
Due to the large impedance mismatch between air tfinal set h tE=Eref 0:07 h 8b
and concrete, a small air gap can already cause the
energy transmission coefficient to decrease signifi- For mixtures with OPC, an average repeatability
cantly. Due to this large sensitivity of the energy error lower than 5% can be obtained for these
measurements to poor sensor contact, care should be threshold times.
taken to limit drying shrinkage (see Sect. 4.3).
In the period ranging from 4 to 48 h after mixing,
an average repeatability error lower than 7% can be 8 Frequency spectrum
obtained.
8.1 Calculation of the transmitted frequency
spectrum
40
38
36
34
32
Temperature [C]
30
28
26
24
22
20 CH 3 [C]-Dewar
CH 4 [C]-Specimen
18 CH 2 [C]-Air
16
0 5 10 15 20 25 30 35 40 45 50
Concrete age [h]
Fig. 10 Contour plots of a fresh concrete sample (w/c = 0.5,
CEM I 52.5): the frequency content is presented by the contour Fig. 11 Example showing the evolution of the hydration
lines in function of the concrete age and the frequency (-54, temperature in a Dewar container and in the specimen
-70, and -100 dB(lV)) compared to the ambient (air) temperature
1058 Materials and Structures (2011) 44:10471062
Fig. 13 Example
describing the AIC
calculations
1060 Materials and Structures (2011) 44:10471062
(DFT) is applied. The fast Fourier transform (FFT) is component is a triple product of vectors, which can
a fast algorithm to calculate the DFT when the be called score vector (sample mode) and loadings
number of samples N is a power of 2. If necessary, (other modes) by analogy with principal component
zeros are added at the end of the signal. analysis (PCA). A slab Xk (I 9 J) of X can be
described by Eq. 16.
Time domain FFT Frequency domain Xk B D AT Ek 16
Signal versus time ? Signal versus frequency
where B is the I 9 R matrix of loadings, A the J 9 R
Sampling interval Frequency resolution matrix of loadings, Dk the R 9 R diagonal matrix
Dt = 1/fs Df = fs/N = 1/(NDt) containing the scores for slab Xk and Ek the I 9 J
Number of samples Number of samples matrix of residuals. The main advantage of PARA-
N N FAC is the uniqueness of the solution.
Real signal Complex frequency content If the peaks in the spectrum measurements shift
x[i] X[i] with |X[i]| = |X[N - i]| and from sample to sample, the trilinearity condition is
phase(X[i]) = phase(X[n - i])
not fulfilled and PARAFAC will give degenerate
solutions with unstable and unreliable parameters. A
The DFT generates the two-sided spectrum which modified version of PARAFAC known as PARA-
is symmetrical in the amplitude. Therefore, only the FAC2 can model data with such shifts in one mode.
first half of this spectrum contains useful information. Unlike PARAFAC, the latter model does not assume
Equation 9 shows the computation from a two-sided parallel proportional profiles from sample to sample
DFT to a single-sided amplitude spectrum. in the first mode and is therefore described by Eq. 17.
p
2 Magnitude DFTxi Xk Bk D AT Ek with Bk BTk constant over k
Xi for i
N 17
N
1 to 1 14
2 which is identical to the PARAFAC model in Eq. 16,
Magnitude DFTxi except that Bk is not fixed for all the slabs Xk.
Xi for i 0 15 To select the optimum number of PARAFAC2
N
components R, the explained variance can be calcu-
Multi-way analysis of the received frequency spectra lated for an increasing number of components
according to Eq. 18.
The dataset with the change of the frequency
spectrum in time consists of a 3-way array: the explained variance % 100 1 Rij e2ij =Rij dij2
frequency content is measured in function of concrete
18
sample (first mode), frequency (second mode) and
concrete age (third mode). In Eq. 18 dij is an element in the raw data and eij is
The data should be centered across the second the corresponding element in the residuals after the
mode, i.e. the frequency mode, to remove possible modelled variation is removed. The explained vari-
off-sets. To reduce the dataset, only the frequency ance indicates how well the model approximates to
band of the spectrum containing the most energy the raw data. The optimum R is achieved if the
should be selected based on the generated pulse explained variance does not increase significantly if
signal (e.g. first 250 kHz for pulse of 5 ls). more components are added.
For each model, also the explained variance after
Exploratory multi-way analysis cross-validation should be calculated which can be
seen as a measure for the ability to predict new data.
