In order to further understand the microstructure and high temperature mechanical behavior of SiC/SiBCN-Si₃N₄ composite， and establish a scientific and reliable quantitative characterization methodology， this paper uses a variety of characterization methods to quantitatively observe SiC/SiBCN-Si₃N₄ material. Firstly， the porosity and density of the material are tested. Then the in-situ mechanical properties of the material at high temperatures were tested and the damage mechanism of the material was analyzed. Finally， an interpretable deep learning model was constructed based on the test data to realize the prediction of the nonlinear constitutive relationship of the material at high temperature. The results show that the average stress prediction error ranges from 0.27% to 0.59%， and the average strain prediction error ranges from 1.96% to 3.41%. Through quantitative analysis， it is also clear that the factors successively affecting the mechanical properties are temperature， off-axis Angle， porosity and density. In this paper， the macroscopic mechanical properties of SiC/SiBCN-Si₃N₄ under different ambient temperature， off-axis angles and external loads are predicted， which provides a new idea for the establishment of high temperature constitutive model of ceramic matrix composites.