Deformation Prediction of 4D-Printed Active Composite Structures Based on Data Mining

Voxelizing active composite structures and controlling voxel-level material properties via 4D printing significantly expand design possibilities. However, as the number of voxels increases, the design space grows exponentially, posing significant challenges for predicting structural deformation. Here, a scalable deformation prediction method based on data mining is proposed. This method constructs a feature database using manually extracted features and employs the proposed curvature-driven sequence point generation (CSPG) algorithm to predict deformations for voxel encodings of arbitrary length. Compared with the traditional finite element (FE) method, this approach significantly improves prediction efficiency, completing a single task within a second. In contrast to deep learning (DL) methods, this method improves prediction accuracy and effectively addresses the limited generalization ability of DL models when applied to structures beyond the training domain. To enhance usability, an interactive web-based platform is developed that allows users to customize voxel encodings and obtain end-to-end predictions. In addition to serving as an efficient tool for deformation prediction of active composite structures, this work introduces a novel pathway for the optimal design of complex intelligent structures.