You are on page 1of 15

Construction and Building Materials 369 (2023) 130569

Contents lists available at ScienceDirect

Construction and Building Materials


journal homepage: www.elsevier.com/locate/conbuildmat

Correlation between compressive strength of concrete and ultrasonic pulse


velocity: A case of study and a new correlation method
Ramón Mata a, Rafael O. Ruiz b, Eduardo Nuñez a, *
a
Department of Civil Engineering, Universidad Católica de la Santísima Concepción, Concepción, Chile
b
Department of Mechanical Engineering, University of Michigan-Dearborn, Dearborn, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Concrete strength assessment is an important topic in evaluating existing structures. Formerly, only destructive
UPV test tests were employed, limiting the number of tests due to their complexity and cost. Nowadays, the application of
Assessment of existing structures non-destructive tests has been booming to determine material strength, offering a more accessible and cheaper
Concrete strength
strategy than its counterpart. Non-destructive strategies are based on two steps: (1) the identification of the
Concrete core
Destructive testing
correlation between the concrete strength and another parameter that is easy to measure in situ, and (2) the use
In-situ concrete strength evaluation of this easy-to-measure parameter to infer the concrete strength in any desired element of the structure. The most
common parameter adopted for this purpose is the Ultrasonic Pulse Velocity (UPV). However, the correlation
between concrete strength and UPV must be determined via destructive experiments. From the research
perspective, attention has focused on determining the correlation coefficient and the range of credibility for
estimating the inferred concrete strength. Despite it, this strategy has remained elusive in the fundamental
understanding and accounting of the joint dispersion of the concrete strength and the UPV. The present work
addresses this knowledge gap by proposing a new correlation method based on probability interpretations to
infer the compressive concrete strength from in-situ UPV measurements and including the dispersion evidenced
in UPV measurements in both steps mentioned. The results demonstrated that it is possible to determine the
confidence interval for the concrete compressive strength given a certain percentile of the UPV measured in situ.
Finally, the application of the proposed method is illustrated through a case study, which is representative of
different building pathologies. This novel proposal is a foundation to deal with the uncertainties involved in non-
destructive tests, inspiring future advances in this field.

1. Introduction extracting concrete cores of the structural system and testing them at the
laboratory. However, the number of concrete cores is typically reduced
The structural intervention of actual buildings represents one of the since their extraction could be complex, slow, and costly. On the other
most important tasks in modern civil engineering practice. These in­ hand, other non-destructive tests have recently gained attention, for
terventions are usually motivated by construction regulations, design example, measuring Ultrasound Pulse Velocities (UPVs) in situ to esti­
codes, changes in the building’s extension and use, demolitions, and mate concrete strength or another mechanical characteristic indirectly.
retrofitting. The current design codes have evolved, demanding a careful However, the results of the non-destructive tests are commonly based on
structural analysis that ensures these interventions’ structural safety and correlations between the lecture of the instrument and the target
feasibility. Here, the standard practice is to develop numerical structural property, which introduces an uncertainty in the mechanical property
models such that the structure’s geometrical and mechanical charac­ identified, especially in concrete structures [1].
teristics are accounted for to identify the structural performance under The use of non-destructive tests based on UPV has been applied in
service and extreme loads. In this sense, an adequate determination of many concrete composites and structures. For example, UPV and con­
the current mechanical properties of the materials involved is crucial for crete strength correlations have been proposed and studied for concrete
a reliable numerical performance prediction. The material’s properties based on recycled waste liquid crystal display glass [2], aggregate
identification is commonly carried out by destructive tests, i.e., seawater [3], rubbers containing silica-fume-zeolite [4], lightweight

* Corresponding author.
E-mail address: enunez@ucsc.cl (E. Nuñez).

https://doi.org/10.1016/j.conbuildmat.2023.130569
Received 19 October 2022; Received in revised form 16 January 2023; Accepted 27 January 2023
Available online 9 February 2023
0950-0618/© 2023 Elsevier Ltd. All rights reserved.
R. Mata et al. Construction and Building Materials 369 (2023) 130569

[5], segregated lightweight [6], and standard concrete [7]. The results the methodology does not allow the evaluation of the concrete strength
obtained in these studies revealed the robust capability of the UPV to variability from non-destructive tests.
infer concrete strength. However, the adequacy of this technique relies The destructive and non-destructive tests have also been applied in
on proper identification of the UPV-concrete strength correlation, which historical buildings. For example, the National Museum of “Magna
in cases of using nonstandard aggregates could be required a nonlinear Grecia” in Reggio Calabria was studied in Pucinotti [8]. Here, 36 con­
regression model [2–4]. Additionally, the use of UPV has also been crete cores were extracted from structural elements to assess the
studied to predict crack patterns and propagations using support vector building’s concrete strength. Additionally, 100 UPV measurements and
machines [7], to track concrete damage based on an ad hoc index [8], 64 rebound hammer tests were realized. Both non-destructive methods
and to track concrete strength degradation due to extreme temperatures (UPV and rebound hammer) were properly calibrated (adopting the
[9]. However, the most common application of this non-destructive method presented in ACI-228.1R [10]) using the concrete strength ob­
technique is oriented to asset the concrete strength in structures where tained in the destructive tests, obtaining a better correlation for the UPV,
standard aggregates are employed and where a linear correlation be­ which agrees with the results found in [1]. The results show a significant
tween UPV and concrete strength is feasible. dispersion in the inferred concrete strength and the measured UPVs.
Many researchers have proposed different alternatives for the cor­ Nevertheless, the assessment of new correlation methods was not
relation between concrete strength and UPV, ranging from correction deemed or compared, and the dispersion evidenced in the UPV was not
factors [10] to correlations based on artificial neural networks [6]. explicitly accounted for. A similar study but in a historic masonry
However, practitioners do not prefer techniques based on artificial building was conducted by Sýkora et al. [18]. Here, 15 historic stone and
neural networks since they require a significantly extensive database of clay masonry brick structures were studied using destructive tests to
destructive tests, making their implementation expensive, slow, and identify the correlation with the rebound hammer test. The results show
unfeasible. Therefore, some relevant studies, from the practitioner’s that a non-destructive test can be used to estimate the masonry strength;
perspective and regarding the correlation between concrete strength however, a more significant number of destructive tests is required to
and UPV, are presented next. improve the concrete strength identification.
In 2013, Pucinotti [8] presented a study using 359 concrete cores As mentioned above, non-destructive tests have been widely
extracted from an existing reinforced concrete structure. The cores were employed for the indirect estimation of concrete strength. However,
subjected to a destructive test to identify the concrete strength. Three after a careful literature review, it is possible to conclude that there is a
approaches according to the NTC 08 Italian Code [11], EN 1990 Euro­ limited number of case studies on real structures containing correlations
code [12], and EN 13791 Eurocode [13] were used to determine the between UPV and concrete strength. Concerning the correlation
characteristic concrete strength, and a new formulation to determine it methods, it is possible to highlight the following aspects: (1) it is
was proposed [11]. The formulation allows the introduction of correc­ observed that assuming a Gaussian PDF for the concrete strength is
tive coefficients to consider the geometry of drilled specimens and the widely accepted [19–24], (2) despite that some works have introduced
disturbance during drilling operations. The results indicate that the spatial correlation for the concrete strength [19,20,25–27], there is no
methodology proposed in EN 1990 [12] offers a lower value for concrete consensus in the adequate model [28], and (3) the dispersion observed
strength in comparison to [13] and [14], indicating a possible under­ in the UPV measurements are not explicitly accounted adopting a sta­
estimation. The formulation proposed by Pucinotti tries to avoid this tistical metric.
underestimation by incorporating correction factors. In the present work, a new correlation method between the
On the other hand, Giannini et al., in 2014, adopted a probabilistic compressive strength of concrete and UPV is proposed and implemented
approach based on Bayesian inference [15]. This approach defines the in an actual structure, presented here as a case study. The proposed
UPV and the concrete strength by probability density functions (PDFs). method assumes the compressive concrete strength and the UPV as a
First, a linear regression with deterministic coefficients between con­ multivariate Gaussian PDF which is calibrated using the results coming
crete strength and UPV is established to estimate the concrete strength from destructive tests. Afterward, the practitioner can use in situ UPV
PDF given the in-situ measurement of the UPV. This strategy guides the measurements to make inferences about the compressive concrete
prior PDF selection for the concrete strength based only on the in situ strength. The most important feature of the proposed method is that the
UPV measurements. Second, a likelihood function is proposed adopting practitioner can incorporate the dispersion of the in situ UPVs in terms of
the same linear regression but conditioning the concrete strength PDF to percentiles while the inference of the compressive concrete strength is
a known UPV. Finally, the measurements of the UPV and the concrete offered based on a probability of exceedance or a particular confidence
strength pairs obtained from the extracted cores allow using the Bayes interval. Ultimately, the method allows the practitioners to select/
theorem to update the concrete strength PDF. The procedure was vali­ identify the compressive concrete strength based on their attitude to­
dated in two real cases with experimental data from more than 20 wards risk. Here, a risk-averse position will look for low UPV percentiles
concrete cores. The primary assumption that justifies the implementa­ and high exceedance probabilities for concrete strength, while a risk-
tion of this approach is that the dispersion of the identified concrete neutral position, for example, will look for medium UPV percentiles
strength in the destructive experiments significantly differs from that and exceedance probabilities for concrete strength. Finally, the imple­
identified in the in-situ tests. Other than that, the approach lacks closed- mentation of the method is illustrated in a case study of an existing
form expressions making difficult its incorporation into the current building and compared with other well-known correlation methods. The
design provisions or the general adoption by the practitioners. Other case study corresponds to an icon close-to-the-sea hotel located in
recent work proposes a new correlation method to analyze the reliability Vargas-Venezuela. The main goal is to study a structure with different
of the concrete strength variability evaluation, considering a core se­ pathologies due to earthquakes, aging, and advanced corrosion pro­
lection based on non-destructive tests (conditional coring) [16]. The risk cesses. Additionally, the details of the destructive and non-destructive
curves concept was adopted to correlate the reliability of the estimation tests are presented, facilitating their use in future studies.
with the considered factors in the analysis. The use of risk curves shows The rest of the paper is organized as follows. Section 2 presents an
that the risk on mean strength assessment decreases as the number of overview of the formulation for the standard correlation method. Sec­
test locations for cores increases. This is because increasing the number tion 3 presents the proposed multivariate correlation model for the
of cores reduces the sampling uncertainty. However, due to its compressive concrete strength and UPV. Section 4 presents the case
complexity, the authors only recommend this approach when the eval­ study and the comparison of the proposed method with well-known
uation of the concrete strength variability is requested. In this sense, correlations. Finally, Section 5 concludes the study.
[17] proposes a methodology that defines the number of cores necessary
to reach a target accuracy of the concrete strength estimation. However,

