Chemical Dosing System Design: Pumps, Sizing and Control

What is a chemical dosing system and what must the design deliver?

A chemical dosing (or metering) system delivers a measured mass flow of a treatment chemical into a process stream at a controlled rate, concentration and point of injection. In water and wastewater treatment the same architecture recurs across duties: coagulants (ferric chloride, ferric sulphate, aluminium sulphate, polyaluminium chloride/PAC), polymers (anionic, cationic and non-ionic polyacrylamide flocculant aids), acids and alkalis for pH correction (sulphuric or hydrochloric acid, sodium hydroxide, lime), and oxidants/biocides (sodium hypochlorite, hydrogen peroxide, sodium bisulphite for dechlorination).

Regardless of chemical, a competent design must guarantee five things: accuracy (the delivered dose matches the setpoint across the full flow range), turndown (usable control at both minimum and maximum plant flow), material integrity (every wetted part resists the chemical for its service life), adequate mixing (the reagent is dispersed into the bulk before the next unit operation), and containment (a spill or leak is caught, not released). A skid that meters perfectly but injects into a dead zone, or that is dimensionally correct but built in the wrong elastomer, is a failed design. The engineering therefore spans hydraulics, reaction and mixing theory, chemistry and safety in one package.

The physical building blocks are a bulk store or IBC, a day/dosing tank (often with make-up and dilution), one or more metering pumps, suction and discharge pipework with isolation, calibration and relief, an injection quill, a mixing device (static or mechanical), instrumentation (flow, pH/ORP, level) and a bund. Design proceeds from the mass balance outward: fix the dose, then the pump, then the make-up, then the mixing, then the materials and containment.

Why the dose must be delivered as a precise mass rate, rather than "about right", comes down to the chemistry each reagent drives. Coagulant dosing destabilises colloidal particles by compressing the electrical double layer and by charge neutralisation of the negative surface charge that keeps colloids dispersed — the classical DLVO picture in which the net inter-particle potential is the sum of van der Waals attraction and electrostatic repulsion. Underdose and the repulsive energy barrier remains, so particles stay dispersed and the downstream clarifier or filter fails; overdose and the surface charge reverses, re-stabilising the colloid and wasting reagent while depressing pH. The dosing system is therefore the instrument that places the water at the narrow optimum on that titration-like curve, and a ±20 % error in delivered dose can be the difference between a clear and a turbid effluent. The same premium on accuracy applies to pH correction, where the neutralisation curve is near-vertical around pH 7, and to oxidant dosing, where residual and by-product formation both scale with dose.

How do you size the dose rate and select the metering pump?

The starting point is a mass balance. The mass of chemical required per unit time equals the plant flow multiplied by the target dose concentration. If the product is not 100 % active, you divide by its strength (mass fraction) and density to convert a required active mass into a volumetric pump duty.

Once the neat duty is known, the pump is selected on three axes: flow range, discharge pressure, and turndown. The nominal duty should sit near the middle of the pump curve so that both low- and high-flow excursions remain controllable. Turndown matters because a metering pump loses accuracy at the extremes of its stroke; a pump quoted at 10:1 turndown holds ±1 % linearity only over the usable band, so specify the duty against the usable range, not the theoretical maximum.

Three pump technologies dominate. Solenoid or motor-driven diaphragm metering pumps are the default for clean chemistries (acids, alkalis, hypochlorite, coagulants) — accurate, cheap, self-priming to a degree, and easily controlled by stroke length and stroke frequency. Peristaltic (hose) pumps excel where the fluid is abrasive, gassing (sodium hypochlorite off-gasses and vapour-locks diaphragm heads) or a shear-sensitive polymer, because the fluid only touches the hose and the pumping action is gentle and self-priming. Progressing-cavity (progressive cavity) pumps suit viscous or high-flow duties such as lime slurry and neat polymer, giving a near-pulseless flow at higher capacity. Choosing the right metering pump and dosing skid package for the chemistry is the single decision that most often determines reliability.

