Modular simulation strategies have gained an increasing interest during the last years. However, most authors considered only lumped systems and simple transfer laws for the connection of phases. The methods developed are not well suited for spatially distributed systems. Some shortcomings are identified and a new concept is introduced. It takes into consideration the difference between normal and orthogonal boundaries of submodels. Each part of the process which may be represented seperately is viewed as a submodel. In addition to their interior representation, the submodels define the state variables and the fluxes at the boundaries as unknowns, thus forming an underdetermined set of equations. The missing equations are given by the connections between the submodels, which follow from the conservation of mass, energy and momentum and the equality of state variables at the boundaries. The new concept can be transformed into an easy and flexible algorithm which forms the basis of the simulation program MODUSIM which is briefly discussed.