The most simple and restricted multi-way model is Samples can be successively left out and the models
parallel factor analysis (PARAFAC) [810] which build upon these reduced datasets are used to predict
decomposes a three-dimensional array X (I 9 J 9 K) the left out elements and calculate the cross-validated
into a summation of R trilinear components. Each explained variance. The optimum R is achieved if the
Materials and Structures (2011) 44:10471062 1061
cross-validated explained variance stagnates or validation set. The model constructed by the cali-
decreases if more components are added. bration set is then used to predict the validation
Classification of concrete mixtures is achieved set. The root-mean-square error of cross-validation
through visualization of the samples in score plots (RMSECV) for a variable is generally defined by
which are the projection of the data onto the Eq. 20.
PARAFAC components. For instance, the scores of p
the second, third and fourth component are plotted in RMSECV Ri y0i;cv yi 2 =N 20
function of those of the first component. The location where y0i;cv is the value of the predicted variable with
of the samples in the score plots are then determined
sample i not included in the model formulation.
by the setting process. The distances between the
Additionally, also the coefficient of determination R2
scores of two samples indicate how much the change
can be used to indicate the accuracy of the prediction.
in frequency spectrum from these samples differs.
The optimum number of components R is achieved
if the prediction errors increase or stagnate if more
Basics of multi-way regression analysis components are added.
Multi-way analysis can be performed to relate the Example of multi-way regression analysis
change in the ultrasonic frequency spectra, compiled
in the predictor three-dimensional array X (I 9 J An example of the change of the received frequency
9 K), to other measured quantities such as the setting spectra in time is shown in Fig. 14a for 3 out of 34
times, compiled in the response matrix Y (I 9 L). concrete samples studied by Robeyst [1]. Initially,
Multi-way partial least square (N-PLS) regression, when the concrete is still a fluid-like suspension, lower
finds a linear regression model by projecting X and Y frequencies of only lower amplitudes can be detected,
to a new space. Analogously to PARAFAC, X and Y while the high frequencies are completely damped. The
are decomposed into R components, which are also high frequencies gradually appear when the concrete
called latent variables (LVs). They contain all the starts to stiffen. The frequency spectrum broadens
relevant information, while redundant data and noise fastest for the Portland cement concrete without
are left in the residuals. additives (Fig. 14a, #2) and slowest for concrete with
To select the optimum R, the errors between the blast-furnace cement type III/C (Fig. 14a, #12) or with
predicted and the actual values can be calculated in high amounts of BFS or FA added separately (Fig. 14a,
several ways. The root-mean-square error of calibra- #23). This change of the frequency bandwidth in time
tion (RMSEC) for a variable is generally defined by can be compared with the evolution of the velocity,
Eq. 19. described in earlier publications.
p To quantify the differences between the spectra of
RMSEC Ri y0i yi 2 =N 19
the samples, multi-way analysis was performed as
where y0i is the value of the predicted variable when described in the following.
all samples are included in the model formulation, N The dataset consists of a three-dimensional array:
the number of yi values. The RMSEC is a measure of the frequency content was measured in function of
how well the model fits the data. concrete sample (first mode), frequency (second
The models ability to predict new samples is mode) and concrete age (third mode). Several
investigated by performing a cross-validation. The multi-way models were tried out, but the best results
calculation is analogous, except y0i is the prediction for were achieved with PARAFAC2. This model is an
sample i if sample i is not included in the model extension of the more common PARAFAC model
formulation. Two methods can be used to cross- which decomposes a three-dimensional array into a
validate the model. In the leave-one-out (loo) method, summation of triple products of vectors (called
every sample is successively excluded. The model components) such that the greatest variance of the
constructed by this reduced dataset is then used to data is captured by the first component [11]. Each
predict the excluded element. In the split-half method, component is the product of a score vector and two
the dataset is split into two parts: a calibration and a loading vectors. The loading vectors can be seen as the
1062 Materials and Structures (2011) 44:10471062
Frequency (kHz)
the frequency content #2: CEM I 42.5
200
versus concrete age and (b) 0.35
150
frequency. b PARAFAC2 100 0.30 11
Score plots for concrete 50
8
7 6 12 23
samples with blast-furnace
12 24 36 0.25
cements (filled square), 10
Frequency (kHz)
#12: CEM III/C 32.5 33
blast-furnace slag addition 21
Component 2
200 9 20
(filled triangle) and fly ash 0.20 22
150 32
addition (filled circle). The 100 34
colour indicates the BFS or 50 0.15 slower
setting
FA content from black (no 26 19 behaviour
12 24 36 29 30
BFS or FA) to white (high 0.10 27
Frequency (kHz)
#23: 85% BFS 4
5
24
content) 200 28 17 14
25 18 2
150 13 3 1
0.05 16
15 31
100
50
0.00
0.05 0.10 0.15 0.20 0.25 0.30
12 24 36
Component 1
Concrete age (h)