Reliability and redundancy are part of the sizing, not a later thought. For any dose that a discharge consent or a downstream process depends on — coagulant ahead of a clarifier, acid or alkali on a final pH-correction stage, dechlorination before discharge — a duty/standby pump pair with automatic changeover is normal practice, so a single check-valve failure or diaphragm rupture does not stop treatment. Where two chemicals must be dosed in a fixed ratio (for example coagulant and polymer, or acid and a scale inhibitor) the pumps are interlocked so that loss of one dose stops or alarms the other, preventing a half-treated stream from passing forward. Sparing the store the same way — keeping a made-up or bulk reserve above the minimum working level — guards against a delivery slipping while the plant keeps running.

The pump hydraulics deserve attention beyond the nameplate flow. A reciprocating diaphragm pump delivers a pulsating flow: each stroke is a sharp displacement followed by a suction refill, so the instantaneous flow can be several times the mean. On long or small-bore discharge lines this raises acceleration head and can starve the pump on suction, so a pulsation damper and adequate suction line sizing are specified where accuracy or dosing into pressurised mains matters. Net positive suction head must be checked for volatile or hot liquors and for viscous polymer; a flooded suction (tank above the pump) is the safest arrangement. Check-valve reliability governs long-term accuracy, because a coagulant that crystallises or a hypochlorite that off-gasses will hold a valve fractionally open and let dose drift — which is precisely why calibration is a designed-in feature rather than a commissioning afterthought.

Control strategy is chosen alongside the pump. Manual/fixed dosing suits a steady flow. Flow-proportional (flow-paced) dosing scales the pump output to a plant flowmeter signal (4–20 mA or pulse), holding a constant mg/L irrespective of flow — essential for coagulant and polymer where the dose is tied to solids load. Feedback (closed-loop) control trims the dose from a measured process variable: pH for acid/alkali neutralisation, ORP for oxidant/dechlorination, residual chlorine for disinfection. The most robust neutralisation designs use feed-forward plus feedback (flow-pace the bulk of the reagent, then trim on pH) because the pH titration curve is steeply non-linear near neutrality and pure feedback alone tends to hunt.

The non-linearity is worth quantifying, because it dictates the control hardware. Since pH is the negative logarithm of hydrogen-ion activity, the reagent demand per pH unit changes by orders of magnitude across the curve: correcting a strong acid effluent from pH 2 to pH 4 consumes far more alkali than moving pH 6 to pH 7, yet the last fraction of a pH unit near neutrality is where the loop gain is highest and overshoot is most likely. Robust neutralisation designs therefore stage the correction — a coarse dosing pump for the bulk of the demand and a fine trim pump for the final approach — and may use two or three tanks in series so each vessel handles only part of the curve. Loop tuning must account for the transport delay between injection, mixing and the pH electrode; too short a mixing zone or a poorly placed electrode injects dead time that forces the controller to be detuned and sluggish. This is the clearest illustration of why the pump, mixing and instrument are one design, not three procurements.

Worked example: metering pump duty from a target dose

Design a ferric chloride coagulant dosing pump for a works treating a peak flow of 120 m³/h, with a jar-test-derived target dose of 45 mg/L as delivered product (a common way coagulant doses are quoted), using 40 % w/w ferric chloride of density 1,410 kg/m³. Work the duty through step by step.

