Advanced hybrid machine learning models for estimating chloride penetration resistance of concrete structures for durability assessment: optimization and hyperparameter tuning

Figures & Tables

Statistics	W/B	Water	Cement	Slag	Fly ash	SF	FA	CA	TA	SP	MT age	D _nssm
Statistics		kg/m³	kg/m³	kg/m³	kg/m³	kg/m³	kg/m³	kg/m³	kg/m³	% (by W_b)	Day	(×10⁻¹² m²/s)
Mean	0.42	162.3	330.07	29.3	33.9	5.28	759.06	818.5	1,636	0.48	83.06	9.1
Median	0.44	164.5	334.61	0	0	0	768	944	1746	0.47	28	7.04
Mode	0.45	162	360	0	0	0	786	463.0	1,960	0	28	8.8
SD	0.09	54.68	134.32	85.6	81.8	19.9	183.63	307.0	323.9	0.53	95.47	9.42
Maximum	0.65	1,049	2,384.97	1,284	735	468	1,574.1	1,240	2,097	4.17	365	133.6
Minimum	0.19	8.46	13.02	0	0	0	27.53	0	54.04	0	3	0.2
Skewness	−0.02	9.74	8.32	8.23	5.14	15.4	−0.47	−0.86	−2.33	1.25	1.83	5.4

Phase	Models	R ²	Adj R ²	RMSE	MAE	RSR	a-10 index	a-20 index
Training	SVR-GTO	0.95	0.95	2.95	1.13	0.25	0.94	0.97
	SVR-GWO	0.94	0.93	3.12	2.07	0.27	0.92	0.95
	SVR-PSO	0.92	0.91	3.45	2.21	0.33	0.91	0.93
	SVR-FFA	0.90	0.90	3.56	2.34	0.45	0.90	0.92
Testing	SVR-GTO	0.97	0.97	0.93	0.33	0.11	0.94	0.98
	SVR-GWO	0.92	0.92	1.55	0.69	0.19	0.92	0.94
	SVR-PSO	0.91	0.90	1.67	0.87	0.24	0.90	0.93
	SVR-FFA	0.89	0.90	2.13	1.12	0.43	0.89	0.91

Reference	Best model	Model interpretation	R ²
[50]	XGBoost		0.99
[51]	PSO-XGBoost	SHAP	0.996
[52]	LGB		0.999
[53]	ANN	Sensitivity analysis	0.99
[54]	BR	SHAP	0.99
[55]	MLP		0.912
Current study	SVR-GTO	SHAP, ICE, and PDP	0.97