research methodology

P resent chapter focuses on the detailed introduction and specification of the research methods used in present studies.

The chapter is structured in accordance with the research signi-ficance and the three main objectives introduced in Chapter 1.

A detailed statistical analysis was performed to study the variability parameters of rebound hardness. A phenomenological model was formulated between the rebound index and the compressive strength of concrete to characterize the relationship of the two time dependent properties. Laboratory verification tests as well as parametric simulation were carried out for the validation of the model. Targeted experiments were designed and conducted to study the relationships between the rebound index and the mechanical properties of concrete.

3.1 Statistical analysis 3.1.1 Normality tests

In mathematical statistics, normality tests are used to determine whether a data set can be modelled by normal distribution or not. The importance of the normality tests concerning the rebound hardness method can be understood since normality is an underlying assumption of many statistical procedures. There are about 40 normality tests available in the technical literature (Dufour et al, 1998), however, the most common normality test procedures of statistical software are the Shapiro-Wilk test, the Kolmogorov-Smirnov test, the Anderson-Darling test and the Lilliefors test. It is demonstrated in the technical literature that the Shapiro-Wilk test is the most powerful normality test from the above four (Razali, Wah, 2011). The analyses provided by present thesis focused on the Shapiro-Wilk normality test. To see if the probability distribution of the rebound index reading set of an individual test area can be described by normal distribution or not, the Shapiro-Wilk normality test was run. From 24 different sources, 4555 test areas were selected (from which 3447 of laboratory testing and 1108 of in-situ testing) where 10 individual rebound index readings were recorded at each test area by N-type original Schmidt rebound hammer.

Considering the rebound hardness method, one can assume that the rebound index reading sets of separate

test areas are independent and identically distributed (i.i.d.) random variables since it can be accepted that

the probability distribution of the rebound index does not change by test area within a concrete structure and

the separate test areas can be considered to be mutually independent. Based on these assumptions, the

central limit theorem applies for the rebound hardness method; i.e. the probability distribution of the sum (or

average) of the rebound index reading sets of separate test areas (each with finite mean and finite variance) approaches a normal distribution if sufficiently large number of i.i.d. random variables are available.

Testing of the central limit theorem for the rebound index reading sets of individual test areas may be a good indicator of the precision of the rebound hammer devices.

The practical application of the central limit theorem was the running of the Shapiro-Wilk test for multiple rebound index reading sets combined. The expected behaviour is the value of the W statistic approaching unity by the increasing number of test areas combined.

3.1.2 Calculation of repeatability parameters

An extended repeatability analysis was made on 8955 data-pairs (own measurements: 2699 laboratory data-pairs, 578 in-situ data-pairs, total 3277 data-pairs) of corresponding average rebound indices and standard deviations of rebound indices that were collected from 48 different sources (in which the number of in-situ test areas was 4785 and the number of laboratory test areas was 4170; resulting more than eighty thousand individual rebound index readings). Range of the studied concrete strengths was f

cm

= 3.3 MPa to 105.7 MPa, and the range of the individual rebound indices was R = 10 to 63. The averages and the standard deviations were calculated by 10 to 20 replicate rebound index readings on the same surface of a concrete specimen during laboratory tests, or at the same test area in the case of in-situ testing. The data were analysed to see the general repeatability (within-test variation) behaviour of the rebound method. Analysis of reproducibility (batch-to-batch variation) was not the aim of the studies. Standard deviation and coefficient of variation was calculated and analysed. The range of the analysed data is from R

m,min

= 12.2 to R

m,max

= 59.0 for the averages and from s

R,min

= 0.23 to s

R,max

= 7.80 for the standard deviations. Coefficient of variation range was found to be as from V

R,min

= 0.43% to V

R,max

= 31.12%.

3.1.3 Goodness of fit tests

An extended statistical analysis has been made on the previously detailed database (8955 test areas) to ascertain the probability distribution of the statistical parameters of the rebound index (i.e. standard deviation, coefficient of variation, range, studentized range).

The goodness of fit (GOF) tests were used to compare test data to the theoretical probability distribution functions. Three tests were run to get the best goodness of fit out of more than 60 different types of distribution functions: Kolmogorov-Smirnov test, Anderson-Darling test and χ

²

test.

