Artificial intelligence simulation of suspended sediment load with different membership functions of ANFIS

Modeling suspended sediment load is a critical element of water resources engineering. In this work, using the ANFIS method, everyday suspended sediment particles were estimated in different categories of the river in US Sediment big data, and various flow rates were utilized for testing and training. The artificial intelligent (AI) method called ANFIS is used to train actual data from the river and provide an AI model with artificial data points. This artificial data point can show the occurrence of disaster for a critical day with different flow rates. The changing parameter in the AI model enables us to make a correct decision about critical time for rivers. This study also concentrates on the sensitivity investigation of ANFIS setting parameters on the accurateness of numerical results in order to find the best ANFIS model for rapid oscillation in the data set. The best performance of the ANFIS method is achieved with the trimf membership function, the number of input membership function = 16, and the number of iteration = 1000. The results also showed that the ANFIS model can provide fast computational calculation, and adding more nodes for the prediction cannot change the overall time of calculation due to the meshless behavior of the model. In addition to this model, we used the ant colony method for training of data set, and we found that the ANFIS method is better in learning and prediction of the dataset.

Automated summary

This study utilized the ANFIS method to estimate and train daily suspended sediment particles in different categories of rivers in the US, while also investigating the impact of setting parameters on result accuracy. The results demonstrated that the ANFIS model excelled in fast calculations and accurate predictions of critical time points.