A hybrid fuzzy inference model based on RBFNN and artificial bee colony for predicting the uplift capacity of suction caissons

• We propose an intelligent fuzzy inference model, abbreviated as IFRIM.
• We examine IFRIM's performance against other models via two experiments.
• Two comparisons of results indicate that IFRIM is the fittest model.
• We recommend IFRIMG as the best choice in terms of performance and efficiency.

The suction caisson is an essential part of the foundation system used in offshore platforms. The failure of a single suction caisson may cause the collapse of an entire offshore system. Hence, accurately predicting the uplift capacity of suction caissons is of critical importance to platform function and reliability. This study proposes the intelligent fuzzy radial basis function neural network inference model (IFRIM) to predict the uplift capacity of suction caissons. IFRIM is a hybrid of the radial basis function neural network (RBFNN), fuzzy logic (FL), and artificial bee colony (ABC) algorithm. In the IFRIM, FL deals with imprecise and uncertain information; RBFNN acts as a supervised learning technique to address fuzzy input–output mapping relationships; and ABC searches for the most appropriate parameter settings for RBFNN and FL. Comparison results show IFRIM to be the fittest model for predicting the uplift capacity of suction caissons in terms of accuracy and reliability. A 10-fold cross-validation approach found that the IFRIM reduced the RMSE and MAPE at least 70% and 90%, respectively, below other tested models.