The goodness of fit tests measure the compatibility of a random sample with a theoretical probability distribution function. In other words, these tests show how well the selected distribution fits to the data.

The general procedure consists of defining a test statistic, which is a function of the data measuring the distance between the hypothesis and the data, and then calculating the probability of obtaining data which have a still larger value of this test statistic than the value observed, assuming the hypothesis is true.

In present analyses 60 different probability distributions were studied by GOF to find the best fit to the

experimental data.

3.1.4 Influences on the repeatability parameters

The governing parameters over the changes of the standard deviation, coefficient of variation, range, and studentized range were analysed based on the available database, with the selection of the following possible influencing parameters: the w/c-ratios of the concretes, the age of the concretes, the cement types used for the concretes, the testing conditions of the concretes (dry/wet), the carbonation depths of the concretes and the impact energy of the rebound hammers (N-type original Schmidt hammer with impact energy of 2207 Nmm or L-type original Schmidt hammer with impact energy of 735 Nmm).

3.2 Modelling

3.2.1. Development of the phenomenological model

The development of the model was induced by the extensive literature survey of the rebound method after the analysis of more than 150 technical publications of the last 60 years.

Deductive principles were followed in the theoretical research. The ideas were based on theoretical considerations, where it was appropriate, while in other cases empirical relationships were considered.

General experimental observations and limitedly available theoretical models were studied for the compressive strength and rebound index. Models were preferred where the degree of hydration was found to be the primary driver of phenomena.

Since the mathematical modelling and experimental determination of the degree of hydration do not satisfy the principle of “intended simplicity for practical use”, therefore, a simplification was applied; the degree of hydration was characterized by three variables: type of cement, water-cement ratio (w/c) and age of concrete. The randomness of the phenomena were not taken into consideration during the theoretical research by focusing mostly on general laws, that is, the particular influencing parameters were not considered as random variables. Revealing of the possible interrelationships has lead to the hypothesis of a phenomenological model for the compressive strength and rebound index of concrete which was able to generate data points of compressive strength and rebound index for concretes made from a given type of cement, with a given water-cement ratio, at a given age, by means of five general relationships. The generator functions are (all of them can be validated empirically): relationship between the water-cement ratio and compressive strength of concrete at the age of 28 days; development of compressive strength in time; relationship between compressive strength and rebound index of concrete at the age of 28 days; development of carbonation depth in time; relationship between carbonation depth and rebound index of concrete.

3.2.2 Robustness study by parametric simulation

The applicability of the model was tested by parametric simulation; by the preliminary selection of

arbitrary function parameters. Series of functions were generated to simulate results that are similar to

real experimental observations. Empirical formulations were selected from the technical literature for the

generator functions of the model for the parametric simulation.

3.2.3 Model verification with laboratory tests

The intention of the experimental part of the research connected to modelling was to verify the applicability of the developed phenomenological model. Inductive principles were followed, i.e. laboratory tests were carried out under strictly controlled experimental conditions, with the introduction of sufficiently large number of test parameters changed on a wide range, on a large number of specimens.

The general performance of the developed phenomenological model was studied by the appropriate graphical representation of the particular observations.

The experimental verification study was carried out at the Budapest University of Technology and Economics (BME), Department of Construction Materials and Engineering Geology. The tested 72 concrete mixes were prepared in accordance with present concrete construction needs during the experiments, i.e. slightly over-saturated mixes with different admixtures were designed. Consistency of the tested concrete mixes was constant: 500±20 mm flow provided by superplasticizer admixture.

Design air content of the compacted fresh concrete for the tested concrete mixes was 1.0 V%. The specimens were stored in water tank for 7 days as curing. After 7 days the specimens were stored at laboratory condition.

Test parameters were:

Water-cement ratio:

0.38 – 0.41 – 0.43 – 0.45 – 0.47 – 0.50 – 0.51 – 0.55 – 0.60 Cement type:

CEM I 42.5 N – CEM III/B 32.5 N

In document Rebound surface hardness and related properties of concrete (Pldal 46-49)