Load-carrying capacity and mode failure simulation of beam-column joint connection: Application of self-tuning machine learning model
Introduction
Reinforced concrete (RC) beam-column joints are normally analyzed for structural integrity using design methods that are equipped with the current codes of practice like EC2, EC8, and ACI 318 [1]. However, there are significant differences in the amount of stirrup reinforcement prescribed by these codes despite the number of studies that have been carried out in this regard. Such differences have been previously reported by [2], where the amount of EC2 and EC8-specified stirrup reinforcement was found to be more than 4 folds of the ACI 318- specified amount. As such, the design of a beam-column joint which safeguards structural performance and meets the specifications of EC2 and EC8 for earthquake-resistant structures may not be possible in such a case; in fact, the joint will not only be prone to considerable cracking before forming a plastic hinge in the adjacent beam but will similarly be prone to such cracking at the early load stages. This violates the assumption of ‘rigid joint’ which is the basis for the adopted structural analysis methods in practice [3]. Other studies have also reported similar findings [4], [5], [6], [7] and based on these reports, there is a need to develop a reliable method for structural assessment and sustainability.
In this study, the major aim is to develop a data intelligence (DI) model for the structural analysis of RC external beam-column joints. Analytical algorithms with an accurate prediction of the behavior of simply-supported RC beams have reportedly been developed using data DI models [8], [9]. Contrary to most of the recent methods, those that are included in the current code methods included; DI models are not developed based on preconceived theories that describe load transfer mechanisms within a structure or a structural element which can later be calibrated using experimental data, rather, they are developed based on their ability to fit the data as close as possible.
This study was performed using the experimental information (comprising of the design details, the properties of the materials, and the structural response of the material to the applied loading history such as the load-carrying capacity and load-deflection curves) on 153 exterior RC beam-column joint sub-assemblages as a database. The linear elements of these sub-assemblages are taken as the areas of the frame-type structures that extend between the interface of the joint elements and the nearest contra-flexure point. The beam-column sub-assemblages are sometimes exposed to transverse loading which is often applied near the end of the beam and rarely near the upper column end. The transverse load in most of the sub assemblages is applied in combination with a relatively constant axial compressive force which is concentrically exerted at the upper column end.
Most recently, the application of data intelligence models have shown a massive progress on the structural engineering related issues [10], [11], [12], [13]. There are several versions of DI models have been investigated for prediction problems. Artificial Neural Network (ANN) one of the predominate model applied extensively over the others [14], [15], [16], [17], [18]. The ANN is an intelligent technique which can solve complex nonlinear problems that may not be possibly solved with the classic parametric approaches [19]. The ANN can be trained with different algorithms like Gradient Descent (GD) but its major drawback is that it requires a long learning time [20]. As new enhanced version of classical ANN, the extreme learning machine (ELM) was proposed by [21], as a single-layer feedforward neural network (SLFN). The ELM algorithm was proposed address the issue of prolonged training time of ANN and has been proven to be efficient in reducing the learning time of ANN, making it faster with good generalization performances. The learning speed of the ELM is faster compared to that of the traditional algorithms like back-propagation (BP). Furthermore, the ELM presents a better performance as it can achieve the least norm of weights and training error. These attributes have endeared the ELM to several fields like time-series prediction, classification, and pattern recognition.
To the best knowledge of the authors, the current study is for the first-time implementation of extreme learning machine predictive model for load-carrying capacity and mode failure (i.e., beam failure and joint failure) simulation of beam-column joint connection. The feasibility of ELM model validated against one of the robust and reliable regression models (i.e., multivariate adaptive regression spline). The models were constructed based on multiple related beam-column joint connection information.
Section snippets
Beam-column joint connection data description
The current modeling procedure was constructed based on various concrete characteristics, beam, column and joint dimensional information. The systematic visualization of beam-column joint connection with its loading presented in Fig. 1. 153 experimental observation collected from several published researched in the literature [2], [4], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48],
Extreme learning machine (ELM)
The ELM was developed as a new soft computing techniques (SCTs) with the aim of addressing the problems of the classical ANN [53]. As the name suggests, “extreme” depicts the strong capability of the ELM to mimic the human brain activity within a short period [54]. Compared to the classical SCTs like artificial neural network and support vector machine where there is a need for human intervention and tuning of the internal parameters, the learning process of the ELM requires no parameter tuning
Simulation results and analysis
This section covers the phase of the modeling application results of the proposed ELM model and the comparable MARS predictive model. The applied ELM model is appraised in comparison with MARS model using statistical metrics and diagnostic plots between the predicted and experimental Pmax as well as the model failure. Table 1 presents the input combinations used to construct the applied ELM and MARS predictive models based on the beam, column, joint and concrete information in the load-carrying
Conclusion and remarks
As a matter of fact, the characteristics of the reinforced concrete external beam-column joint is nonlinear and nonstationary. The is because of the involvement of several dimensional and concrete possessions. Therefore, structural engineers are highly motivated to establish a reliable predictive model where to determine the significant engineering prospects related to design and set up a reliable structural model. The current investigation was implemented a newly data intelligence model called
Declaration of Competing Interest
The authors have no conflict of interest to any party.
