Experimental and numerical study on FRP-rehabilitated RC beam-column joints at high temperature with artificial neural network

Abstract

This research explores the structural integrity and thermal durability of reinforced concrete beam—column joints rehabilitated with fiber-reinforced polymer (FRP) laminates under elevated temperatures reaching 800 °C. The work addresses a significant knowledge deficiency by combining computational, experimental, and machine learning methodologies to thoroughly assess FRP performance under thermal stress, a subject inadequately explored in previous literature. A total of 36 models, including 9 conventional and 27 rehabilitated configurations, were analyzed through coupled thermo-mechanical simulations using finite element software. The novelty lies in the pre-experimental identification of critical regions and optimal materials via numerical analysis. At 500 °C, CFRP-rehabilitated joints reduced deflection by up to 42.8% and stress by 37.2% compared to GFRP. In contrast, AFRP showed over 60% higher deflection. The most vulnerable area was identified as the joint core, especially on the column face adjacent to the beam. Experimental tests confirmed CFRP’s superiority; with specimens showing a 28.5% higher load capacity and 31.6% lower core temperature at failure than GFRP-enhanced specimens. Artificial neural network (ANN) regression models were developed to predict deflection and nodal temperature based on input parameters. These models yielded high accuracy (R² > 0.99), closely matching experimental and numerical results. However, generalizing predictions beyond the studied range may introduce over fitting risks, and the model remains sensitive to data quality. In summary, CFRP demonstrated optimal performance, particularly at 400 °C before rehabilitation and 500 °C afterward, making it the most effective choice for high-temperature FRP-based RC joint rehabilitation. This integrated methodology presents an all comprehensive structure for performance-oriented FRP restoration of reinforced concrete joints.

Introduction

Reinforced concrete (RC) structures play a pivotal role in ensuring the integrity and overall stability of buildings and infrastructure, with beam-column joints being especially crucial components. External joints, in particular, face significant risks of deterioration stemming from environmental influences, seismic activity, and prolonged service conditions¹. In recent decades, extensive research has focused on exploring the behavior of RC joints when subjected to different loading conditions, alongside the development of effective rehabilitation techniques. FRP laminates stand out for their exceptional mechanical properties, impressive corrosion resistance, and straightforward application process, making them a focal point of interest in various fields^2,3. This research focuses on exploring the rehabilitation of reinforced concrete external joints Utilizing fiber-reinforced polymer laminates, employing a thorough experimental, numerical, and regression-based methodology. This study combines experimental investigation, finite element modelling (FEM) with ABAQUS/CAE version 6.17 (www.3ds.com/products/simulia/abaqus), and regression analysis Utilizing ANN to create predictive models for the performance of rehabilitated joints⁴.

In the realm of structural engineering, it is crucial to recognize that RC beam-column joints, especially those located on the exterior, face a heightened risk of substantial damage when subjected to cyclic, seismic, and thermal loading conditions⁵. Extensive research has underscored the susceptibility of these joints, emphasizing the importance of implementing effective strengthening and rehabilitation approaches. Numerous studies have investigated the behavior of reinforced concrete joints when subjected to seismic forces, revealing that insufficient detailing, inadequate confinement, and material degradation contribute to brittle failures⁶. Recent experimental investigations into damaged reinforced concrete joints have highlighted the effectiveness of various rehabilitation techniques, including steel jacketing, epoxy injection, and externally bonded fiber-reinforced polymer composites^7,8. In the realm of advanced materials, FRP laminates stand out as a compelling option, celebrated for their exceptional strength-to-weight ratio, straightforward installation process, and corrosion resistance⁹. Extensive research has been conducted on the application of carbon fiber-reinforced polymer (CFRP) and glass fiber-reinforced polymer (GFRP) laminates to strengthen RC joints. Research has demonstrated that CFRP wraps enhance joint ductility, shear capacity, and load vs. deflection characteristics when subjected to mechanical loading¹⁰. Furthermore, the combination of hybrid strengthening techniques that incorporate FRP alongside other rehabilitation methods, epoxy-based injections, has shown enhanced effectiveness in restoring the structural integrity of compromised joints¹¹. Nonetheless, in light of these advancements, it is important to note the absence of thorough studies that combine experimental, numerical, and regression-based methods to enhance the optimization of FRP rehabilitation for RC joints¹².

This study examines the rehabilitation of RC external beam-column joints utilizing FRP laminates, incorporating experimental testing, FEA, and ANN regression methodologies. The study examines enhancements the load, defection, crack and nodal temperature of CFRP and GFRP laminates under diverse loading circumstances. This study offers a thorough methodology by verifying numerical simulations with experimental data and utilizing artificial neural networks for predictive modelling. The results enhance the development of sustainable, economical FRP-based reinforcement techniques for degraded concrete structures.

To date, limited research has comprehensively integrated experimental validation, finite element modelling, and machine learning-based regression to assess the rehabilitation of RC beam–column joints using FRP laminates under elevated temperatures. The novelty of this study lies in its multi-method approach identifying critical temperatures and optimal materials numerically, validating through experiments, and predicting outcomes using ANN. The main objective is to evaluate the performance of FRP-strengthened RC joints under thermal and mechanical loads, and to develop predictive tools for deflection and core temperature. This study addresses the performance gap in existing rehabilitation techniques and provides practical design guidance for engineers and researchers.

Methodology

This flowchart illustrates a systematic approach for examining the rehabilitation of RC structures Utilizing FRP laminates under thermo-mechanical loading as shown in Fig. 1. The research process initiates with a review of existing literature, establishing a foundation for comprehending prior studies, pinpointing gaps in knowledge, and outlining research goals. The materials properties section encompasses two essential components: the collection of materials, including cement, fine aggregate, coarse aggregate and FRP laminates, and the preliminary testing phase, which involves conducting material characterization tests such as specific gravity, sieve analysis, and water absorption. Considering these properties, a carefully crafted mix design and detailing phase guarantees the required strength and workability of the concrete mixture prior to moving on to experimental and numerical analyses. The experimental process includes the casting and curing of RC specimens, subsequently subjected to mechanical and thermo-mechanical loading tests to assess structural behavior until the onset of initial crack formation. Following the initial damage, rehabilitation is conducted through the application of resin and FRP laminates, thereby improving the structural capacity. Following this, additional mechanical and thermo-mechanical loading is implemented to evaluate performance until failure occurs. The numerical analysis is conducted in a structured manner, encompassing the definition of boundary conditions and the simulation of loading conditions in two distinct phases: Phase 1 (prior to laminates) and Phase 2 (subsequent to laminates), during which both pressure loads and the effects of elevated temperatures are assessed. The concluding phase of meshing and analysis confirms the experimental results by assessing stress distribution and deflection characteristics. The results obtained (load and deflection) are analyzed to assess the effectiveness of FRP rehabilitation. This is followed by a comparison and discussion phase aimed at interpreting the findings and drawing conclusions regarding the structural performance of rehabilitated RC members when exposed to elevated temperatures.

Fig. 1 [Images not available. See PDF.]

Flowchart for the rehabilitation of RC joint using FRP laminates under thermo-mechanical loading.

Rehabilitation of RC external joints using FRP laminates

FRP laminates, especially CFRP and GFRP, find extensive application in the strengthening and rehabilitation of RC joints. This is largely attributed to their impressive tensile strength, durability, and ability to withstand environmental degradation¹³. The performance enhancement of RC joints through FRP is influenced by various factors, including the type of fiber used, its orientation, the bonding method employed, and the technique applied during the process¹⁴. Furthermore, Surya Prakash¹⁵ discovered that FRP confinement improves shear resistance and deformation capacity, which helps to postpone joint failure. Nonetheless, issues like debonding, stress concentration, and the long-term durability of materials continue to be significant topics for investigation¹⁶.

This study explores the structural performance of RC external joints, examining the effects of rehabilitation with CFRP and GFRP laminates, both prior to and following the intervention. RC beam-column specimens, crafted according to ACI 318 guidelines, will experience axial and lateral loads to replicate actual and seismic scenarios. Results are compared control specimens with those that have been damaged, focusing on the rehabilitation methods full-wrap FRP configurations. The load-deflection response, crack patterns, Stiffness Degradation, Energy Dissipation and failure modes will be documented for further analysis. A detailed finite element analysis will be carried out concurrently. A 3D finite element model that integrates concrete damage plasticity, cohesive zone modelling, and surface interactions is set to simulate nonlinear behavior and the debonding of fiber-reinforced polymers^17,18. The model will undergo validation with experimental data and will facilitate parametric studies focusing on FRP type, orientation, and bonding. This integrated experimental and numerical method seeks to offer detailed insights into the behavior and performance of FRP-rehabilitated RC joints.

The CFRP laminate utilized in the experimental study was a unidirectional carbon fibre sheet with a modulus of elasticity of around 230 GPa and a tensile strength of about 3500 MPa. A two-part epoxy resin was used to apply CFRP. The structural adhesive epoxy resin was created especially to adhere CFRP sheets to concrete surfaces. Because of its superior mechanical qualities and tolerance to heat, it may be used in high-temperature rehabilitation applications. To guarantee good adherence and long-term durability, the resin was applied in accordance with the manufacturer’s instructions and blended in a 10:1 binder to hardener ratio.

Artificial neural network

This study aims to explore the development of predictive models that assess the performance of rehabilitated RC joints through the application of ANN regression analysis. Artificial Neural Networks represent a formidable approach in the realm of machine learning, adept at capturing intricate connections between input variables and their corresponding outputs. The input data for training the ANN model will consist of both experimental and numerical results. In this discussion, we will take into account various factors, including joint dimensions, reinforcement details, FRP type, fiber orientation, and loading conditions. In this endeavor, we will embark on the design of a multi-layer perceptron (MLP) neural network, focusing on the optimization of hidden layers and activation functions to enhance its performance¹⁹. In this process, the ANN model will undergo training with a dataset that has been divided into distinct training and validation subsets. In this discussion, we will explore the application of back propagation and optimization techniques aimed at Minimizing prediction errors. In this analysis, we will assess the accuracy of the ANN model through various statistical measures, including R-squared (R²), mean squared error (MSE), and root mean square error (RMSE)²⁰. This study focuses on Utilizing ANN-based regression analysis to forecast the performance of rehabilitated RC joints across different conditions, thereby contributing to the enhancement of rehabilitation strategies.

