1. Introduction
Reinforced concrete with steel bars, which can be considered to be a composite material, is widely used in civil engineering structures because of its high load-carrying capacity and low maintenance cost. In a reinforced concrete structure, rebars are protected from the outside by a few centimeters of concrete called coating (refer to Eurocode 2 EN 1992 (section 4)), a crucial protective layer for the service life. They meet the needs of mechanical strength thanks to the reinforcements and bonding with concrete. However, concrete is a material that shows changes over the period of operation. Thus, several factors, such as the loading of the structure, the environment, or the attacks sustained over time cause degradation of concrete material. Therefore, due to environmental and mechanical aggression, which lead to the penetrating the material leading to electrochemical reaction (corrosion) that eventually attack the steel, result in a material degradation [
1]. This corrosion is a critical factor in the mechanical strength of the structure. To prevent corrosion, it is crucial to detect and characterize the cracks that may appear on the surface of the concrete as early as possible, and, in particular, to be able to estimate the sufficient depth and opening of these cracks. Therefore, monitoring of changes in the condition of an RC structure and detecting microcracks before they develop into macrocracks, and by timely intervention, could lead to a longer life of the structure. However, the degradation encountered in the concrete structures appears at different stages of the service life. In addition to, sometimes, destructive testing is not allowed to determine a quality of concrete (also taking sample is dangerous for structure). For this reason, there is an increased demand for more precise non-destructive testing (NDT) techniques and, at the same time, more flexible evaluation in the ability to detect the quality of the concrete.
The main interest of non-destructive testing (NDT) as a tool of auscultation is that it allows the carrying out of investigations (even repeated over time) without affecting the operations. The NDT are popular measuring methods, which is increasingly applied to the structures and structural parts in reinforced concrete as part of the maintenance and inspection [
2]. Presently, NDT used in civil engineering structures, are easier to handle due to the continuous development of measuring technology and, taking into account the “correct” evaluation of the results, provide a reasonable basis for a reliable assessment of the actual condition of the structure. There are various NDT techniques have been used for decades, currently more than 70 types of standardized testing methods can be applicable to the evaluation of concrete structures, e.g., acoustic emission [
3,
4,
5,
6], infrared thermography [
7,
8,
9], ground-penetrating radar [
10,
11,
12,
13], Fiber Bragg grating (FBG) [
14,
15], or digital image correlation (DIC) [
16,
17,
18].
The traditional methods used for detecting cracks are mainly based on the measurement of global deformations. Unfortunately, the sensitivity and precision of this type of measurement are quite weak, the density of cracking must reach comparatively high values before its impacts start to be detectable in terms of deformations [
19,
20,
21]. Acoustic emission (AE) can be a promising technique [
8,
10]. Listening to AE events gives information earlier than the visible opening of cracks, but the interpretation of results is always a difficult matter. This is because most AE events occur just before the propagation of microcracks [
4]. The lack of significant AE activity at the initial stages of loading causes difficulty in distinguishing between background noise and acoustic events related to the crack. The techniques using ultrasonic waves velocity are particularly interesting because of the direct relationship between characteristics of wave propagation and the stage of damage to the material [
22].
