Date of Award

January 2015

Document Type


Degree Name

Doctor of Philosophy (PhD)


Civil Engineering

First Advisor

Harvey Gullicks


Reverse osmosis (RO) is increasingly being used for water treatment because of its small ecological footprint and improved membrane technology. However, a major challenge to the application of this technology in water treatment is the irreversible fouling observed in RO membranes. Fouling, mainly caused by dissolved organic matter (DOM) and colloidal materials (CM) in water, can increase the energy and maintenance costs and decrease the permeation flux and membrane life. Different pretreatments, such as coagulation, flocculation, sedimentation, and membrane-filtration, need to be applied upstream of the RO system to remove potential RO foulants. Membrane remediation by chemical cleaning also needs to be conducted to restore the membrane water flux. The purpose of the models constructed for the treatment trains in this pilot study is to investigate and identify system-specific performance parameters. The following paragraphs will discuss the findings from the investigations conducted during the Grand Forks Water Treatment Plant pilot study.

The pilot study on pretreatment indicated that DOM and turbidity could be effectively removed using ferric chloride (FeCl3) or polyaluminum chloride (PACl) as coagulants if the pH and chemical coagulant dose were optimized. Under the optimized pretreatment conditions, the irreversible fouling of RO membranes could be reduced or mitigated. This research showed that pretreatment, including coagulation, flocculation, sedimentation, and ultrafiltration, lead to the removal of 42.2% and 59.44% of DOM on

using PACl and FeCl3 as coagulants, respectively, indicating improvement over the average baseline removal of 30% under non-optimized conditions. In addition, the removal of more than 90% turbidity (with PACl, at temperatures >20 °C; with FeCl3, at temperatures <4 °C) was achieved. PACl and FeCl3 exhibited very good removal efficiency for DOM and turbidity at doses of 40 and 50 mg/L, respectively, at pH 6.5.

In this study, a new testable neural platform prediction model was constructed for the removal of turbidity and total organic carbon (TOC) in the pilot pretreatment study at the Grand Forks Water Treatment Plant. The model accurately predicted the quantitative dependence of the effluent TOC on coagulant dose, acid dose, temperature, influent-TOC, conductivity, and total dissolved solids (TDS). Similarly, it predicted the quantitative dependence of effluent turbidity on flow rate, coagulant dose, acid dose, temperature, influent-TOC, conductivity, TDS, and total suspended solids. These analyses investigate and identify system-specific performance parameters in the pretreatment unit that are responsible for turbidity and TOC removal.

A new testable mathematical model of normalized permeability and normalized system salt passage was developed to predict the quantity and quality of the product water during the pilot study on RO systems A and D. The model constructed from RO system A data accurately predicts the quantitative dependence of normalized permeability on temperature, feed flow, system recovery, net driving pressure, and system water flux. The model constructed from RO system D data accurately predicts the quantitative dependence of normalized system salt passage on temperature, feed flow, post-recycle feed conductivity, system recovery, permeate TDS, manufacturer’s rated membrane salt passage, and system water flux. This analysis explains the manner in which fouling is caused by both physical and chemical interactions between the membrane and fouling agents.

The strong interdependence of these fundamental operating conditions and the correlation between permeability and system salt passage were confirmed when the models were tested on data collected from RO systems A, B, C, and D. Although reasonable agreement between the results was obtained when the model was tested on these four RO systems, the models slightly overestimated the permeability values and underestimated the system salt passage values for RO system B. This discrepancy may be attributed to fouling, concentration polarization, the morphology and structure of the RO membrane. Additionally, system recovery (RO B ran at 75%, RO systems A and D ran at 82%) and the increase in membrane water flux for RO systems A and D from 11 gallons/ft2/day (gfd) to 12 gfd may also be important.

An effective cleaning sequence that restores 100% of membrane performance has been demonstrated for the RO membranes. The effects of fouling on RO permeability and salt rejection were studied by comparing the permeabilities of clean and fouled membranes, and by relating the values to the cleaning sequence used for recovery. The reported results indicate that the recovery of RO membrane performance depends on the physicochemical properties of the membrane foulant, the cleaners, and the sequence in which the cleaners are used. Caustic cleaning, followed by acid cleaning, was very effective, leading to a permeability recovery of more than 100%. On the contrary, acid cleaning followed by caustic cleaning only caused partial restoration of the membrane’s ion retention ability. The use of either acid cleaning or caustic cleaning resulted in partial water flux recovery, while a combination of the two led to complete water flux recovery.