Using Neural Networks and Analytical Methods to Calculate Stress Concentration Factors Based on Equivalent Defect Parameters in Offshore Fixed Platforms

The study addresses the problem of stress assessment in structural elements of offshore fixed platforms exposed to corrosive damage. Corrosive caverns generated on the metal surface are local stress concentrators, which reduces the strength and reliability of facilities. To solve the problem of prompt calculation of stress concentration factor in corroded elements of offshore fixed platforms, a hybrid method combining numerical modeling and artificial intelligence has been proposed. At the first stage, using the SolidWorks CAD system, modeling of sectors of offshore structural steel elements with conic corrosive defects of varying geometry has been conducted. In the course of calculation using the finite element method, the impact of parameters such as the depth and dimensions of semiaxes of a studied defect on the distribution of equivalent stresses under applied tensile load has been studied. The obtained dataset has represented a selection to train a neural network capable of forecasting the stress concentration factor based on the preset geometric properties of the equivalent defect. For the simplified prompt calculation based on the neural network output data, analytical dependencies on the cavern key parameters have been derived. Model accuracy validation has shown the average absolute error of around 4 % compared to the results of numerical modelling in SolidWorks. For the detailed analysis, heat maps of error distribution and curves of partial dependencies have been built to visualize the impact of each geometrical parameter of a cavern on the stress concentration factor and help identify the most critical combinations of defect dimensions. The proposed approach combines high accuracy and the promptness of calculations. It can be integrated into offshore platform technical monitoring systems, enabling timely assessment of the criticality of corrosion damage and making decisions on the repair and reinforcement of structures.