Optimization of Adaptive Neuro-Fuzzy Inference System using Differential Evolution Algorithm for Scour Prediction around Submerged Pipes

Nowadays, a huge amount of natural resources such as gases and oil are exploited from offshore oil fields and transported by pipes located at seabed. The pipelines are exposed to waves and currents and scour may occur around them. Subsequently, stability of the pipes can be threatened, so estimation and simulation of scouring around the pipes are quite vital. In this study, a hybrid method for simulating the scour depth in the vicinity of submerged pipes was developed. In other words, the adaptive neuro-fuzzy inference system (ANFIS) and the differential algorithm were combined with each other to simulate the scour depth. In general, ANFIS is an artificial neural network acts based on the Takagi-Sugeno inference system. This model is a set of if-then rules which is able to approximate non-linear functions. In addition, the differential algorithm is a powerful evolutionary algorithm among optimization algorithms which have many applications in scientific fields. In this study, the Monte-Carlo simulation was employed for examining the ability of numerical models. To validate the modeling results, the k-fold cross validation approach was also utilized with k=6. Then, the parameters affecting the scour depth were detected and six ANFIS and hybrid models were developed for scour estimation. After that, the results of the mentioned models were examined and this analysis showed that the superior model predicts scour values in terms of all input parameters. This model has reasonable accuracy. For example, the values of R and RMSE for this model were calculated 0.974 and 0.079, respectively. Furthermore, the analysis of the modeling results indicated that the ratio of the pipe distance from the sedimentary bed to the pipe diameter (e/D) was identified as the most effective parameter.