Two Models for Estimating the Effective Thermal Conductivity (kE) of Multi-Phase Materials Are Comparatively Investigated. the First Model Is the Effective Medium Approximation (EMA), which Is Based on the Extension of the Percolation Theory. the Second Is the Randomly Mixed Model (RMM), a Numerical Method in which All Components Are Seen as Cubech_cubecucube in Shape and Are Randomly Dispersed inside the Space. Two Models Can Be Directly Applied to Multi-Phase Media without Empirical Parameters. Compared with Experimental Data of Food Materials in the Literature, Two Models both Give Good Estimations of the Temperature-Dependent KE.