The decomposition of a carbon fibre reinforced epoxy resin with an acetone/water solvent supplied in the ratio of 80:20 v/v has been investigated using a non-stirred batch reactor. Temperatures and pressures in the range of 300–380 °C and 16–30 MPa were studied over a reaction time of up to 150 min using an initial reactor loading of 30 g_resin L_solvent⁻¹. The minimum conditions necessary for effective fibre recovery were identified as 320 °C and 20 MPa. After processing for 120 min at this temperature and pressure, up to 95 wt.% of the resin was stripped from the fibre surface and the original weave architecture was retained. Increasing the reactor loading to 90 g_resin L_solvent⁻¹ demonstrates no significant effect on the extent of resin decomposition. This, combined with imaging using X-ray computer tomography, suggests that the solvent is always in excess and that the mass transfer of the solvent into the composite and degradation products away from the surface is fast. A first order reaction rate equation based on the Arrhenius expression and a reaction-rate limited shrinking core model (SCM) have both been successfully fitted to the experimental data. From this, the parameters activation energy and frequency factor have been calculated to be 222.3 kJ mol⁻¹ and 7.64 × 10¹⁷ min⁻¹ respectively for the Arrhenius model and 164.3 kJ mol⁻¹ and 1.29 × 10¹² min⁻¹ for the SCM. The results presented in this work demonstrate that either approach can be used to predict the resin decomposition to within ±5% which is similar to the margin of experimental error.

已经使用非搅拌的间歇反应器研究了碳纤维增强环氧树脂与以80:20 v / v的比例提供的丙酮/水溶剂的分解。使用最初的30 g_树脂L_溶剂^-1加载反应器，在长达150分钟的反应时间内研究了300–380°C和16–30 MPa范围内的温度和压力。有效回收纤维所需的最低条件为320°C和20 MPa。在此温度和压力下处理120分钟后，从纤维表面剥离了高达95 wt。％的树脂，并保留了原始的编织结构。将反应器负荷增加至90 g_树脂L_溶剂^-1证明对树脂分解程度没有明显影响。这与使用X射线计算机断层摄影术的成像相结合，表明溶剂始终过量，并且溶剂向表面复合物和降解产物的质量转移很快。基于Arrhenius表达式的一阶反应速率方程和反应速率受限的收缩核模型（SCM）均已成功地拟合到实验数据中。据此，对于阿伦尼乌斯模型，参数激活能和频率因数已分别计算为222.3 kJ mol ^-1和7.64×10 ¹⁷ min ^-1，以及164.3 kJ mol ^-1和1.29×10 ¹² min ^-1用于SCM。在这项工作中提出的结果表明，这两种方法都可以用来预测树脂的分解率在±5％以内，这与实验误差的余量相似。