Back to search

Democritus University of Thrace (DUTH)

Ενίσχυση και επισκευή δομικών στοιχείων κόμβων από ωπλισμένο σκυρόδεμα με σύνθετα υλικά

Abstract

dc:description

The failure of beam-column joints in Reinforced Concrete (RC) structures in high-seismic-risk regions is critical, as it not only causes significant damage but often leads to immediate partial or total collapse of the structure, resulting in significant human losses. As modern structures constitute only a small percentage of the global building stock, many existing structures exhibit low load-bearing capacity during seismic events because they were constructed under older design and analysis methods, as well as inferior materials and construction techniques. During the last decades, a significant global effort has been devoted to upgrading and/or rehabilitating those structures. A large part of this effort involves the use of new materials, such as Fiber Reinforced Polymers (FRPs). These materials offer significant advantages and certain disadvantages, and their applicability for strengthening RC members still requires further research. In this doctoral dissertation, the mechanical behavior of RC beam-column joints under seismic loading is investigated, along with the effectiveness of FRP rehabilitation/strengthening techniques. The predictions of the relationships proposed by international codes for assessing the capacity of joints, with and without strengthening, are evaluated. A simplified analytical model of the mechanical behavior of FRP-strengthened RC joints is proposed, whose reliability is evaluated using experimental data. Furthermore, appropriate Machine Learning algorithms are developed to determine the joint's failure mode. Finally, through the application of Explainable Machine Learning methods, analytical relationships for predicting failure mode are derived, demonstrating significantly greater predictive reliability than conventional analytical predictions. Specifically, the dissertation includes: 1. The literature review and existing knowledge: i. In the field of predicting the mechanical behavior of reinforced concrete joints under cyclic loading, as well as the techniques for their repair and strengthening with fiber-reinforced polymers (FRP). The parameters governing the joint's behavior are also analyzed, and the failure modes are classified. ii. In the collection and presentation of the main existing models for calculating joint strength of BC joints without strengthening under seismic loading (empirical or semi-empirical and diagonal strut-and-tie models). In the presentation, there are also analytical models for predicting failure mode critical values, including 2D and 3D finite element models used by various researchers. iii. In the presentation of modern codes’ provisions for assessing the (seismic) load-bearing capacity of RC beam–column joints without strengthening.Specifically, the regulations of the following are presented in detail: (a) Eurocode 8 (EC8-1 and -3), (b) Greek Code of Interventions (KAN.EPE.), (c) fib90, (d) Italian CNR DT 200-R2, and (e) American code (ACI 369). iv. Τhe type of the required interventions, classified according to the deficiency they address (diagonal compression of the joint web or insufficient joint reinforcement), to increase the load-bearing capacity and ductility of joints (strengthening/seismic upgrade), or to restore them to their initial capacity (repair/restoration/rehabilitation). v. In the objectives of upgrading RC beam-column joints using FRP, which include: (a) increasing the capacity of the structural members framing into the joint (beam, column), (b) increasing the shear capacity of unreinforced or under-reinforced BC joints, (c) improving the bond between beam reinforcement bars with insufficient anchorage and the joint, to enhance the flexural capacity of the member, and (d) shifting the potential failure from the joint or the column to the adjoining beams. vi. In the internationally prevalent techniques for the application of fiber- reinforced polymers (FRP), both for the strengthening and rehabilitation of beam–column joints, with characteristic cases including: (a) strengthening of beams or column framing into the joint in flexure and/or shear, (b) strengthening of the exterior or interior joint in shear, and (c) combined strengthening (shear strengthening of the joint together with flexural and shear strengthening of the column and/or beam), as well as in the results from the application of the above techniques. Special emphasis is placed on the NSM (Near Surface Mounted) technique for FRP installation, which is a relatively recent method, in which FRP fibers (rope), are placed inside grooves cut into the surface of the member. The advantages and disadvantages of using FRPs for joint strengthening are identified and described. vii.In the assumptions and provisions for the design of strengthening using FRP materials, as introduced and proposed by modern codes related to interventions and strengthening. Specifically, the design philosophy, the failure models of the individual mechanisms and the safety approach of the most widely adopted codes in this field are presented: (a) Eurocode 8 (EC8- 3), (b) Greek Code of Interventions (KAN.EPE.), (c) fib90, (d) Italian code (CNR DT 200-R2), and (e) American code (ACI 440-2R). 