Mass Exchange Networks (MENs) are used in the chemical industry to reduce the waste generated by a plant to an acceptable level at the cheapest cost. Finding the optimal network, however, is often difficult due to the non-convexity of the mathematical representation of the problem. This paper describes a novel approach for the synthesis of MENs and MENs with regeneration using Genetic Algorithms (GA), a stochastic optimisation technique based on the concepts of natural evolution. We present an encoding for a genetic algorithm which describes a rich search space, considering both stream splitting and in-series exchangers. For a certain class of problems, all encoded solutions are feasible and require a simple evaluation to yield a cost, resulting in an efficient genetic algorithm. For other problems, the number of infeasible solutions is small, having little effect on the convergence of the genetic algorithm. In comparison with other methods, the GA presented herein is able to find better networks than have been reported elsewhere.