Laboratory and pilot testing

Pilot the integrated process, not isolated hero equipment.

Controlled principle A test is valuable only when it is designed around a named project decision, uses a representative physical and chemical basis, produces auditable data, and changes or defends a design, guarantee, cost, residual or stage-gate conclusion. A successful isolated unit operation is not evidence that the complete MLD, ZLD or valorization train will operate reliably.

Test decision claims—not a generic technology

Chapter 32 hands this chapter a preferred concept and a register of exact claims that must be closed. The laboratory or pilot programme should therefore begin with the decision that the owner intends to make: reject a mechanism, select an option, define a scale-up factor, establish a cleaning basis, qualify a product or residual, narrow a cost range, support a guarantee, or authorize the next development stage. “Test the technology” is not a decision claim.

EPA treatability-study guidance uses a tiered laboratory-screening, bench-scale and pilot-scale approach, but also allows unnecessary tiers to be skipped when the data gap and decision do not require them. The transferable lesson is not the CERCLA process itself; it is that test scale follows the evidence needed, not a ceremonial sequence. [6]

Decision claim	Evidence currently missing	Minimum useful test	Output that changes the project	Failure condition
A precipitation step can protect the next concentrator	Kinetics, co-precipitation, dewaterability and filtrate chemistry on real matrix	Controlled jar/bench reactor followed by integrated solids separation	Dose, residence time, cake load, filtrate spec and downstream scaling basis	Required removal is achieved only with unacceptable impurity, reagent or sludge burden
A membrane block can sustain the required duty	Element-scale flux, passage, pressure drop, fouling and cleanability under representative recycle	Element or multi-element loop with standardized operating data	Design flux, recovery, staging, cleaning interval and replacement basis	Performance cannot be normalized to a stable condition or cannot be restored by an acceptable cleaning regime
A crystallizer can make an acceptable solid	Phase, growth, retained liquor, wash yield, fines and mother-liquor purge	Batch then continuous/semicontinuous crystallization with solids handling	Crystal duty, yield, wash/dry requirements, purge and product/residual route	Mixed phase, unstable product or unmanageable mother-liquor accumulation
The integrated train can survive feed and operational variability	Coupled behaviour across pretreatment, concentration, recycle, purge and controls	Representative integrated pilot with defined normal, design and upset cases	Control philosophy, buffers, redundancy, consumables, availability and guarantee basis	Water or residual service fails during a credible case or depends on continuous product sales

Table 33.1. Every test must close a named decision claim and identify the consequence of failure.

Figure 33.1. Evidence must expand from mechanism screening to integrated, design-actionable scale-up. The ladder is decision-based, not a universal sequence or TRL claim.

Select the lowest test scale that can reproduce the controlling risk

Laboratory screening is efficient for mechanism rejection, parameter windows and analytical development. Bench work adds integration, repeated cycles and better control of residence time, mixing and solids. Pilot work becomes necessary when the decision depends on representative hydraulics, heat and mass transfer, commercial-format components, continuous recycle, operator interventions, cleaning, control logic or the interaction of multiple blocks. Scale should be increased only when the unresolved risk cannot be represented credibly at the smaller scale.

Reclamation describes laboratory projects as small-flow studies used to determine viability, while pilot projects are intended to test practicality at a scale that supports operational requirements and finished-water goals. Its Pitch-to-Pilot programme also requires natural rather than synthetic feed for that specific programme. These are useful programme examples, not universal hydraulic thresholds for industrial pilots.

Test tier	Primary question	What it can establish	What normally remains unresolved	Advance when
Analytical / mechanism screen	Is the proposed reaction, separation or analytical method plausible?	Speciation, equilibrium window, initial selectivity, method detection and sample stability	Hydrodynamics, continuous accumulation, commercial component behaviour and operability	A plausible mechanism exists and the next decision requires controlled process testing
Bench batch	What response surface and failure mechanism exist?	Dose, kinetics, phase, equilibrium, sensitivity, repeatability and first solids properties	Steady recycle, long-term deposition, full controls and production-scale interfaces	The preferred operating window and critical variables are defined
Bench continuous / cyclic	Can the integrated duty repeat under controlled flow and recycle?	Residence time, cyclic capacity, accumulation, cleaning response and balance closure	Commercial geometry, full-scale transfer coefficients, operator logistics and availability	Scale-dependent risks are identified and the pilot can be designed around them
Component / element	Does a commercial-format component behave as required?	Pressure drop, flux/passage, integrity, normalized performance and cleaning response	Complete train interaction, long-duration reliability and plant design	The component duty is stable enough to test in the integrated configuration
Integrated pilot	Can the selected process operate as a system on representative conditions?	Control, recycle, purge, solids, cleaning, upset recovery, consumables and scale-up factors	Lifetime beyond the demonstrated envelope and construction/commissioning risks	Acceptance criteria are met and unresolved effects are bounded for feasibility/pre-FEED

Table 33.2. Test scale should reproduce the controlling decision risk without paying for unnecessary complexity.

