Caustic Concentration Control in Bayer Digestion Liquor Circuits

In Bayer digestion circuits the term caustic concentration refers to free sodium hydroxide in the circulating process liquor, expressed in alumina refinery practice as grams of sodium oxide equivalent per litre, with the unit written as Na₂O. Spent liquor entering the digestion train typically holds 130 g/L to 220 g/L free Na₂O and an alumina-to-caustic mass ratio of 0.30 to 0.45. After bauxite slurry preparation and digestion at temperature, pregnant liquor advances to clarification with an A/C ratio of 0.60 to 0.72. The control problem is not simply maintaining a fixed sodium hydroxide titre. The liquor contains sodium aluminate, sodium carbonate, sodium sulfate, sodium chloride, silicate species, humate and oxalate degradation products, and suspended red mud. Each dissolved component changes density, conductivity, refractive index, and boiling point rise to a different extent. Consequently the caustic concentration controller must be designed around a specific measurement hierarchy, a defined sampling interface, and a mass-balance structure that reconciles evaporation load, bauxite moisture, wash water, and fresh caustic injection.

The digestion loop is a recirculating inventory with two primary accumulation zones: the spent liquor storage and the slurry mixing tanks. A caustic excursion of 5 g/L to 10 g/L Na₂O in spent liquor may shift pregnant liquor A/C by 0.02 to 0.04 unless dissolved alumina moves in the opposite direction. This coupling creates a process conflict. Elevated caustic improves the driving force for gibbsite, boehmite, and diaspore dissolution, but it reduces the subsequent precipitation productivity per unit volume because the stable alumina concentration rises. Low caustic conserves precipitation yield but depresses extraction efficiency and can increase scale formation in preheaters. The control objective is therefore a narrow setpoint band defined by bauxite grade, digestion temperature, impurity inventory, and evaporator capacity.

What Distinguishes Free Caustic from Total Alkali in Digestion Circuits?

Free caustic is commonly differentiated from total alkali by barium chloride precipitation of carbonate and sulfate, followed by acidimetric titration. The procedure separates hydroxide from carbonate because the barium salt precipitation removes CO₃²⁻ and SO₄²⁻ from the reaction path. In manual or automatic titrators, the sample is diluted to prevent sodium aluminate hydrolysis interference, cooled to 25 °C, and titrated potentiometrically to pH 8.3 and pH 4.5. ASTM E291-22 provides the test methods for sodium hydroxide and caustic potash; ISO 979 specifies industrial sodium hydroxide assay procedures. Plant laboratories usually report total caustic, carbonate, and sulfate on a Na₂O or Na₂CO₃ equivalent basis.

Total alkali is not equivalent to free caustic when carbonate and aluminate are present. The carbonate fraction consumes acid in a different inflection region, and the aluminate ion hydrolyzes to aluminum hydroxide if the sample is diluted with water too rapidly. For process control, the free caustic measurement must be corrected for dissolved alumina. In high-temperature digestion liquors with A/C ratios above 0.60, the aluminate background can depress the apparent caustic titre if the titration endpoint is evaluated without prior barium chloride addition. Automatic titrators using two-stage pH endpoints reduce operator bias but require daily verification against a certified 1.0 mol/L sodium hydroxide standard. Reagent consumption is 10 mL to 25 mL barium chloride solution per sample depending on sulfate and carbonate loading.

In digestion circuits the free caustic value is used as the primary setpoint because it defines alumina solubility. Total alkali is retained as a diagnostic for soda balance and causticization efficiency. When the difference between total alkali and free caustic exceeds 15 g/L Na₂O equivalent, carbonate or sulfate accumulation is usually the cause. The control response may be a purge stream to the lake or an increase in lime causticization. Without this distinction, a conductivity-only analyzer can misrepresent a carbonate excursion as a caustic rise.