2
R. Mata et al. Construction and Building Materials 369 (2023) 130569

2. Standard procedure to estimate the correlation between model is the most popular approach, other alternatives can better suit
concrete compression resistance and the UPV particular applications. Breysse in [30] presents an extensive review in
this regard.
The correlation between compression strength and UPV is obtained
by adopting the procedure presented by the ACI228.1R [10], which 2.3. Identification of the in situ UPV
relies on four steps: (i) experimental characterization of the concrete
cores, (ii) identification of the concrete strength and UPV correlation, Once a and b are identified, it is possible to use Eq. (3) to predict the
(iii) identification of the in situ UPV, and (iv) corrections to the corre­
value of fc in any desired element of the structure. In that sense,

lation function to estimate the in situ concrete strength indirectly. A


recording the UPV in the elements of interest is required. In other words,
brief description of this procedure is presented next.
this step involves identifying a set of m measurements of the ultrasound
pulse velocity V = [ V1 V2 ⋯ Vm ]. Then, Eq. (3) could be
2.1. Experimental characterization of the concrete cores
rewritten to estimate the set f c = [ fc1 fc2 ⋯ fcm ] such that:
′ ′ ′ ′

After extracting the hardened concrete cores from the structural el­ ′
f c = a + Vb (5)
ements, the concrete cores should be prepared to obtain the concrete
strength employing a compression test protocol. First, the dimension The regression model presented above is not perfect due to the un­
and weights of the specimens should be established according to [29], certainties associated with fc . In this sense, it is necessary to introduce a

and the edges of the concrete cores should be cut to remove bumps. After correction factor that accounts for the uncertainties in the regression
obtaining the dimensional and physical characteristics of the concrete model. Furthermore, this factor makes it possible to decide whether to
cores, the ultrasound test is applied to take three readings per probe consider the estimate as representative or not. The correction factor
[29]. Subsequently, the capping is applied to the cores at the edges introduced for the estimation of fc is based on the application of the

(upper and lower), enforcing a uniform distributed load on the entire Danish method described in the next section.
surface of the core at the moment of the compression test. Finally, the
compressive strength test to the concrete cores is performed, obtaining a 2.4. Correction to the correlation function
set of n pairs of ultimate compression strength fc and ultrasound pulse

velocity V. For the sake of simplicity, the set of n pairs of fc − V is denoted The Danish method [31] is used here to correct the correlation

as follows: function. This method identifies the ultimate compression threshold



fc− threshold , so the probability of obtaining fc < f c− will be low
′ ′
(1)
′ ′ ′ ′
f c− sample = [ fc− fc− ⋯ fc− ] threshold
1 2 n
(typically lower than 10 %). The concrete strength fc is treated as a

V sample = [ Vc− 1 Vc− 2 ⋯ Vc− n ] (2) random variable defined by a log-normal distribution. In other words, if
Y = log(f c ), then Y follows a Gaussian distribution with mean Y and

2.2. Identification of the correlation between concrete strength and UPV standard deviation σ y . Note that Y and σ y can be computed directly from
the mean and the standard deviation of log(f c ). Here, a finite-sample-

According to [10], the correlation between concrete compression size-population correction factor can be introduced to improve the
strength and UPV is typically established by the equation: method [1]. Additionally, it is possible to define a tolerance interval

fc = a + Vb (3) between Y − kσ y and Y + kσ y . The parameter k is related to the proba­
bility that Y lies inside the interval. The suggested confidence level
where a and b are constant coefficients that define a linear regression. values are 75 % for ordinary structures, 90 % for buildings of excep­
The identification of the coefficients a and b is performed by adopting tional interest, and 95 % for nuclear power plants. For more detailed
the least square method employing the n pairs fc − V identified in step 1:
′ information on how to define k, please refers to [10]. Ultimately, the
concrete strength value used as reference is just fc− threshold , which is

( (( )2 ) )

a, b = min sum f c− sample − a − V sample b (4) defined as follows
( )
(6)

fc− threshold = exp Y − kσy
It is important to note that the correlation quality depends on the
number of samples (number of concrete cores tested), then it is impor­ Although the methodology is widely adopted, it is very restrictive
tant to select cores that represent the entire structure, more details on since it does not entirely offer control over the tolerance interval asso­
this can be found in ACI-228.1R [10]. Although the linear correlation ciated with both parameters: the ultimate compression resistance and

( ′ )
Fig. 1. Behavior of p fc , V for a strong correlation coefficient. a) shows the bi-variate Gaussian distribution for the concrete strength and the UPV assuming a
correlation coefficient of 0.8. b) shows the marginal distribution (blue line) and a conditional distribution (assuming Vo = μV 0.9) for the concrete strength.