Ultrasonic Pulse Velocity (UPV) method is most commonly used to detect the quality of concrete, the position of crack or deepness inside both reinforced and masonry structures. In the ultrasonic inspection, the most widely used modes are longitudinal and shear waves for the propagation. Zhong and Yao et al. [
23] identified the self-healing capability of normal and high-strength concrete damaged under compressive loads at several periods using UPV measurements. In [
24], the application of ultrasound diffusion to detect damage in aluminum plates was used. The characterization of concrete by the propagation of ultrasonic waves is a usual way to evaluate the potential resistance of a structure. The use of proven techniques such as transmission, echo pulse or surface waves can identify areas with weaken mechanical characteristics, or even to detect the presence of cracks. Usually, these techniques applied from the component surface. Here, the contact between the material and the ultrasonic sensors is not always of equal, stable quality, as they are usually coupled with water or glycerin. However, it is important to note that many experimental and environmental parameters can influence measurements due to surface connection. One should be very careful about the constant coupling of ultrasonic sensors. Otherwise, the use of evaluation methods will evaluate the slightest changes in the signal caused by changes in the state of the component. The ultrasonic velocity—compressive strength correlations that are generally used can only be applied under specific circumstances. Therefore, Bundesanstalt für Materialforschung und -prüfung (BAM) developed a novel ultrasonic transducer, which can be permanently embedded into concrete structure [
25]. It allows the constant coupling of the embedded ultrasonic sensors to the concrete and the embedding in deeper areas inside a tested structure. The sensors are also suitable for the permanent investigations of concrete structures with the pulse velocity method. Moreover, embedding provides the ability to monitor areas that are conventionally no longer accessible from the component surface. Since acquired signals from these embedded sensors need to be analyzed to characterize the concrete, it is essential to use primary signal processing methods and approaches, such as features and simple statistical measures.
Another commonly used technique for detection of cracks in structures (e.g., plate or rod) is the group of ultrasonic guided waves (Rayleigh and Lamb waves) approach. In this technique, where guided wave modes are preferred to obtain a clear response from damage via a single-mode. However, if the structure is made of heterogeneous and strongly scattering material like concrete, guided waves are difficult to interpret (methods are restricted to components). For these situations, diffuse ultrasonic waves can be created by an impulse excitation, allowing many reflections to occur and resulting in a similar diffuse wave in the structure [
26]. The challenges linked with diffuse waves is the complexity of the waveforms, because it allows many modes, as the structure can support during the propagation (like a random walk).
Addressing these challenges and extracting damage/change information from complex diffuse waves have been the subjects of the vast number of studies. For example, in [
26], investigated small cracks under environmental changes. In [
27], presented the efficacy of the ultrasonic technique in discerning healing from its failure. In [
28], studied real crack and influence on the diffusion parameters (degradation of the signal scattered from structural deformation). The diagnosis of large cracks/notches and the monitoring of crack propagation using diffuse ultrasonic wave can be found in [
28,
29]. In [
30], Michaels and Michaels have presented the structural change in a simple aluminum specimen using short-time cross-correlation of two diffuse ultrasonic signals recorded from the same transmitter and receiver, before and after damage. Anugonda et al. [
31] investigate the propagation and scattering of ultrasound in concrete structure and determined the diffusion parameters. In [
32], Won presented the measurement of the artificial cracks varying depth in the concrete specimens with diffuse ultrasound. Eunjong et al. [
33] examined the water permeability and chloride ion penetrability of cracked concrete sample using diffuse ultrasonic signal and shown that the relations between crack width, water flow, and diffuse ultrasound parameters. Considering this, the diagnosis of propagation of microcracks in reinforcement concrete remains a significant challenge for NDT techniques, despite the special interest in making such degradation since these cracks may lead to undesirable premature failure. Advanced signal processing techniques, such as time-frequency domain analysis, statistical, matching pursuit, and other, could be useful for determination of damage-sensitive features. In most of the cases, different NDT techniques produce multiple decisions, often conflicting about the integrity of the monitored structure. In [
34], the distributed fiber optic and coda wave techniques for damage investigation in concrete structure are presented, and they showed that both techniques achieved earlier damage detection than standard sensors. However, no statistical methods have been used to compare all the techniques. The above challenge led researchers to use fusion techniques at different levels of data processing. Information-level fusion has been used after data transferred into abstractions. In NDT techniques, distracted decisions at a high level of abstractions may be produced by several techniques about the integrity of the structure/material [
35]. Typically, decision fusion is applied at the final stage of the process of evaluation. Ideally, decision fusion reduces the level of uncertainty in the decision made by different techniques and produce more trusted decisions with high level of confidence.