• Active vs product basis: the jar test here fixed the dose as 45 mg/L of the 40 % product, so w_a is already embedded in the dose figure. Compute the product mass rate directly: 120 m³/h × 45 g/m³ = 5,400 g/h = 5.4 kg/h of product.
• Volumetric duty: Q_dose = 5.4 kg/h ÷ 1,410 kg/m³ = 3.83 × 10⁻³ m³/h = 3.83 L/h of neat 40 % ferric.
• Pump selection: a duty of 3.83 L/h at, say, 3 bar discharge suggests a metering pump with a nominal 8–10 L/h maximum, so the duty sits near mid-range (~45 % of stroke) with headroom for higher dose or flow. With 10:1 turndown the usable minimum is ~1 L/h, covering low-flow periods down to ~30 m³/h at the same 45 mg/L.
• Cross-check the active dose: 45 mg/L of 40 % product = 18 mg/L as FeCl₃, i.e. ~6.2 mg/L as Fe. That is a sensible coagulant dose for a moderately loaded water, confirming the jar-test figure is physically reasonable.

The result: a nominal 8–10 L/h diaphragm metering pump, flow-paced to the works flowmeter so the 45 mg/L holds across the diurnal flow swing, delivering 3.83 L/h at design peak. This is the first-pass duty; confirm the coagulant dose seasonally by jar testing, because raw-water quality drives the required dose more than any other variable. For the coagulation chemistry that sets this dose, see our coagulation and flocculation guide.

Two operational cross-checks finish the calculation. First, examine the turndown demand the plant actually imposes: if the works flow swings from 30 to 120 m³/h at a constant 45 mg/L, the pump must run from 0.96 to 3.83 L/h — a 4:1 range that sits inside a 10:1 pump but would fail a pump of only 3:1 turndown, so the flow envelope, not just the peak, sets the model. Second, sanity-check the injection concentration profile: a neat product dosed as a thin thread into a fast pipe disperses only if the mixing zone is adequate, which is why the pump duty and the static-mixer design (next worked example) are settled together. Where the neat dose were much smaller — say a trim polymer at fractions of a litre per hour — the answer would be to dilute the product so the pump runs in an accurate part of its range and the reagent disperses without a dead thread down the pipe wall.

How do you design solution make-up, dilution and day-tank storage?

Many reagents are dosed diluted rather than neat, and polymers must be made up from powder or emulsion. Dilution improves dosing accuracy (a larger volumetric rate is easier to meter precisely and to disperse), but it also introduces a make-up system and, for polymers, an ageing/maturation requirement.

The distinction between neat and diluted dosing is worth making explicit before looking at polymer. Neat dosing injects the product at delivery strength and is chosen for compactness and to avoid a make-up water supply; it suits stable, non-viscous reagents whose neat dose rate is already large enough to meter accurately, such as ferric or hypochlorite on a mid-sized works. Diluted dosing pre-mixes the reagent to a lower working strength in a make-up tank, which raises the volumetric pump duty into a more repeatable range and disperses the reagent more readily, at the cost of an extra tank, water treatment for the make-up, and a shorter usable life for the dilute solution. The rule of thumb is to keep the neat pump duty above roughly a fifth of the pump maximum; below that, dilute so that a metering error of one stroke is a small fraction of the dose rather than a large one.

Polymer make-up is the exacting case. Dry polyacrylamide must be wetted without forming gel "fish-eyes", then given time for the coiled chains to hydrate and extend so the charge sites become available. Typical make-up concentrations are 0.1–0.5 % w/v for dry polymer and around 0.5–1 % for emulsions, with an ageing time of 30–60 minutes in a two- or three-compartment maturation tank before use. Made-up solution has a limited shelf life (often <24 h) because the polymer slowly degrades; oversized batches waste product. Emulsion polymers additionally need an in-line disperser to invert the emulsion. Dilution water hardness and chlorine can damage some polymers, so dechlorinated, low-hardness water is preferred.

Acid and alkali are frequently dosed neat for compactness, but concentrated sulphuric acid diluted in a make-up tank generates significant heat and must be added acid-to-water, never water-to-acid; caustic soda similarly liberates heat and can crystallise below ~12 °C at 50 % strength, so trace heating or a lower delivered strength is specified for cold plant rooms. The dilution decision is a genuine trade-off. Diluting a reagent multiplies the volumetric dose rate, which pushes the pump duty into a more accurate and repeatable part of its stroke range and makes the reagent far easier to disperse in the mixing zone, but it adds a make-up water supply, a larger tank, and a fresh contamination and biofouling risk in the dilute solution. As a rule, dilute where the neat dose would be so small that the pump would run near the bottom of its turndown, or where the neat product is too viscous or aggressive to inject as a thin thread into the flow.

