In the present paper an algorithmic procedure for generating innovative reactor network designs for complex reaction schemes is presented. Computational methodology is derived from the basic principles of reaction and mixing. Candidate reactor networks are generated by expanding the solution space into regions where different levels of mixing are required. Directions of extending the design space are identified by using formal optimization theory. A set of candidate reactor networks is generated in a finite number of the solution space extensions. The derived methodology eliminates the requirement for an postulated superstructure and the use of qualitative knowledge for finding the optimal solution. With certain modifications the proposed design procedure can also be used to support tile design of batch processes. The generated reactor structures are truly optimal. The proposed methodology is illustrated by an example of integrating chemical reactors into the overall process.