Numerical simulation of joints

Finite element model

FE modelling serves as an effective method for assessing the structural performance of RC joints under intricate thermo-mechanical loading scenarios. This section presents an overview of the creation of finite element models aimed at stimulating the behavior of reinforced concrete joints that have been rehabilitated with FRP laminates. The commercial software ABAQUS/CAE version 6.17 (www.3ds.com/products/simulia/abaqus), was utilized to develop these models, guaranteeing a precise depiction of actual experimental conditions²¹.

To ensure the accuracy of the FE models, experimental data was employed, revealing a significant alignment between the simulated outcomes and laboratory findings. After validation, the models provided a basis for parametric studies, allowing for an in-depth evaluation of rehabilitated RC joints under different loading conditions and FRP configurations²². To ensure an accurate representation of experimental conditions, it was essential to carefully select the appropriate finite element types and boundary conditions. Furthermore, accurately defining the material properties of concrete, steel reinforcement, FRP laminates, and adhesive layers was crucial for understanding the interaction and mechanical response of each component when subjected to applied stresses. This modelling approach offers essential insights into the structural efficiency of FRP rehabilitation, establishing a framework for enhancing reinforcement techniques in RC joints subjected to various thermal and mechanical conditions.

Modeling

The illustration depicts a finite element model of a reinforced concrete beam-column joint, which includes the beam, column, steel reinforcement, and FRP strengthening elements as shown in Fig. 2. The beam and column components are represented as beam elements to effectively capture their flexural and shear behavior. The steel reinforcement is designed as a truss element to effectively represent the transfer of axial forces within the reinforced concrete structure. FRP laminates are represented as shell elements to capture their thin-layered characteristics and their interaction with the concrete surface.

Table 1. Description of numerical analysis

S.No	Description	Modeling properties
1	Length (span)	1300 mm for column and 520 mm for beam
	Cross section	100 mm × 75 mm for column and 75 × 75 mm2 forbeam
2	Reinforcement	Main reinforcement 4#8 mm diameter bar and 6mm stirrups at 50 mm c/c (Column) and Mainreinforcement 4 #6 mm diameter bar and 6 mmstirrups at 50 mm c/c (Beam)
3	FRP laminates	Shell member with cross section of 400 mm × 900mm in column and 300 mm × 300 mm in beam in allfour sides
4	Element type	C3D8T—Coupled Temperature Displacement T2D3—Two-node truss elements S4R—Three-dimensional shell components
5	Interaction	Embedded connection (reinforcement to concrete) Tie connection (Concrete to FRP laminates)
6	Materialproperties—concrete	EN 1992-1-2 (2004) (English): Eurocode 2: Design of concrete structures—Part 1-2: General rules—Structural fire design24
7	Materialproperties—steel	EN 1993-1-2 (2005) (English): Eurocode 3: Design of steel structures—Part 1-2: General rules—Structural fire design25

Fig. 2 [Images not available. See PDF.]

Modelling of RC joints for FEM analysis.

Fig. 3 [Images not available. See PDF.]

Loading and boundary condition of RC joints for FEM analysis.

Each component is assigned material properties like density, poisson’s ratio, and elastic modulus to facilitate a precise simulation of mechanical behavior. Concrete exhibits nonlinear material properties to effectively address issues of cracking and crushing, whereas the reinforcement steel adheres to an elastic-plastic model. FRP laminates are characterized as orthotropic materials, exhibiting significant tensile strength along the fiber direction. The beam, column, and reinforcement are developed separately and then brought together to create the entire joint structure as explained in Table 1. To simulate the bond between reinforcement and concrete, as well as the adhesion of FRP laminates, surface-to-surface interaction and embedded constraints are utilized. This model supports the assessment of joint behavior when subjected to mechanical and thermal loads, contributing to the exploration of rehabilitation strategies.

Loading and boundary conditions

When considering loading conditions, vertical loads are introduced at the free end of the beam, reflecting the external forces that occur during execution. Extra axial loads are added to the column to simulate real-world structural scenarios, like the gravity loads that are conveyed through the column in a building framework. The boundary condition is established at the base of the column, where a fixed support is implemented, restricting all degrees of freedom for both translation and rotation to emulate the effect of a rigid the basis restraint. Alongside mechanical loading, a thermal boundary condition has been introduced as shown in Fig. 3. The temperature is consistently applied throughout the beam-column joint area, varying from ambient room temperature up to 800 °C, effectively simulating the impacts of elevated temperatures akin to those encountered during fire exposure. The maximum temperature of 800 °C was selected based on standard fire exposure curves such as ISO 834²³ and Eurocode EN 1991-1-2²⁴, which are widely accepted for structural fire design. These standards indicate that temperatures can rise to approximately 800 °C maximum of 4 h of exposure during a typical compartment fire.

The thermo-mechanical loading regime is clearly defined by applying a sequentially coupled analysis. Initially, a transient thermal load is applied, simulating elevated temperatures up to 800 °C with realistic time-dependent heating curves. Subsequently, mechanical loads such as comprising axial and lateral forces are applied to replicate seismic conditions. This regime effectively captures structural response under combined stresses. This temperature range provides a realistic benchmark for evaluating the thermal degradation and rehabilitation performance of structural elements, ensuring practical relevance to real-world fire scenarios. This thermal-mechanical coupling allows for a precise evaluation of stress distribution, material degradation, and the overall performance of joints when subjected to both thermal and structural loading scenarios.

Material properties for numerical analysis

For the purpose of to achieve accurate and realistic numerical modelling, the mechanical and thermal properties of all constituent materials, including concrete, reinforcing steel, and fiber-reinforced polymers (CFRP, GFRP, and AFRP), were determined through a blend of experimental data and recognized literature values. The finite element model incorporated these properties to simulate the structural response of the RC joints when subjected to elevated temperature conditions.

Table 2. Mechanical and thermal properties of concrete and steel for numerical analysis.

Temp (℃)	Concrete				Steel
	Modulus of elasticity (MPa)	Thermal conductivity (W/m K)	Linear expansion	Specific heat (J/kg K)	Modulus of elasticity (MPa)	Thermal conductivity (W/m K)	Linear expansion	Specific heat (J/kg K)
20	27386.16	1.95	9.2E−09	900	200,000	53.33	0	439.8
100	27386.16	1.77	9.3E−06	900	200,000	50.67	0.00001248	487.6
200	26016.82	1.35	1.0E−05	1000	180,000	47.34	0.00001288	529.8
300	23278.20	1.36	1.1E−05	1050	160,000	44.0I	0.00001328	564.7
400	20539.59	1.t9	1.3E−05	1100	140,000	40.68	0.00001368	605.9
500	12431.67	1.04	1.5E−05	1100	120,000	37.35	0.00001408	666.5
600	12323.75	0.91	1.8E−05	1100	62,000	34.02	0.00001448	759.9
700	8215.83	0.51	2.1E−05	1100	26,000	30.69	0.00001488	1005.2
800	4107.91	0.72	1.8E−05	1100	18,000	27.30	l.4102E-05	503.3

*EN 1992-1-2 (2004)²⁴, EN 1993-1-2 (2005)²⁵.

In Table 2, the mechanical and thermal characteristics that were utilized in the model for concrete and steel reinforcement are shown. In Table 3, the mechanical and thermal values of various FRP laminates are presented.

Table 3. Mechanical and thermal properties of FRP for numerical analysis²⁶.

Properties	CFRP	GFRP	AFRP
Young’s modulus E1 (MPa)	130,000	30,900	67,000
Young’s modulus E2 (MPa)	8000	8300	4700
Young’s modulus E3 (MPa)	8600	8300	4700
Poissons ratio Nu12	0.28	0.26	0.26
Poissons ratio Nu13	0.28	0.2	0.26
Poissons ratio Nu23	0.33	0.2	0.28
Shear modulus G12 (MPa)	4500	3000	2000
Shear modulus G13 (MPa)	4500	3000	1586
Shear modulus G23 (MPa)	3600	3000	1586
Density (tonne/mm3)	1.56E−09	1.80E−09	1.44E−09
Conductivity W/mk	0.912	0.938	0.892
Expansion	1.90E−06	4.43E−06	− 5.20E−06
Specific heat J/kg	760	840	760

The stiffness, strength, and stress distribution in the system are all directly influenced by the mechanical characteristics, which include compressive, and tensile strength, Young’s modulus, and Poisson’s ratio, among others. On the other hand, the thermal characteristics, which include thermal conductivity, specific heat, and the coefficient of thermal expansion, are extremely important when it comes to precisely capturing the deterioration of materials when they are exposed to high temperatures and when it comes to modelling coupled thermo-mechanical behavior.

Temperature bond modeling for FRP and epoxy

The thermal degradation of FRP and epoxy resin was modeled using temperature-dependent material properties based on empirical data from the literature. Specifically, the reduction in elastic modulus and tensile strength of CFRP with temperature was adopted from Firmo et al.²⁷ and Bisby et al.²⁸, who reported quantitative degradation trends for CFRP systems subjected to high temperatures. Degradation occurs as temperature rises beyond the epoxy’s glass transition temperature (180–200 °C), leading to softening and loss of bond strength. The FRP’s tensile strength and stiffness also reduce due to thermal damage, weakening the fiber–matrix interface and causing debonding and loss of confinement.