Storage volume is sized from consumption and a target autonomy (days between deliveries or make-up batches). The design rule is straightforward but must include working freeboard and a minimum residual so the pump never runs the tank dry.

Level instrumentation, low-level pump cut-out, high-level overfill protection and an anti-siphon/loss-of-containment strategy on the fill line complete the storage design. A dosing pump that continues to stroke against an empty tank loses prime and, on gassing chemicals, can vapour-lock — another reason low-level interlocks are not optional.

Autonomy is an economic as much as a technical choice. Too small a store means frequent deliveries or make-up batches, more manual handling, and greater exposure to a supply interruption; too large a store ties up capital, occupies bunded floor area, and — critically for reactive or biologically active reagents — risks the chemical ageing or supporting microbial growth before it is used. Sodium hypochlorite is the archetype: it decomposes steadily, losing available chlorine faster when warm or exposed to light, so a large hypochlorite store sized for convenience can deliver a materially weaker solution than the label strength, quietly shifting the true dose. For such reagents the storage design should favour shorter autonomy, cool and shaded siting, and a strength check built into the dosing calculation. For stable reagents such as caustic or ferric, longer autonomy is usually the more economic answer, subject to bund area and delivery logistics.

Worked example: day-tank sizing from consumption

Size a sodium hydroxide day tank for a pH-correction duty consuming, on average, 4.5 L/h of 30 % caustic, with a target autonomy of 3 days (72 h) between bulk top-ups and a usable fraction of 0.85.

• Consumed volume over autonomy: 4.5 L/h × 72 h = 324 L.
• Gross tank volume: V_tank = 324 ÷ 0.85 = 381 L, so select the next standard size up — a 500 L tank — giving ~4 days real autonomy plus margin.
• Bund: at 110 % of the 500 L tank, the bund must hold ≥ 550 L. If a second 500 L acid tank shares the area, the aggregate rule (25 % of 1,000 L = 250 L) is less onerous than 110 % of the largest single tank, so 550 L still governs. Acid and alkali should not, in any case, share a bund.
• Delivery check: 324 L consumed per 72 h corresponds to ~7.8 m³/month, informing whether IBC delivery or a small bulk tank is more economic.

The outcome: a 500 L polyethylene day tank in a 550 L dedicated bund, with low-level cut-out set to leave a small heel above the pump suction. Segregating the acid and alkali stores in separate bunds prevents an accidental mix that would generate heat and off-gas. Containment sizing and segregation are covered further in our pollution prevention and containment guide.

How much mixing energy is needed at the injection point?

Dosing accuracy is wasted if the reagent is not dispersed into the bulk flow before the next process step. The mixing intensity is quantified by the velocity gradient G (s⁻¹), the root of power dissipated per unit volume over the dynamic viscosity. Coagulant addition needs vigorous rapid mixing to distribute the reagent before the fast hydrolysis and destabilisation reactions occur; polymer flocculation needs gentle mixing to grow flocs without shearing them apart.

The choice between a static mixer and a mechanical (flash) mixer follows from the flow and the head budget. A static mixer draws its mixing energy from the process flow itself, so it needs no motor and gives a fixed, flow-dependent G — ideal for a steady, pressurised line but weak at very low flow. A mechanical rapid-mix tank with an impeller sets G independently of flow, which suits highly variable flows and long contact times, at the cost of a motor, a vessel and a larger footprint. For a compact dosing skid injecting into a pipeline, the static mixer is almost always the right answer; for a coagulation stage ahead of a clarifier with widely swinging flow, a stirred flash-mix chamber may be preferred.

