Shape memory polymers (SMPs) are currently one of the most attractive smart materials expected to replace traditional shape memory alloys and ceramics (SMAs and SMCs, respectively) in some fields because of their unique properties of high deformability, low density, easy processing, and low cost. As one of the most popular SMPs, shape memory polyurethane (SMPU) has received extensive attention in the fields of biomedicine and smart textiles due to its biocompatibility and adjustable thermal transition temperature. However, its laborious synthesis, limitation to thermal response, poor conductivity, and low modulus limit its wider application. In this work, biocompatible poly(ε-caprolactone) diol (PCL-2OH) is used as the soft segment, isophorone diisocyanate (IPDI) is used as the hard segment, and glycerol (GL) is used as the crosslinking agent to prepare thermoset SMPU with a thermal transition temperature close to body temperature for convenient medical applications. The effects of different soft-chain molecular weights and crosslinking densities on the SMPU’s properties are studied. It is determined that the SMPU has the best comprehensive performance when the molar ratio of IPDI:PCL-2OH:GL is 2:1.5:0.33, which can trigger shape memory recovery at body temperature and maintain 450% recoverable strain. Such materials are excellent candidates for medical devices and can make great contributions to human health.

The inevitable pore defects generated in the preparation process have a great impact on the mechanical properties of the ceramic matrix composites. However, the pore defects on the composites were ignored to a large extent in models established in the previous research. In this study, in order to investigate the bending damage behaviors of SiC_f/SiC (SiC fiber-reinforced SiC matrix) angle-interlock (2.5D) woven composites prepared by the precursor immersion pyrolysis (PIP) method, a more precise full-scale model of composites was established by finite element (FE) method with taking into account of random pore defects generated by Monte Carlo algorithm. Micro-computed tomography (Micro-CT) was employed to acquire the statistical data of the yarns and pores of SiC_f/SiC 2.5D woven composites. A bending test was conducted to study the damage behaviors of the composite and compared with the prediction of the FE model. The result shows that the proposed model with random pores can predict the mechanical damage behavior of SiC_f/SiC 2.5D woven composites effectively under three-point bending. The simulated bending strength shows a good agreement with the experimental data, with a relative error of approximately 4.6%.