Abstract: After entering deep mine,with the influence of high stress and strong disturbance,various failure modes such as breaking,shearing and compaction yielding are symbiotic in rock mass,the characteristics of nonlinear behavior are obvious,and rock failure mode is transformed in various failure mechanisms. It is necessary to analyze the influence of the intermediate principal stress on the strength theory of deep rocks and rock masses. Based on the Paul-Mohr-Cou- lomb (PMC) failure criteria,considering the transformation of rock failure mode of porous rock and the pore collapse characteristics,the Back-Paul-Mohr-Coulomb (BPMC) model for characterizing the compression-shear yielding under high stress is proposed. Representation method of the friction angle envelope are derived. According to the piecewise linear principle,a 12-sided BPMC linear failure criterion is obtained in the p-q plane. Based on the transformation re- lationship between the stress invariants in the p-q plane,the parameters of the model features are obtained with the least-square method. In order to verify the rationality of the PMC & BPMC criterion and the corresponding failure mode transformation theory,a PMC & BPMC model is fitted and compared with the experimental data. Sichuan sandstone and Jinchuan marble are used for high-pressure triaxial compression tests with the portable self-sealing rock triaxial test high pressure chamber system. Fitting the experimental data,the obtained damage envelope and the experimental data have a good fitting degree. The Sichuan yellow sandstone has a shear stress peak inflection point after the confi- ning pressure reaches 90 MPa,resulting in pore collapse type damage,which is the compaction yielding. Experiments are conducted to study the rock failure characteristics and strength characteristics under different confining pressures. The results show that the PMC & BPMC model can predict the cap yield surface of porous rock and the transition of rock failure mode from shear failure to compaction yielding,which can predict the failure strength under different con- fining pressures.