An in-line static mixer is the usual rapid-mix device for a dosing point: it converts pumped head loss into turbulent mixing energy with no moving parts. The design task is to select enough mixing elements to reach the target G at the design flow while accepting the resulting head loss, and to ensure the pipe velocity keeps the flow turbulent (Reynolds number Re = ρvD/μ well above ~4,000, and typically 1–3 m/s) so the elements actually mix rather than laminate. The injection quill should introduce the reagent into the mixer inlet or the centre of the pipe, not dribble it down the wall.

The trade-off is real: more elements and higher velocity give a higher G and better dispersion but cost pumping head. For a coagulant, err toward vigorous mixing over a short contact time; the destabilisation reactions of hydrolysing metal salts complete within roughly 0.5–3 seconds, so the reagent must be everywhere in the flow within that window or part of the dose is wasted.

The competition between mixing and reaction can be framed with a Damköhler number, the ratio of the mixing (or residence) time to the reaction time. When metal-salt hydrolysis is far faster than the local blending of reagent into water (Da >> 1), the reaction is micromixing-limited: freshly injected coagulant reacts with whatever water is immediately around the injection point before it has spread, producing localised over- and under-dosing and poorly formed hydroxide precipitate. Vigorous rapid mixing pushes the mixing time down so that the reagent is distributed on the same timescale as, or faster than, the reaction — the physical justification for the high G-values and the requirement that the injection quill discharge into the turbulent core of the mixer, not down the pipe wall. For flocculation the logic inverts: the reaction (floc growth by orthokinetic collision) is slow and the danger is floc break-up, so G is deliberately kept low and the tapered-flocculation approach reduces G stage by stage as flocs enlarge and become more fragile.

A last practical point on static mixers: their head loss scales roughly with the square of velocity, so a mixer sized comfortably at average flow can impose a punishing head loss at peak, and one sized for peak may drop below its turbulent-mixing threshold at minimum flow. Confirm both the maximum head loss (against the available pump head) and the minimum-flow Reynolds number (against the ~4,000 turbulent transition) before fixing the element count.

Worked example: static-mixer velocity gradient and head loss

Check the rapid-mix intensity for a coagulant injection into a DN150 (0.15 m) pipe carrying 120 m³/h of water at 20 °C, using a static mixer whose manufacturer quotes a head loss of 0.35 m at this flow. Water: ρ = 998 kg/m³, μ = 1.0 × 10⁻³ Pa·s.

• Flow in SI: Q = 120 ÷ 3,600 = 0.0333 m³/s.
• Pipe velocity: A = π/4 × 0.15² = 0.01767 m²; v = 0.0333 ÷ 0.01767 = 1.89 m/s — comfortably turbulent and in the ideal 1–3 m/s band.
• Reynolds number: Re = ρvD/μ = 998 × 1.89 × 0.15 ÷ 1.0×10⁻³ ≈ 283,000 — fully turbulent, so the elements mix effectively.
• Power dissipated: P = ρ·g·h_L·Q = 998 × 9.81 × 0.35 × 0.0333 = 114 W.
• Mixing volume: take the mixer as ~3 pipe-diameters long, L ≈ 0.45 m, so V = A·L = 0.01767 × 0.45 = 0.00795 m³.
• Velocity gradient: G = √(P / (μ·V)) = √(114 / (1.0×10⁻³ × 0.00795)) = √(1.43×10⁷) ≈ 3,790 s⁻¹.
• Contact time: t = V/Q = 0.00795 ÷ 0.0333 = 0.24 s, giving G·t ≈ 910.

A G of ~3,800 s⁻¹ is a vigorous rapid mix — at or above the upper end of the 300–1,000+ s⁻¹ guideline — which is exactly what a fast-hydrolysing metal coagulant wants over a sub-second contact time. If the resulting 0.35 m head loss is unwelcome, a shorter mixer or fewer elements would lower both G and h_L; the design sits comfortably within acceptable bounds here, so this mixer is fit for the duty.