A toroidal conductivity probe installed in a spent liquor transfer line does not respond only to NaOH. The measured electrical conductivity is a sum of contributions from Na⁺, OH⁻, Al(OH)₄⁻, CO₃²⁻, SO₄²⁻, Cl⁻, and organic acid anions. Temperature compensation is performed by the analyser electronics using an internal Pt100 or Pt1000 reference element, but the compensation curve is matrix-specific. A Bayer liquor with a high aluminate background has a conductivity-temperature slope different from dilute sodium hydroxide. Conductivity calibration therefore requires at least 5 to 10 plant titration samples across the working range. The correlation may be linear only over a limited span, for example from 180 g/L to 240 g/L Na₂O caustic. Outside that span, carbonate and sulfate interactions create non-linear curvature. Conductivity analyzers are often calibrated using standards referenced to ASTM D1125-23 or ISO 7888:1985 for water, but Bayer liquors require matrix-specific standards.

Sensor fouling is the dominant field failure. Toroidal conductivity sensors are preferred over contacting two-electrode cells because the toroidal design tolerates a thin scale film. However, when sodalite or sodium sulfate deposits form on the toroid bore, the measured conductivity decreases. Automatic retractable probes with steam or dilute caustic wash cycles of 10 s to 30 s every 2 h to 8 h reduce drift. The calibration check frequency advised by analyzer suppliers is usually 7 days to 14 days, but high-carbonate Bayer liquors may require weekly verification. If carbonate exceeds 15 g/L Na₂O equivalent, conductivity-derived caustic should not be used as the sole control input.

Coriolis Density Inference Under Bayer Liquor Solids Loading

Inference of NaOH concentration from density requires a fixed or modelled sodium aluminate background. An oscillation-type density meter based on ISO 15212-1 can resolve density differences of 0.0002 g/cm³ to 0.0005 g/cm³ under single-phase conditions. In Bayer liquors the density change associated with a 1 g/L NaOH change is approximately 0.0004 g/cm³ to 0.0006 g/cm³, but the exact sensitivity depends on temperature, alumina concentration, and impurity inventory. At a digestion liquor temperature of 150 °C the density signal must be compensated for expansion. A 5 °C temperature error can produce a caustic error of 3 g/L to 5 g/L Na₂O if the compensation polynomial is incorrectly configured.

Suspended red mud solids cause a density bias proportional to the solids volume fraction. Because the density meter cannot distinguish dissolved solids from fine hematite and quartz particles, the inferential model must include a suspended solids correction from a turbidity or differential pressure measurement. Entrained gas from autoclave blow-off or pump cavitation causes negative density transients that can be misread as caustic loss. A Coriolis mass flow and density meter with a straight-tube geometry may tolerate some solids, but viscous Bayer slurries with solids above 5 g/L can damp the oscillation amplitude and trigger meter fault alarms. The meter should be installed after a degassing vessel or in a vertical upward line to prevent vapor accumulation.

Density-based caustic inference is often acceptable for spent liquor because the composition is relatively stable and the solids loading is low. It is less suitable for raw slurry before digestion because the bauxite solids loading is high and variable. When density is paired with toroidal conductivity, a two-input inferential model can separate caustic and alumina concentration changes more reliably than either signal alone. The model must be retrained whenever the bauxite source changes or when the organic carbon concentration moves by more than 5 g/L.

Because high-temperature autoclave liquor at 250 °C must be cooled before any conventional electrochemical or optical measurement, the sample conditioning system often becomes the dominant maintenance burden. The sample is withdrawn through a nozzle, depressurized across a letdown valve, and cooled in a double-pipe or coil-in-shell cooler from 250 °C to 25 °C to 40 °C. If the pressure reduction occurs before cooling, flashing can cause caustic concentration increases at the cooler wall and produce sodium aluminate scale. The preferred sequence is indirect cooling followed by pressure reduction, with a sample flow of 0.5 L/min to 2.0 L/min. A high-flow bypass loop reduces transport lag but increases the erosion rate of the letdown valve trim. Sample coolers in caustic aluminate service require scheduled descaling with hot dilute acid or by mechanical rodding. Electrical conductivity and titrator sample lines should be flushed with hot process liquor rather than water, because water flushing precipitates gibbsite and blocks narrow-bore tubing.