3
R. Mata et al. Construction and Building Materials 369 (2023) 130569

fc
fc

V
Po =

V
Po =

V V

Fig. 2. Illustrative implementation of the proposed method for: a) data with strong correlation (0.8), and b) a week correlation (0.2). The data presented was
artificially created. Blue circles represent data from destructive tests. Black and red stars represent the recommended fc for a 0.5 and 0.9 probability of exceedance

flowing the proposed procedure, respectively. Black solid lines represent the Danish recommended fc . The proposed procedure is more conservative than the Danish

method for data with small coefficient of correlation. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

( ′ ) ( ′ )
the UPV. Next, a methodology is proposed that provides a robust esti­ p fc |V =Vo . If p fc , V is assumed as a Gaussian distribution, the con­
mation of the ultimate concrete strength by adopting a specific proba­ ( ′ )
ditional distribution p fc |V =Vo is also Gaussian with mean and vari­
bility of exceedance and an ad hoc percentile for the UPV. ance given by [32]:
( )
3. Proposed method based on exceedance probability μo = μfc′ + ρσfc′ o
V − μV
σV
(9)
The proposed procedure is based on identifying the statistical cor­ √̅̅̅̅̅̅̅̅̅̅̅̅̅
σo = σV 1 − ρ2
relation between fc and V rather than imposing a linear regression.

Previous works have already determined that the UPV and the ultimate Note that the normality assumption (Gaussian distribution) must be
concrete strength could be assumed as random variables normally checked, for example using the Shapiro-Wilk test [33] for fc and V

distributed [15,19–24]. Based on these observations, it is assumed that fc



independently or using the Royston’s test [34] for the multivariate
and V follow a normally distributed probability density function fc ∼ distribution. Also, it is important to mention that the proposed proced­

(μfc′ , σ 2f ′ ) and V ∼ (μV , σ2V ). Here, μ and σ2 denote the mean and the ure is easy extendable to log-normal distribution. The only special
c
consideration is that the analysis must be conducted over the natural
variance, respectively. Additionally, it is expected an important corre­
logarithm of fc and V. Back to equation (9), different scenarios can be

lation coefficient ρ between fc and V; then, it is convenient to define a


evaluated; for example, it is possible to identify the distribution of fc


bivariate Gaussian distribution for fc and V:


assuming that V could reach any value Vo. This analysis is illustrated in
( )
Fig. 1b, where the blue line indicates the unconditional distribution of fc

( ′ ) 1 1 ∑
∑ ̅ exp − (x − μ)
p fc , V = √̅̅̅̅̅̅̅̅ T
(x − μ) (7)
while the red line indicates the distribution of fc for a Vo = μV 0.9 (Vo is

2π | | 2
arbitrarily selected as the 90 % of the mean only to illustrate the point).
with:
In other words, the blue line is the distribution of the concrete strength
x = [ fc

V]
T values assuming that the UPV is unknown. In contrast, the red line is the
[ ]T distribution of the concrete strength value when the UPV takes a specific
μ = μfc′ μV known value.
⎡ ⎤ (8)
∑ σ 2f ′ σ fc′ σV ρ Now, the question here is how to select the Vo properly. First, recall
=⎣
c
⎦ that two different data sets are available. One corresponds to the
σfc′ σ V ρ σ2V destructive test (described in section 2.1), which is used to define the
bivariate Gaussian distribution presented in Eq. (7). The other data set
Eq. (7) can be used to describe the relationship between fc and V

corresponds to the UPV recorded across the building, previously iden­
obtained experimentally in the compression test (explained in section tified as V in Section 2.3. This data set could be used to select Vo based on

2.1). Specifically, the mean vector μ and the covariance matrix are a specific percentile. Note that the worst-case scenario is to use the
obtained using the data set identified in Eq. (1) and (2). Fig. 1a shows the lowest value reported in V (lower V also implies lower values of fc ), then

( ′ )
behavior of p fc , V under a strong correlation coefficient. A correlation the percentiles used should also be lower values.
ρ = 0.8 is arbitrarily selected here only for an illustrative purpose. The Once Vo is defined, it is possible to estimate the conditional proba­
important issue here is that the information associated with fc strongly

bility distribution of fc . Ultimately, the concrete compression strength to

depends on the available information of V. In this sense, it is essential to be used in the building retrofit (fc− recommended ) can be established by

identify the conditional probability distribution for fc given a known



adopting a specific probability of exceedance (Po) over the distribution
value of the ultrasound pulse velocity Vo, which can be expressed as ( ′ )
p fc |V =Vo . Formally, the problem is posed as the identification of

4
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Po
Po

V V

Fig. 3. Illustrative isocurves for fc (units in MPa). The curves were obtained employing data generated artificially. The isocurves represent the recommended fc given
′ ′

by the proposed method but adopting different values for the probability of exceedance Po and different percentiles for the UPV. Figure a) represents a case with a
significant correlation between fc and UPV (ρ = 0.8). Figure b) represent a case with a poor correlation between fc and UPV (ρ = 0.2). The Danish recommendation
′ ′

corresponds to fc = 26.5 MPa and fc = 34.9 MPa, for case a) and b), respectively.
′ ′

fc−

recommended such that desired value of Po.

∫∞
( ′ ) ( ′ ) ′ 3.1. Effect of the variance and the correlation coefficient in the
(10)

P fc > fc− recommended |V =Vo = p fc |V =Vo dfc = Po
fc− recommended


fc− recommended

( ′ ) The primary assumption adopted in the standard approach described


Note that as p fc |V =Vo is Gaussian, the previous integral is
( ′ ) in ACI228.1R [10] (which is based on the Danish Method [31]), deals
computed using the cumulative distribution function for p fc |V =Vo .
with the expectation of a significant correlation between the ultimate
Ultimately, the user has to iterate fc− until it achieves the

recommended concrete strength and the UPV. However, this assumption could not

Fig. 4. Caribbean Hotel located in Venezuela. a) the limits of the hotel are marked in red. It is observed a near-to-the-sea location (picture taken from google maps).
b) picture of the hotel’s pools before the closeness in 1999. c) picture of the hotel’s pool in 2014 after more than 10 years out of service. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

5
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Fig. 5. Plan view of buildings distribution. The Hotel comprises 12 independent buildings. Buildings are identified with a number and shadowed in colors. The two
main sections of the hotel are shown in red dotted lines and identified as sections A and B. Concrete cores for destructive tests were extracted from buildings 1 to 4,
together with buildings 6 and 9. Nondestructive tests with UPV were performed in buildings 3 to 8. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

always be adequate since the correlation depends on uncontrolled var­ circles) are below that line. On the other hand, the Scenario B (Fig. 2b),
iables, i.e., concrete homogeneity, crack dimension, and orientation. In the Danish method recommendation of fc is inadequate since almost half

this sense, finding low correlation coefficients between fc and V is of the tested specimens are below the recommended value. Therefore, if

plausible. In cases where a low correlation is identified, the Danish the proposed method is used instead, it is necessary to define the
method estimates a linear regression with a slope close to zero, i.e., small percentile of interest for the UPV and the probability of exceedance (Po)
b in equations (3) and (4). As a consequence, the recommended fc is

for the ultimate concrete strength. In this particular case, two conditions
almost independent of the UPV leading to a non-conservative estimation are presented: (1) Vo defined as the 10th percentile of V together with a
of fc .