The bond interface between concrete and FRP was modeled using surface-based cohesive zone modeling (CZM) with a traction-separation law. Teng et al.²⁹ and Lu et al.³⁰ these works provide empirical relations and reduction factors that relate shear bond strength and interface stiffness to increasing temperatures, particularly around and beyond the glass transition temperature (T_g 180–200 °C) of epoxy. In the model, the traction stiffness and peak bond stress were progressively reduced according to these empirical relations as temperature increased from room temperature to 800 °C. Beyond T_g, rapid degradation of epoxy was implemented, leading to significant reduction in interface capacity, effectively capturing the bond failure and de-bonding observed in experiments.

Meshing

The finite element model of a FRP-rehabilitated reinforced includes detailed meshing details as shown in Fig. 4. This illustration showcases the entire finite element model, encompassing the RC beam, RC column, and CFRP laminates. All structural components are divided into finite elements to enable precise numerical simulation. The meshing size for this model is set at 25 × 25 mm². This choice was made following a mesh convergence study aimed at achieving a good balance between accuracy and computational efficiency. The beam and column are interconnected through solid elements, usually 3D hexahedral elements, which are well-suited for accurately representing the intricate stress distribution. This meshing configuration effectively addresses the interactions among concrete, steel reinforcement (represented as truss elements), and CFRP (depicted as shell elements), ensuring clarity and precision in the analysis.

Fig. 4 [Images not available. See PDF.]

Meshing of whole RC joint model for FEM analysis.

Fig. 5 [Images not available. See PDF.]

Meshing of FRP laminates.

Figure 5 specifically addresses the meshing of CFRP laminate. The modelling of CFRP wraps around the joint region utilizes shell elements, which effectively represent thin composite laminates. The elements presented here effectively illustrate how CFRP interacts with the concrete substrate and its surface behavior. The same mesh size of 25 × 25 mm² is utilized here to keep things consistent and to ensure a proper fit with the surrounding concrete elements. This refined meshing effectively captures the delamination, stress transfer, and confinement effects that CFRP experiences under mechanical and thermal loads.

A fully linked thermal-stress solver for transient analysis and an implicit solver for static analysis are two examples of the explicitly described solver parameters. To guarantee solution stability and precise depiction of linear behavior, convergence criteria, time step control, and numerical approaches have been established.

Repairing algorithm

A Two-step procedure was carried out with the Step Manager in analysis to model the rehabilitation process of a reinforced concrete beam-column joint subjected to high temperature conditions. This method clearly illustrates the sequential application of thermal and mechanical loads together with CFRP strengthening³¹. In Step 1, carries a coupled temperature-displacement analysis, applying beam and column loads at the same time. Furthermore, a thermal load was introduced at the lower section of the column to replicate pre-damage heating scenarios. This step focuses on assessing how the joint performs structurally prior to the application of CFRP, considering both thermal and mechanical loads together. After completing the analysis, CFRP laminates are applied around the joint region, which is modelled using shell elements to represent external wrapping.

In Step 2, we defined another coupled temperature-displacement step. The identical beam and column loads were utilized; however, the thermal load was shifted above the wrapped area to replicate post-rehabilitation conditions with increased temperatures. This step evaluates how well CFRP protects and contains the joint from additional harm caused by thermal influences. The two-step algorithm illustrates a practical situation of repair and reloading: first, there is initial damage caused by heat and load, and then comes the phase of rehabilitation and re-exposure. This method facilitates an in-depth examination of how CFRP reinforcement performs structurally, especially when subjected to a series of damage and repair scenarios.

Result and discussion

The focus of this numerical analysis is to pinpoint the crucial temperature range and determine the optimal strengthening material for RC beam-column joints when exposed to high temperatures. Three types of FRP laminates are CFRP, GFRP, and AFRP were examined under thermal exposure ranging from room temperature to 800 °C. A total of 36 models were created, comprising 27 FRP-rehabilitated models across nine temperature intervals and 9 conventional unwrapped models as shown in the Table 4.

Table 4. Details of number of models to analysis in numerical investigation.

S.No	Model	Number of model for analysis	Purpose
1	Conventional model	9 (Room temperature to 800 °C)	Identify the critical temperature
2	Rehabilitation model	9 for CFRP (Room temperature to 800 °C)	Identify the perfect material with critical temperature
		9 for GFRP (Room temperature to 800 °C)
		9 for AFRP (Room temperature to 800 °C)

The models underwent mechanical loads from beams and columns, as well as thermal loading, through a coupled temperature-displacement analysis. the main response parameters examined included the maximum Deflection and the maximum principal stress at the joint region. the performance of each material was assessed using these parameters across various thermal conditions. the results were analyzed alongside experimental outcomes to confirm the accuracy of the numerical model. the analysis revealed the top-performing FRP material along with the critical temperature that leads to the most substantial degradation. the chosen conditions optimal material and key temperature were carefully applied in the experimental study to confirm and support the numerical results, thereby ensuring the dependability and relevance of the suggested repair approach. A Deflection result for single model is shown in Fig. 6 and categorizes CFRP-strengthened joint 6(a), GFRP-strengthened joint 6(b) and Un-strengthened joint 6(c).

Fig. 6 [Images not available. See PDF.]

Deflection (mm) in numerical analysis before and after rehabilitation (a) CFRP-strengthened joint, (b) GFRP-strengthened joint and (c) Un-strengthened joint.

Temperature displacement characteristics

The graph showcases the temperature-deflection characteristics, highlighting the behavior of various FRP strengthening systems CFRP, GFRP, AFRP, and a conventional RC specimen when subjected to increasing thermal exposure in numerical simulations. This research plays a crucial role in evaluating serviceability limits and understanding failure mechanisms under thermo-mechanical loading conditions. The deflection starts to rise with temperature because of thermal expansion and softening effects, hitting a maximum around 500–700 °C, which varies based on the material. CFRP shows a unique deflection pattern, featuring an initial peak near 250 °C, which is then followed by a significant drop, suggesting material degradation. GFRP and AFRP exhibit impressive stability up to 500 °C, but they start to show signs of deterioration beyond that point. The traditional specimen exhibits a comparable pattern, yet it demonstrates reduced overall deflection when contrasted with FRP-wrapped specimens.

Fig. 7 [Images not available. See PDF.]

Temperature vs. deflection characteristics of FRP based RC joint.

At 300 °C, CFRP experiences about 30–40% more deflection compared to conventional RC, but by the time it reaches 700 °C, its performance declines by more than 50%. GFRP and AFRP demonstrate a consistent level of performance, exhibiting a resistance that is 10–15% greater than that of CFRP when temperatures exceed 400 °C as explained in Fig. 7. This numerical study explores the effectiveness of FRP reinforcement at intermediate temperatures.

Principal stresses

The analysis of principal stresses offers valuable insights into the load-bearing capacity and failure mechanisms of reinforced concrete joints subjected to thermo-mechanical loading. This analysis highlights key areas of stress where the material is prone to degradation or failure as a result of temperature influences. This study evaluates various FRP to identify which material maintains superior stress retention at high temperatures, thereby ensuring structural integrity.

Fig. 8 [Images not available. See PDF.]

Comparison of various FRPs in principal stress.

As the temperature goes up, the main stress first rises because of thermal expansion and material hardening, then it hits a peak and later decreases due to thermal degradation. Across the temperature range from 27 °C to around 600 °C, all FRP specimens demonstrate a consistent rise in stress, with GFRP and AFRP displaying marginally better stress retention compared to CFRP. The maximum stress observed for CFRP, AFRP, and GFRP is approximately 8.5 MPa at 600 °C, while the conventional specimen achieves 9 MPa at 700 °C, indicating a strength retention that is 5–10% greater than that of CFRP as shown in Fig. 8. Once temperatures exceed 600 °C, CFRP experiences a significant drop, losing nearly 50% of its stress capacity. In contrast, AFRP and GFRP retain approximately 80% of their maximum stress levels even at 800 °C. The traditional specimen holds onto more residual stress, demonstrating a strength that is 20–30% higher than that of FRP-based materials when temperatures exceed 800 °C. Interestingly, CFRP shows an erratic variation past 900 °C, suggesting instability caused by significant thermal decomposition. This numerical study reveals that FRP reinforcement enhances strength at intermediate temperatures, whereas conventional RC demonstrates superior performance at very high temperatures owing to its natural thermal resistance.

The numerical analysis indicates that thermo-mechanical loading significantly influenced the identification of the optimal strengthening material and the critical thermal exposure range for RC beam-column joints. In the analysis of the 27 FRP-rehabilitated models, CFRP stood out as the most effective material, demonstrating reduced deflection and principal stress when compared to GFRP and AFRP. The performance of GFRP was moderate, whereas AFRP proved to be the least effective, exhibiting notably higher deflection and stress, which led to its disqualification from further consideration. The analysis revealed that 400 °C serves as the critical temperature prior to rehabilitation, beyond which structural degradation intensifies significantly. After rehabilitation with CFRP laminates, the critical temperature threshold improved to 500 °C, showing enhanced thermal resistance and load-bearing capacity. The validation of these findings was achieved by comparing them with experimental results. The top-performing material (CFRP) along with the relevant critical temperature zones were chosen and applied during the experimental phase to maintain consistency and to confirm the effectiveness of the suggested strengthening method in elevated temperature scenarios.

Experimental investigation

Specimen details

For the purpose of experimental analysis, this Fig. 9 illustrates the intricate design of a RC specimen, which includes its dimensions, reinforcement details, and CFRP enveloping configuration. The specimen is designed with a T-joint geometry, which is commonly employed to investigate the behavior of RC beam-column joints under loading conditions, particularly when they are rehabilitated using CFRP laminates. The specimen’s specifics can be divided into three categories: (A) Dimensions of the Specimen, (B) Reinforcement Details, and (C) Wrapping Dimensions.

1. Dimension of the specimen.

The RC specimen is composed of a beam-column joint, in which the column extends vertically and the beam extends horizontally from the column. The critical dimensions are as follows: Column height: 1300 mm, Column cross-section: 75 × 75 mm², Beam length: 100 mm from the column face, Beam cross-section: 75 × 75 mm² and the concrete cover is 15 mm in all sides. This geometry enables the simulation of load transfer between structural elements and guarantees an accurate representation of a joint in an RC frame. The design ensures that the actual structural behavior is accurately replicated under thermal and mechanical stresses by maintaining an appropriate aspect ratio.