Acknowledgement
The authors would like to acknowledge their gratitude and appreciation to all the authors of the previous published researches where their experimental data was used for the current research modeling. The appreciation is also extended to the respected editors and reviewers for giving a valid concerns to enhance the visualization of the manuscript context.
References (80)
- et al.
Modeling corrosion currents of reinforced concrete using ANN
Autom Constr
(2009) - et al.
On the complexity of artificial neural networks for smart structures monitoring
Eng Struct
(2006) - et al.
ANN prediction of cement mortar compressive strength, influence of cement strength class
Constr Build Mater
(2017) Prediction of fracture parameters of concrete by Artificial Neural Networks
Eng Fract Mech
(2004)- et al.
Extreme learning machine: theory and applications
Neurocomputing
(2006) - et al.
Trends in extreme learning machines: a review
Neural Networks
(2015) - et al.
Burning state recognition of rotary kiln using ELMs with heterogeneous features
Neurocomputing
(2013) - et al.
Predicting compressive strength of lightweight foamed concrete using extreme learning machine model
Adv Eng Softw
(2018) - et al.
Non-tuned data intelligent model for soil temperature estimation: a new approach
Geoderma
(2018) - et al.
Accurately predicting building energy performance using evolutionary multivariate adaptive regression splines
Appl Soft Comput
(2014)
Pan evaporation modeling using six different heuristic computing methods in different climates of China
J Hydrol
The whale optimization algorithm
Adv Eng Softw
Analysis of FRP-strengthened RC beam-column joints
J Compos Constr
Seismic behaviour of RC external joints
Mag Concr Res
Improving the design of earthquake resistant reinforced concrete exterior beam-column joints
Struct Eng
Experimental investigation of FRP-strengthened RC beam-column joints
J Compos Constr
Exterior beam column joints – shear strength model and design formula
Eng Struct
Seismic response of steel fibre reinforced concrete beam-column joints
Eng Struct
Effect of transverse beams and slab on behavior of reinforced concrete beam-to-column connections
J Proc
Neural network modeling of the load-carrying capacity of eccentrically-loaded single-angle struts
J Constr Steel Res
Assessment of RC exterior beam-column Joints based on artificial neural networks and other methods
Eng Struct
Multiple linear regression, artificial neural network, and fuzzy logic prediction of 28 days compressive strength of concrete
Front Struct Civ Eng
Neural network modeling of confined compressive strength and strain of circular concrete columns
J Struct Eng
Optimizing an ANN model with ICA for estimating bearing capacity of driven pile in cohesionless soil
Eng Comput
Feasibility of PSO-ANN model for predicting surface settlement caused by tunneling
Eng Comput
Prediction of shear strength for squat RC walls using a hybrid ANN–PSO model
Eng Comput
RBFNN versus FFNN for daily river flow forecasting at Johor River, Malaysia
Neural Comput Appl
ANN models optimized using swarm intelligence algorithms
WSEAS Trans Comput
Comparison between interior and exterior RC beam-column joint behavior
Spec Publ
High-strength RC connections subjected to inelastic cyclic loading
J Struct Eng
Recent studies on reinforced concrete beam column joints in Japan. Phil M
Response of reinforced high-strength concrete beamcolumn joints under load reversals
Mag Concr Res
Effectiveness of RC beam-column connection repair using epoxy resin injections
J Earthq Eng
Cyclic behavior of reinforced concrete eccentric beam-column corner joints connecting spread-ended beams
ACI Struct J
Seismic load tests on interior and exterior beam-column joints with substandard reinforcing details
ACI Struct J
Performance-based evaluation of reinforced concrete building exterior joints for seismic excitation
Earthq Spectra
Application of steel fiber reinforced concrete in seismic beamcolumn joints
Seismic rehabilitation of beam-column joint using GFRP sheets
Eng Struct
Seismic resistance of reinforced concrete beam-column joints
J Struct Div
Cited by (65)
Development of optimized machine learning models for predicting flat plate solar collectors thermal efficiency associated with Al<inf>2</inf>O<inf>3</inf>-water nanofluids
2024, Engineering Applications of Artificial IntelligencePrediction on the seismic performance limits of reinforced concrete columns based on machine learning method
2024, Soil Dynamics and Earthquake EngineeringMachine learning prediction of structural dynamic responses using graph neural networks
2023, Computers and StructuresHybrid mechanical and data-driven probabilistic model for shear strength of RC beam-column joints
2023, Probabilistic Engineering Mechanics