2. The construction and experimental investigation of beam–column joints, with and without FRP strengthening, in order to experimentally examine the following cases: (a) Strengthening using the new NSM technique, in which FRP (ropes) are embedded in grooves. It should be noted that the vast majority of relevant publications concern the strengthening of joints using externally bonded FRP fibers (EBR), (b) the experimental investigation of FRP mechanical anchors with fan-shaped ends in EBR strengthening, and (c) the behavior of corner joints (presence of transverse beam) with plain bars, strengthened with FRP.Specifically: i. At the Laboratory of Reinforced Concrete and Seismic Design of Structures of Democritus University of Thrace (DUTH), four full-scale (1:1) reinforced concrete beam–column joints were constructed. Two of them were strengthened with FRP continuous rope using the NSM method, while one unstrengthened and one strengthened specimen had conventional shear reinforcement in the joint (stirrups). The specimens were tested under (pseudo-)seismic loading, with imposed (controlled) displacement at the beam tip. This experimental investigation aimed to evaluate the seismic behavior of unreinforced or under-reinforced concrete joints, as well as the influence of using continuous carbon fiber rope as a strengthening method (NSM) to improve their seismic response. The main conclusions include: (a) For the joint specimens without strengthening, the presence of even minimal transverse reinforcement (stirrup) in the joint, improved the mechanical behavior in the post-peak (descending) branch under all loading cycles. Similarly, NSM strengthening with CFRP ropes improved both load-bearing capacity and ductility, particularly at large displacement levels. Moreover, comparing the conventionally reinforced specimen without strengthening with the unreinforced but strengthened specimen showed comparable mechanical response, with this strengthening technique acting as a quasi-restoration of the reinforcement lacking in the unstrengthened joint. (b) In terms of total energy dissipation, the strengthened specimens exhibited improved performance at large deformations. The use of FRP in the strengthening of the unreinforced specimen enhanced its energy dissipation more than that of the corresponding of the conventionally reinforced one. (c) The specimens strengthened with diagonally placed rope into grooves (NSM), showed improved behavior, with lower shear deformation values for the same relative displacement level of the beam tip. The best performance in terms of strength and shear deformation was observed in the specimen with joint stirrup and FRP strengthening, which maintained an almost elastic response at the beam tip for relative displacements up to δ≤3%. (d) For large deformation levels, the FRP-strengthened specimens exhibited a lower Park and Ang damage index compared to the corresponding unstrengthened ones. ii. At the Department of Structures for Engineering and Architecture, University of Napoli “Federico II”, in collaboration with the University of Sannio, four full-scale unreinforced RC corner beam–column joints, using plain reinforcing bars and two different longitudinal reinforcement ratios in the beam (0.4% and 0.67%), were fabricated. Two of the joints were strengthened with quadriaxial CFRP fabric, anchored with mechanical CFRP spike anchors with fan-shaped ends. The specimens were subjected to (pseudo-)seismic loading, with imposed (controlled) displacement at the beam tip. The design of the specimens was according to older Italian seismic provisions (D.M. 1982, D.M. 1986), representing a first-floor corner joint of an RC building, designed for gravity loads in a region of moderate seismic hazard (approximately 0.07g), and with details similar to those found in buildings constructed in Greece. The main results of this study include: (a) The specimens without strengthening exhibited significant deficiencies under pseudo-seismic loading. In TP_0.40%_as-built, although the beam reached flexural yielding, extensive cracking within the joint panel prevented the development of its full flexural overstrength, leading to joint shear failure. Specimen with the higher geometric ratio of beam longitudinal reinforcement (TP_0.67%_as-built) experienced even more severe cracking in the joint, which inhibited beam yielding entirely. It also caused premature shear failure of the joint, accompanied by a significant strength and stiffness degradation, as well as bar buckling of the column’s reinforcement into the joint panel. (b) The FRP- strengthened specimens showed improved performance compared to the corresponding unstrengthened ones. Specimen TP_0.40%_1L_12A, strengthened with one layer of CFRP fabric and 12 anchors, exhibited a 