Po = 50 % (black star in Fig. 2), and (2) Vo defined as the 10th percentile
To illustrate the impact of the correlation coefficient on the fc esti­
′ of V together with a Po = 90 % (red star in Fig. 2). Remember that Po
indicates the probability of fc being greater than fc− recommended given Vo.
′ ′

mation, two scenarios are studied, one with a strong correlation (Sce­
nario A) and the other under a weak correlation (Scenario B). For both Here, it is possible to observe that the proposed method provides a
conservative recommendation (fc− recommended ) even in weak correlation

scenarios, a set of 500 pairs of fc − V is artificially created following a

Gaussian distribution, where the mean of fc and V are 36 MPa and 4500
′ scenarios.
m/s, respectively. Additionally, it is imposed a coefficient of variation
equal to 20 % for both parameters, while the correlation coefficient is 3.2. Risk-informed decision making
ρ = 0.8 for Scenario A and ρ = 0.2 for Scenario B. These data sets
simulate the experimental characterization of the core drills explained in The proposed method requires the selection of the percentile asso­
ciated with the UPV and the probability of exceeding fc− recommended given

section 2.1. The distributions of the data set with strong and weak cor­
relations are presented in blue circles in Fig. 2a and b, respectively. Vo. This formulation facilitates the generation of maps where
Additionally, it is required the data set of the UPV recorded across the fc− recommended depends on the in situ UPV percentile and the probability of

building (vector V defined in Section 2.3); this set is also generated exceedance for the concrete strength. Fig. 3 shows the isocurves of
artificially using the same mean and coefficient of variation employed fc− recommended generated by using different Vo and Po; in particular, the left

before. and right figures show the results for strong and weak correlations,
The recommended fc obtained by the Danish method [10] is identi­ respectively. The proposed method represents a powerful tool since the

fied in Fig. 2a and b by a solid black line. Here, it is important to note isocurves include relevant information that can be used in decision-
that for Scenario A (Fig. 2a), the recommended fc represents a conser­ making. Additionally, the decision maker can incorporate the attitude

vative scenario since only a small number of tested specimens (blue towards risk in the selection of the ultimate concrete resistance. These
concepts are applied in a case study presented in the following section.

6
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Fig. 6. Elevation view of buildings 5 and 8 according to Section A-A presented in Fig. 5. Only nondestructive tests were performed in this section of the building.

Fig. 7. Elevation view of buildings 2, 6, and 12 according to Section B-B presented in Fig. 5. Only destructive tests thought concrete core extraction were performed
in this section of the building.

4. Case study comprehensive characterization of the concrete strength assessment for


the mentioned structure, including destructive and non-destructive tests
The Caribbean Hotel (Fig. 4) was an icon hotel located in Vargas- and the indirect characterization methods presented in previous sections.
Venezuela, incorporating the learned lessons from the 1967 Caracas
Earthquake [35], which was the most destructive seismic event during 4.1. General description
the 20th century according to [36] in Venezuela. This structure was built
in the 70s, and the construction system used was characterized by con­ The mechanical properties specified by the original structural design
crete moment frames without interaction between masonry (following engineer and the project documents were unavailable. Therefore, the
recommendations to avoid the short-column effect), ties closed in col­ building’s configuration and elements’ dimensions were obtained by as-
umns and beams, and ribbed slabs in one direction. Posteriorly in 1999, built information. The hotel comprises 12 independent structures
the structure was hit by the most important debris flow recorded in separated by structural joints, numbered from 1 to 12, according to the
Venezuela’s history [37], causing the hotel’s closure. In Fig. 4a, it is original project (see Fig. 5). Each building is presented and identified in
possible to appreciate the near-to-the-sea hotel’s location (marked in a colored shadow. Additionally, the figure presents the two most
red), which also suggests a deterioration caused by salinity. In 2014, the important structural sections identified as Sections A-A and B-B. The
recovery of the hotel was mandated, requiring an extensive assessment elevation view according to these structural sections can be seen in
process. For the sake of comparison, a picture of the hotel previous and Figs. 6 and 7, respectively. Structures 1, 2, 3, 4, and 6 were used for
posterior to the closure in 1999 is presented in Fig. 4b and 4c, respec­ rooms; structures 7-to-11 were used for offices; structure 5 was used for
tively. The evaluation due to the current seismic hazard and the structural events; structure 12 for the reception hall; and structure 9 for access
assessment according to standards as ACI-318 [38] was performed, stairs and elevators.
paying special attention to the elevated deterioration of structural ele­ The structural condition showed advanced deterioration, mainly
ments due to environmental factors. This section presents a concentrated in the modules near the beach, reaching columns, beams,

7
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Fig. 8. View of generalized damage. Damage in the ceiling, partition walls and structural members are shown in a). Damages in beam-column joints are shown in b).
Damages in the column and ribbed slab are shown in c).

Fig. 9. A general view of the foundation. a) Shows a representative pile of the whole structure. b) Shows a pile cap.

Fig. 10. Structural members after the removal of the concrete cover. a) part of a beam used to extract steel samples, b) column after extracting concrete cores, and c)
view of core extraction procedure.

and ribbed slabs. The most common damages were corrosion of the up to 11–14 m. A water table measured from the surface of 3.8 m was
reinforcing steel and cracking of the concrete sections due to the detected in the six borings carried out. The type of foundation identified
corrosion process’s expansion of reinforcing steel bars. Examples of is a deep foundation with piles and pile caps (see Fig. 9a and 9b). The
these damages are shown in Fig. 8 where it is observed a damage in the piles have a diameter of 500 mm and a distribution of 4 piles per column.
ceiling, partition walls, structural member, and column-beam joints. However, the length of the piles was not reported. On the other hand,
Although the structure was designed in the 1970 s, it does not show any isolated reinforced concrete footing foundations were reported in the
damage associated with cracks or fissures due to other actions during the structural modules up to 2-floor levels. To obtain the distribution of the
visual inspection. The subsoil profile is constituted by a sequence of reinforcing steel, the demolition of concrete cover in the face of beams,
detrital sediments of alluvial and torrential origin, formed by the columns, and slabs was performed. Fig. 10 shows an example of the
interdigitation of silty sands intercalated with pebbles and cobbles. At naked structural members studied. Note that the removal of the concrete
the surface level, variable thicknesses of fill materials between 3 m and cover also helps to extract cleaner concrete cores used later in the
5 m thick were detected. Subsequently, silty sands and silty sands with compressive tests. Although the information presented in this paragraph
gravel were detected up to depths of no more than 8 m. Additionally, is not required to perform the concrete strength identification via UPV, it
gneissic shales were reported in diameters from 10 cm to 60 cm with is decided to present it since has the potential to add value in future
variable thicknesses from 1 m to 2 m and depths ranging from 4 to 9 m analysis.

8
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Fig. 11. Values of compression resistance for the concrete cores extracted for different L/D ratios and different floor levels. Red circles indicate the fc− mod , red dotted

lines indicate fc− mod /0.75, and the blue dotted lines represent fc− mod /0.85. Results for level 8, 5, lobby, beach, and foundation are presented in a), b), c), d), and e),
′ ′

respectively. Note that some points are overlapped. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

