A hot water sanitizable reverse osmosis (RO)/electrodeionization (EDI) system installed at a pharmaceutical manufacturing facility was experiencing premature failures of their EDI module. A review of the process data revealed the cause of the failures was likely due to oxidation of the EDI module. Although the USP Purified Water System continued to produce acceptable quality water, the pressure drop across the EDI module increased daily (approximately 2 psid increase per day) and there was no indication of fouling or scaling.
The upstream RO membranes were performing well, with no signs of oxidation forming. Additionally, the RO permeate quality and flow were consistent, with no changes in operating pressures. The phenomenon of oxidation of EDI modules downstream without any visible oxidation of upstream RO membranes has been witnessed at other installations in the past. The RO membranes are rated for 1000 ppm hours of chlorine exposure, while the EDI modules have minimal, if any, tolerance to chlorine or similar oxidative compounds. It is not uncommon for the EDI modules to fail prior to the RO membranes when both are exposed to chlorine. What made this case unique was that free and total chlorine levels taken prior to the RO unit, analyzed and recorded daily, were all reported as less than detectable.
What made this case unique was that free and total chlorine levels taken prior to the RO unit, analyzed and recorded daily, were all reported as less than detectable.
This particular system was designed to use sodium sulfite injection to remove residual chlorine from the feed water prior to the RO unit. The chemical was injected via a metering pump upstream of the RO prefilter. There was no in-line monitoring for chlorine or ORP and the injection was set during system commissioning based on the chlorine levels in the feed water. As mentioned, grab samples were taken prior to the RO unit and analyzed for free and total chlorine once per day.
As the investigation expanded, free and total chlorine samples were taken at other points in the system including throughout the pretreatment, downstream of the RO and prior to the EDI, and lastly downstream of the EDI unit. These additional samples were taken several times a day over a period of two weeks. The results, as discovered, were alarming. Pretreatment total chlorine values were steady at values in the range of 1.0 mg/l. The sulfite system appeared to be effective in reducing the total chlorine levels to less than 0.05 mg/ml. However, total chlorine levels downstream of the RO unit as high as 2.5 mg/ml were observed. Most concerning was that chlorine was even detected in the product water downstream of the EDI unit.
Solution and Root Cause
Illogically, the total chlorine values downstream of the RO unit could be, at times, higher than the levels in the feed water. Indeed, these same observations were made on multiple days of sampling. Yet on other days, chlorine levels in the RO product were non-detectable. It was as if the RO membranes were generating chlorine on occasion.
As an immediate and definitive fix, hot water sanitizable, service exchange activated carbon vessels were added to the system as a replacement to the sulfite injection system. The EDI module was replaced again, and no subsequent oxidation of the EDI module was observed. Chlorine values were non-detectable, over weeks of sampling, throughout all sampling locations downstream of the new activated carbon units including downstream of the RO membranes.
It was reasoned that the sulfite injection system was not working properly at times allowing for chlorine breakthrough. The sulfite pump would turn on and off subject to whether the RO was in operation. Most likely, at times, the sulfite injection pump would lose its prime when power was cycled. As sampling only occurred once per day, periodic high chlorine levels to RO membranes would go undetected.
It was reasoned that the sulfite injection system was not working properly at times allowing for chlorine breakthrough... As sampling only occurred once per day, periodic high chlorine levels to RO membranes would go undetected.
Discussions with membrane manufacturers confirmed that their research had also shown that thin-film composite polyamide membranes were capable of adsorbing free chlorine. When dechlorinated water was again fed to the membranes, the free chlorine would elute off the membranes and oxidize the downstream EDI unit. This resulted in feed water to the EDI that could be unexplainably high.
The issue here could have been avoided with on-line monitoring of chlorine or ORP in real-time downstream of the sulfite injection system or using downstream levels to automatically control the sulfite injection rate. However de-chlorination with activated carbon offered an absolute, robust process with minimal risk of chlorine breakthrough. The system has operated well since the process change was made with no loss of performance from the EDI after months of operation.