20.6% increase in strength and fully developed the beam’s flexural overstrength. The anchors effectively interrupted FRP debonding propagation, resulting in a 176% increase in energy dissipation and a shift in failure mode, from brittle joint shear failure to a more ductile beam-hinge mechanism, with significantly increased ductility. Specimen TP_0.67%_2L_16A, with the higher beam reinforcement ratio and strengthened with two layers of CFRP fabric and 16 anchors, exhibited a 30% increase in strength and a 52% improvement in energy dissipation. However, the increased bending moment demand in the column led to failure of column anchors, followed by a significant stiffness degradation, anchor failure on the beam and ultimately joint shear failure.This strengthening configuration, using mechanical anchors, delayed FRP debonding and enabled a better utilization of the material. Proper extension of the CFRP fabric and its secure anchorage further optimizes the strengthening performance and mitigates the risk of debonding. Additionally, it increases both the strength and energy dissipation of the specimens and shifts the failure mode from joint shear failure to a more ductile mechanism, forming a plastic hinge in the beam. 3. The creation of two comprehensive experimental databases, which incorporate internationally published data on joints without strengthening, and on joints strengthened and/or repaired with composite materials (FRP). The database for joints without strengthening, includes data and experimental results from 652 RC beam–column joints and is one of the largest, recently updated collections of experimental data on joints subjected to high-intensity cyclic loading until failure. Due to the large number of specimens included, it can be used to draw reliable conclusions regarding joint failure, to compare with existing analytical relationships and to evaluate the reliability of the relevant provisions in modern assessment codes. From the analysis of the data, the following conclusions were drawn: i. For joints without strengthening:· Exterior joints exhibited shear failure (S) in 49% of cases, while the percentage of combined flexural and shear failure (S-F) was 29%. Failure due to insufficient bond of reinforcement (B) occurred in 3% of cases. The corresponding percentages for interior joints are 32% for shear failure (S), and 56% for combined flexural and shear failure (S-F), while no cases of bond-related failure (B) were observed in the database. · The presence of transverse reinforcement within the joint (even in small ratios) generally improves the mechanical behavior, as it reduces the percentage of brittle shear failure (S) to 37% compared to 57% in unreinforced joints. In comparison, it increases the percentage of combined flexural and shear failure (S-F) from 23% to 45%. No cases of bond-related failure were presented clearly. · From the results of the investigation on the influence of joint reinforcement on its shear capacity, it was found that the maximum obtained shear strength (τexp) exhibits an upper bound, which depends on the compressive strength of the concrete (fc), independently of the existing shear reinforcement ratio (ρjh %).For exterior joints without reinforcement, this limit corresponds to 35% of the concrete compressive strength (τexp<0.35fc). For values of reinforcement ratios of the joint in the range 0<ρjh<0.4%, the average experimental values of τexp/fc increase almost proportionally to the increase in joint shear reinforcement ratio (ρjh). On the contrary, for ρjh>0.4%, the addition of shear reinforcement in the joint does not significantly affect the increase of the maximum shear stress capacity, which is mainly governed by the concrete strength.For interior joints, the ratio τexp/fc ranges between 0.1 and 0.4 (0.1<τexp/fc<0.4), exhibiting a clear lower bound. It is also observed that the increase in shear reinforcement has a significant effect on joint shear strength for reinforcement ratios ρjh≤0.3%. In contrast, for higher reinforcement ratios, it does not substantially influence the increase in the maximum shear stress capacity, which is also governed by the concrete strength.The above conclusions, formulated for the first time, were derived from a large number of joint specimens, including both exterior and interior joints. · For joints with shear reinforcement (ρjh>0), the correlation between the mechanical reinforcement ratio ω (ω=ρjhfy/fc) and the normalized experimental shear stress (τexp/fc) exhibits a relatively weak correlation pattern. On the contrary, plotting the square root of the concrete compressive strength of the joint (√fc) against the experimental shear stress at joint failure (τexp) reveals an improved correlation between these two parameters. For exterior joints at failure, the lower bound of the maximum shear stress capacity can be taken as τexp=√fc-4.0 MPa. For failure in the joint panel of interior joints, the corresponding lower bounds are given by τexp=0.5√fc+2.0MPa and τexp=1.6√fc-4.0MPa, for unreinforced and reinforced joints, respectively. · From