4.2. Destructive tests identification of the concrete strength by adopting a non-destructive


strategy. This indirect identification of the concrete strength (sup­
Sixty-six concrete cylinders (cores) were extracted from structural ported by ASTM-C597 [29]) is desirable since the core extraction for
elements considering the extraction feasibility and structural impor­ destructive tests is not economically efficient and, in some cases, diffi­
tance, removing, in all cases, the non-structural mortars from the cult to access.
structural element studied. Pictures of the core extracting process with The details of the concrete cores extracted in terms of location,
the drilling tool are shown in Fig. 10c, while Fig. 10a and b show the dimension, weight, density, UPV, cross-section area, volume, and fc− mod

place of extraction for a column and beam, respectively. Before per­ are presented in the Appendix Table 1. Note that all buildings are
forming the destructive compression test, the UPV for each concrete core studied except buildings 5, 7, 8, 10, 11, and 12. First, the length L and
was identified. In this test, a vibrating pulse is transmitted between two diameter D of each concrete core was identified. The dimensional values
transducers, and the wave travel time is recorded to obtain a propaga­ reported in the table corresponds to average values. Similarly, the
tion velocity. From the principles of elastic wave propagation and ACI- density was estimated using average dimensions and the measured
228.1R [10], the UPV is proportional to the square root of the elastic weight. Next, the pulse time was identified using an ultrasonic pulse
modulus of the hardened concrete. Therefore, identifying this velocity transducer PROCEQ PUNDIT LAB+, serial number PL02-003-0503.
and the concrete strength identified later in the destructive test is Finally, the value of fc is identified by employing a universal compres­

mandatory to establish a correlation that enables the indirect sion machine (IBERTEST, MEH model) with a capacity of 5000 kN

9
R. Mata et al. Construction and Building Materials 369 (2023) 130569

70

60

50

[MPa]
f'c (MPa) 40

30

20

y = 0.0171x - 40.86
10 R² = 0.3063

0
3000 3500 4000 4500 5000 5500 6000

V [m/s]

Fig. 12. Linear regression applied to the results obtained in the concrete core destructive tests. A poor R2 coefficient of determination is shown..fc−

mod

following the procedure indicated by ASTM-C42 [39]. According to selected according to the judgment of the structural engineer from the
ASTM-C42 [39], the values of fc’ must be corrected by slenderness ratio results obtained in the tests performed. In the case study, the building
L/D, obtaining a modified resistance value fc− mod . Moreover, the con­

was built by floor levels, i.e., the same concrete mix was used per floor
crete is considered structurally adequate (structural verification) ac­ level constructed. Therefore, it is decided to suggest a fc− d value per

cording to ACI-318 [32] if it satisfies two conditions: (i) floor, allowing a direct comparison between the concrete strengths
fc−

≥ 0.85fc− d , and (ii) fc−
′ ′
≥ 0.75fc− d . Here, fc−
′ ′
corresponds to identified and their variation across the building’s levels. Fig. 11 pre­
mod mod mod
sents in red circles the values of fc− mod , together fc− in blue

the compression strength average for the concrete cores tested, fc− mod
′ ′
mod /0.85
corresponds to the compression strength value for the core with the dotted lines, and fc− mod /0.75 in red dotted lines. Here, it is observed that

smallest compression strength, while fc− d corresponds to the design fc− d is controlled by fc− mod /0.75, being the most unfavorable condition
′ ′ ′

strength. For this case study, the design strength fc− d is not available,

for the case study. At the 8th, 5th, Lobby, Beach, and Foundations levels,
being impossible to perform a structural verification. However, the two the value of the concrete design strength fc− d obtained are 33.8 MPa,

conditions presented above can be used to suggest a value for fc− d such

34.6 MPa, 29.3 MPa, 33.3 MPa, and 48.5 MPa, respectively. Notably, the
that it can be used as a reference to perform any new numerical foundations levels present a remarkable high concrete strength, which
assessment of the structure’s performance. The simplest way to suggest a can be attributed to the insulation respect to the saline environment or
value for fc−

is to select the minimum between fc−

and to the deliberate use of a stronger concrete mix. Unfortunately, there is
d mod /0.85
not available information to corroborate the later.
fc− Currently, the requirements for defining the most repre­

mod /0.75. The concrete strength of each core is also compared with its
sentative design strength fc in a structural evaluation are not established

respective UPV, this comparison is presented in Fig. 12. Here, the data
in the design codes. As a common practice, the design strength fc is coming from the five levels studied are used (levels presented in Fig. 11).

Po

V V

Fig. 13. Isocurves for the inferred compressive concrete strength (units correspond to MPa) obtained via the proposed method for buildings 3 and 4. The inference
are made with UPV measurements exclusivity recorded within each building. Additionally, results of the Standard method (Section 2.2), Danish [34], General
Tolerance Factor [35], Rigorous [36], and Alternative [37] are also presented.

10
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Po

V V

Fig. 14. Isocurves for the inferred compressive concrete strength (units correspond to MPa) obtained via the proposed method for buildings 5 and 6. The inference
are made with UPV measurements exclusivity recorded within each building. Additionally, results of the Standard method (Section 2.2), Danish [34], General
Tolerance Factor [35], Rigorous [36], and Alternative [37] are also presented.

The data related to the foundation level is incorporated despite it has a 4.3. Non-destructive tests
significant higher strength compare with other levels. The main reason
to use the whole data set is to avoid a discretional data selection. The The data set obtained in the destructive test, composed of the UPV and
coefficient of variations for the concrete strength and the UPV are 26 % the concrete compression resistance, could be used to establish a corre­
and 7 %, respectively; with a correlation coefficient between them of lation model. Then, the correlation model is used to infer the concrete
0.55, which is considered a low value. Additionally, a linear fit was compression resistance in situ for any desired element of the studied
imposed, obtaining a low coefficient of determination (R2 = 0.3063) structure. The only requirements to perform this task are to correctly
which is expected due to the significant disparity on the coefficient of account for the correlation model and have the UPV for the structural
variations. Given the analysis presented in Section 3 (low correlation element of interest. The most standard model corresponds to the model
coefficients), it is expected that this data set will be not adequate to presented in Section 2; however, provision ACI-228.1R [10] contains
calibrate a standard correlation model for the concrete strength and detailed information about its implementation and the available alter­
UPV. natives. The case study recorded 120 UPV in situ across different
Po

V V

Fig. 15. Isocurves for the inferred compressive concrete strength (units correspond to MPa) obtained via the proposed method for buildings 7 and 8. The inference
are made with UPV measurements exclusivity recorded within each building. Additionally, results of the Standard method (Section 2.2), Danish [34], General
Tolerance Factor [35], Rigorous [36], and Alternative [37] are also presented.

11
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Table 1
Distribution of tests in structural elements and results of physical properties of specimens.
Building N◦ Story Level Dimensions UPV Area Volume Load fc−

mod
Weight Density
2 3
D [mm] L [mm] [m/s] [cm ] [cm ] [kgf] [Mpa] [g] [g/cm3]

