Tanrıver, KürşatEtyemez, AyhanAy, Mustafa2025-11-072025-11-072025Tanrıver, K., Etyemez, A., & Ay, M. (2025). Integrated use of finite element analysis and gaussian process regression in the structural analysis of AISI 316 stainless steel chimney systems. Scientific Reports, 15, pp. 1-16. https://doi.org/10.1038/s41598-025-21678-z2045-2322https://doi.org/10.1038/s41598-025-21678-zhttps://hdl.handle.net/20.500.13055/1171This study aimed to conduct a comprehensive structural analysis and machine learning-assisted predictive modelling of a chimney system manufactured from 2 mm thick AISI 316 stainless steel with a diameter of Ø500 mm. The primary motivation of this work was to examine, in detail, the structural behavior of chimney modules under various force and pressure conditions using conventional methods, and to develop a reliable model capable of performing parametric predictions for new scenarios based on the acquired data. The scope of the study encompassed finite element analyses of both the entire chimney system and 3-meter-long intermediate modules, field tests, and the application of the Gaussian Process Regression (GPR) machine learning model. In the analysis of the entire chimney system under an applied force of 22,000 N, a maximum stress of 28 MPa and a safety factor of 8.39 were observed in the chimney clamps. The total deformation was found to be 0.58 mm, which is within acceptable limits. In the structural analysis of the intermediate chimney modules under a force of 1000 N and an internal pressure of 5 MPa, a maximum stress of 11,984 MPa, a safety factor of 1.71, and a total deformation of 0.46 mm were determined, all of which are consistent with the literature. The accuracy of these analyses was validated through pressure and leakage tests conducted in accordance with the EN 1859 standard. The developed GPR machine learning model demonstrated exceptionally high accuracy (R² > 0.999) in predicting Von Mises stress values, providing reliable forecasts with an error rate of less than 3% when compared to ANSYS simulation outputs. However, in predicting total deformation values, error rates exceeded 70%, indicating that the model was less sensitive in low-amplitude deformation cases. These findings suggest that the GPR model can generate reliable predictions for Von Mises stress a more critical parameter than total deformation in chimney design. By integrating conventional structural analysis methods with advanced machine learning techniques, this study demonstrates the potential of predictive modeling as an efficient and reliable tool in engineering design processes, making a significant contribution to the field’s body of knowledge.eninfo:eu-repo/semantics/openAccessChimney SystemFEM AnalysisGaussian Process Regression (GPR) Stress PredictionStructural AnalysisArticle10.1038/s41598-025-21678-z15116Q12-s2.0-105020279975PMID: 41162472Q1