Representative feed is a controlled distribution, not one drum

A test feed must represent the process states that control design: normal production, peak load, cleaning or campaign transitions, recycled streams, additive carryover, temperature and the species that accumulate under concentration. A single composite can erase the event that actually causes scaling, foaming, corrosion or product rejection. Conversely, a worst-case synthetic mixture can create combinations that never occur and may reject a viable process for the wrong reason.

ISO 5667-3:2024 covers preservation, handling, transport and storage of water samples, while ISO 5667-14 addresses sampling QA/QC. These controls matter because ageing, temperature change, gas exchange, precipitation, biological activity, adsorption and container interaction can alter the test article before the experiment begins. The programme must record collection time, preservation, storage, homogenization, filtration and any reconstruction or spiking. [2,3]

Feed control	Required record	Why it matters	Invalid shortcut	Test response
Source and operating state	Unit, campaign, load, cleaning state, upstream treatment and sample time	Links the sample to a real process condition	“Representative brine” without production context	Separate normal, design and event feeds or justify blending
Ageing and preservation	Container, headspace, temperature, light, time, mixing, filtration and preservative	Prevents the sample from changing unnoticed before testing	Using stored brine as if it were fresh	Validate holding period or report the test as aged-feed evidence
Chemistry closure	Major ions, alkalinity/inorganic carbon, density, TDS/conductivity, solids and relevant organics/additives	Detects missing or transformed mass and supports modelling	Accepting a vendor analysis without charge/density checks	Reconcile, reanalyse or explicitly carry the gap
Variability design	Frequency distribution, minimum/maximum, event frequency and correlation	Distinguishes a rare event from a persistent design case	Testing only the average and the independent maxima	Create coherent normal, design and upset cases
Inventory sufficiency	Volume required for conditioning, repeats, pilot operation, cleaning and retained samples	Avoids changing feed halfway through the programme	Assuming one batch will cover every task	Secure campaign samples, on-site side stream or controlled make-up strategy

Table 33.3. Representative-feed control must preserve both chemistry and the process context that created it.

Synthetic and real feeds answer different questions

Synthetic feed is useful when the programme needs controlled mechanistic comparisons, a clean mass balance, safe parameter screening or a deliberate challenge concentration. It is weak evidence when trace organics, oils, polymers, colloids, minor ions, biological activity, corrosion products or unknown interactions may govern performance. Real feed is therefore required before claiming representative fouling, scaling, cleaning, product quality or residual behaviour. A hybrid programme may use synthetic feed to map the mechanism and real feed to validate the selected window.

Use case	Synthetic feed strength	Synthetic feed limitation	Real-feed requirement	Controlled conclusion
Reaction or selectivity screen	Independent control of target and competing species	May omit inhibitors, colloids and background ionic-strength effects	Confirm selected window in representative matrix	Mechanism appears viable; no scale-up claim yet
Analytical method development	Known additions support recovery and interference checks	Does not prove field sampling or matrix stability	Matrix spike, duplicates and retained real samples	Method is fit for the stated matrix and range
Membrane material comparison	Repeatable comparison under one defined solution	Does not reproduce actual foulant layer or additive history	Real-feed element/loop test before design duty	Relative ranking only until real-feed validation
Precipitation / crystallization	Clear stoichiometry and phase mapping	May miss nucleation surfaces, inhibitors, mixed solids and entrainment	Real-feed kinetics, mineralogy and dewatering/washing	Candidate operating window and solids route
Integrated pilot	Useful only for commissioning, controls or deliberate challenges	Cannot prove representative accumulation or cleaning burden	Normal/design/upset real feed or justified side stream	Project-specific integrated evidence

Table 33.4. Synthetic feed can isolate mechanisms; real feed is needed to support representative scale-up claims.

Batch tests must not be mistaken for continuous operation

Batch experiments can define reaction windows, equilibrium, phase formation and preliminary kinetics. They do not automatically reproduce steady feed, nucleation history, surface conditioning, recycle accumulation, bleed control, cleaning frequency or the interaction between residence-time distribution and solids inventory. Continuous or cyclic testing is needed when the plant decision depends on these time-dependent effects.

The most important time scale is not the calendar duration by itself. It is the time needed to expose the controlling phenomenon: induction, inventory turnover, membrane or resin cycles, crystal residence, cleaning-and-recovery cycles, feed campaigns, material attack, mother-liquor accumulation or operator response. No universal number of days or cycles is imposed here; the protocol must link duration to the mechanism and decision threshold.