1 8 74.1 106.5 4278.3 43.1 458.8 11230.0 25.3 1097.2 2.4


1 74.1 133.5 4291.0 43.1 574.9 13710.0 31.9 1347.4 2.3
1 74.0 132.8 4338.6 43.1 571.6 14860.0 34.6 1352.9 2.4
2 74.1 132.4 4427.8 43.2 571.5 17100.0 39.7 1342.9 2.3
2 74.0 104.3 4458.5 43.0 448.5 18220.0 41.0 1077.3 2.4
2 74.0 107.4 4588.5 43.0 462.0 20960.0 47.4 1090.8 2.4
2 74.1 115.2 4251.7 43.1 497.0 11640.0 26.5 1156.5 2.3
2 74.1 102.2 4369.2 43.1 440.3 18470.0 41.3 1029.8 2.3
2 74.1 93.1 4351.9 43.1 401.5 12640.0 27.8 956.4 2.4
3 74.1 89.0 4974.3 43.1 383.9 18530.0 40.2 930.2 2.4
3 73.9 115.3 4451.7 42.9 495.1 14270.0 32.7 1215.6 2.5
3 74.1 139.9 4572.9 43.1 602.7 12740.0 29.9 1434.6 2.4
1 5 74.2 130.1 4184.2 43.2 562.6 16140.0 37.3 1300.6 2.3
1 74.2 133.6 4423.5 43.2 577.1 16330.0 37.9 1340.1 2.3
1 74.0 130.1 4424.8 43.0 559.5 14710.0 34.2 1359.1 2.4
2 74.3 131.2 4139.4 43.3 568.5 13850.0 32.0 1331.5 2.3
2 74.1 129.2 4013.4 43.2 557.8 11220.0 25.9 1285.9 2.3
2 74.1 122.7 4172.4 43.1 529.1 13390.0 30.8 1214.9 2.3
2 74.3 132.7 3834.4 43.4 575.5 16410.0 37.9 1305.1 2.3
2 74.2 133.1 4008.1 43.2 575.1 14550.0 33.7 1309.4 2.3
2 74.2 112.5 4033.7 43.3 486.8 15200.0 34.5 1103.9 2.3
3 74.0 103.5 4621.4 43.0 445.2 16650.0 37.4 1082.9 2.4
3 74.2 126.2 4396.5 43.2 545.5 16240.0 37.4 1293.4 2.4
3 74.0 117.0 4270.1 43.0 503.2 15730.0 36.0 1206.8 2.4
6 Lobby Level 74.2 116.8 4184.9 43.2 504.7 10710.0 24.4 1180.9 2.3
6 74.3 101.0 4510.3 43.4 438.3 15970.0 35.4 1044.2 2.4
6 74.3 94.5 4218.8 43.3 409.3 14110.0 31.0 962.8 2.4
2 73.8 87.7 4641.8 42.8 375.3 14340.0 31.3 893.5 2.4
2 74.2 109.5 4310.6 43.2 473.4 10810.0 24.4 1107.5 2.3
2 74.2 94.3 4406.1 43.2 407.4 17880.0 39.3 990.2 2.4
4 74.3 87.9 4310.3 43.3 380.9 13250.0 28.5 881.6 2.3
4 74.3 108.5 4445.9 43.3 470.0 13480.0 30.3 1083.9 2.3
4 74.3 84.6 4361.3 43.4 366.9 10350.0 22.0 842.8 2.3
4 72.8 72.9 4288.2 41.7 303.7 13200.0 28.1 690.9 2.3
4 74.3 102.0 4881.8 43.4 442.7 13440.0 29.8 1033.1 2.3
4 74.0 74.0 4378.7 43.0 318.2 16010.0 33.0 743.2 2.3
2 Beach Level 74.3 89.5 4864.7 43.3 387.9 15420.0 33.3 917.6 2.4
2 74.3 94.7 4756.3 43.4 410.4 12840.0 28.2 960.1 2.3
2 74.2 113.4 4343.3 43.2 490.1 11010.0 25.0 1167.6 2.4
2 74.2 110.1 4511.1 43.3 476.2 21070.0 47.6 1156.3 2.4
2 74.3 96.7 4984.0 43.4 419.2 15510.0 34.1 1003.0 2.4
2 74.4 97.4 5154.5 43.5 423.6 18460.0 40.6 1248.9 2.9
9 74.3 127.1 4208.3 43.3 550.4 17000.0 39.1 1253.1 2.3
9 74.0 125.8 4279.6 43.0 541.4 14110.0 32.6 1280.2 2.4
9 73.8 109.1 4134.1 42.7 466.4 16270.0 37.2 1123.9 2.4
4 74.3 133.5 4319.1 43.3 577.9 16850.0 39.0 806.8 1.4
4 74.3 79.0 4411.7 43.3 342.0 12270.0 25.6 1357.3 4.0
4 74.3 97.6 4669.4 43.4 423.1 16320.0 36.0 1006.8 2.4
4 74.3 115.5 4324.3 43.4 500.7 19380.0 43.9 1176.3 2.3
4 74.3 124.5 4382.0 43.3 539.4 15190.0 34.8 1279.4 2.4
4 74.3 97.0 4530.8 43.4 420.8 15630.0 34.4 989.9 2.4
1 Found. Level 74.2 83.6 5099.4 43.3 361.7 31910.0 67.7 871.3 2.4
1 74.3 121.8 4890.0 43.4 528.3 20420.0 46.6 1260.8 2.4
1 74.2 102.3 4893.3 43.3 442.5 19790.0 44.1 1071.0 2.4
1 74.3 81.2 4665.5 43.3 351.7 29630.0 62.2 836.2 2.4
1 74.3 104.4 4766.7 43.3 452.4 18440.0 41.1 1082.0 2.4
1 74.2 74.2 4979.9 43.2 320.8 22350.0 45.9 769.5 2.4
4 74.4 108.0 4822.8 43.5 469.7 21390.0 47.9 1128.1 2.4
4 74.2 92.0 4867.7 43.3 398.2 22050.0 48.2 959.6 2.4
4 74.3 91.9 4738.7 43.3 398.3 27500.0 60.0 968.0 2.4
3 74.3 93.1 5059.2 43.4 403.6 24110.0 52.8 969.6 2.4
3 74.3 104.1 4862.1 43.3 450.8 19720.0 44.0 1088.7 2.4
3 74.4 116.6 4777.5 43.5 506.6 21190.0 48.0 1225.3 2.4
4 74.4 99.5 4759.8 43.5 432.5 16510.0 36.4 1031.3 2.4
4 74.2 121.1 4962.3 43.3 524.2 29610.0 67.6 1286.4 2.5
4 74.2 132.0 4905.6 43.3 570.8 17180.0 39.8 1381.7 2.4

structural elements. The results are presented in Tables 2 and 3, where it is [31], General tolerance factor method [40], Rigorous method [41], and
possible to find the building identification number, the level at which the Alternative method [42]. Additionally, for the case studied, it is decided
test was performed, the type of structural component, and the UPV. to include the standard method presented in Section 2.2 and the pro­
According to ACI-228.1R [10], the interpretation and reporting of posed method introduced in Section 3. The standard method is benefi­
results can be performed using the following methods: Danish method cial to perform a fast in situ estimation of fc . Note that all mentioned

12
R. Mata et al. Construction and Building Materials 369 (2023) 130569

Table 2 measurements with 10, while building 6 has the more significant
Results of UPV tests in structural elements for buildings 3, 4, and 5 ac­ number of measurements with 38.
cording to ASTM-C597 [29]. The proposed method presented in Section 3 is first implemented.
Building 3 Building 4 Building 5 The Shapiro-Wilk test is applied to the concrete strength and UPV,
Level UPV (m/s) Level UPV (m/s) Level UPV (m/s)
obtaining that UPV can be considered normal distributed but the con­
crete strength not (adopting a p-value of 0.05). However, when the
8 4561 8 4357 Lobby 3927
Shapiro-Wilk test is repeated using the natural logarithm of fc− mod and

8 4089 8 4137 Lobby 3673


UPV, the results indicate that ln(fc− mod ) and ln(UPV) can be considered

8 4476 8 4294 Lobby 4184
8 4020 8 4184 Lobby 3996 normally distributed. Additionally, the Royston’s test is also applied to
5 3788 8 4136 Lobby 4094
the natural logarithm of fc− mod and UPV, indicating that the multivariate

5 4163 5 4541 Beach 3895
5 4175 5 4141 Beach 4260 distribution can be considered log-normal distributed. Thus, the equa­
5 3992 5 4541 Beach 4090 tion (9) is employed adopting the natural logarithm of fc− mod and UPV.

