Prediction of Ultimate Strength of Locally Corroded Plates Using ANFIS Model

In recent years, increasing attention has been paid to the effects of pitting corrosion in ships’ hull structures. Unlike general corrosion, where the thickness loss is normally assumed to be uniform, pitting corrosion is extremely localised which poses difficulties in the ultimate strength assessment of structures suffering from it. The use of an Adaptive Neuro-Fuzzy Inference System (ANFIS) method is proposed in this paper to predict the ultimate strength reduction of locally corroded plates. Published ultimate strength data sets for unstiffened plates with pitting corrosion subjected to uniaxial inplane compressive loads were used to create and test ANFIS models. A number of input variables were used including the ratios of the plate slenderness, pit breadth to plate width, pit length to plate length and pit depth to plate thickness. Rule-based fuzzy sets were used for mapping the inputs to the output which is the reduction in the ultimate strength. Different types of membership functions were tested to find the best accurate model. The two-sided Gaussian-type function was found to be more effective and less sensitive to the sample size than other functions tested in this study. It was concluded that the developed model can predict the ultimate strength of structures locally corroded.