1. Reinforcement details of the specimen.

The RC specimen’s ductility and strength are guaranteed by the reinforcement configuration under loaded conditions. The reinforcement details are delineated in two sections (A–A and B–B): Column section (A–A): Four bars with a diameter of 6 mm are situated at the extremities. Stirrups with two legs (45 × 45 mm²) that are spaced at 50 mm c/c for transverse confinement. Beam section (B–B): The principal flexural reinforcement consists of four bars with a diameter of 8 mm. Two-legged links (70 mm × 45 mm) that are spaced at 50 mm c/c and provide shear reinforcement. In order to prevent premature failure, the reinforcement detailing adheres to the standard practices for seismic-resistant RC joints, ensuring appropriate anchorage and confinement.

Fig. 9 [Images not available. See PDF.]

Prototype details of RC joint (a) Dimensions of specimen, (b) Reinforcement details of specimen, and (c) Wrapping dimensions for experimental work.

1. Wrapping dimensions (CFRP rehabilitation).

The joint region is marked with black shading to denote the CFRP enveloping configuration. The dimensions for wrapping are as follows: The height of the CFRP sheathing is 275 mm from the joint region, and it covers a portion of the column. The CFRP wrapping is 200 mm wide at the beam-column junction. The CFRP sheathing is applied to the joint to improve the joint’s performance under thermo-mechanical loading by increasing shear strength and ductility. CFRP laminates are essential for the rehabilitation of RC structures that have been damaged by fire or exposed to high temperatures. This method restores the structure’s rigidity and load-carrying capacity, thereby mitigating the effects of deterioration.

1. Concrete ingredients and mix proportions.

OPC of Grade 53 was used in the construction of the frame, which was made out of reinforced concrete. For the purpose of fine aggregate, river sand that had a specific gravity of 2.65 and conformed to Zone II was utilized. A coarse aggregate with a specific gravity of 2.82 and a nominal size of 10 mm was used in the construction project. It was mixed with water that may be consumed. Based on the guidelines provided by IS 10262 (2009)³², the proportion of cement to sand to coarse aggregate in the concrete mix was 1:1.57:1.42 and the ratio of water to cement were 0.4.

The purpose of this specimen is to examine the structural behavior of RC beam-column joints under mechanical loading conditions and elevated temperatures, particularly before and after CFRP rehabilitation. The reinforcement detailing is intended to ensure strength and ductility, while the CFRP wrapping is intended to strengthen the joint. The study will evaluate the load-deflection behavior, crack propagation, and strength recovery post-rehabilitation through experimental and numerical analysis.

Rehabilitation method

The procedure for CFRP wrapping to enhance RC beam-column joints consists of four essential stages: preparing the surface, applying the resin, placing the CFRP, drying, and conducting tests. Combining and Implementing The surface of the damaged or exposed concrete joint is initially cleaned to eliminate loose particles, dust, and grease. The two-part epoxy resin is combined in the appropriate ratio and then applied to the surface with a brush. This resin functions as a bonding agent, guaranteeing effective adhesion between the CFRP laminate and the concrete surface shown in Fig. 10. Applying CFRP: The CFRP sheets are meticulously cut to the specified dimensions and subsequently positioned onto the resin-coated surface. Extra resin is applied over the CFRP to thoroughly saturate the fibers, guaranteeing proper adhesion and removing any air voids.

Fig. 10 [Images not available. See PDF.]

Stages of CFRP wrapping for rehabilitaion.

The wrapped joint remains undisturbed for a designated period, enabling the resin to solidify and attain maximum bonding strength. In certain instances, external confinement, like the use of wires, is employed to sustain pressure throughout the drying process for 48 h. Evaluation: The rehabilitated RC joint undergoes mechanical loading in a controlled setup to assess its load capacity, deflection, and failure modes under structural conditions. This approach greatly improves shear strength, ductility, and thermal resistance in reinforced concrete joints exposed to elevated temperatures and loads.

Test setup

This image presents a bi-axial testing setup featuring a self-straining 3D loading frame, specifically engineered for the evaluation of RC columns and beam-column joints under regulated loading conditions as seen in Fig. 11. This system, with a capacity of 2000 kN, is designed to apply lateral and axial loads, effectively simulating the structural forces encountered in real-world scenarios. The RC column or joint is positioned centrally on the testing platform and is firmly secured to ensure it remains in place during the testing process. Instrumentation and data acquisition involve the use of sensors, including LVDTs (Linear Variable Differential Transformers), and load cells, which are employed to capture load-deflection behavior, strain variations, and failure modes. Load Application: A hydraulic actuator exerts lateral force to replicate the effects of seismic or wind conditions. The axial load simulates the vertical forces exerted by the weight of the building. The assessment of structural response is conducted through monotonic loading. In the observation of failure, the load is systematically increased until the formation of cracks, yielding of reinforcement, or ultimate failure becomes evident. This testing method plays a vital role in assessing the load-bearing capacity, ductility, and effectiveness of rehabilitation for reinforced concrete structures. Figure 12 shows schematic diagram of the experimental test setup the arrangement of load cell, LVDT, dial gauge and heating oven.

Fig. 11 [Images not available. See PDF.]

Loading setup for experimental work.

Fig. 12 [Images not available. See PDF.]

Schematic diagram of the experimental test setup.

Loading condition

The image presents two different loading scenarios imposed on a RC beam-column joint: mechanical loading and thermo-mechanical loading. The analysis of structural response in RC joints is conducted under both normal and elevated temperature conditions. The mechanical loading condition features a hydraulic jack applying force to the beam, producing lateral loads akin to those experienced during seismic or wind occurrences. A load cell is positioned between the jack and the beam to ensure precise measurement of the applied force. The column is secured at both ends, establishing a fixed boundary condition that reflects the actual constraints found in reinforced concrete structures. The lower section of the column was tightly fixed to eliminate any possibility of translation or rotation. The upper portion of the column was intentionally left free to allow for vertical displacement and to facilitate axial loading through a hydraulic jack. The free end of the beam was permitted to move vertically, facilitating deflection in response to the applied load. The loading conditions were illustrated in Fig. 13a,b. This also offers foundational information for comprehending the performance of the RC joint prior to any rehabilitation or reinforcement methods, including CFRP wrapping.

Fig. 13 [Images not available. See PDF.]

loading categories of RC joint for experimental work (a) Mechanical loading and (b) Thermo mechanical loading.

The thermo-mechanical loading condition incorporates heat exposure as a supplementary factor to the mechanical loading scenario. During the casting process, a thermocouple was attached to the steel reinforcement. The temperature was monitored with the thermocouple and the temperature indicator during the testing process. The thermocouple was placed in two different places: one end was attached to the reinforcement, while the other end was attached to the temperature indicator. A heating device (Muffle furnace) is strategically positioned along the lower section of the column to replicate elevated temperatures, mirroring scenarios like fire incidents or high-temperature environments encountered in practical applications.

Result and discussion

This study involved the casting of two RC specimens Utilizing M30 grade concrete and Fe415 grade steel, aimed at examining their structural behavior when subjected to mechanical and thermo-mechanical loading. The specimens underwent a curing period of 28 days to guarantee adequate strength development prior to testing. One specimen was identified as the standard case, where solely mechanical loading was implemented, whereas the second specimen experienced both thermal exposure and mechanical loading.

In Case 1, the conventional specimen was first subjected to a load of 70% of its ultimate capacity, resulting in the emergence of minor cracks in the joint region and at the top of the column. The rehabilitation of the cracks was accomplished through the application of CFRP laminates. Following rehabilitation, the specimen underwent re-testing under full loading conditions until failure, facilitating a comparative evaluation of strength, deflection, and failure patterns. The specimen underwent exposure to a specified critical temperature of 400 °C for duration of 240 min prior to rehabilitation, effectively simulating damage caused by fire. Following the initial loading, CFRP wrapping was implemented to regain its structural integrity. The rehabilitated specimen underwent a re-evaluation under mechanical loading at an elevated temperature of 500 °C for duration of 240 min. Data was collected on the failure pattern, load-carrying capacity, deflection response, and variations in core temperature for both scenarios. The findings shed light on how effective CFRP rehabilitation is in regaining cracks and core temperature when subjected to extreme loading conditions, emphasizing the influence of high temperatures on the performance of RC joints.

General observations

In the early stages of loading, the specimen having some small cracks in the joint area and at the top of the column, suggesting the beginning of structural issues. Furthermore, a slit crack emerged in the beam, indicating a concentration of localized stress. Temperature readings were taken at the core region with thermocouples to observe the thermal effects. In order to ensure the structural integrity, all minor cracks are repaired with cement paste. After this, the rehabilitated specimen underwent mechanical and thermo-mechanical loading to evaluate its performance after rehabilitation. One specimen was subjected to mechanical loading in the traditional manner, whereas another was assessed considering both thermal and mechanical influences. The load was gradually increased until failure occurred, while temperature readings were taken in the core area for analysis purposes. Cracks before rehabilitation shows in Fig. 14a and de-bonding the surface and wrapping after rehabilitation as shown in Fig. 14b.

Fig. 14 [Images not available. See PDF.]

Failure pattern (a) before rehabilitation and (b) after rehabilitation.

Load displacement characteristics before and after rehabilitation

Examining load vs. deflection provides insights into how a structure responds to applied loads, which is crucial for evaluating both its serviceability and ultimate load capacity. The load-deflection curve offers important insights into the various stages of a material or structure, including its elastic, plastic, and failure phases. This study explores the load versus deflection behavior, focusing on the changes observed before and after the rehabilitation of a reinforced concrete beam-column joint, which has been treated with CFRP laminates and subjected to thermo-mechanical loading. The numerical analysis aimed to establish the overall load-carrying capacity of the RC joint. Prior to rehabilitation, merely 70% of the overall load was utilized. Following rehabilitation, the joint underwent a full application of the calculated load until it reached failure. This method allows for a thoughtful examination of how effective rehabilitation techniques are in both restoring and improving structural performance.

