Process Systems Design Framework for Resource Recovery from Wastewater

A surrogate-based superstructure optimisation framework is presented and applied to the design of optimal flowsheets for the recovery of resources from wastewater. The process systems design framework involves training artificial neural networks (ANNs) using data sampled from commercial simulation software, where the sampling strategy incorporates Sobol sequences and support vector machines, ensuring good feasible design space coverage. A mixed integer linear programming (MILP) problem is formulated to solve the design problem for a set of optimal flowsheets highlighting the trade-offs between economic and environmentally focussed objective functions. However, despite the formulation of the MILP problem guaranteeing globally optimal solutions, this assurance comes at the expense of errors in the ANNs. These errors could become considerable for large design spaces, so exploration of the trade-off between optimality and accuracy is highlighted as a direction for future work.