(610c) Real-Time Image Analysis for Automated Scale Detection and Cleaning in a Reverse Osmosis Desalination System

Bartman, A., University of California Los Angeles
Cohen, Y., UCLA

In recent years, reverse osmosis (RO) desalination has emerged as a leading method for desalting seawater, inland brackish water and water for water reuse applications. In inland desalination of brackish water and water reuse, concentrate (brine) management is a major challenge given the limited options for concentrate disposal. With increasing product water recovery, the volume of the residual concentrate stream is reduced, increasing the available options for management of this stream (i.e., treatment and disposal). Optimal product water recovery levels in brackish water desalting are highly dependent on feed water quality, target production capacity, and locally available options for concentrate disposal. Because the costs associated with managing residual desalination concentrate are typically high (especially at inland locations), high levels of product water recovery (85%-95%) are often required for optimal inland desalting operation. As the permeate recovery level increases, the level of concentration polarization (i.e., increased solute concentration at the membrane surface relative to the bulk) rises, increasing the propensity for membrane fouling and scaling. Mineral scaling can occur when the concentrations of sparingly soluble dissolved mineral salts (e.g., gypsum (CaSO4∙2H2O), BaSO4, SrSO4, CaCO3, SiO2, etc.) near membrane surfaces rise above their solubility limits. As a consequence, sparingly soluble mineral salts can precipitate in the bulk and subsequently deposit onto the membrane surface as well as crystallize directly on the membrane. Mineral scaling can lead to a significant reduction in membrane performance (e.g., flux reduction and salt rejection impairment) and shortening of membrane life, thereby increasing process cost and imposing operational limits on the achievable product water recoveries. If scale formation is detected at an early stage, membrane cleaning can be effectively accomplished via chemical cleaning, osmotic backwash or feed flow reversal. Also, if the onset of mineral scaling is detected, conservative operation (i.e., low recovery) can be imposed to ensure that mineral salt concentrations at the membrane surface are below saturation. In full-scale RO operation with fed water of high mineral scaling propensity, it is imperative to establish when mineral scale may occur and/or directly detect the onset of mineral scaling in order to determine the appropriate frequency of needed membrane cleaning. Early mineral scale detection is also instrumental in determining required adjustments in operating conditions (e.g., recovery level and antiscalant dose) to ensure process operation below the scaling threshold. In this work, the application of a novel on-line mineral scale detection system was demonstrated for early scale detection and application to RO system control. The present ex-situ membrane monitor (MeMo), with its real-time image analysis software, enables measurement of the fractional area of the membrane surface that is covered by mineral scaling, as well as the number of crystals present on the membrane surface. In order to assess the accuracy of real-time analysis of mineral scaling on the MeMo membrane surface, sequentially recorded images were analyzed by the on-line image analysis algorithm to determine the surface mineral scale coverage profile and crystal count and then were compared with analysis done "manually" off-line using the Fovea plug-in for Adobe Photoshop (Fovea Pro, Version 3.0, Reindeer Graphics, Asheville, NC). The automated image analysis program (operating either on-line, or in a post-processing mode after the experiment) provided accurate measure of the membrane surface mineral coverage and the evolution of the crystal number density on the surface, especially in the early stages of scale growth. The automated crystal detection software was designed to send a control signal after the fractional surface coverage or the crystal count reached a user-defined threshold in order to actuate an appropriate RO system control action. Through early detection of membrane scaling (i.e., before permeate flux decline is observed), the system operator can prevent irreversible membrane damage and loss of system productivity. The utility of the MeMo system was evaluated using a pilot RO system demonstrating continuous RO system operation in feed-flow reversal mode, while avoiding mineral scaling, without the use of antiscalant additives to the RO feed. The MeMo detector was also demonstrated for field use to determine recovery limits and arrive at the optimal antiscalant dose for safe scale-free brackish water RO desalting operation.