Abstract Under low-velocity impact, ammunition may suffer damage, fracture, and localized heating ignition, which can further induce combustion, deflagration, explosion, or even detonation with different intensities. Previous studies have shown that the explosive impact ignition process is a multi-scale and multi-physical field coupling problem, which is jointly determined by macro-scale explosive fragmentation and meso-scale processes such as shear crack interface friction and slip, endothermic melting, viscous rheology, thermal decomposition, and heat conduction. However, existing impact ignition models fail to effectively connect macro-scale and meso-scale processes; especially in meso-scale interface effects, there exists a problem that the fluid after explosive melting is still calculated according to solid friction, making it difficult to accurately predict the ignition time and location. To solve this problem, a meso-scale crack interface force-thermal-chemical coupling ignition model considering viscous rheology after melting was proposed in this study. First, the model improved the classical Frank-Kamenetskii model by introducing the endothermic effect of explosive melting and the viscous shear rheology of the fluid phase after melting: when the explosive temperature reaches the melting point, the heat accumulation is used for melting until the latent heat of fusion is satisfied to complete the phase transition; after complete melting, the viscous shear rheology of the fluid replaces solid friction as the main heat source, and the velocity distribution of the solid-liquid two-phase after melting is determined by the balance between viscous shear stress and frictional stress. Furthermore, the meso-scale model was secondary developed and connected with the discrete element model to realize the iterative solution of the macro-mesoscopic ignition model. The proposed model was verified by simulating the Steven experiment, and the results showed that the ignition threshold predicted by the model was 48 m/s; the ignition delay time at 50 m/s was 286 μs (error about 5.92% compared with the experimental value of 270 μs), and the ignition delay time at 100 m/s was 55 μs (error about 8.33% compared with the experimental value of 60 μs), indicating good experimental comparability. Meso-scale parameter analysis showed that: compared with the traditional friction model, the model considering viscous rheology delayed the ignition time in most cases, but accelerated the ignition in the specific range of low pressure and high shear; for narrow cracks, the ignition delay time first advanced and then delayed with the increase of crack thickness due to the competitive relationship between heat generation and heat dissipation; the increase of pressure and shear velocity shortened the ignition delay time, while the increase of crack thickness prolonged the ignition delay time. The research results provide a new method for accurately predicting the non-impact ignition time and location of explosives, and offer a new understanding of the crack friction ignition mechanism.