Cracks in rebar-reinforced concrete beams provide a very useful first warning for the monitoring of structures in risk environments. In this paper, the cracks are caused by static load application on a reinforced concrete beam equipped with four embedded ultrasonic sensors in a four-point arrangement. On the transmitted ultrasonic signals, the different features are extracted as a function of the load. To evaluate the accuracy of the crack detection with the embedded ultrasonic sensors, the state change of the beam is monitored with additional NDT methods. Moreover, the implementation of decision fusion method may significantly reduce the level of uncertainty and enhance damage detection ability.
As it can be concluded from above literature review, the information fusion on various levels is a very helpful technique, which can provide earlier crack detection in undamaged structures due to the appropriately constructed decision-making algorithm. Moreover, for successful detection of damage one need to apply the processing algorithms with enough sensitivity to detect early cracks in a structure. Finally, the appropriate threshold should be established to distinguish between healthy and damaged structure and minimize the possibility of false damage indications. In this paper, the authors combined all these approaches in order to develop a damage detection system based on embedded sensors, which is sensitive to microcracks and robust to false damage indications, simultaneously. The originality of this paper covers the application of newly developed features based on advanced signal processing measures and functions, selection of the most sensitive ones to cracks, and their implementation into the fusion algorithm, which resulted in increasing of overall sensitivity to damage in considered benchmark structure.
The rest of the paper is organized as follows. In
Section 1, the test object and experimental setups are described, then the acquired signal is explained along with proposed features. In
Section 2, these features are evaluated with the different NDT techniques. The role of information fusion method is introduced, and the comparison of different NDT techniques is highlighted using receiver operating characteristic (ROC) curves also in
Section 2. The whole paper is summarized and remarked in the end of
Section 3.
4. Conclusions
The presented study was aimed to evaluate various structural testing techniques to detect early cracking behavior in RC structures. For this purpose, various testing techniques and processing algorithms were used. Apart from external sensors, the primary attention was paid to embedded ultrasonic sensors. The ability of diffuse ultrasonic technique in monitoring the cracking behavior of the tested beam specimens was verified. Similar results were observed when the load-crack opening curves obtained from ultrasonic features and traditional sensors were compared. For the detection of changes in the RC benchmark structure, the most sensitive NDT method is the diffusing ultrasonic sensors. From the ultrasonic features, the formation of microcracks inside the beam is detected and localized. This result clearly implied the great ability of ultrasonic features to detect the crack opening and crack propagation. The peak to peak amplitude and AR coefficient indicator is interesting since it both continuously evolves and follows well the three different phases which describe the failure mechanisms that are the microcracks initiation, the propagation of cracks, and the final failure. Thanks to the use of the CWT coefficient feature, the damage could even be detected before it reaches the surface. The evaluation of the ultrasonic signal attenuation leads to the early crack detection, even before this crosses the direct wave path.
The DIC, on the other hand, detects the smallest deformations of the surface caused by the bending of the beam. Crack developments through the section depth were also monitored by DIC, and it was found that the width of the crack sustained for a certain stage of loading. When multiple cracking behavior is observed, the advantage of the DIC method over other measurement techniques was noticed, since cracks appear on concrete face parallel to the load application lead to registering of increased values. The exact cracking locations cannot be predicted with other methods. The study has furthermore shown that the use of features and feature-based fusion improves the overall decision based on the detection capability of a multisensor system located in different places of the structure.
Finally, it can be concluded that ultrasonic measurements have the potential to be used as an alternative to more traditional sensors, and, offer significant advantages over traditional measurement techniques because it can provide full-field surface strain measurements, as well as be advantageous in determining crack opening and propagation. Most importantly, the ultrasonic feature can monitor even tiny strains/cracks during loading. Therefore, this part of our work is in progress. The effect of signal-level fusion on all transducer pairs to find localization will be investigated. Also, the effectiveness of ultrasonic sensors for long term monitoring in the real structure will be in the focus of further studies.