Fig. 15 [Images not available. See PDF.]

Load-displacement responses of FRP-rehabilitated RC beam-column joints before and after rehabilitation.

The load vs. deflection graph seen in Fig. 15 illustrates two distinct scenarios: the pre-rehabilitation phase represented by the black curve and the post-rehabilitation phase shown by the blue curve. During the pre-rehabilitation phase, the RC joint showed a restricted ability to bear loads and experienced greater load vs. deflection, as evidenced by the early appearance of deflection with rising loads. At the beginning, under lower loads (0–6 kN), the deflection response showed a fairly linear pattern, suggesting elastic behavior. As the applied load surpassed 6 kN, the rate of deflection began to increase more rapidly, attributed to the development of cracks, which resulted in considerable structural deterioration. The maximum load prior to rehabilitation reached around 8 kN, at which point significant deflections were observed, signaling failure.

The load-carrying capacity showed a notable enhancement following rehabilitation with CFRP laminates, as illustrated by the blue curve. The elastic region stretched past 6 kN, suggesting an improvement in structural stiffness. The final load reached about 11 kN, accompanied by a deflection of 51.6 mm prior to failure. This indicates that CFRP wrapping successfully enhanced the joint’s load resistance, postponing the onset and spread of cracks. Additionally, the behavior observed after the peak suggests an increased ability to absorb energy, as the structure is capable of enduring larger deflections prior to a catastrophic failure.

Upon examining both curves, it becomes clear that the use of CFRP in rehabilitation has significantly improved both load capacity and ductility. Prior to rehabilitation, the structure encountered early cracks and failure when subjected to 70% of the total load. After rehabilitation, the full load-carrying capacity was successfully restored, and ultimate failure was observed at 100% of the load determined through numerical analysis. The results of this experimental validation indicate that CFRP wrapping greatly improves structural integrity, establishing it as a dependable method for rehabilitating RC joints subjected to thermo-mechanical loading.

Nodal temperature

Particularly when temperatures are elevated, the behavior of RC connections is significantly influenced by nodal temperature. The mechanical properties of steel reinforcement and concrete deteriorate when exposed to high temperatures, resulting in increased deformation, reduced load-bearing capacity, and the potential for structural failure. The thermal expansion, tension redistribution, and overall performance of the structure are influenced by the temperature distribution across various regions of the RC joint. Monitoring nodal temperatures is beneficial for evaluating the efficacy of rehabilitation techniques, such as CFRP laminates, predicting failure mechanisms, and assessing thermal gradients. Three thermocouples were strategically positioned in order to measure nodal temperatures at critical locations in this study:

1. Joint centre—This region is the most susceptible to damage as a result of the convergence of load paths and heat accumulation, which impacts both the rigidity and strength of the bond.

2. Beam region—This section is subject to temperature-induced expansion and bending stresses, which affect the flexural strength and fracture propagation.

3. Column region—The axial load-bearing component, where temperature effects can result in the fracturing of reinforcement and a loss of compressive strength.

The temperature at the joint’s Centre was the highest prior to rehabilitation, as a result of limited thermal dissipation and heat concentration. This resulted in the reduction of rigidity, the deterioration of the bond between concrete and reinforcement, and the formation of micro cracks. The beam and column regions also experienced a substantial temperature increase, which resulted in thermal expansion and differential movement. This, in turn, accelerated the damage caused by mechanical loading. The core joint temperature exceeded the critical threshold, resulting in the spalling of concrete and the deterioration of the load transfer mechanism, as indicated by the thermocouple readings. Furthermore, the likelihood of failure was elevated by the thermal-induced tension that resulted from the high-temperature gradient between the beam and column regions.

The application of CFRP laminates and resin improved thermal resistance and confined the core region after rehabilitation, thereby minimizing the rate of heat penetration. The bond strength and structural integrity were preserved as a result of the substantially reduced temperature of the Centre joint. The beam and column regions also demonstrated improved thermal stability, which resulted in a decrease in differential thermal expansion and the associated stress concentrations. Thermocouple data verified that the CFRP-wrapped RC joint experienced more uniform heat dissipation and lower peak temperatures, which delayed fracture initiation and improved post-fire mechanical performance. This further substantiated the efficacy of CFRP laminates in preventing temperature-induced degradation, thereby regaining the strength and ductility of RC joints that are subjected to thermo-mechanical loading.

Fig. 16 [Images not available. See PDF.]

Comparison graph of Core temperature for all three regions before wrapping.

The temperature variation at three distinct locations (beam, column, and joint) in relation to deflection is depicted in the two graphs, both before and after rehabilitation as shown in Figs. 16 and 17. Particularly when temperatures are elevated, the temperature measurements are indispensable for comprehending the thermal response of the RC joint.

In the beam region, the greatest temperature recorded was approximately 80 °C at a deflection of 13.75 mm prior to rehabilitation. The temperature remained essentially constant at 75–78 °C following rehabilitation; however, the deflection decreased to 8.8 mm. Deflection reductions by approximately 36% while maintaining a comparable temperature profile suggests that rehabilitation has resulted in increased their value in temperature.

Fig. 17 [Images not available. See PDF.]

Comparison graph of Core temperature for all three regions after wrapping.

Column Region: Prior to rehabilitation, the column’s greatest temperature was 44 °C, which decreased to approximately 26 °C as deflection increased. The column’s maximal temperature was 30 °C following rehabilitation, and it decreased to 20 °C within a comparable deflection range. The temperature in the column region decreased by approximately 31.8%, indicating improved thermal resistance following rehabilitation.

In the Joint Region: Prior to rehabilitation, the maximal joint temperature was approximately 400 °C with a deflection of 13.85 mm. This temperature subsequently decreased as failure occurred. The maximal joint temperature was 350 °C at 8.8 mm deflection following rehabilitation. The efficacy of rehabilitation in minimizing thermal degradation was demonstrated by the approximately 12.5% reduction in temperature at the joint.

The thermal and mechanical performance of the RC joint was substantially enhanced by the rehabilitation process, as evidenced by the comparison of these two graphs. The decrease in deflection suggests an increase in structural rigidity, while the decreased temperatures, particularly at the column and joint, suggest enhanced thermal insulation. The column region exhibited the most substantial improvement, with temperature reductions exceeding 30%. The beam and joint regions also experienced moderate improvements. These results underscore the critical role of rehabilitation in enhancing the structural integrity and fire resistance of RC joints in extreme conditions.

Predicting thermo-mechanical behavior of CFRP-retrofitted RC joints using artificial neural networks (ANNs)

ANNs have become a significant computational resource for forecasting structural behavior in response to different loading and environmental scenarios. In structural engineering, artificial neural network models play a significant role in simulating intricate relationships among various input parameters³³. These include applied load, temperature, material properties, and the resulting structural responses, such as deflection, stress distribution, and failure modes. In contrast to traditional numerical methods such as the FEM, ANNs offer swift predictions while minimizing the need for significant computational resources³⁴.

Deflection serves as an essential factor in evaluating the structural integrity of reinforced concrete joints when exposed to mechanical and thermo-mechanical loading. Research indicates that when subjected to pure mechanical loading, deflection values for conventional RC joints stay within acceptable limits. When exposed to higher temperatures, such as 400 °C and 500 °C, notable alterations take place in the deflection behavior, driven by thermal expansion and a decline in material deflection. Artificial Neural Networks that are trained on experimental data have the capability to predict deflection variations with precision³⁵. They achieve this by understanding the complex, non-linear relationships that exist between applied temperature, load levels, and structural property. An ANN model, when trained on data from CFRP, AFRP, and GFRP-rehabilitated specimens, demonstrates the capability to accurately predict deflection trends across various loading scenarios, thereby assisting in performance evaluation.

In the utilization of artificial neural networks for forecasting structural reactions, including deflection and nodal temperature, many limits must be recognized despite attaining elevated correlation coefficients (R-values). One of the primary concerns in the restricted generalizability of the ANN model outside the scope of the trained data. The model may exhibit accuracy within the confines of the experimental and numerical datasets utilized for training; however, its predictive efficiency may markedly diminish when confronted with novel scenarios featuring parameters beyond the studied domain (e.g., elevated temperatures, varying loading rates, or alternative materials not incorporated in training). Moreover, elevated R-values may not ensure strong generalization, since they may obscure over fitting problems, wherein the model assimilates noise and particular patterns from the training data instead of accurately representing the fundamental physics-based connections. This danger is particularly pronounced in models trained on restricted or narrowly dispersed datasets. The precision and dependability of ANN outputs are significantly influenced by the quality, completeness, and representativeness of the input data. Insufficient sampling of critical influencing factors or the incorporation of biased data might undermine the model’s validity. Consequently, although ANN is an effective instrument for delineating intricate connections, prudence is necessary when interpreting its predictions and findings should be corroborated by experimental or numerical validations.

This study utilized an ANN regression model to explore the structural behavior of CFRP-rehabilitated RC joints when subjected to both thermal and mechanical conditions. The ANN model was developed by utilizing both experimental and numerical data, focusing on the following input parameters:

• Temperature (applied nodal/core temperature), and

• Loading (mechanical force applied on the RC joint).

The parameters predicted by the ANN model included:

• Deflection (the structural displacement when under load), and

• Core (Nodal) Temperature (the temperature response observed at the joint region).

This configuration enabled the ANN to grasp the intricate nonlinear connections between the thermal-mechanical inputs and the resulting structural responses. The ANN architecture engages in a fascinating process where inputs are channeled through layers of interconnected neurons, with internal weights being adjusted to reduce prediction errors effectively.

Regression analysis

The section provides insights into the results of the ANN regression analysis concerning two essential structural parameters obtained from the experimental study of rehabilitated RC joints Utilizing CFRP laminates. This regression was created to forecast structural behavior using experimental data obtained under high-temperature conditions. The two sets of plots presented are: (i) Load versus Deflection and (ii) Nodal Temperature at Joint. Both outputs demonstrate how well ANN can model complex thermo-mechanical responses.