the comparative assessment of the reliability of the results for estimating shear demand at the joint using the analytical models (EC8-1 and Greek Code of Interventions-KAN.EPE.), it was found that, for a significant number of joints, the predicted values are lower than the corresponding experimental values. Specifically, it was observed that Vmax,exp-Vjh,analyt<0 in 41% of the cases. This fact indicates that, for most specimens, the joint panel failed at a level above the calculated shear demand, as predicted by analytical equations and transferred from the beam to the joint. · Comparative results derived based on the provisions of the most widely employed international codes (KAN.EPE., fib90, CNR DT 200-R2, EC2 and EC8, and ACI 369) for the maximum shear capacity of the joint, showed that the code predictions deviate from the corresponding experimental results. Specifically: (a) Results from the application of KAN.EPE. (Vj,cap,KAN.EPE.) underestimate the shear demand required for joint failure in 56.7% of exterior joints (Vj,cap,KAN.EPE.<Vj,exp) and in 27.8% of interior ones. (b) The provisions of Eurocode (EC8-3 and EC8-1) for the assessment of joint shear capacity (Vjcap,EC8) show significant deviation from the corresponding experimental values (Vjcap,exp). Specifically, concerning exterior joints, in 94.9% of cases, the analytical value of the maximum shear capacity is higher than the corresponding experimental value at which the joint core failed. Similar results are observed for interior joints, where analytical values are higher than the corresponding experimental ones in 95.5% of cases. This discrepancy between experimental and analytical values is attributed to the fact that EC8-3 provisions apply to reinforced concrete structures with high ductility (Ductility Class High – DCH), for which, strict requirements for local ductility in critical structural elements are imposed. For beam–column joints designed for DCH, the joint core is assumed to be confined, ensuring high shear and compressive strength. The application of the above provisions, using a reduced value of the concrete strength reduction factor due to transverse tensile strains, with η=0.35(1-fck/250), instead of the value η=0.60(1-fck/250) given in EC8-1, shows significant improvement, overestimating the joint shear capacity in 34.4% of the cases for exterior joints and in 32.2% of the cases for interior joints. It is proposed that, for the assessment of joint capacity in reinforced concrete structures which have not been designed according to EC8-1 provisions and do not belong to DCH class, the reduction factor for the concrete compressive strength of the joint due to transverse tensile strains to be taken within the range η=0.30(1-fck/250) to η=0.40(1- fck/250). (c) The assessment of joint strength using the provisions of fib90 and CNR DT 200-R2 yields more conservative results, with the analytical values Vj,fib,CNR being lower than the corresponding experimental values (Vj,exp) in 96.6% of exterior joints. In particular, for interior joints, all analytical values are lower than the corresponding experimental values. (d) The assessment methodology proposed by ACI 369, based on statistical processing of experimental data, shows relatively satisfactory reliability for estimating joint shear failure, with its predictions exceeding the corresponding experimental values (Vn,ACI>Vj,exp) in 41.1% of exterior joints and 12.8% of interior ones.ii. For FRP-strengthened joints: · A corresponding database was developed, which constitutes the largest in this field and includes data for 180 RC beam–column joints (160 T-type and 20 corner joints). As can be observed, the investigation of the mechanical behavior of joints strengthened and/or repaired using FRP, remains a complex problem. · The strengthening of joints with FRP included: (a) strengthening of beams framing into the joint in flexure and/or shear, (b) strengthening of the column in flexure and/or shear, (c) strengthening of the joint core of exterior joints in shear, and (d) combined strengthening (strengthening of the joint core together with strengthening of the column and/or beam in flexure and/or shear). · The classification based on the strengthening material showed a clear predominance of CFRP (carbon fiber) at 91%, compared to 9% for GFRP (glass fiber). The most prevalent FRP strengthening techniques are externally bonded strips/sheets (Externally Bonded Reinforcement – EBR) and ropes embedded in grooves (Near Surface Mounted – NSM). The employed configurations include X, L, U, T, and full wrapping, with 1–4 layers. The main anchorage types are overlapping (for sheets and ropes) and the use of mechanical anchors. · The distribution of concrete compressive strength spans a wide range, corresponding to older reinforced concrete structures with low seismic capacity, with fc≤30 MPa (52%) (structures built up to the 1970s). However, there are also specimens (48%) with relatively high concrete strength (fc>30 MPa), reflecting researchers’ interest in the seismic upgrading