| This thesis develops design strategies for chemical reactor network synthesis. The strategies chosen are optimization based. It addresses the problem: "given the kinetics of a chemical reaction, what network of idealized reactor models (plug flow, continuous stirred tank and recycle reactors) will maximize a performance criterion, such as yield of a species, selectivity and the operating cost of a reactor?"; Heuristic approaches to this problem do not guarantee optimality. On the other hand, geometric or graphical techniques are limited to systems that have no more than about three reacting chemical species. Hence, there is a strong motivation for an algorithmic approach.; In this work, two approaches have been taken to address the synthesis problem, namely the superstructure and the target approaches. In the superstructure approach, a generalized network of reactors is postulated. The structural parameters (for example, recycle ratios, residence times, feed splits, etc), which provide the degrees of freedom in the system, are used as decision variables in a gradient based optimization algorithm. This algorithm employs successive quadratic programming, with adjoint variables providing gradient information. The optimal values of the decision variables are then used to extract the optimal subnetwork of reactors from the superstructure.; The second approach, namely the target approach, uses the systems or population balance approach to define a residence time distribution function (RTD) and a micromixing function (h) using an intermediate mixed reactor model. The strategy then is to formulate the reactor synthesis problem as an optimal control problem in which the RTD and the h profiles are determined in an optimal fashion, in order to maximize the given objective function. Using the RTD and h profiles, one may in certain cases, derive the optimal reactor network in terms of PFR's CSTR's and recycle reactors.; The main contributions of this research are: (i) posing the reactor synthesis problem as a superstructure based optimization problem that can handle complex reaction kinetics and a general objective function, (ii) constructing all inclusive superstructures, (iii) postulating the target approach which provides an independent measure of the objective function and, (iv) deriving adjoints, an effective way for getting the gradients. |
