Ansys Employee
As a guess you've got a load of particles stuck in the boundary layer somewhere and the energy/mass source is causing hot & cold spots. That'll eventually confuse the solver and diverge.  
Assuming you're looking at wet particles in a medical throat model use less refinement near the wall (you may want deeper near wall cells for DPM) and review the mesh: there's no need to use an O-grid in a modern solver, there's a reason we did all the work on the Fluent Meshing poly & poly-hexcore methods.