Presentation: 2025 ND EPSCoR Annual conference 

October 21, 2025, NDSU Memorial Union, Fargo, North Dakota

Buckling Prediction of Stiffened Cylindrical Shells Through the Integration of Finite Element and Machine Learning Techniques

Stiffened cylindrical shell buckling strength mainly depends on the unstiffened geometric and stiffener properties. A detailed parametric study was conducted to investigate the influence of these properties on the stiffened aluminium cylindrical shell buckling strength. The proposed framework involves an integration of finite element methods and various machine learning techniques. The dataset was obtained from the validated finite element models of 350 numerical simulations using ANSYS workbench 2022. 350 sample specimens were categorized into seven groups based on the no. of stiffeners varying from 3 to 17 while their optimum sizes were obtained from an optimization study. Each group consists of 50 samples with ten distinct values of tank radius and five distinct values of shell thickness. Datasets were trained (80%) and tested (20%) with various simple to complex machine learning algorithms. The predicted buckling strength obtained from each ML technique was compared to the numerical buckling strength. R2 and Mean Square Error were considered cost functions to evaluate the performance of each ML algorithm. A comparison between the proposed algorithms revealed that the artificial neural networks performed excellently followed by Random Forest and polynomial regression respectively. KNN and linear regression models are the least-performing models for the present dataset. Furthermore, the SHapley Additive exPlanations analysis is employed to analyze the contribution of each input parameter to predict the buckling strength, both on a global and local scale. shell thickness and stiffener area to tank radius ratio are the most critical parameters influencing the buckling strength prediction compared to the other input parameters.