TY - JOUR
AU - Patel, Dilip
AB - Miter bends have a major impact on the efficiency of piping systems and are essential in a variety of sectors, such as chemical, nuclear, and hydropower plants. It is imperative to forecast their failure under different loading scenarios, and Von Mises stress turns out to be a critical element. This study investigates the correlation between geometrical parameters, such as pipe diameter, thickness, and weld count, and miter bend lifespan. Von Mises stress was calculated using finite element analysis in a variety of configurations, creating an extensive dataset for creating Machine Learning (ML) models. Comparative analysis revealed that random forest regression outperformed other ML models, such as multiple linear regression and decision tree regression, achieving an impressive R-squared score of 0.98 for accurate stress prediction of Miter bends. Furthermore, the study benchmarked these ML models against conventional methods, including Taguchi regression and response surface regression, where ML models demonstrated superior predictive power, particularly in handling complex, nonlinear relationships and maintaining robustness against multi-collinearity. It is advised to use sophisticated methods such as SMOTE for dataset augmentation and Generative Adversarial Networks for creating synthetic data. Subsequent research paths entail investigating advanced machine learning algorithms to augment predicted accuracy, emphasizing dataset augmentation, managing anomalies, refining feature selection techniques, and executing stringent cross-validation. These discoveries lay a strong basis for the development of Miter bend performance prediction, which will ultimately optimize industrial pipe systems for increased dependability and effectiveness. The study emphasizes that it is possible to further improve the forecast of Miter bend performance for continuous operational efficiency in industrial environments.
TI - Development of Machine Learning Regression Models to Assess Von Mises Stresses of Miter Bend
JO - Journal of The Institution of Engineers (India): Series C
DO - 10.1007/s40032-025-01181-0
DA - 2025-06-01
UR - https://www.deepdyve.com/lp/springer-journals/development-of-machine-learning-regression-models-to-assess-von-mises-Zyf4R390V0
SP - 749
EP - 764
VL - 106
IS - 3
DP - DeepDyve
ER -