How do you get materials compatibility, calibration and safety right?

Materials compatibility is where dosing skids most often fail in service. Every wetted component — tank, pipework, pump head, valves, seals, injection quill — must resist the chemical at the service concentration and temperature. The elastomer choice (seals, diaphragms, hose) is usually the weak link because a polymer that suits acids may be attacked by oxidants. The table below gives defensible first-pass selections; always confirm against a manufacturer compatibility chart for the exact concentration and temperature.

The failure mechanisms behind the table are worth naming so the selection is defensible rather than table-copied. Ferric and aluminium coagulants are strongly acidic and chloride-bearing, so they pit and crack stainless steels and dissolve mild steel; thermoplastics and fluoropolymers are chosen because they are chemically inert to the halide attack. Concentrated sulphuric acid is an oxidising acid that passivates 316L at high strength but attacks it at intermediate dilutions (the classic "dilute acid is worse than concentrated" trap), which is why PTFE/PVDF are the safe default. Sodium hypochlorite liberates active chlorine and oxygen, oxidising most metals and degrading many elastomers, and its off-gassing is itself a design load — hence degassing valves and generous vent paths. Sodium hydroxide attacks aluminium and glass and, at strength, embrittles some fluoroelastomers, so polyolefins and EPDM are preferred. In every case, confirm the specific grade against temperature, because a material rated at 20 °C may fail at 40 °C, and the same reagent can be handled in PVC when cold but needs CPVC or PVDF when warm.

Calibration is designed in, not added later. A calibration (draw-down) cylinder is a graduated transparent tube on the pump suction: to verify output you isolate the tank, let the pump draw from the cylinder, and time the volume drawn down over a set interval to get the true delivered rate in L/h. This should be a standard fitting on every dosing skid because pump output drifts as check valves foul, diaphragms age and back-pressure changes. The measured draw-down is compared with the theoretical Q_dose and the stroke setting trimmed to match.

Safety and containment are integral to the design, not bolt-ons. Provide a bund (secondary containment) sized to at least 110 % of the largest tank, with incompatible chemicals (acid/alkali, oxidiser/reductant, hypochlorite/acid which liberates chlorine gas) in separate bunds. Include a pressure relief/back-pressure valve to protect the pump and pipework, an anti-siphon valve at the injection quill to stop the process syphoning the tank, degassing/priming provision on gassing chemicals, local eyewash and safety shower, appropriate PPE and correct COSHH labelling. These containment and pH-handling requirements dovetail with dedicated dosing skid and containment design support and with the neutralisation control described in our pH neutralisation systems guide. Designing the pump, materials, mixing and containment as one integrated package — rather than sizing a pump in isolation — is what separates a reliable dosing installation from a maintenance liability.

Chemical dosing system design sequence

1. Fix the dose from a mass balance. Set the target dose (mg/L) from jar/bench testing, then compute product duty Q_dose = Q_plant x C_dose / (density x active fraction).
2. Select the pump technology and turndown. Match diaphragm, peristaltic or progressing-cavity to the chemistry; size so the duty sits mid-curve with usable turndown across the flow range.
3. Choose the control mode. Pick manual, flow-proportional or feedback (pH/ORP) control; use feed-forward plus feedback for neutralisation to avoid hunting.
4. Design make-up and storage. Set solution strength and (for polymer) ageing time; size the day tank from consumption and autonomy with usable-volume and bund margins.
5. Design the injection and mixing. Select a static mixer to reach the target G-value at turbulent pipe velocity; place an injection quill at the mixer inlet with an anti-siphon valve.
6. Confirm materials, calibration and containment. Verify every wetted material against the chemical, fit a calibration cylinder, and provide bunding, relief and segregation of incompatible chemicals.