The construction industry relies heavily on heavy machinery for the construction of facilities, real estate, and infrastructure. However, the design of heavy machinery poses significant challenges, including inaccuracies in finite element method (FEM) models and the high computational cost of multi-scale topology design. In this presentation, we propose a hybrid data-driven framework that addresses these challenges and allows for the selection and optimization of key features to improve the overall performance of heavy machinery systems.

To overcome the inaccuracy of FEM models, we use a hybrid data-driven FEM model that combines experimental data and physical simulations. We also use feature engineering and parametric modeling to simplify the original model while maintaining the most important features. To reduce the computational cost associated with contact regions, we have developed a data-driven framework that uses predictive machine learning algorithms to efficiently design the contact regions between different parts. To enhance the design of the contact region, we use the association rule mining algorithm as a post-processing method for the predictive models.

To validate our proposed approach, we apply it to the design of a telehandler product manufactured by Xuzhou Construction Machinery Group American Research Corporation (XCMG-ARC). We use the hybrid data-driven FEM model to simulate the telehandler’s performance and optimize its key features, such as total weight and maximum deformation. We also use MC-SCA to concurrently optimize the material characterization and selection for the telehandler design.

In conclusion, our proposed data-driven framework offers a comprehensive optimization and design process from engineering experiments to reduced-order simulations in the heavy machinery industry. It is applicable to different heavy industry designs and can provide guidance for the industry.