Knowledge-driven reactor network synthesis and optimisation

The limitations of existing methods for reactor network synthesis, including the more robust stochastic optimisation based methods, to cope with complex reaction schemes involving highly non-linear kinetics and multiple reactions, requires a novel approach to the problem. This paper uses knowledge derived from fundamental kinetic information to compose design rules representing the dominant design trends that lead to high system performance. This is the basis of a customised optimisation algorithm that features rule-based move selection to guide optimisation towards the most promising spaces, achieving more effective knowledge-based decision making. Results show optimal solutions obtained for an illustrative example agree with published literature whilst achieving better convergence compared to standard stochastic optimisation-based methods.