An ANFIS-based approach for predicting the bed load for moderately sized rivers

A total of 346 sets of bed-load data obtained from the Kinta River, Pari River, Kerayong River and Langat River were analyzed using four common bed-load equations. These assessments, based on the median sediment size (d₅₀), show that the existing equations were unable to predict the measured bed load accurately. All existing equations over-predicted the measured values, and none of the existing bed-load equations gave satisfactory performance when tested on local river data. Therefore, the present study applies a new soft computing technique, i.e. an adaptive neuro-fuzzy inference system (ANFIS), to better predict measured bed-load data. Validation of the developed network (ANFIS) was performed using a new set of bed-load data collected at Kulim River. The results show that the recommended network can more accurately predict the measured bed-load data when compared to an equation based on a regression method.