PSI - Issue 70

N. Vignesh Kumar et al. / Procedia Structural Integrity 70 (2025) 453–460

458

b

Fig. 2. (a) predicted fcc vs actual f cc for CFRP; (b) predicted fcc vs actual f cc for GFRP

Scatter plot (Fig. 2 (a) and (b)) confirm GPR’s high accuracy, comparing predicted vs. experimental fcc. The model captures the non-linear behavior effectively. Our findings align with recent studies highlighting complex relationships in FRP-confined concrete.

4.1 Derived Prediction Equations

The prediction equations obtained from LASSO and STEPWISE regression models are as follows: CFRP: STEPWISE Regression = 19.997 − 0.47299 + 107.14 + 0.024102 + 0.06268 − 0.89942 · − 0.00070104 · + 0.71653 · + 0.089784 · + 0.0010431 · − 0.0044125 2 − 0.0036412 2 (4) GFRP: STEPWISE Regression = −16.835 + 3.0753 + 1.3991 + 0.050484 − 0.12102 · + 0.05512 · − 0.001397 · (5) CFRP: LASSO Regression f cc = 22.98229 − 0.14486D + 0.63238f co + 0.00107D · f co + 0.02975( H D ) · f co + 0.18956t frp · f co + 0.00501( H D ) · f frp + 0.07428t frp (6) GFRP: LASSO Regression f cc = 56.86298 + 0.00874f frp − 0.00110D 2 − 0.01920D · ( H D ) − 0.04700t frp · f co + 0.00776f c 2 o + 0.00009( H D ) · f frp + 0.05403t frp . f frp − 0.00029f co · f frp (7) The coefficients in the equations of the parametric models irrespective of the CFRP and the GFRP type indicate the FRP thickness has substantiate priority followed by other parameters given as input in the dataset.