Evaluation of membrane distillation for the treatment of surface water during emergency situations

Abstract

The continued pollution of water bodies and the possibility of acute pollution events such as the collapse of tailing dams are problems faced by the modern society. Aside the environmental effects of it, this severe contamination can difficult or even impair the surface water treatment to public supply, which leads to a state of emergency during certain situations. As a short-term solution to at least diminish the chaos provoked by the water shortage, the establishment of advanced water treatment routes based on sustainable principles is a feasible alternative. Membrane distillation (MD) is a technology reported by the literature as a robust and low energy demand (when associated to alternatives sources of heating) separation process. Having in sight the Brazilian context of wide solar availability and the potentiality of solar-driven membrane distillation systems, this work aimed to evaluate technical aspects of MD treating surface water contaminated with tailings from an iron mining dam. Aspects such as feed pretreatment, temperature and flow rate were assessed in bench-scale, as well as the impacts of different cleaning agents and membrane ageing. In these cases, it was observed that for feeds whose turbidity was above 300 NTU, a pretreatment should be carried out, and ultrafiltration is the most suitable technology from the technical point of view. In addition, at 70°C and 0.18 m/s, the system presented its best energy efficiency without compromising the permeate quality. Regarding the cleaning agents, deionized water at 60°C demonstrated the best cleaning efficiency among the agents tested, being able to restore the membrane initial permeate flux without compromising its integrity. Furthermore, the mass and heat phenomena that took place in the system were mathematically modelled, in order to assist the design of a semi-pilot membrane distillation module. The models presented an accuracy of at least 80%, and similar results were found the the semipilot simulation, which indicates that the model remained valid for a module with a total active area of 0.3m2