Fast-loop analyser enclosures are typically purged with instrument air and maintained at 20 °C to 25 °C. Analyzer availability in Bayer plants is frequently limited by sample cooler plugging, not sensor failure. A retractable conductivity sensor with a manual insertion length of 100 mm to 300 mm can be extracted for cleaning without shutting down the sample loop. The sample system must include a pressure relief device downstream of the letdown valve because a closed analyser loop can be overpressured by thermal expansion of process liquor. Published data for specific Bayer sample conditioner availability across different refineries is limited, but field reports consistently identify sample transport delay and cooler fouling as the largest sources of online caustic measurement discrepancy.

When Carbonate and Sulfate Impurities Shift Conductivity Baselines

When carbonate accumulates in spent liquor, the conductivity baseline shifts upward even if free caustic remains constant. Carbonate enters the Bayer circuit through bauxite containing siderite or organic carbon, through atmospheric absorption in storage tanks, and through make-up caustic impurities. Sulfate originates from sulfide minerals in bauxite. In spent liquor, carbonate concentrations can range from 5 g/L to 30 g/L Na₂O equivalent depending on refinery design and purge policy. Sulfate concentrations may reach 10 g/L to 40 g/L. Both ions increase solution density and conductivity, but neither participates in alumina dissolution. If the caustic analyser is a single-variable conductivity model, a carbonate excursion of 5 g/L Na₂O equivalent can add an apparent caustic reading error of 3 g/L to 8 g/L.

The corrective strategy is to combine conductivity with automatic titration or density. The titration-based automatic analyser can report free NaOH, carbonate, and sulfate in 10 min to 25 min per cycle. The cycle time is slower than conductivity, so a cascade structure is used: the titrator updates the conductivity model bias every 30 min to 60 min, while the conductivity signal supplies continuous caustic trim. This combination holds caustic setpoint within ±2 g/L Na₂O in spent liquor service when the model bias is updated frequently. In high-sulfate liquors, densitometer compensation is also required because sulfate contributes strongly to density but not to alumina solubility.

Toroidal conductivity cells are often selected for continuous service because they tolerate a thin scale film, but their raw signal is not specific. Raman immersion probes can measure aluminate and water bands and provide a more direct estimate of free caustic in clear liquors, but high background fluorescence from humates can reduce signal-to-noise. The Raman calibrations are site-specific and require chemometric models built from at least 50 to 100 plant samples. NIR process analyzers similarly require multivariate calibration and are affected by bubbles and turbidity. Oscillation-type density meters are robust for clear spent liquor but fail to separate dissolved caustic from dissolved sulfate or chloride.

Feed-Forward Caustic Trim and Dead-Time Compensation in Slurry Mixing

In a digestion liquor circuit, spent liquor flow to the grinding mills and slurry mix tanks is the main caustic feed to digestion. The fresh caustic injection point is often upstream of the mills or into the dilution liquor header. The manipulated variable for concentration control is either the fresh 50% NaOH flow or the ratio of spent liquor to bauxite slurry water. The process gain from fresh caustic flow to downstream caustic titre depends on the circulating liquor inventory, which can be 500 m³ to 5,000 m³ per train. With a mixing dead time of 5 min to 20 min and a time constant of 30 min to 90 min, a simple PID controller tuned aggressively can amplify concentration cycles. The control loop should use either a Smith predictor or a feed-forward ratio structure.

Feed-forward compensation is based on the bauxite mass flow, free moisture, reactive silica, and the target digestion liquor caustic concentration. The required caustic addition can be expressed as m_NaOH = V_slurry × C_target − V_spent × C_spent. The C_target term is not fixed; it is adjusted by the metallurgical model as the predicted extraction and A/C ratio change. When reactive silica in bauxite rises from 3% to 5%, desilication caustic demand increases by roughly 5 g/L to 12 g/L Na₂O depending on the desilication product. The feed-forward controller must receive the ore analysis update at least every 2 h to prevent sustained offset.

At the digestion train, the control target is commonly the pregnant liquor A/C ratio and the caustic concentration in the blow-off stream. The blow-off stream is sampled after flash cooling and sand removal. The analytical dead time includes sample transport of 5 min to 15 min, analyzer cycle time of 10 min to 25 min, and controller execution delay. A cascaded loop with inner flow control on the dilution liquor and outer composition control should be designed with anti-windup. The outer loop reset time is usually set between 20 min and 60 min in industrial practice. High-frequency noise from the conductivity signal is filtered with a first-order lag of 30 s to 120 s.