This section presents the ANN regression findings, focusing on how deflection is forecasted based on applied loads across different thermal and structural scenarios. The dataset is divided into segments for training (a), validation (b), testing (c), and overall (d) performance visualizations as shown in Figs. 18 and 19:

• Training: A correlation coefficient R = 1 signifies a flawless linear relationship between the predicted and target outputs, demonstrating that the model has effectively grasped the training data.

• Validation: The model demonstrates impressive predictive accuracy on unseen validation data with a R value of 0.99626, indicating that over fitting is minimal.

• Testing: The R-value of 0.92191 indicates strong generalization ability. While it is somewhat lower than the training and validation results, it nonetheless demonstrates a solid level of agreement.

• Overall Performance: The combined dataset yields R = 0.99036, highlighting the ANN’s remarkable capability to capture the nonlinear behavior of RC joints regarding load and deflection, particularly in post-rehabilitation scenarios Utilizing CFRP laminates.

Fig. 18 [Images not available. See PDF.]

Regression plot for load vs. deflection (a) Training, (b) Validation, (c) Testing, and (d) Overall.

Fig. 19 [Images not available. See PDF.]

Regression plot for nodal temperature (a) Training, (b) Validation, (c) Testing, and (d) Overall.

This collection of regression plots illustrates the ANN model’s ability to forecast nodal temperatures at the beam-column joint, a crucial aspect in scenarios involving fire or elevated temperatures:

• Training: The model demonstrates an R value of 1, signifying flawless assimilation of temperature data from the training set.

• Validation: An R-value of 0.89657 indicates that the model maintains strong performance on validation data, albeit with a slight decrease attributed to variability in experimental thermal profiles.

• Testing: The model demonstrates a strong prediction capability for new, unseen data with an R value of 0.97993, confirming its reliability in thermal prediction tasks.

• Overall: The regression coefficient R = 0.98537 indicates a strong alignment between the predicted and actual nodal temperatures throughout the dataset.

The ANN regression analysis effectively captures the load-deflection behavior and nodal temperature distribution for CFRP-rehabilitated RC joints when subjected to elevated temperatures. The consistently high R-values in all datasets indicate that the ANN delivers precise predictions, establishing it as an effective instrument for parametric optimization and forecasting material performance³⁶. The results align with experimental findings and contribute to the decision-making process regarding the selection of CFRP as a suitable retrofitting material, especially in terms of thermal resilience and structural safety.

Performance analysis

The plots provide insights into how effectively the network learns and its ability to generalize throughout the training process. The left plot illustrates the performance of the ANN in relation to load and deflection, whereas the right plot depicts its performance for predicting nodal temperature at the joint under elevated temperature conditions. The vertical axis in both graphs indicates the MSE on a logarithmic scale, whereas the horizontal axis shows the number of training epochs.

This plot illustrates how effectively the ANN can forecast the deflection behavior in response to applied loads for CFRP-strengthened RC joints. The training process unfolded over four epochs, culminating in the best validation performance of 0.01435 MSE at epoch 3, as highlighted by the green circle.

• The training results demonstrate a notable and steady decrease in MSE, suggesting successful learning while avoiding over fitting.

• The validation closely tracks the training error up to epoch 3, after which the validation error levels off, indicating that this is the ideal stopping point to prevent over fitting.

• The stability of testing continues to support the idea of model generalization.

• The final MSE value is remarkably low (~ 10^-15 on training), indicating a highly accurate model for predicting deflection based on applied load.

This plot shown in Figs. 20 and 21 illustrates how well the ANN predicts the temperature distribution at the beam-column joint, which is essential for assessing the thermal degradation resistance in CFRP-retrofitted structures.

• The optimal validation performance recorded is 0.84097 MSE at epoch 3, marked by the green circle.

• The training curve shows a significant decline, much like the earlier plot, suggesting that learning is occurring effectively. The final training error achieves an impressively small magnitude, approximately ~ 10^-15.

• The validation and testing errors exhibit slight variations but quickly stabilize, indicating that the model is not substantially over fitting and retains its predictive capability.

• The increased validation error in comparison to the deflection model could be linked to greater variability in thermal responses, stemming from experimental uncertainties like material heterogeneity, heat transfer dynamics, and sensor placement.

Fig. 20 [Images not available. See PDF.]

Performance plot for load vs. deflection.

Fig. 21 [Images not available. See PDF.]

Performance plot for nodal temperature.

The performance plots of these ANN models showcase their reliability and efficiency in learning from experimental data related to rehabilitate RC joints. The low MSE values observed in both the load-deflection and nodal temperature models indicate that the ANN has successfully captured the underlying nonlinear relationships. Additionally, implementing early stopping at epoch 3, guided by validation performance, effectively mitigates over fitting, thereby improving the model’s robustness and generalizability. The findings support the use of ANN as an effective predictive tool for assessing structural rehabilitation strategies with CFRP when subjected to mechanical and thermal stresses.

In the ANN component of this study, while the regression plots and performance curves demonstrate high correlation coefficients (R-values) for deflection and nodal temperature predictions, several limitations must be acknowledged. One primary limitation is the risk of generalizing results beyond the studied range of input parameters. The ANN model is trained specifically on data confined to the experimental and numerical scenarios investigated in this research such as particular load levels, temperature ranges, and material configurations. Therefore, applying this trained model to predict behavior for significantly different conditions may result in unreliable or inaccurate outcomes.

Another critical concern is the potential for over fitting, especially in models that achieve very high R-values (close to 1) during training and validation. Over fitting occurs when the ANN learns the noise or specific patterns of the training data rather than the underlying general relationships. This can lead to poor performance on unseen or new datasets, thereby limiting the robustness and applicability of the model in broader structural engineering problems.

Furthermore, the accuracy of ANN predictions is highly dependent on the quality and representativeness of the input data. If the input dataset lacks sufficient variability, contains measurement noise, or does not cover all critical conditions, the trained model may fail to capture essential structural responses. For reliable prediction, it is crucial that the data used for training, validation, and testing comprehensively represents the range of physical behaviors expected in practical rehabilitation scenarios. These limitations highlight the importance of careful data selection, validation procedures, and cautious interpretation of ANN-based results.

Discussion

This study offers a comprehensive examination of the rehabilitation of RC joints through the use of three varieties of FRP laminates: Carbon FRP, Glass FRP, and Aramid FRP. The assessment utilizes a tri-model strategy that includes finite element numerical analysis, experimental validation, and ANN regression modelling, concentrating on temperature-induced degradation, mechanical performance, and predictive abilities.

Effect of investigated parameters based on numerical

The models took into account transient thermal loading, which was then followed by mechanical loading, effectively simulating conditions after a fire or at elevated temperatures. The analysis of temperature versus deflection indicated that all materials exhibited a decline in performance as the temperature rose³⁷. Nonetheless, the joints that were rehabilitated with CFRP consistently showcased better performance compared with previous studies^8,11. At 400 °C prior to rehabilitation, un-strengthened specimens showed a significant rise in deflection. Following rehabilitation at 500 °C, CFRP was able to decrease this deflection by almost 35% in comparison to AFRP, and by 22% when compared to GFRP. The analysis of principal stress distribution across the joint revealed that AFRP exhibited an earlier concentration of tensile stresses at the beam-column interface, which resulted in diagonal shear failure. GFRP demonstrated a fair level of performance; however, it was susceptible to stress redistribution and early debonding issues.

On the other hand, CFRP laminates demonstrated an effective redistribution of stresses and provided confinement to the joint core, leading to improved shear resistance. The reduced magnitude and improved distribution of principal stresses clearly demonstrated their role in limiting crack propagation and delaying the onset of crushing. The critical region was pinpointed at the joint core and the beam-column intersection, where thermal and mechanical stresses converge. At higher temperatures, the deterioration of concrete in this area became noticeable, especially for joints rehabilitated with AFRP, which showed localized crushing and deboning failures at lower load levels when compared to CFRP and GFRP.

Effect of investigated parameters based on experimental

The experimental investigation confirmed the numerical findings through the testing of RC joint specimens, including both control and rehabilitated versions, under combined axial and lateral loads to simulate seismic conditions.

The load-deflection behavior clearly showcased how CFRP significantly improves both load-bearing capacity and ductility. The peak load supported by CFRP-rehabilitated joints showed an increase of up to 40% when compared to the unwrapped control specimens, and also demonstrated enhancements of 25% and 15% over AFRP and GFRP, respectively. AFRP showed promise at first because of its strong tensile properties, but it encountered early failures due to inadequate bond strength and its inability to handle mechanical loading at higher temperatures.

Table 5. Summarizing key results before and after rehabilitation.

Parameter	Conventional specimen in high temperature	CFRP-rehabilitated specimen in high temperature
Ultimate load (kN)	85	128
Deflection at peak load (mm)	15.2	8.5
Ductility ratio	2.1	4.7
Crack width at failure (mm)	2.8	1.2
Crack pattern observation	Wide diagonal shear cracks at joint core	Minor diagonal cracks with CFRP confinement and de-bonding the wrapping

The patterns of crack propagation seen in CFRP-strengthened specimens revealed smaller, more evenly distributed cracks, whereas AFRP resulted in larger, localized cracks that ultimately led to brittle failure. The performance of GFRP was found to be intermediate, providing sufficient confinement; however, its resistance to mechanical loads diminished after thermal exposure. Nodal temperature measurements taken during loading indicated that CFRP laminates exhibited enhanced thermal resistance, resulting in reduced internal core temperatures in rehabilitated joints as detailed in Table 5. For example, when exposed externally to 500 °C, the CFRP-laminated specimens measured internal nodal temperatures of approximately 340 °C, whereas the AFRP and GFRP exhibited values surpassing 380 °C and 360 °C, respectively. This highlights CFRP’s reduced thermal conductivity and superior thermal insulation properties, which enhance its suitability for rehabilitation applications following fire incidents or exposure to high temperatures.

