Effect of MLSS on flux – MLSS paradox

  1. In immersed membrane filtration

It had been thought in early days of MBR that flux could be raised by lowering MLSS. Based on this assumption, groups of researchers tested membrane performance at a reduce MLSS. The most extreme case of low MLSS test was performed by filtering secondary effluent that contained very low MLSS such as <50 mg/L. However, it was concluded that solids settlings before membrane filtration had no significant benefits for increasing flux.

In one experiment  (Chang, 2005), side-by-side tests were performed with and without a clarifier in between aeration tank and membrane tank. The MLSS of the feed water to the membranes tanks were 100 mg/L and 6,000 mg/L, respectively, with and without a clarifier. Immersed hollow fiber membranes made of polysulfone were used for both trains and scouring air was supplied underneath the membrane bundles. During the 7-day experiment, significantly more permeability loss was observed in the low MLSS train as shown in Fig. 1. The rfrc (see equations here), which represents the filtration resistance caused by membrane fouling, was estimated at 94.5×1012/m and 46.2×1012/m, respectively, with and without a clarifier. Virtually identical observations were also made by Hong (2002).

MLSS e16
Fig. 1. Flux declines as a function of operation time; Reactor 1 indicates the control experiment without settling while reactor 2 indicates experiment with settling. The y-axis indicates normalized flux. (Chang, 2005)

Similar phenomenon was also observed when the effects of suspended growth and attached growth were compared (Lee, 2001). While the MLSS of control MBR was ~3,000 mg/L (suspended growth), the MLSS of the MBR with media (attached growth) was maintained at only ~100mg/L. The TMP increasing rate at lower MLSS was 7 times higher than that in the control MBR. The scanning electron microscopy (SEM) images confirmed the membrane in the control reactor had thicker cake layer although it showed higher permeability than that in the MBR with media. It was concluded that the lower flux at lower MLSS was caused by a thin and dense cake layer formed by small biopolymers.

The unusual relation between MLSS and flux can be explained by dynamic membrane theory. As Imasaka (1993) suggested, initially formed cake layer with large particles can form relatively porous cake layers that can hinder further particle deposition. In above experiment, the relative scarcity of large particles in settled mixed liquor is likely responsible for the quicker flux decline in spite of the low MLSS. Other potential explanation is that the large amount of particles in high MLSS condition can directly scour membrane surface while moving upward, but evidences are yet to be found.

It is not clear what is the lower limit of MLSS that does not affect membrane performance negatively. Perhaps the lower MLSS limit is affected by other operational parameters such as F/M ratio, SRT, DO, etc, but it appears somewhere around 3 – 5 g MLSS/L. Below this range, MLSS might influence membrane performance negatively.

  1. Rational of using dead-end membrane filtration for tertiary treatment

As discussed above, the performance of immersed membranes rarely improves by lowering MLSS assuming other operational parameters such as DO, pH, SRT, F/M ratio, etc. are in reasonable ranges. It is partly because the major membrane foulants are not the over-micron sized particles that can be scoured easily by aeration, but the sub-micron sized fines originated from dead microorganisms and biopolymers. Since the fines concentrations are rather independent from MLSS, low MLSS does not guarantee low membrane fouling.

If MLSS is sufficiently low, e.g. 10-30mg/L or lower such as in secondary effluent and surface water, permeability loss of membrane occurs mainly by fines rather than by large particles even in dead end filtration. Since fines have inherently low back-transport velocity as discussed in Fig. 1 here, the efficacy of aeration in preventing membrane fouling is not high. As a consequence, aeration becomes less and less effective as MLSS decreases. Instead, allowing all the fines and large particles deposit on membrane surface by operating the membrane unit at dead-end mode and backwashing it periodically might be more efficient way.

Due to the low level suspended solids in secondary effluent, i.e. typically 10-30 mg/L or lower,  cake layer build up is slow enough to sustain the filtration for 5-20 minutes without air scouring. Membranes are backwashed whenever TMP reaches a preset level, e.g. 0.3-0.6 bar, or whenever cycle time ends. The typical average flux in the tertiary filtration runs at dead-end mode is reported at 22-45 LMH (Côté, 2004, 2005), which is around 30-50% higher than the flux in iMBR.