During a high-silica bauxite campaign, the desilication product incorporates sodium, alumina, and silica; the caustic loss measured as Na₂O can be 0.8 g to 1.2 g per gram of reactive silica entering the circuit. If the caustic controller does not compensate for this loss, the free caustic concentration drifts downward after a feed change to high-silica bauxite. The drift is not immediate because the desilication reaction occurs partly in the preheater and partly in the digestion train. A high-silica bauxite campaign can create a caustic deficit of 5 g/L to 10 g/L Na₂O over 8 h to 24 h. The correction is made by raising fresh caustic flow upstream of the mills or by increasing evaporated spent liquor caustic concentration. Overcorrection should be avoided during the first hour because the desilication reaction consumes caustic over a prolonged residence time.

Preheater scale in Bayer digestion is commonly observed in indirect steam-heated shell-and-tube units with tube outside diameters of 19 mm to 25 mm and liquor velocities of 1.0 m/s to 2.5 m/s. As sodalite scale thickness reaches 0.5 mm to 1.0 mm, the tube-side heat transfer coefficient drops and the required steam pressure rises by 0.2 MPa to 0.5 MPa at constant throughput. This is a process conflict: raising caustic setpoint can improve alumina extraction but may accelerate desilication product supersaturation at the wall and shorten the preheater run length. The inverse risk occurs when bauxite changes to a low-silica ore. The caustic inventory then rises because less sodium is fixed in desilication products. If the setpoint is not adjusted, the spent liquor caustic concentration can overshoot by 5 g/L to 8 g/L. The resulting high-caustic liquor increases alumina solubility, reduces precipitation productivity, and can increase the boiling point rise in evaporation. For this reason the feed-forward model must include a reactive silica input and a dynamic desilication caustic consumption term, not just a steady-state ratio.

Reconciling Grab Titrations with Online Analyzer Drift

Because grab sample titration remains the reference method for caustic concentration in most alumina refineries, the online analyzer must be continuously reconciled against laboratory results. A sample is collected from a recirculating liquor line, cooled to 25 °C, filtered, and analysed by manual or automatic titration. ASTM E291-22 provides caustic soda assay procedures; ISO 979 provides industrial sodium hydroxide assay. The manual titration repeatability is approximately ±0.5 g/L to ±1.0 g/L Na₂O when performed by a trained analyst. The automatic online titrator repeatability is often ±0.2 g/L to ±0.5 g/L, but its accuracy is limited by sampling delay, incomplete carbonate precipitation, and calibration standard degradation.

Online analyzer drift should be tracked with a daily quality control sample of known concentration. The control limit for a conductivity model bias is typically ±2 g/L Na₂O. If the bias exceeds this limit, the analyzer is recalibrated against plant titrations. A CUSUM chart on the bias detects slow drift earlier than a Shewhart chart. The bias correction should not be applied as a step change to the control loop because a sudden caustic setpoint shift can upset precipitation. Instead the bias is ramped over 30 min to 60 min at a rate not exceeding 0.1 g/L Na₂O per minute.

The online titrator cycle is often scheduled for 15 min to 25 min. During that period the conductivity loop runs at 1 s to 5 s scan intervals. The two measurements are reconciled in a DCS calculation block. The final caustic value is used by the spent liquor flow controller or the fresh caustic trim valve. The sample point location should be selected to minimize dead time but must be placed after solids removal to prevent analyser plugging. Sampling after the flash tank but before the sand clarifier may reduce lag but increases suspended solids load. Many plants sample the clarified pregnant liquor after sand removal, accepting an additional 10 min to 20 min delay for a cleaner sample.

Although model predictive control has been applied to Bayer liquor circuits to manage the interaction between caustic concentration, evaporation duty, and precipitation feed, the achievable performance depends on analyzer availability. The controller uses a dynamic matrix model identified from step tests on fresh caustic flow, spent liquor flow, and evaporator steam pressure. The controlled variables are blow-off caustic concentration, blow-off A/C ratio, and spent liquor inventory level. The manipulated variables are fresh caustic valve position, dilution water flow, and evaporator feed flow. Time constants range from 20 min to 180 min depending on the mixing volume. The MPC sampling period is typically 1 min to 5 min. Operators retain low-level flow controllers in auto.

