PSI - Issue 64

Amir Mofidi et al. / Procedia Structural Integrity 64 (2024) 999– 1008 Mofidi et al./ Structural Integrity Procedia 00 (2019) 000 – 000

1007

9

coefficient of efficiency, the best models are Mofidi et al. (2023), Bianco et al. (2014), and Dias and Barros (2013) with R 2 1:1 equal to 0.41, 0.33, and 0.19, respectively. Lastly, the most accurate models when it comes to the values of the index of agreement are Bianco et al. (2014), Mofidi et al. (2023), and modified Mofidi et al. (2016) with d values equal to 0.84, 0.81, and 0.79, respectively. From the results of this study, it can be concluded that mechanics-based models proposed by Mofidi et al. (2023), Bianco et al. (2014) and the modified Mofidi et al. (2016) were among the best, if not the best, in most measured metrics. The models produced by regressions of data or neural networks such as Dias and Barros (2013) and Perera et al. (2014) performed better under the statistical parameter for which they were fitted, e.g., CoV . Such models may not perform well when compared with the data that was not used to calibrate the models or when assessed by a metric that they are not calibrated with. On the other hand, for the mechanics-based models, due to the existence of the principles of shear mechanics and bonding in RC beams in the development of such models, the mechanics-based models can perform well with existing and incoming experimental test data and through different statistical test parameters. It is worth noting that Mofidi et al. (2023) produced predictive results, with 10 simple and easy-to-use equations, that are comparable to, if not more accurate than; results produced by Bianco et al. (2014), produced using 27 equations in a rather complex design process. 4. Conclusions In this study, the state-of-the-art design models for shear-strengthened RC beams with NSM FRP bars and laminates were compared using the experimental results on 131 test specimens from 24 studies. A multi-metric comparative analysis was conducted on the existing NSM predictive models for shear strengthening of RC beams to compare their accuracy in the predicted results. Unlike, most existing analytical models, Mofidi et al. model (2023) benefits from a state-of-the-art bond model with a very practical and straightforward approach with a total of 10 simple equations. This makes the model an excellent candidate to be implemented in a standard code or guidelines to be used by engineers in practice with a hassle-free approach. From the multi-metric comparative study presented in this article, it can be concluded that the model is not only practical but also one of the most accurate existing models for the shear-strengthening of RC beams using the NSM technique. The following conclusions can be drawn from this comparative study: • Mofidi et al. (2023) model had the superior R 2 1:1 equal to 0.41 followed by Bianco et al. (2014) and Dias and Barros (2013) models with R 2 1:1 equal to 0.33 and 0.19, respectively. • Bianco et al. (2014), Mofidi et al. (2023), and modified Mofidi et al. (2016) models were the best in the other correlation-based metrics considered in this study, i.e. , R 2 and d . • The best produced CoV values belong to Perera et al. (2014) ( CoV = 0.44), Dias and Barros (2013) ( CoV = 0.46), Mofidi et al. (2023) ( CoV = 0.55), and Bianco et al. (2014) ( CoV = 0.55). • With respect to the produced errors, Mofidi et al. model (2023) was able to produce the most precise values with the least errors compared to the rest of the models in all error-based metrics ( MAE = 19.11 kN, NRMSE = 0.43, and RMSLE = 0.27). • Rizzo and De Lorenzis (2009) upper GIP, Parreti and Nanni (2004), Rizzo and De Lorenzis (2009) Lower GIP produced extremely safe results with Means of / equal to 2.04, 1.90, and 1.76, respectively. • Perera et al. (2014) and Dias and Barros (2013) produced the most unsafe results with Means equal to 0.74 and 0.95, while Bianco et al. (2014), modified Mofidi et al. (2016), Mofidi et al. (2023) models produced reasonably safe results with Means equal to 1.55, 1.22, 1.12, respectively. • From the results of this comparative study, it can be concluded that the easy-to-use approach of Mofidi et al. model (2023) led to simple steps in the calculation of without compromising the proposed model’s accuracy. • For future research, a reliability-based assessment of the Mofidi et al. (2023) model can be conducted based on different existing standard codes and guidelines’ reliability approaches .