of joints in more modern structures as well. · For the majority of strengthened and/or repaired members using externally bonded FRP, failure was due to partial or total failure of the FRP substrate and FRP debonding (bond and/or anchorage failure). The correlation between the maximum experimental shear strength of the FRP-strengthened joint (τexp) and the concrete compressive strength (fc) is relatively weak, with significant scatter in shear-strength values. · The use of FRP for joint strengthening increased the shear capacity (Vs) of the strengthened member compared to the corresponding unstrengthened (as-built) member (Vv) by up to 140% [(Vs-Vv)/Vv ≤ 1.40]. On the contrary, for joints that had been previously loaded and subsequently repaired, the increase in shear capacity was significantly lower, with ratios of (VR+S- Vv)/Vv ≤0.9. · It is proposed that the upper limit of the maximum shear capacity of the strengthened member be taken as the maximum shear force of the as- built joint (Vv) increased by 100% (Vs≤2Vv). Beyond this limit, intervention in the joint geometry is required, with an increase in its geometric dimensions. Similarly, for joints that have been previously subjected to loading and, if damage in the joint core has been successfully repaired, the upper limit of shear capacity increase could be taken as 60% of the shear capacity of the as-built joint (VR+S≤1.6Vv). · For most experimental data, the total tensile capacity of the installed FRP was, by an order of magnitude, significantly greater than its actual contribution to resisting tensile forces, which remained largely underutilized in both strengthened and repaired joints. This fact was mainly due to the failure mode, which was governed by FRP debonding and/or anchorage failure rather than rupture of the FRP itself, highlighting the need for a more rational application of the strengthening material. · From the comparison of the maximum shear capacity of the FRP- strengthened joint, as obtained from the analytical expressions proposed by the main international codes (Van), with the corresponding experimental shear value at joint failure (Vexp), it was found that: – Provisions of EC8-3 for calculating the maximum horizontal shear capacity of the FRP-strengthened joint yield analytical values higher than the corresponding experimental ones for the entire set of joints, with the ratio VEC8/Vexp ranging from 1.4 to 6.69. The application of EC8-3, in conjunction with EC2 provisions, also overestimates shear capacity by 41.8%, with a VEC2/Vexp ratio ranging from 0.24 to 2.56. – The application of provisions of the Greek Code of Interventions (KAN.EPE.) overestimated the joint shear failure value by 51%, with theratio of analytical to experimental values ranging between VKAN.EPE./Vexp =0.15-2.78.– From the application of the relationships proposed by fib90 and CNR DT200-R2, the percentage of joints for which the shear capacity (Vfib,CNR) was overestimated is 41.8%, while the ratio of analytical to experimental shear failure values ranged between Vfib,CNR/Vexp =0.38- 2.08. The provisions of the new Italian CNR DT 200-R2/2025 for the prediction of strengthening with the use of mechanical FRP fan-anchors provide a satisfactory approximation of the performance of anchored fabric. – The analytical expressions of ACI, yielded analytical values greater than the corresponding experimental ones in 58% of cases, with the ratio of analytical to experimental values ranging between VACI/Vexp=0.19-6.55. · In summary: (a) the application of EC8-3 provisions exhibits the largest deviations compared to all other codes, (b) fib and CNR represent the most conservative approach, with analytical shear capacity values lower than those of the other codes and, for the majority of specimens, also lower than the corresponding experimental values, (c) ACI provisions, in their simplicity and ease of use, provide relatively reliable analytical values, and (d) the results from the application of the Greek Code of Interventions (KAN.EPE.) provide intermediate assessment values compared the other codes.4. In the context of analytical investigation, a simplified analytical model for the inelastic mechanical behavior of FRP-strengthened RC joints is proposed and shown to exhibit satisfactory reliability. This model can be applied to predict the time-history response of structural systems with a large number of joints, without the need for 3D finite element analysis, thereby reducing complexity and computational cost. The joint response is correlated with the two main mechanisms: shear behavior and rotation due to the slip of the beam longitudinal reinforcement. The contribution of FRP strengthening is achieved through the use of an extra-rotational spring. The model also incorporates, in addition to the contribution of FRP strengthening, the effect of FRP debonding.A comparative investigation was carried out between the predictions of the proposed analytical model applied to joint specimens and the corresponding experimental results. The investigation pertained to