SoftwareX (Dec 2023)
Pressure vessel nozzle local stress prediction software based on ABAQUS- machine learning
Abstract
The process of calculating local stresses at pressure vessel nozzle connections using the finite element analysis method is intricate and significantly restricts the analysis speed. To address this challenge, this study undertakes secondary development of an ABAQUS plugin to generate a dataset, utilizing machine learning techniques to construct a predictive model. Based on the model, a GUI has been added, transforming the model into a user-friendly software that can be easily installed and utilized. This software enables the rapid prediction of local stresses at pressure vessel nozzle connections, achieving an accuracy level exceeding 0.999 and a mean squared error of only 1.639. This approach contrasts with traditional finite element analysis design methods, as it circumvents numerous complex analysis steps. It provides a reliable and convenient platform for swiftly evaluating and optimizing pressure vessel designs. Users can refer to the user manual to develop specific stress prediction software for different working conditions and material properties. (The plugin, dataset, software, code, user manual, and related resources are all publicly available at https://github.com/Fan-Tank/Local_Stress_Prediction.git).
