Adiabatic reactor network synthesis using coupled genetic algorithm with quasi linear programming method

In this paper a new approach is presented for synthesis of adiabatic reactor networks (ARNs) using genetic algorithm (GA) coupled with quasi linear programming (LP) method. In most cases reactor network (RN) synthesis leads to mixed integer non-linear programming (MINLP) problems or complicated non-linear programming (NLP) problems. In the suggested method to prevent complexity, GA is used to find the best configuration while continuous variables are handled using a quasi LP formulation. In quasi LP model continuous variables are regrouped into (a) normal variables such as concentration, temperature and flow rate (b) variables which make the equations nonlinear such as conversion in each reactor, split ratio of streams, recycles and exit temperatures from coolers or heaters. To optimize continuous variables the quasi LP consists of two nested loops. In the outer loop variables of group (b) are optimized while variables of group (a) are obtained using known values of group (b) in the inner loop. Therefore the complex MINLP or NLP model is converted to a quasi LP which is much easier to solve. Although in the RN only continuous stirred tank reactors (CSTR) and plug flow reactors (PFR) are used. The cases studied showed that this method can reach near optimum and in some cases better solutions compared to the literature.