Effect of ANN regression parameter

The ANN regression model uses a feed-forward back propagation neural network to relate input variables like temperature and mechanical loads to output parameters like deflection and nodal temperature. For training, validation, and testing, we used experimental and numerical datasets. The ANN model predicted load-deflection with an MSE of 0.01435 during validation³⁸. The regression coefficient (R) values frequently surpassed 0.99, showing that the model well reflected the nonlinear load-deformation relationship across material kinds and temperatures. CFRP had the highest prediction accuracy, followed by GFRP, whereas AFRP had worse predictive consistency because to its unpredictable performance at higher temperatures.

The ANN regression for nodal temperature prediction had a significantly increased MSE of 0.84097, indicating difficulties in effectively forecasting thermal behavior due to material conductivity, bond deterioration, and heat transport characteristics. A significant connection with FEM and experimental results supported the trend projection. ANN models showed CFRP had the most accurate prediction behavior, which matches its stable performance in physical and numerical settings. Due of its heat sensitivity and irregular deflection data, AFRP has higher prediction errors, making it unsuitable for high-temperature rehabilitation.

Comparison of experimental and numerical results

The numerical analysis using finite element method successfully replicated the experimental behavior with minimal deviations in both load and thermal parameters. The deflection trend showed slightly lower values in simulations, likely due to idealized boundary conditions and perfect material assumptions in the model. Nodal temperatures observed at the joint core confirmed that CFRP wrapping enhanced thermal insulation, delaying internal temperature rise during elevated exposure. The difference in results remained within 1–6%, indicating high reliability of the simulation model and proper calibration with experimental data. Table 6 shows the comparison results of numerical and experimental investigation.

Table 6. Comparison on numerical and experimental work before and after Rehabilitation.

Parameter	Condition	Numerical result	Experimental Result	Difference (%)	Remarks
Ultimate load (kN)	Before rehabilitation	72.5	70.3	3.1%	Good correlation; validates numerical calibration
	After rehabilitation (CFRP)	93.2	90.1	3.4%	Load capacity improved by ~ 28.6% (Exp.), ~ 28.5% (Num.)
Deflection at peak load (mm)	Before rehabilitation	10.8	10.2	5.8%	Acceptable variance; due to local cracking and support flexibility
	After rehabilitation (CFRP)	8.3	8.6	3.5%	Deflection reduced due to confinement and stiffness enhancement
Nodal temperature (°C)	Before rehabilitation	400 °C	395 °C	1.25%	Close match; confirms temperature profile modelling accuracy
	After rehabilitation (CFRP)	500 °C	495 °C	1.0%	FRP wrapping improved thermal resistance and delayed core temperature rise

A close comparison was performed between the failure mechanisms observed in the FE simulations and those documented during the experimental testing of both conventional and CFRP-strengthened RC beam-column joints under elevated temperatures.

Conventional RC joints:

• Experimental: The un-strengthened specimens exhibited significant cracking at the beam-column interface, progressing into wider diagonal shear cracks. At temperatures beyond 350 °C, rapid spalling of concrete and bar exposure was observed, leading to brittle shear failure.

• Numerical: The FE model simulated initial flexural cracking, followed by diagonal tensile cracks along the joint region. Beyond 400 °C, concrete damage localization and stiffness degradation were evident in the joint core, with reinforcement yielding, matching experimental patterns.

CFRP-strengthened joints:

• Experimental: The CFRP-laminated joints demonstrated delayed cracking, reduced crack widths, and greater confinement. At temperatures approaching 650 °C, delamination of CFRP wraps and eventual crushing of the concrete in the joint core occurred. The failure was more ductile compared to the conventional joints.

• Numerical: The simulations showed improved stress redistribution in CFRP models, with delayed damage initiation. The cohesive zone model captured debonding initiation near 600–650 °C, followed by localized concrete crushing in the joint core, closely mirroring the experimental failure mode.

The numerical failure progression exhibited strong agreement with experimental observations, particularly in crack formation zones, stiffness degradation trends, FRP debonding, and ultimate failure regions. This validates the reliability of the FE model in predicting both thermal and mechanical degradation phenomena in RC joints under extreme conditions.

Stiffness degradation

Stiffness degradation is a gradual reduction in a structural element’s capacity to withstand deformation when subjected to external loads. In reinforced concrete structures, particularly when exposed to high temperatures, degradation can take place as a result of micro cracking, softening of materials, deterioration of bonds, or thermal damage. The reduction in stiffness affects the ability to carry loads and maintain serviceability, highlighting its importance in evaluating structural health and performance. Grasping the concept of stiffness degradation is crucial in rehabilitation studies, as it indicates the ability of a reinforced system to maintain its mechanical integrity over time or when faced with challenging conditions like fire or seismic loading.

The Fig. 22 illustrates the differences in stiffness degradation between CFRP-laminated RC joints and traditional RC joints as temperatures rise to 1000 °C. At first, both joints show comparable stiffness up to about 100 °C, but as the temperature rises, significant differences start to appear. The traditional RC joint exhibits a notable stiffness peak at approximately 300 °C, achieving around 32 kN/mm. However, it experiences a significant decline, falling by nearly 87% to about 4 kN/mm as it approaches 400 °C. On the other hand, the CFRP-laminated joint shows a more gradual rise in stiffness, reaching its maximum near 500 °C at approximately 29 kN/mm before slowly declining. At 500 °C, the CFRP joint maintains approximately 90% of its maximum stiffness, whereas the conventional joint holds onto less than 15%. At elevated temperatures ranging from 800 to 1000 °C, the joint rehabilitated with CFRP exhibits a stiffness that is roughly 25–40% greater than that of the conventional joint. This clearly shows how CFRP-laminated joints succeed in thermal resistance and maintain stiffness, making them a more dependable choice in high-temperature situations.

Fig. 22 [Images not available. See PDF.]

Stiffness comparison of before and after rehabilitation.

Up to 300 °C, the conventional RC joint exhibits higher stiffness than the CFRP-strengthened counterpart. This behavior can be attributed to the thermal inertia of concrete and the absence of polymeric degradation in conventional joints during this initial heating range. Meanwhile, the CFRP system begins to experience matrix softening and minor degradation of the adhesive used for bonding, reducing its early-stage stiffness contribution despite its superior tensile capacity. However, at around 350 °C, the conventional joint undergoes a sudden drop in stiffness, which is primarily due to the dehydration of cement paste and thermal degradation of the steel reinforcement-concrete bond, leading to a weakened internal structure. After 390 °C, a slight increase in stiffness is observed in the conventional joint. This rise is typically attributed to thermal expansion of the concrete and steel, which can temporarily improve confinement and increase apparent stiffness. Nonetheless, this effect is short-lived and does not represent structural recovery. In contrast, the CFRP-strengthened joint maintains relatively higher stiffness beyond 400 °C, showcasing the benefits of external reinforcement. However, as the temperature reaches approximately 650 °C, the CFRP begins to lose its structural effectiveness due to resin degradation and fiber deterioration, leading to a progressive decline in stiffness. This temperature marks the threshold of thermal failure for CFRP-strengthened systems, highlighting the importance of thermal compatibility and insulation in fire-prone environments.

Energy dissipation

Energy dissipation serves as an essential factor that indicates how well a structure can absorb and release energy during dynamic occurrences like seismic loading. In RC joints, effective energy dissipation plays a crucial role in minimizing the risk of sudden failure. It enables the structure to experience inelastic deformations while still preserving stability. Greater energy dissipation suggests enhanced ductility and superior performance under thermo mechanical loading conditions. In this study, the non-rehabilitated joint showed an energy dissipation value of 768.43 kN.mm, while the CFRP-rehabilitated joint reached a notably higher value of 1368.93 kN.mm. This indicates a notable enhancement of around 78% in energy dissipation resulting from CFRP rehabilitation. This significant rise underscores how CFRP laminates effectively improve the ductility and seismic resilience of RC joints.

Conclusion

This research thoroughly investigated the rehabilitation performance of reinforced concrete joints exposed to high temperatures, utilizing various fiber-reinforced polymer laminates, including CFRP, GFRP, and AFRP. This research undertook a comprehensive approach, Utilizing experimental testing, advanced finite element modelling, and ANN regression analysis to assess the mechanical and thermal behavior of rehabilitated RC joints when subjected to severe loading conditions. This research examined essential performance metrics such as load-deflection behavior, principal stress distribution, nodal temperature response, energy dissipation, stiffness and the predictive accuracy of ANN models. Focusing from the combined insights of numerical simulations, laboratory experiments, and machine learning results, the subsequent key conclusions are presented to evaluate the relative effectiveness of the chosen FRP materials.

• Of the three rehabilitation materials CFRP, GFRP, and AFRP. CFRP demonstrated superior performance.

• The load carrying capacity of CFRP-rehabilitated joints demonstrated a significant enhancement, with peak loads rising by around 40% when compared to the unwrapped control specimens. The results indicate that GFRP exhibited an increase of approximately 25%, whereas AFRP demonstrated a capacity improvement of merely 15%, suggesting a restricted effectiveness when subjected to elevated temperatures and mechanical loading.

• Deflection control: The use of CFRP led to a 35% reduction in peak deflection when compared to AFRP, and a 22% reduction in comparison to GFRP under post-thermal loading conditions. CFRP exhibited superior when subjected to temperatures of 500 °C.

• Principal stress distribution: CFRP successfully redistributed principal tensile stresses, postponing the onset and spread of cracks. AFRP exhibited initial stress concentration and brittle failure, whereas GFRP demonstrated moderate redistribution but experienced earlier softening.

• Nodal temperature: Under an external thermal exposure of 500 °C, CFRP sustained an internal nodal temperature of around 340 °C. The core temperatures for GFRP and AFRP peaked at 360 °C and 380 °C, respectively, suggesting poorer insulation and increased thermal penetration.

