An optimal Adaptive Neural Fuzzy Inference System (ANFIS) model and regression relations to predict stable channel geometry in rivers gravel bed

Hydraulic geometry of a river has primary importance in the design, planning, management and river training in river engineering science. In investigation of stable channels dimensions, the most presented relations are based on statistical and theoretical methods that don’t have more accuracy. In last decades, using soft computing methods or artificial neural methods because of high accuracy and fewer time and cost are interested by different science researches. In the present paper, using Adaptive Neural Fuzzy Inference System (ANFIS) model, the accuracy of regression relations to predict width, depth and slope of stable channels are improved. A set of observed data (including 85 cross section data) are used to train and test ANFIS models and also to fit regression relations. The two models efficiency are evaluated and compared with observed data. Results show that ANFIS models with R2 values of 0.9224, 0.7464 and 0.9264 show a high accuracy to predict width, depth and slope of stable channels, respectively. Also, the mean absolute relative error (MARE) values in regression relation are 73, 57 and 50 times higher than ANFIS models in predicting width, depth and slope, respectively. Therefore, using ANFIS model causes to improve regression equations performance and its results can be used in the design of Executive channels.

کلیدواژه ها:

ANFIS model ، regression relations ، models efficiency ، observed data ، stable channel geometry