Lobby 4923 5 4289 Beach 3914


The proposed method requires identifying fc− recommended for different

Lobby 4175 5 4467 Beach 4442
Lobby 3913 Lobby 4451 – – probabilities of exceedance Po and difference ultrasonic pulse velocities
Lobby 4913 Lobby 4012 – – Vo by solving Eq. (10). Hence, it is required to define Vo based on a
Beach 4670 Lobby 4333 – –
particular percentile obtained from the in situ UPV measurements. The
Beach 4382 Lobby 4337 – –
Beach 4816 Lobby 4373 – – range selected to evaluate Vo is from the 10th to 50th percentile. Note
Beach 4447 Beach 4050 – – that lower percentiles cannot be accounted for since the minimum
– – Beach 4198 – – number of measurements for the UPV is 10. The range for Po is estab­
Beach 4333
– – – –
lished between 0.5 and 0.95. The isocurves of fc− recommended for different

– – Beach 4437 – –
– – Beach 4249 – – Po and velocity percentiles are shown in Figs. 13, 14 and 15 for the
buildings studied. These isocurves allow selecting the concrete strength
based on the risk attitude of the engineer. For example, the concrete
strength value located at the lower left corner of the map (Po = 0.95 and
Table 3
Results of UPV tests in structural elements for buildings 6, 7, and 8 ac­
the 10th percentile for the UPV) represents a lower risk identification
cording to ASTM-C597 [29]. since the corresponding velocity is low (compared to the set of measured
velocities), and the probability that the current concrete strength ex­
Building 6 Building 6 Building 7 Building 8
ceeds fc− recommended is 95 %. The concrete strength value identification

Level UPV Level UPV Level UPV Level UPV


ultimately depends on the engineer’s attitude toward risk. However, for
(m/s) (m/s) (m/s) (m/s)
engineers with a risk-averse attitude, the combination of a 10th
8 4396 Beach 4374 8 4397 8 3751
percentile for the UPV and Po = 0.90 seems to be an adequate recom­
8 4205 Beach 4374 8 4184 8 4391
8 4211 Beach 4345 8 4611 8 4228
mendation. For the mentioned risk-averse attitude, the concrete strength
8 4585 Beach 4313 8 4531 8 4318 value identified corresponds to 21.2 MPa, 22.7 MPa, 18.8 MPa, 24.0
8 4278 Beach 4371 5 4692 8 4321 MPa, 23.3 MPa, and 20.8 MPa, for buildings 3 to 8, respectively.
8 4177 Beach 4464 5 4481 5 4499 In the case presented (Figs. 13, 14 and 15), it is possible to observe
8 4158 Beach 4328 5 4115 5 4518
that all buildings present similar isocurves, except for building 5, which
5 4608 Beach 4145 5 4458 5 4331
5 4787 Beach 4215 Lobby 5083 5 4276 shows isocurves with significantly reduced values. A possible reason to
5 4889 Beach 4267 Lobby 4162 5 3933 explain this accelerated deterioration is its location since this building
5 4520 Beach 4307 Lobby 4107 Lobby 4263 receives straight salinity wind from the sea. Additionally, the isocurves
5 4362 Beach 4379 Lobby 4171 Lobby 4322 also present the concrete strength values identified by adopting other
5 4561 Beach 4404 Beach 4564 Lobby 4246
5 4016 Beach 4307 Beach 4299 Lobby 3873
methods. In this regard, the Danish [31] and General tolerance factor
Lobby 4790 Beach 4211 Beach 4032 Lobby 4184 methods [40] offer a similar identification for the concrete strength
Lobby 4404 Beach 4352 Beach 4098 Beach 4314 value. The least conservative identification (larger values of concrete
Lobby 4328 Beach 4143 – – Beach 4337 strength) is obtained from the Rigorous [41] and the standard method
Lobby 4455 Beach 4314
followed by the Alternative method [42]. If the identified concrete
– – – –
Lobby 4431 – – – – Beach 4257
Lobby 4395 – – – – Beach 4171 strength is compared to the destructive test results, it is possible to
Lobby 4357 – – – – – – observe that the non-destructive tests offered lower values, 22-to-34
MPa and ~ 36 MPa, for the non-destructive and destructive tests,
respectively.
methods offer either a single scalar value or a value associated with
reliability or probability of exceedance for the concrete strength. How­ 5. Conclusions
ever, the proposed method is the only method that explicitly in­
corporates the dispersion evidenced in the in situ UPV measurements. This research proposes a new method of correlation between con­
The identification of the UPV and the concrete strength obtained crete compressive strength and ultrasonic pulse velocity using a statis­
from the destructive test are used here to establish the correlation tical method. In order to compare the existing methods and the proposed
described in Section 2.2. This correlation is assumed valid for all method, a case study is analyzed and the results are compared. The main
buildings presented in Fig. 5. However, the indirect identification of the conclusions of the present work are listed as follows:
concrete strength is performed using the UPV recorded within each
building, i.e., there is a different concrete strength identification in each • A case study is presented towards the identification of the
building. In other words, Section 2.3 is applied independently in each compressive concrete strength via destructive and non-destructive
building. Although the hotel comprises 12 buildings, the UPV presented tests based on UPV. All data was presented facilitating future
in the appendix Tables 2 and 3 only show the measurements for build­ studies and comparisons.
ings 3 to 8; the remaining buildings were not included because the • The proposed method allows the identification of the concrete
contractor catalog them as an unsafe and difficult-to-access area. In strength by considering its probability of exceedance and the
particular, building 5 has the lowest number of in situ UPV

13
R. Mata et al. Construction and Building Materials 369 (2023) 130569

percentile associated to the in situ UPV measurements. The method • An interesting future direction is to incorporate the concept of con­
generates concrete strength isocurves facilitating the incorporation ditional coring within the proposed procedure, i.e., to use UPV to
of the practitioner’s attitude towards risk. This represents a contri­ select the core extraction location such that the uncertainty in the
bution compared to the methods currently established in the current concrete strength identification will be reduced.
standards.
• The methods established in ACI228.1R for correlation between UPV CRediT authorship contribution statement
and destructive tests do not allow for rationally established scenarios
for decision-making, sometimes reaching an overestimation of con­ Ramón Mata: Software. Rafael O. Ruiz: Conceptualization, Formal
crete strength values. However, the proposed method (concrete analysis, Methodology, Project administration, Supervision, Validation,
strength isocurves) could be used jointly with the standardized Visualization, Writing – original draft, Writing – review & editing.
method to extend the interpretation based on risk. Eduardo Nuñez: Conceptualization, Formal analysis, Investigation,
• The standard non-destructive methods based on UPV improves the Methodology, Project administration, Supervision, Software, Validation,
characterization of concrete strength using an incremental of studied Visualization, Writing – original draft, Writing – review & editing.
zones. However, the correction of data generates a decrease in the
concrete resistance of up to 30 % without further justification. This Declaration of Competing Interest
confirms the importance and the need to build an appropriate cor­
relation curve or to use methods that justify the results obtained. The authors declare that they have no known competing financial
• Finally, the models are valid in the domain where it was built. interests or personal relationships that could have appeared to influence
Therefore, using the presented models for different types of concrete the work reported in this paper.
is not recommended without a previous extensive study. The corre­
lations with concrete resistance obtained from cores should be per­ Data availability
formed when non-destructive tests are required.
The data that has been used is confidential.

Appendix

This section contains the details of the different tests performed in the case study. In particular, Table 1 presents the information of the physical
properties and characteristic of the concrete core extracted, including the compressive concrete strength fc and the UPV. On the other hand, Tables 2

and 3 present the information of the UPV measured in different structural elements of buildings 3 to 8.

