Boiler feed, process duty, semiconductor, pharma — the cleanest water in the plant. Each stage has multiple technology paths. Toggle the configuration; click any node to open it.
Typical operating envelopes from raw-water inlet through to point-of-use distribution. Polishing-stage targets get aggressive — measured in parts per billion, not parts per million.
Six failure families. The rule that sticks: most polishing-stage problems are really upstream feed-water problems in disguise. Test cheap upstream hypotheses first.
Tells: rising ΔP across stages, normalized permeate flow drop, salt rejection drop. Layer: feed-water quality. Fix: SDI check, antiscalant audit, CIP with acid then caustic, autopsy a lead module if persistent.
Tells: shorter run lengths between regens, conductivity creep at outlet. Layer: feed-water (organics, iron) or regen quality. Fix: resin sample analysis, increase regen dose / time, NaCl / NaOH brine wash for organic foulant.
Tells: sodium analyser climbs from ppb to ppm. Layer: cation resin or MB regen. Fix: check cation regen efficiency, MB resin separation procedure, consider EDI upgrade if recurring.
Tells: silica analyser trending up at point-of-use. Layer: anion resin or EDI feed quality. Fix: anion regen with hot caustic, EDI feed conductivity check, downstream carbon dioxide check.
Tells: distribution conductivity creeps up overnight, lower at high turnover. Layer: tank vent. Fix: install / restore N₂ blanket, vent filter integrity check, air-gap design audit.
Tells: poor regen efficiency, conductivity bump on return-to-service, pH spike at wastewater. Layer: SOP / DCS. Fix: stroke-test regen valves, verify sequence in logic, retrain on regen procedure.
Almost every stage of a demin train has competing technologies that produce roughly the same outcome. The right pick depends on feed-water variability, capex vs opex posture, HSE constraints (chemical handling), water scarcity (recovery), and the existing plant footprint. Below: every redundant technology I have seen in the field, tabulated.
UF dominates new builds; MMF still owns most installed capacity. ABF and disc filters are niche. Cartridge filters appear in every plant as a 5 µm guard before the RO high-pressure pumps regardless of upstream choice.
Required when raw water hardness is high enough to scale the RO concentrate side despite antiscalant. Often skipped on low-hardness feeds — antiscalant alone handles it. Lime + IX combined for very hard brackish feeds.
Free chlorine destroys polyamide RO membranes (lifetime in hours, not years) and oxidises IX resin. One of these MUST sit upstream of any membrane / resin contact.
New builds default to RO. IX still wins where feed TDS is genuinely low and water scarcity / reject disposal is a hard constraint. Modern plants run RO + EDI. Hybrid RO + IX appears for very high TDS feeds where IX is too costly alone and RO doesn't hit purity in two passes.
EDI has overtaken mixed bed in new builds for high-purity boiler feed water — no chemical regen, continuous service, smaller footprint. Mixed bed remains common in plants with existing IX pre-treatment, where the regen plant is already built. Separate cation / anion polishers are an older variant, occasionally retained for condensate polishing in power plants.
A demin train is a chemical-handling plant in disguise. Coagulants and dechlorinators enter the feed; antiscalants and biocides protect the membranes; regen acids and caustics regenerate the resins; CIP chemicals clean everything. This is the full list by stage with typical doses — useful for sizing, audits, and HSE conversations.
Names that show up on most major projects. Membrane and resin markets have consolidated hard over the last decade — DuPont absorbed FilmTec and Dow Water; Suez became Veolia Water Technologies.