the prediction of the behavior of six joints without strengthening, based on experiments conducted within the framework of the present doctoral dissertation, as well as corresponding cases reported in the literature, and also four FRP-strengthened joints, with different strengthening configurations and layouts, most of which represented the strengthened counterparts of the unstrengthened ones.The main conclusions from this part of the research include: · The proposed model represents an intermediate solution combining ease of application, with a satisfactory level of reliability. Using readily available data, the model assesses the overall mechanical behavior of the joint, not only in terms of maximum load (as provided by the relevant code relationships), but also in terms of load-bearing capacity, displacement, stiffness, and total energy dissipation. The model accounts for: (a) the shear behavior of the joint, (b) its rotation due to slip of the reinforcing bars, (c) the FRP strengthening and the associated debonding failure mechanism. The quantification of the strengthening contribution is assessed using specialized formulations for fib90 and the Italian code for strengthening (CNR), both incorporated into the proposed model. · The model applied to joints without strengthening yielded satisfactory accuracy, particularly for deformations beyond the peak load, where it reliably predicted the load at each step. In addition, the model’s ability to adequately capture the variation in stiffness during unloading and reloading, led to a realistic prediction of the pinching effect in the hysteretic response. · From the evaluation of the results of the proposed model, it was found that: – Despite its simplicity, the model satisfactorily predicted the mechanical behavior of the joints for most specimens, both in terms of stiffness and load-bearing capacity, with the ratio of analytical to experimental peak load values ranging from Pmax,sc/Pmax,exp = 0.87–0.9. – In terms of total energy dissipation, the ratio Etot,sc/Etot,exp=0.77–1.06 corresponds to a level of predictive reliability similar to that of the unstrengthened specimens, the behavior of which represents the largest part of the dissipated energy. – The results can be further improved through a more detailed calibration of the input parameters and their adjustment to the type of specimen, without, however, compromising the model's simplicity and ease of use. 5. Suitable Machine Learning (ML) algorithms were developed to predict the failure mode of RC joints without strengthening under cyclic loading. The models predicted the joint failure mode with high reliability, without overfitting.Specifically, they achieved successful predictions with an overall accuracy of over 80%. XGBoost achieved the best performance, with a prediction accuracy of 85%. The Gradient Boosting and Bagging ensemble with Artificial Neural Networks (Bagging ANN) also achieved notable predictive accuracy of approximately 82.5%. · The results of applying the best-performing model (XGBoost) to predict the joint failure mode were compared with those obtained from classical analytical expressions. As shown by the comparison, the overall prediction accuracy of XGBoost was 85%, compared to 67% for the classical analytical approach. · The other performance metrics obtained from the ML model application were also significantly superior compared to classical analytical predictions for shear failure mode (JS). Specifically, Precision reached 86% compared to 65% for the classical analytical predictions, Recall was 85% compared to 82%, and F1-score was 86% compared to 73%, respectively. Larger discrepancies were observed for the flexure–shear failure mode (BY-JS), where Precision was 84% versus 71% for the classical approach, Recall 85% versus 49%, and F1-score 84% versus 58%. · Due to the demonstrated ability of Machine Learning algorithms to achieve very high predictive accuracy, as well as the need for relatively simple numerical expressions with greater reliability, the application of Explainable Machine Learning methods was investigated to derive predictive relationships for the failure mode of RC joints. The overall accuracy of successful predictions of the joint failure mode, obtained by applying the proposed numerical expressions to the entire dataset, reached 78%, compared to 84% for the full XGBoost machine learning model and 67% for the classical analytical predictions. It should be noted that the above results refer to the complete set of joints (interior + exterior).

Degree

thesis:*
Grantor dc:publisher
Democritus University of Thrace (DUTH)
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Karabini, Martha
  • Καραμπίνη, Μάρθα

Subjects

dc:subject × 25

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/62546

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Karabini, Martha; Καραμπίνη, Μάρθα. Ενίσχυση και επισκευή δομικών στοιχείων κόμβων από ωπλισμένο σκυρόδεμα με σύνθετα υλικά. Democritus University of Thrace (DUTH), 2026. http://hdl.handle.net/10442/hedi/62546