The prediction of fiber reinforced ceramic is one of the most important procedure when investigating the application of ceramic composite. Numerical simulations were applied and a novel model was brought out in this paper. Firstly, four different models for predicting thermal conductivities of unidirectional fiber reinforced materials were compared, which include the Rayleigh,LN,ST and TE model,. It shows that Rayleigh model and LN model have good precision only in low fiber volume content cases. There existed big differences between the experimental and numerical results if predicted the high fiber volume content with either these four models. Then a novel model based on LN model was studied with the correction of the representative volume element method. Further comparison results indicate that the error can be reduced as 55.6% with this novel model. At the same time, the longitudinal (k11) and transverse (k22) thermal conductivities predicted by the novel model were also analyzed. It was found thatk11had a linear relationship with fiber volume fraction and thermal conductivity ratio (p). Butk22had a nonlinear relationship with fiber volume fraction, which increased much greatly when fiber volume fraction increasing at high fiber volume fraction andp>1.
