A new method called KESOPS - Knowledge-based Evolutionary Structure-Optimizing Process Synthesis - is presented, aiming at an automated generation of structure-optimized solutions for a given chemical engineering task treated on an abstract thermodynamic level. It is based on the use of process synthesis knowledge combined with an optimizing technique belonging to the field of evolutionary algorithms. Randomly chosen search steps are guided by a thermodynamic likelihood derived from a knowledge-based system containing process synthesis heuristics and thermodynamic knowledge, In contrast to other approaches dealing with process optimization this does not impose absolute limits to the solution space. Although in this conceptual design phase a detailed cost analysis cannot be performed for each process idea, evaluations of the generated process alternatives have to take place during the synthesis and optimization process. Therefore an assessment technique is presented which relies only on thermodynamical means and physical boundary conditions. The results achieved with the derived target function reveal a good correspondence to economical analysis made by other authors.