.png){kind=icon}
.png){kind=icon}
Return
Multi-layer collaborative stacking optimization framework for sprocket stress prediction
.png){kind=icon}
PRE
AI
DOI：doi:10.1088/1361-6501/adf910.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In response to the current challenges of low computational efficiency in stress analysis for critical mechanical components and the limited generalizability of predictive models, this study proposes a multi-level collaborative stacking optimization framework to investigate the stress characteristics of scraper conveyor sprockets. Dimensionality reduction of sprocket mesh nodes is performed through a mesh-based divide-and-conquer strategy combined with the K-nearest neighbors interpolation method, thereby enhancing computational efficiency. To optimize model performance, the particle swarm optimization algorithm is employed for hyperparameter tuning of six heterogeneous base models. The Technique for Order Preference by Similarity to Ideal Solution is utilized to conduct an objective evaluation and to identify the three most optimal base models. A multi-layer collaborative stacking framework is subsequently constructed, incorporating performance-based weighting, feature augmentation, and ten-fold cross-validation to achieve generalized prediction. The predictive performance of the MCSOF is validated through comparative analysis of multiple decision fusion methods and evaluation on the test dataset. The results indicate that the MCSOF achieves the highest prediction accuracy (R2 = 0.9739, MSE = 0.1859,MAE = 0.0703). Additionally, stress rendering using Unity demonstrates a strong consistency between the predicted results and the simulation outcomes. This study provides a novel methodological framework for accurate stress field prediction of sprockets and offers theoretical support for the application of machine learning in stress prediction of mechanical components.
Keywords:
stress analysis
machine learning
multi-level collaborative stacking
sprocket
computational efficiency
Journal
IF:
3.4
Papers:
2.6K
Citations:
2.3W
Organization
No organization information available