[Paper Review] Comparative Analysis of Shear Strength Prediction Models for Reinforced Concrete Slab-Column Connections

Sarmed Wahab, Nasim Shakouri Mahmoudabadi+5 more|arXiv (Cornell University)|Sep 29, 2023

Structural Behavior of Reinforced Concrete4 citations

TL;DR

This study compares shear strength prediction models for reinforced concrete slab-column connections using machine learning (PSOFNN, BATFNN, FNN), design codes (ACI 318-19, EC2, CFP), and finite element analysis (FEA). PSOFNN outperformed all models with an R² of 99.37%, MSE of 0.0275%, and MAE of 1.214%, demonstrating superior accuracy in predicting shear strength.

ABSTRACT

This research aims at comparative analysis of shear strength prediction at slab-column connection, unifying machine learning, design codes and Finite Element Analysis. Current design codes (CDCs) of ACI 318-19 (ACI), Eurocode 2 (EC2), Compressive Force Path (CFP) method, Feed Forward Neural Network (FNN) based Artificial Neural Network (ANN), PSO-based FNN (PSOFNN), and BAT algorithm-based BATFNN are used. The study is complemented with FEA of slab for validating the experimental results and machine learning predictions.In the case of hybrid models of PSOFNN and BATFNN, mean square error is used as an objective function to obtain the optimized values of the weights, that are used by Feed Forward Neural Network to perform predictions on the slab data. Seven different models of PSOFNN, BATFNN, and FNN are trained on this data and the results exhibited that PSOFNN is the best model overall. PSOFNN has the best results for SCS=1 with highest value of R as 99.37% and lowest of MSE, and MAE values of 0.0275%, and 1.214% respectively which are better than the best FNN model for SCS=4 having the values of R, MSE, and MAE as 97.464%, 0.0492%, and 1.43%, respectively.

Motivation & Objective

• To evaluate and compare the accuracy of shear strength prediction models for reinforced concrete slab-column connections.
• To integrate machine learning models (FNN, PSOFNN, BATFNN) with established design codes (ACI 318-19, EC2, CFP) and FEA for improved prediction.
• To validate experimental results using finite element analysis and optimize machine learning models via metaheuristic algorithms.
• To identify the most accurate predictive model for practical structural design applications.
• To quantify the performance of hybrid models using statistical metrics like R², MSE, and MAE.

Proposed method

• Trained three types of feedforward neural networks: standard FNN, PSO-optimized FNN (PSOFNN), and BAT algorithm-optimized FNN (BATFNN) on experimental slab-column connection data.
• Used mean square error (MSE) as the objective function to optimize network weights in PSOFNN and BATFNN models.
• Conducted finite element analysis (FEA) of slab-column connections to validate experimental results and support model training.
• Compared model performance across seven configurations of PSOFNN, BATFNN, and FNN using statistical metrics: R², MSE, and MAE.
• Applied metaheuristic optimization (PSO and BAT) to enhance the convergence and accuracy of the neural network weights.
• Evaluated models under different shear capacity scenarios (SCS=1 to SCS=4) to assess robustness across loading conditions.

Experimental results

Research questions

• RQ1 How do PSOFNN, BATFNN, and FNN models compare in predicting the shear strength of RC slab-column connections?
• RQ2 What is the performance of machine learning models relative to established design codes (ACI 318-19, EC2, CFP) in shear strength prediction?
• RQ3 To what extent does finite element analysis improve the reliability of experimental data and model validation?
• RQ4 Which model achieves the highest R², lowest MSE, and lowest MAE across different shear capacity scenarios?
• RQ5 How do metaheuristic optimization techniques (PSO and BAT) enhance the predictive accuracy of neural networks in this context?

Key findings

• PSOFNN achieved the highest R² value of 99.37% for SCS=1, significantly outperforming the best FNN model (R² = 97.464%) for SCS=4.
• PSOFNN recorded the lowest mean squared error (MSE) of 0.0275% and mean absolute error (MAE) of 1.214% across all models and scenarios.
• The FNN model for SCS=4 showed an R² of 97.464%, MSE of 0.0492%, and MAE of 1.43%, indicating lower accuracy than PSOFNN.
• PSOFNN consistently outperformed BATFNN and FNN across all performance metrics, demonstrating superior generalization and optimization.
• Finite element analysis successfully validated experimental results, confirming the reliability of the dataset used for model training.
• The integration of metaheuristic algorithms (PSO and BAT) with FNN significantly improved prediction accuracy compared to standard FNN.