Published data on MPC performance in Bayer digestion is limited because alumina refinery control systems are seldom described in peer-reviewed detail. The available technical literature indicates that caustic concentration variance can be reduced by 20% to 50% relative to PID-only control when the model is maintained and the analyzers are reliable. The largest failure mode is not the control algorithm but the analyzer sample system. When analyzer availability drops below 90%, MPC benefits deteriorate and the loop should be switched to a conservative PID mode with tighter output limits.

Evaporator Duty Modulates Spent Liquor Caustic Inventory

The evaporator train controls water balance and therefore caustic concentration in the spent liquor. If the evaporation rate is less than the water entering the circuit from bauxite moisture, wash water, and make-up caustic, the spent liquor volume expands and caustic concentration falls. The relationship is C_final = C_initial × V_initial / V_final for a fixed sodium hydroxide mass. A 2% volume reduction raises caustic concentration by approximately 2%, for example from 200 g/L to 204 g/L Na₂O. The evaporator setpoint is therefore manipulated by the caustic inventory controller.

The evaporator has a nonlinear response. As caustic concentration rises from 180 g/L to 250 g/L Na₂O, the boiling point rise increases and the vacuum system must provide a lower absolute pressure to maintain heat transfer. Scale deposition on the evaporator tubes reduces the overall heat transfer coefficient by 20% to 40% over a production cycle. The cleaning cycle for a falling-film evaporator in Bayer service is typically 7 days to 30 days depending on sulfate and silica load. The steam pressure must then be raised. At some point the steam pressure reaches the available header limit and evaporation capacity becomes constrained. Published data for high-organic Bayer liquor viscosity in the 250 g/L to 280 g/L Na₂O range is limited, but the increase in dissolved solids and boiling point rise consistently reduces available thermal driving force.

The caustic controller should not demand rapid evaporator changes because the evaporator response time is 20 min to 60 min for a vacuum disturbance and 2 h to 6 h for a significant inventory shift. In industrial practice, the fresh caustic valve handles fast caustic deviations, while the evaporator setpoint corrects the long-term water balance. If the spent liquor caustic concentration is below 170 g/L Na₂O for more than 4 h, evaporation capacity may be insufficient and the plant must reduce wash water or increase fresh caustic.

In caustic concentration control circuits, materials selection must account for high pH, dissolved alumina, and temperature. Carbon steel is used extensively in Bayer circuits but has stress corrosion cracking limitations at high caustic concentration and temperature. The sample cooler and analyzer wetted parts in concentrated caustic service above 150 °C are often specified in nickel-containing alloys or stainless steel grades with controlled ferrite. Instrument air purge lines must be dried to prevent condensation from carrying carbonate into the analyzer enclosure. Water rinses upstream of the titrator should be avoided; 5% to 10% NaOH solution is used for flushing.

The lower caustic concentration boundary is set by extraction efficiency and scale risk. In gibbsitic bauxite digestion at 145 °C to 150 °C, a free caustic value below 170 g/L Na₂O can reduce alumina extraction by 2% to 5% for typical ores with reactive silica above 3%. The upper boundary is set by precipitation productivity and physical handling: above 280 g/L Na₂O the stable alumina inventory increases and evaporator and heat exchanger scaling often accelerates. The operational band in many high-temperature digestion circuits is therefore 180 g/L to 260 g/L Na₂O, with the exact band shifted by bauxite mineralogy and impurity balance.

Control of caustic concentration is not satisfied by a single analyzer. The minimum instrumentation set consists of a continuous conductivity or density channel, an automatic titrator for periodic reference, sample conditioning with a cooled fast loop, and a feed-forward caustic trim tied to bauxite mass flow and reactive silica. When carbonate exceeds 15 g/L Na₂O equivalent, titration-based caustic measurement becomes mandatory rather than optional. When the pregnant liquor A/C ratio leaves the 0.60 to 0.72 band, the caustic setpoint must be reviewed against the digestion temperature profile and the evaporation water balance. These boundaries define the operating envelope for the caustic concentration control strategy, rather than a single universal setpoint.