References [17] D. Breysse, X. Romao, M. Alwash, Z.M. Sbartai, V.A. Luprano, Risk evaluation on
concrete strength assessment with NDT technique and conditional coring
approach, J. Build. Eng. 32 (2020), 101541.
[1] N. Pereira, X. Romão, Assessing concrete strength variability in existing structures
[18] M. Sýkora, D. Diamantidis, M. Holický, J. Marková, Á. Rózás, Assessment of
based on the results of NDTs, Constr. Build. Mater 173 (2018) 786–800.
compressive strength of historic masonry using non-destructive and destructive
[2] C. Wang, H. Wang, Assessment of the compressive strength of recycled waste LCD
techniques, Constr. Build. Mater 193 (2018) 196–210.
glass concrete using the ultrasonic pulse velocity, Constr. Build. Mater 137 (2017)
[19] Vu KAT, Stewart MG. Service life prediction of reinforced concrete structures
345–353.
exposed to aggressive environments. In: 9th International Conference on Durability
[3] Y. Tan, H. Yu, R. Mi, Y. Zhang, Compressive strength evaluation of coral aggregate
of Building Materials and Components, CSIRO BCE, 2002.
seawater concrete (CAC) by non-destructive techniques, Eng. Struct 176 (2018)
[20] K.A.T. Vu, M.G. Stewart Predicting the likelihood and extent of reinforced concrete
293–302.
corrosion-induced cracking J Struct Eng, 131 (11) (2005), pp. 1681-1689.
[4] M. Jalal, N. Nassir, H. Jalal, P. Arabali, On the strength and pulse velocity of
[21] M.G. Stewart, Q. Suo, Extent of spatially variable corrosion damage as an indicator
rubberized concrete containing silica fume and zeolite: prediction using
of strength and time-dependent reliability of RC beams, Eng Struct 31 (2009)
multivariable regression models, Constr. Build. Mater 223 (2019) 530–543.
198–207.
[5] J. Bogas, M. Gomes, A. Gomes, Compressive strength evaluation of structural
[22] M.G. Stewart, J.A. Mullard, Spatial time-dependent reliability analysis of corrosion
lightweight concrete by non-destructive ultrasonic pulse velocity method,
damage and the timing of first repair for RC structures, Eng Struct 29 (2007)
Ultrasonics 53 (2013) 962–972.
1457–1464.
[6] A.J. Tenza-Abril, Y. Villacampa, A.M. Solak, F. Baeza-Brotons, Prediction and
[23] A. Firouzi, A. Rahai, Prediction of extent and likelihood of corrosion-induced
sensitivity analysis of compressive strength in segregated lightweight concrete
cracking in reinforced concrete bridge decks, Int. J. Civil Eng. 9 (3) (2011)
based on artificial neural network using ultrasonic pulse velocity, Constr. Build.
183–192.
Mater 189 (2018) 1173–1183.
[24] O.-S. Kwon, A. Elnashai, The effect of material and ground motion uncertainty on
[7] G. Bayar, T. Bilir, A novel study for the estimation of crack propagation in concrete
the seismic vulnerability curves of RC structure, Eng. Struct. 28 (2006) 289–303.
using machine learning algorithms, Constr. Build. Mater 215 (2019) 670–685.
[25] A. Firouzi, A. Rahai, Prediction of extent and likelihood of corrosion-induced
[8] R. Pucinotti, Reinforced concrete structure: non-destructive in situ strength
cracking in reinforced concrete bridge decks, Int. J. Civil Eng. 9 (3) (2011)
assessment of concrete, Constr. Build. Mater 75 (2015) 331–341.
183–192.
[9] E. Hwang, G. Kim, G. Choe, M. Yoon, N. Gucunski, J. Nam, Evaluation of concrete
[26] X. Tang, et al., Study on the heterogeneity of concrete and its failure behavior using
degradation depending on heating conditions by ultrasonic pulse velocity, Constr.
the equivalent probabilistic model, J. Mater. Civil Eng. 23 (4) (2011) 402–413.
Build. Mater 171 (2018) 511–520.
[27] R. Rackwitz, K.F. Müller. On the correlation and autocorrelation of in situ strength
[10] ACI 228.1R-19. Report on In-Place Methods to Estimate Concrete Strength
and standard strength test results. In: Quality control of concrete structures –
STANDARD by American Concrete Institute.
Volume 1, Vol. 21, RILEM; 1979, pp. 159–164.
[11] NTC (2008) Norme tecniche per le costruzioni. D.M. Ministero Infrastrutture e
[28] M. Bonfigli, A. Materazzi, M. Breccolotti, Influence of spatial correlation of core
Trasporti 14 gennaio 2008, G.U.R.I. 4 Febbraio 2008, Roma (in Italian).
strength measurements on the assessment of in situ concrete strength, Struct.
[12] European Committee for Standardization. EN 1990:2002 – Eurocode – Basis of
Safety 68 (2017) 43–53.
Structural Design, 2002.
[29] ASTM C597–2 (2003), standard test method for pulse velocity through concrete,
[13] BS EN 13791: 2007. Assessment of in-situ compressive strength in structures and
2003.
precast concrete components. BSI, London; 2007. p. 29. ISBN 978-0-58050271-2.
[30] D. Breysse, Nondestructive evaluation of concrete strength: an historical review
[14] R. Pucinotti, Assessment of in situ characteristic concrete strength, Constr. Build.
and a new perspective by combining NDT methods, Constr. Build. Mater. 33 (2012)
Mater 44 (2013) 63–73.
139–163.
[15] R. Giannini, L. Sguerri, F. Paolacci, S. Alessandri, Assessment of concrete strength
[31] J.A. Bickley, Variability of Pullout Tests and InPlace Concrete Strength, Concr. Int.
combining direct and NDT measures via Bayesian inference, Eng. Struct 64 (2014)
4 (4) (1982) 44–51.
68–77.
[32] S. Kocherlakota, The bivariate inverse Gaussian distribution: an introduction,
[16] Z.M. Sbartai, M. Alwash, D. Breysse, X. Romao, V.A. Luprano, Combining the
Commun. Stat. - Theory Methods 15 (4) (1986) 1081–1112, https://doi.org/
biobjective approach and conditional coring for a reliable estimation of on-site
10.1080/03610928608829171.
concrete strength variability, Mater. Struct. 54 (6) (2013) 1–12.

14
R. Mata et al. Construction and Building Materials 369 (2023) 130569

[33] S.S. Shapiro, M.B. Wilk, An analysis of variance test for normality (complete [38] ACI 318-19. Building Code Requirements for Structural Concrete and Commentary
samples), Biometrika 52 (3/4) (1965) 591–611. STANDARD by American Concrete Institute.
[34] J.P. Royston, Algorithm AS 181: the W test for normality, Appl. Stat. (1982) [39] ASTM C42–12 (2004), standard test method for obtaining and testing drilled cores
176–180. and sawed beams of concrete, 2004.
[35] G. Suarez, J. Nabelek, The 1967 Caracas earthquake: fault geometry, direction of [40] K.R. Hindo, W.R. Bergstrom, Statistical evaluation of the in-place compressive
rupture propagation and Seismotectonic implication, J. Geophys. Res. 95 (1990) strength of concrete, Concr. Int. 7 (2) (1985) 44–48.
459–474. [41] W.C. Stone, C.P. Reeve, New statistical method for prediction of concrete strength
[36] F. Audemard, Surface rupture of the Cariaco July 09, 1997 earthquake on the El from in-place tests, J. Cem. Concr. Aggregates, ASTM 8 (1) (1986) 3–12.
Pilar fault, northeastern Venezuela, Tectonophysics 424 (2006) 19–39. [42] N.J. Carino. 1993, “Statistical Methods to Evaluate In-Place Test Results,” New
[37] W. Méndez, Análisis cuantitativo del relieve en cuencas de drenaje de la vertiente Concrete Technology: Robert E. Philleo Symposium, SP-141, T. C. Liu and G. C.
norte del macizo “El Ávila” (estado Vargas, Venezuela) y su significado Hoff, eds., American Concrete Institute, Farmington Hills, Mich., pp. 39-64.
hidrogeomorfológico, Investigaciones Geográficas, Boletín del Instituto de
Geografía, UNAM 91 (2016) 25–42.

15

You might also like