• The performance of the ANN regression model in predicting deflection was impressive, with an R value exceeding 0.99 and a mean squared error of 0.01435, effectively reflecting the nonlinear behavior. The nodal temperature prediction achieved an MSE of 0.84097, demonstrating reasonable accuracy, particularly for CFRP-based models. The ANN model demonstrated a prediction deviation of 5–10% for CFRP highlighting the sensitivity of materials in the context of machine learning generalization.

• Temperature impact: All materials exhibited a decline in performance beyond 400 °C. However, CFRP maintained 80–85% of its value at 500 °C, while GFRP held onto 65–70%. In contrast, AFRP fell below 60%, indicating its reduced reliability for high-temperature rehabilitation applications.

• The CFRP-rehabilitated joints showcased a remarkable 42.8% increase in stiffness retention at higher temperatures, along with a 78% improvement in energy dissipation capacity when compared to their non-rehabilitated counterparts. This highlights their superior resilience and structural efficiency when subjected to thermal and mechanical stresses.

• The combined experimental, numerical, and ANN-based study demonstrates that CFRP stands out as the most effective and reliable material for post-fire rehabilitation of RC joints.

Acknowledgements

This research work is done at the Structural Testing Laboratory, Department of Civil Engineering SRM University, Kattankulathur, Tamil Nadu, and India. The authors wish to thank the SRM management for allowing them to use laboratory and technical support and those who were directly or indirectly involved in this study.

Author contributions

R.S.P.: Original draft writing, methodology, data curation, casting, and testing. N.P.: Conceptualization, supervision, methodology, validation. All authors reviewed, authorized, and agreed to the submission of the manuscript.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Golias, E. et al. Effectiveness of the novel rehabilitation method of seismically damaged Rc joints using c-frp ropes and comparison with widely applied method using c-frp sheets—experimental investigation. Sustain. 13, (2021).

2. Antonopoulos, CP; Triantafillou, TC. Experimental investigation of FRP-Strengthened RC Beam-Column joints. J. Compos. Constr.; 2003; 7, pp. 39-49.

3. Al Zand, AW; Badaruzzaman, WHW; Mutalib, AA; Hilo, SJ. Rehabilitation and strengthening of high-strength rectangular CFST beams using a partial wrapping scheme of CFRP sheets: experimental and numerical study. Thin-Walled Struct.; 2017; 114, pp. 80-91.

4. Nie, X. F., Zhang, S. S., Chen, G. M. & Yu, T. Strengthening of RC beams with rectangular web openings using externally bonded FRP: numerical simulation. Compos. Struct. 248, (2020).

5. Kim, M. S. & Lee, Y. H. Evaluation of seismic performance of RC beam-column system using Slit Brace retrofitting. Structures. 71, (2025).

6. Dehghani, E; Daneshjoo, F; Aghakouchak, AA. Khaji, N. A new bond-slip model for adhesive in CFRP-steel composite systems. Eng. Struct.; 2012; 34, pp. 447-454.

7. Mosallam, A; Allam, K; Salama, M. Analytical and numerical modeling of RC beam-column joints retrofitted with FRP laminates and hybrid composite connectors. Compos. Struct.; 2019; 214, pp. 486-503.

8. Zarandi, S; Maheri, MR. Seismic performance of Rc frames retrofitted by FRP at joints using a flange-bonded scheme. Iran. J. Sci. Technol. Trans. Civ. Eng.; 2015; 39, pp. 103-123.

9. Tangadagi, R. B. & Ravichandran, P. T. Performance evaluation of self-compacting concrete prepared using waste foundry sand on engineering properties and life cycle assessment. Recycling. 9, (2024).

10. Rajkohila, A., Prakash Chandar, S. & Ravichandran, P. T. Influence of natural fiber derived from agricultural waste on durability and micro-morphological analysis of high-strength concrete. Buildings. 13, (2023).

11. Chandar, A. R. & Ravichandran, P. T. S. P. Flexural performance of HSC beams containing natural fibers. J. Build. Pathol. Rehabil9, (2024).

12. Naser, MZ; Hawileh, RA; Rasheed, HA. Performance of RC T-beams externally strengthened with CFRP laminates under elevated temperatures. J. Struct. Fire Eng.; 2014; 5, pp. 1-24.

13. Abu Tahnat, YB; Dwaikat, MMS; Samaaneh, MA. Effect of using CFRP wraps on the strength and ductility behaviors of exterior reinforced concrete joint. Compos. Struct.; 2018; 201, pp. 721-739.

14. Allam, K; Mosallam, AS; Salama, MA. Experimental evaluation of seismic performance of interior RC beam-column joints strengthened with FRP composites. Eng. Struct.; 2019; 196, 109308.

15. Surya Prakash, R. P. R. & Parthasarathi, P. N. Transient state analysis of rehabilitated RC beams using finite element modelling and prediction using an artificial neural network. Eng. Res. Express6, (2024).

16. Al-Rousan, RZ; Alkhawaldeh, A. Numerical simulation of the influence of bond strength degradation on the behavior of reinforced concrete beam-column joints externally strengthened with FRP sheets. Case Stud. Constr. Mater.; 2021; 15, e00567.

17. Niroomandi, A; Maheri, A; Maheri, MR; Mahini, SS. Seismic performance of ordinary RC frames retrofitted at joints by FRP sheets. Eng. Struct.; 2010; 32, pp. 2326-2336.

18. Calderón, PA; Adam, JM; Ivorra, S; Pallarés, FJ; Giménez, E. Design strength of axially loaded RC columns strengthened by steel caging. Mater. Des.; 2009; 30, pp. 4069-4080.

19. Berradia, M et al. Prediction of ultimate strain and strength of CFRP-wrapped normal and high-strength concrete compressive members using ANN approach. Mech. Adv. Mater. Struct.; 2024; 31, pp. 5737-5759.1:CAS:528:DC%2BB3sXht1Kis77N

20. Elhag, AB; Raza, A; Kahla, N; Ben,; Arshad, M. Reliability analysis of various modeling techniques for the prediction of axial strain of FRP-confined concrete. Multidiscip. Model. Mater. Struct.; 2024; 20, pp. 869-890.

21. Heydari, P; Mostofinejad, D; Mostafaei, H; Ahmadi, H. Strengthening of deep RC coupling beams with FRP composites: A numerical study. Structures; 2023; 51, pp. 435-454.

22. Surya Prakash, R; Parthasarathi, N. Numerical analysis of FRP retrofitting in RC beam-column exterior joints at high temperatures and predictive modeling using artificial neural networks. J. Struct. Fire Eng.; 2024; [DOI: https://dx.doi.org/10.1108/JSFE-05-2024-0012]

23. ISO. ISO 834 – 12: Fire resistance tests —Elements of building construction. Int. Stand. Organ. (2012).

24. EN 1992-1-2. EN 1992-1-2: Eurocode 2: Design of Concrete Structures. Part 1.2 Gen. rules-structural fire Des. Brussels (2004).

25. EN 1993-1-2. EN 1993-1-2: Eurocode 3: Design of steel structures. Part 1.2 Gen. rules-structural fire Des. Brussels (2005).

26. Ashrafi, H; Bazli, M; Najafabadi, EP; Vatani Oskouei, A. The effect of mechanical and thermal properties of FRP bars on their tensile performance under elevated temperatures. Constr. Build. Mater.; 2017; 157, pp. 1001-1010.1:CAS:528:DC%2BC2sXhs1ygsrzO

27. Firmo, JP; Correia, JR; Bisby, LA. Fire behaviour of FRP-strengthened reinforced concrete structural elements: A state-of-the-art review. Compos. Part. B Eng.; 2015; 80, pp. 198-216.1:CAS:528:DC%2BC2MXhtVejtLzP

28. Bisby, LA et al. Strengthening fire-damaged concrete by confinement with fibre-reinforced polymer wraps. Eng. Struct.; 2011; 33, pp. 3381-3391.

29. Teng, J; Smith, S. FRP-strengthened RC beams. I: review of debonding strength models. Eng. Struct.; 2002; 24, pp. 385-395.

30. Lu, XZ; Teng, JG; Ye, LP; Jiang, JJ. Bond-slip models for FRP sheets/plates bonded to concrete. Eng. Struct.; 2005; 27, pp. 920-937.

31. Surya Prakash, S. P. R. & Parthasarathi, P. N. Transient state analysis of rehabilitated RC beams using finite element modelling and prediction using an artificial neural network. Eng. Res. Express6, (2024).

32. Bureau of Indian Standards (BIS). Concrete Mix Proportioning Guidelines. IS 10262-2019, New Delhi, India (2019).

33. Vishnupriyan, M; Annadurai, R. A study on the macro-properties of PCB fiber-reinforced concrete from recycled electronic waste and validation of results using RSM and ANN. Asian J. Civ. Eng.; 2023; 24, pp. 1667-1680.

34. Jayakeerti, M; Nakkeeran, G; Aravindh, MD; Krishnaraj, L. Predicting an energy use intensity and cost of residential energy-efficient buildings using various parameters: ANN analysis. Asian J. Civ. Eng.; 2023; 24, pp. 3345-3361.

35. Alagundi, S; Palanisamy, T. ANN model for joint shear strength of RC interior beam-column joint. Lect Notes Civ. Eng.; 2022; 203, pp. 1235-1243.

36. Dey, A; Miyani, G; Sil, A. Application of artificial neural network (ANN) for estimating reliable service life of reinforced concrete (RC) structure bookkeeping factors responsible for deterioration mechanism. Soft Comput.; 2020; 24, pp. 2109-2123.

37. Huang, X et al. Finite element method of bond-based peridynamics and its ABAQUS implementation. Eng. Fract. Mech.; 2019; 206, pp. 408-426.

38. Nakkeeran, G; Krishnaraj, L. Developing lightweight concrete bricks by replacing fine aggregate with vermiculite: a regression analysis prediction approach. Asian J. Civ. Eng.; 2023; 24, pp. 1529-1537.