CAD Technique for Microwave Chemistry Reactors with Energy Efficiency Optimized for Different Reactants

─ Upgrading successful processes of microwave-assisted organic synthesis to the level of industrial technology is currently slowed by difficulties in experimental development of largescale and highly-productive reactors. This paper proposes to address this issue by developing microwave chemistry reactors as microwave systems, rather than as black-box-type units for chemical reactions. We suggest an approach based on the application of a neural network optimization technique to a microwave system in order to improve its coupling (and thus energy efficiency). The RBF network optimization with CORS sampling introduced in our earlier work and capable of exceptionally quick convergence to the optima due to a dramatically reduced number of underlying 3D FDTD analyses, is upgraded here to account for an additional practically important condition requiring optimal design of the reactor for different reactants. Viability of the approach is illustrated by three examples of finding the geometry of a conventional 99% energy efficient microwave reactor for 3/3/6 different materials; with 1/5/1 liter reactants, seven-parameter optimization yields the best configurations taking only 16/38/115 hours of CPU time of a regular PC. Index Terms ─ FDTD simulation, microwave reactor, neural network, optimization, scaling up.