A globally convergent mathematical model for synthesizing topologically constrained water recycle networks

This paper develops a mixed-integer linear program (MILP) to minimize fresh-water consumption in process plants by optimal recycle/reuse of process water. Prior work in this area ranged from insight-based graphical methods (pinch analysis) to mathematical programming. However, both approaches have drawbacks in incorporating topological constraints, e.g. constraints arising from capital cost considerations to limit network complexities, interconnections or forbidden source–sink matches. Mathematical programs often render themselves to MINLPs when incorporating topological constraints. To obtain reliable solution to the MINLPs, researchers often employed evolutionary search-based (stochastic) methods. Stochastic methods are computationally tedious and only render ‘near-optimal’ solutions to topologically constrained water-recycle networks. This paper will use some basic linearization of products of binary–continuous variables to cast the MINLP problem into an MILP formulation, thereby guaranteeing global optimality at significantly reduced computational burden. The computational advantage of the proposed MILP model over Genetic Algorithm (GA) and Particle Swarm Optimisation (PSO) will be demonstrated with the help of industrial case studies.