Estimation of hydraulic conductivity using gradation information through larsen fuzzy logic hybrid wavelet artificial neural network and combined artificial intelligence models

Hydraulic conductivity is a critical parameter in geotechnical studies, though determining it with field and laboratory methods is remarkably costly and time-consuming and suffers from innate uncertainty. Over the past few years, various AI models with higher accuracy have been used to determine the parameter. In the present study, two distinct AI models, including Larsen’s Fuzzy Logic (LFL) and the Hybrid Wavelet-Artificial Neural Network (WANN), were implemented to predict hydraulic conductivity based on gradation information in Lines 1 and 2 of Tabriz Metro System, Tabriz, Iran. To enjoy the combined benefits and capabilities of the above models, their output (i.e., hydraulic conductivity) was combined using Sugeno’s Fuzzy Logic (SFL) model and presented as the Combined Model of Artificial Intelligence (CMAI). The findings of the study showed that the CMAI model was more successful than individual models in predicting hydraulic conductivity. In the experimental stage, the model increased the evaluation criterion R2 compared to the LFL and WANN models by 31 and 22 percent, respectively. More specifically, compared to the LFL model, it reduced RMSE and MAE by 33 and 24 percent, respectively. Moreover, compared to the WANN model, it reduced RMSE and MAE by 29 and 28 percent, respectively. In addition to significantly increasing R2 and reducing RMSE and MAE, the combined model had a significant impact on approximating the majority of the calculated values to the values observed during the experimental stage point-by-point.