This work presents a new synthesis approach for the design of chemical reactors. The proposed method is based upon a general system representation which assumes the form of a superstructure network with different types of reactors and interconnections among the units. The superstructure provides the domain of a stochastic optimisation search which is achieved using Simulated Annealing. The search is based upon the successive generation and evaluation of reactor configurations in which changes of stream locations, unit sizes and operating conditions are addressed within a statistical cooling schedule. The application of the annealing algorithm enables uphill and downhill moves and represents a reliable optimisation procedure to bypass local extrema and limitations arising from nonconvexities and multiplicities which regularly appear in complex reaction processes.