The recovery of backwash water from sand filters by ultrafiltration

Groundwater is still one of the main sources for the production of drinking water. In the preparation of drinking water, the groundwater is first aerated and then filtered through a sand filter in order to remove Fe, Mn, NH4 and methane. After saturation, these sand filters have to be backwashed periodically. In the past, the backwash water was discharged in the sewage either directly or after sedimentation. However, the first pilot trials with ultrafiltration showed already in 1995 that this technique was very promising for the recovery of backwash water. As a result, the first full-scale ultrafiltration installation was built at the drinking water company NRE (Nutsbedrijf Regio Eindhoven, now WOB) in Eindhoven, Holland, in 1997. A Belgian drinking water company started in the beginning of 1998, in collaboration with Vito, with pilot testing on this subject. Long-term experiments were carried out at four different locations, with different water qualities. At one location the backwash water coming directly from the sand filters (bulk fraction) was filtered; at the other locations the supernatant, which is formed after sedimentation in a settler, was used as feed for the UF pilot. As a result of the sedimentation process, the Fe content is lower in the supernatant than in the bulk fraction. The experiments were done with two different membranes: X-Flow and Stork. Both membranes are operating in dead-end mode with a standard backwash procedure in a very reliable and stable way. Special attention was given to the optimization of the chemical cleaning procedure and the permeate quality in respect of colony forming units (CFU). At the last location, a set of short-term experiments was carried out in order to get more insight in the fouling behaviour of the backwash water. Therefore, different operation modes were compared (low–high fluxes, short–long filtration time), keeping the dirt load either constant or variable. These results will also be discussed in this paper.

The feasibility of ultrafiltration for the recovery of backwash water was demonstrated on three of the four locations. • At one of the locations, i.e. Hoogstraten, the presence of polyelectrolyte in the feed water in combination with the high degree of biological contamination resulted in an unstable operation mode. • At three other locations, ultrafiltration seemed to be a very reliable technique, producing a permeate of good microbiological and inorganic quality. However, if the soluble fraction of As and Mn is high in comparison with the insoluble fraction, the content of both elements in the permeate may exceed the drinking water norms since ultrafiltration removes only the insoluble fraction. Short-term experiments on the different locations will give the necessary information on the removal rate of both elements. The content of Fe and Ca in the permeate is always far beneath the drinking water norm and will in most cases be no problem. As for the microbiological quality of the permeate, it may be concluded that all harmful micro- organisms (Aeromonas, Coliforms, Streptococcus) are totally removed. However, CFU are measured in the permeate, especially after a stand still of the installation. Therefore, an additional disinfection technique, e.g. UV, may be advisable. • Stable conditions were reached on the three locations when fluxes were fixed at 100 l/h.m². Short-term experiments proved that higher fluxes (150, 200 l/h.m²) and higher recovery rates (>99,5%) are possible. However, to reach this high recovery, it is better to enlarge the filtration time than the flux rate. The former way of operation will result in a lower fouling potency towards the membranes. • The optimum cleaning procedure is hard to define and depends on the feed water quality and the way of operation. However, the best results were always obtained by using NaOCl, eventually in combination with HCl. When the membranes are fouled to a large extent, a soaking for one night in NaOCl 250 ppm will result in a total cleaning of the membranes. In the subsequent filtration process, the initial TMP will be restored. The higher the frequency of cleaning however, the larger the resulting fouling potency of the membranes. • Hardly any difference was observed between the X-Flow and Stork membranes. Both membranes showed stable fluxes and good permeate quality.