Completed Projects

One Step Membrane Filtration: A Fundamental Study

Amir Hoseen Haidari (TUD)

Supervisors: Walter van der Meer (TUD/UT, promotor), Bas Heijman (TUD, copromotor)

Funded by Vitens Water Company

Defense date: September 13th, 2017

The PhD thesis can be downloaded here

Summary of the PhD thesis

This study focuses on spiral-wound membrane (SWM) modules, which are the most common commercially available membrane modules for reverse osmosis (RO) and nanofiltration (NF).

While RO membranes can remove almost all kinds of substances from the feed water, they are usually equipped with pretreatment steps for conditioning and modifying the feed water to prevent clogging and fouling of these modules. Energy consumption, fouling and concentration polarization are considered as the primary challenges in these types of RO modules. These challenging factors depend of the feed water quality and they are related directly or indirectly to the design of applied feed spacer in SWM modules. A feed spacer provides a channel between two envelopes from entrance to outlet of a module to let the water flows tangentially over the membrane surfaces from the feed to the concentrate side. Additionally, feed spacers are designed to destabilize the concentration polarization layer; and thereby increase the mass transfer in SWM modules of RO. However, application of feed spacers to efficiently mix the flow comes at the expense of higher energy consumption and fouling formation. Therefore, it is important to understand the hydraulic conditions inside the spacer-filled channels such as those encountered in SWM-modules of RO.

Previous RO-studies related to production of drinking water are primarily performed with the assumption that RO-elements are used for desalination. In contrast to this assumption, the most commercially available configuration RO elements, SWM modules, gained more attention to be used for purification of freshwater resources currently. The main reason of using RO for freshwater purification is that it provides an effective barrier against the continuously emerging micro- and nano-contaminants, which cannot be (easily) removed by conventional treatment technologies. The energy use, scaling and retention in RO are influenced by the concentration polarization. When RO is applied on freshwater, the effects of concentration polarization on the energy use become significantly smaller because the osmotic pressure difference is negligible.

This thesis is divided into two parts. In the first part, this thesis describes a unique brackish water pilot study, which operates without chemical pretreatment. Such pilot study is referred to as one step membrane filtration (OSMF) system. An OSMF-system is an NF/RO-system in which membranes are applied directly on the feed water and operate without chemical pretreatment. In the second part, this thesis focuses on the hydraulic conditions of spacer-filled channels and the role of spacer geometry and orientation thereon. The effect of feed spacers on hydraulic conditions is investigated by studying the actual velocity profiles occurring in the spacer-filled channel and the relation between energy losses and spacer geometry.

Particle image velocimetry (PIV) technique is used to determine the orientation and configuration effects of spacers on the actual velocity profiles. PIV is a non-invasive and powerful tool to achieve high-resolution velocity profiles experimentally, which can be used for verification and validation of numerical studies related to spacer-filled channels. 2D and 3D numerical studies have contributed significantly to our understanding of hydraulic conditions in SWM modules of RO. However, these numerical simulations are usually validated with low-resolution experimental methods. PIV measurements from this study, thus, can provide high-resolution experimental data (7.4£7.4¹m2) for validation of these numerical studies.

The first results in PIV technique is a simultaneous velocity profile, which is obtained by using each two taken frames at a determined time interval. In this thesis, the simultaneous velocity profiles are used to investigate the variation of temporal velocity at certain locations (points) inside a mesh of a spacer. The spatial velocity profiles that are discussed in this thesis obtained by computing the average of related simultaneous velocity profiles. Chapter 3 describes a detailed explanation about the experimental methods used in this thesis. Summary of chapter 3 is repeated in the experimental section of chapters 5, 6 and 7.

The effect of feed spacers on hydraulic conditions is investigated by comparing an empty channel with a spacer-filled channel, which was filled with a commercial feed spacer (chapter 5). The flow in the empty channel was in a straight line from inlet to outlet and it was steady compared to the flow in the spacer-filled channel. The spatial velocity profiles of spacer-filled channels showed a bimodal shape with a peak at low-velocity ranges and a peak at high-velocity ranges. The low-velocity regimes occurred mostly in regions close to the filaments and a high-velocity regimes occurred at the narrowed parts of the channel or directly after the narrowing parts. The difference between low and high velocity regimes with regard to the velocity magnitude and frequency was higher at a higher flow. Although the increase in velocity magnitude causes generation of a higher shear, it is not necessarily beneficial. That is because the optimal flux as the results of destabilization of the concentration polarization layer will be achieved at a specific velocity. A further increase of the velocity from this optimal velocity will have only a marginal effect on destabilization of the concentration polarization layer, enhancing the flux and consequently production increase.

The hydraulic conditions inside a spacer-filled channel are influenced by configuration as well as orientation of the spacer. The pressure drop, which is measured during this thesis inside the channels with commercial feed spacers was in a good agreement with pressure drop from previous mathematical models. Previous studies reported a lower pressure inside the channels with the cavity spacers than the channels with zigzag spacers. The zigzag or net-type configuration is the common configuration used in SWM of RO. In the zigzag configuration, two layers of filaments with equal average diameter lay on top of each other and make an angle of 45o with each other (hydraulic angle) and with the flow (flow attack angle). In the cavity configuration, the diameter of transverse filaments is smaller than longitudinal filaments. Cavity spacers in this study had a flow attack angle of 135o. The biggest ratio of transverse filaments’ diameter to channel height was about 0.6 with cavity spacers. With cavity spacer used in this thesis, the flow was mainly in a straight line from inlet to outlet. The flow disturbance in channels with cavity spacers was at down- and upstream of transverse filaments. The flow acceleration over the transverse filaments of examined spacers was greater for cavity spacers with bigger transverse filaments. In channels with cavity spacers, the velocity was clearly higher at the channel side without transverse filaments than the channel side with transverse filaments. In channels with zigzag spacers, the flow close to the membrane was in the direction of filaments attached to the membrane, i.e. the direction of flow at the top of channel was perpendicular to the direction of flow at the bottom. The flow pattern at the middle of the channel was a combination of flow patterns at the top and bottom. The greater friction losses that were found for spacers with bigger relative height and smaller aspect ratio indicate that the orientation and geometry of transverse filaments contribute to pressure losses. However, it was not possible to find a reliable correlation for predicting the pressure losses based on the geometric characteristics of feed spacers in experimental conditions.

Effect of feed spacer orientation on the flow investigated by using the same commercial spacer at two different flow attack angles. For this purpose, three feed spacers are used with different thickness. The thickness of the top and bottom filaments was the same for each spacer. Pressure drop in the channel with the thinnest spacer was clearly higher at normal orientation (with a flow attack angle of 450) than at ladder orientation (with a flow attack angle of 90o). The difference between two orientations in the channels with thicker spacers was insignificant. The difference between the lowest and highest the velocity was greater in ladder orientation than zigzag orientation. Commercial ladder spacers with the characteristics described in this thesis are more sensitive to fouling than zigzag spacers because in ladder spacers the velocity becomes virtually zero at the side of the channel where transverse filaments are attached to the membrane.

Copyright (c) 2017 by A.H. Haidari