Premixed and non-premixed flames are clearly defined situations. In (fully) premixed flames a flame front propagates into a burnable mixture of spatially homogeneous fuel-air-ratio, while in non-premixed combustion the fuel-air mixing is assumed to be much slower than combustion, so combustion is purely mixing-dominated.

If those situations are not present, a partially premixed combustion model needs to be used. There are at present no universally accepted accurate combustion models for partially premixed combustion. In your case the combustion process will presumably start with auto-ignition of a diesel-premixed gas mixture at the edge of the diesel jet, it may continue as a diffusion flame around the diesel jet and a premixed flame will propagate into the premixed region. Note that autoiginition in most available combustion models will not be properly predicted.

If your RANS cannot resolve the details of the diesel jet combustion, I would suggest you use a surrogate autoignition model for the initial heat release phase and a premixed combustion model for flame propagation into the mixture.

Some research groups have achieved good results with the conditional moment closures (CMC) and also with transported PDF methods in LES simulations for autoigniting situations. Those methods are very expensive though and not as accurate when coupled with RANS methods.