Metal-organic frameworks (MOFs) have attracted significant attention in catalysis due to their high surface area, tunable porosity, and flexible structures. These properties make MOFs promising candidates for catalytic reactions including hydrogen evolution, oxygen evolution, carbon dioxide reduction, and organic transformations. In parallel, advances in artificial intelligence (AI) and machine learning (ML) are increasingly influencing catalytic research by accelerating the discovery and optimization of MOF-based catalysts. This review discusses the development of MOF catalysts and their unique structural and electronic features, followed by an overview of ML applications in materials science. It outlines common ML workflows, available databases, widely used algorithms, and software tools employed in catalyst prediction and screening. While previous reviews have highlighted the growing integration of ML and MOF research, this article provides a catalysis-oriented perspective by critically examining the descriptors, datasets, and ML workflows for discovery and optimization of MOF-based catalysts. Recent progress demonstrating the role of ML in accelerating catalyst discovery and performance prediction is highlighted, along with current challenges and limitations. Emerging trends in AI-driven catalyst design and future research directions are also discussed. Overall, this review emphasizes the considerable potential of AI and ML in advancing MOF-based catalytic applications.
Materials capable of accommodating large deformation while maintaining recoverability or functional stability are attracting increasing attention in aerospace structures, biomedical devices, soft robotics, and wearable electronics. Among these materials, shape memory alloys (SMAs), shape memory polymers and their composites (SMPs/SMPCs), liquid crystal elastomers (LCEs), and conductive polymer composites (CPCs) exhibit distinct large-deformation responses arising from different physical mechanisms. SMAs achieve large recoverable deformation through reversible martensitic transformation, whereas SMPs/SMPCs rely on stimulus-responsive polymer networks for shape fixation and recovery. LCEs realize programmable deformation through the coupling between liquid-crystalline ordering and elastic polymer networks, while CPCs accommodate large deformation through flexible polymer matrices, where the dynamic reconstruction of conductive networks during deformation enables electromechanical coupling and multifunctional sensing responses. Building on these distinctions, this review examines the material characteristics, response mechanisms, preparation strategies, and engineering applications of these systems. The applications of machine learning in material performance prediction, structural optimization, process regulation, and intelligent sensing are also reviewed. Finally, the key challenges and future perspectives are discussed to provide insights into the further development of large-deformation functional materials.
When bending and twisting occur simultaneously, triboelectric responses from a single sensing position can become coupled, making the two deformation components difficult to distinguish. Here, we develop a flexible three-unit triboelectric sensor array that converts position-dependent response differences into usable spatial information for discrete bending–twisting decoding. Three spatially separated units with the same material system and basic triboelectric configuration provide complementary position-dependent responses. Unit-level waveform features are extracted from each position, and explicit spatial-relation features are further constructed to represent interposition relationships. Separate classifiers identify discrete bending and twisting states within the measured state space, while multi-cycle probability fusion improves prediction stability. Across 24 mechanical states, bending, twisting, and joint-state recognition accuracies reached 97.40
Bone diseases, bone injuries, congenital deformities and other diseases seriously affect the patients’ quality of life, while total knee arthroplasty (TKA) represents an effective solution for these issues. However, the knee products utilized in the current knee arthroplasties are often mass-produced and poorly matched to individual patients. Furthermore, the three-dimensional (3D) printing technology opens new opportunities for personalized high-performance knee joints, which can address drawbacks of conventional solid implants including poor biocompatibility and high risk of loosening. In this paper, we introduce the design principle, biomechanical analysis and manufacturing progress of 3D-printed solid and porous knee joint, illustrate the advantages and limitations of 3D-printed knee joint implants in clinical settings, and look forward to the development trend of 3D-printed knee joints, including technological innovation, material innovation, and expanded clinical applications, laying a theoretical foundation for the use of 3D-printed total knee prostheses.
This study addresses the mechanical fragility, chemical oxidative damage, and environmental concerns of existing superhydrophobic coatings by developing a waterborne, fluorine-free, self-healing superhydrophobic coating. Biomass tannic acid (TA) and nonionic surfactant Triton X-100 are used to non-covalently modify carbon nanotubes (TCNTs), which are incorporated into a waterborne epoxy resin (WEP) matrix as reinforcing fillers to enhance the mechanical robustness of the framework. Together with a waterborne polydimethylsiloxane (PDMS) emulsion and octadecyltrimethoxysilane-modified silica (O-SiO2), an O-SiO2/PDMS/TWEP composite coating is constructed. The coating achieves a water contact angle of 158.6° ± 1.3° without fluorinated reagents or organic solvents in the final formulation. It maintains superhydrophobicity over at least ten damage–repair cycles and retains a contact angle of 151.1° ± 1.5° after 300 sandpaper abrasion cycles. When applied to polyurethane sponges, the resulting material exhibits an oil adsorption capacity of 47.11–86.71 g·g−1, an oil–water separation efficiency above 97