Time-dependent risk	Batch evidence	Continuous/cyclic evidence needed	Minimum tracked outputs	Failure signal
Scaling induction	Supersaturation and initial precipitate	Time to nucleation/deposition under representative surfaces, shear and concentration path	Induction time, deposit location, mass, composition and cleanability	Stable batch liquor but progressive equipment deposition
Fouling and conditioning	Initial flux or adsorption	Performance trend through feed changes and cleaning cycles	Normalized flux/passage, pressure drop, foulant mass and cleaning recovery	Short test is stable but baseline does not recover after repeated cycles
Recycle accumulation	Single-pass product and residual	Multiple inventory turnovers with controlled purge	Ion/additive/organic inventory, purge composition and balance closure	Impurity or hazard rises without a stable purge endpoint
Solids growth and separation	Initial precipitate or crystal	Steady solids inventory, ageing, breakage and classification	Particle size, phase, moisture, filtration/settling and wash yield	Acceptable chemistry but unmanageable fines or retained liquor
Control and operability	Manual set point response	Automated transitions, alarms, shutdown/restart and buffer logic	Control deviation, response time, operator action and recovery	Service cannot be maintained during a credible transition

Table 33.5. Duration and mode must be justified by the time-dependent mechanism that controls scale-up.

Figure 33.2. Scenario-only teaching trace showing how a short stable window can miss cumulative degradation, feed upset and incomplete cleaning recovery. The values and duration are illustrative, not typical.

Membrane evidence must progress from coupon to commercial geometry

Flat-sheet and small-cell tests are appropriate for material screening, intrinsic selectivity and controlled fouling studies. They do not reproduce feed spacers, channel pressure drop, concentration polarization, element staging, vessel hydraulics or the cleaning distribution of commercial assemblies. ASTM D7601 explicitly describes short-term element or bundle evaluation and states that such testing is not intended for plant design. ASTM D4516 standardizes RO performance because pressure, temperature, conversion and feed concentration alter observed permeate flow and salt passage; it also warns that its applicability to wastewaters is not assured.

Evidence level	Useful outputs	Scale effects still missing	Normalization / controls	Permitted design use
Coupon / flat sheet	Material compatibility, intrinsic passage/rejection, foulant response	Spacer, channel, element manufacturing and pressure drop	Temperature, pressure, concentration, crossflow and area	Material or mechanism screening only
Small cell / module	Flux, passage, mass-transfer trend and cleaning chemistry	Commercial element geometry, staging and full hydraulic distribution	Feed analysis, recovery, recirculation, concentration and mass balance	Define element-test window and failure mechanisms
Commercial element	Element flux/passage, pressure drop, integrity and cleaning response	Multi-element staging, interstage conditions and plant controls	Standardized performance and stage-specific data where applicable	Preliminary design factor, not complete plant design
Multi-element / integrated pilot	Staging, recovery, pressure drop, concentrate quality, cleaning sequence and pretreatment interaction	Full plant redundancy, manifold scale, long-term lifetime and construction variability	Calibrated instruments, normalized performance, membrane history and balance closure	Feasibility/pre-FEED inputs with explicit unresolved lifetime factor

Table 33.6. Membrane evidence must reproduce commercial geometry and integrated duty before it supports plant design.

Precipitation and crystallization are solids-management tests

Removal percentage is only one output. The programme must also determine reaction kinetics, co-precipitation, phase identity, nucleation behaviour, particle growth, settling or filtration, cake moisture, wash demand, entrained mother liquor, filtrate composition, reagent-derived ions and the destination of every solid and liquid. A precipitation result that protects a membrane but creates an unstable sludge route has not passed the system test.

Test question	Bench evidence	Pilot / continuous evidence	Scale-up parameter	Downstream consequence
Where is the operating window?	pH/reagent path, equilibrium, temperature and competing-ion response	Dose control and response to feed variability	Dose, excess factor, residence and control limits	Residual scaling species and reagent salt load
What solid forms?	Mineralogy/phase, impurity and initial morphology	Phase consistency, ageing, attrition and secondary precipitation	Expected phase distribution and off-spec allowance	Product/residual classification and handling
Can it be separated?	Settling, filtration, centrifugation and preliminary moisture	Continuous solids loading, blinding, wash and dewatering cycle	Area/flux, cake resistance, wash ratio and equipment duty	Recycle water quality, sludge volume and disposal cost
Can crystals meet specification?	Phase, size, assay, moisture and retained liquor	Growth distribution, fines, washing, drying and storage stability	Yield, on-spec fraction, recycle and purge basis	Qualification, packaging and off-spec route
Can mother liquor remain controlled?	Composition after a batch or evaporation step	Multiple turnovers, impurity build-up and purge response	Purge rate and maximum impurity inventory	Water recovery, product purity and residual endpoint

Table 33.7. Precipitation and crystallization testing must close the solids, filtrate, wash and mother-liquor system.

Pilot the recycle, purge and residual network—not only the main equipment

Integrated pilots often fail to represent the streams that determine full-scale stability: membrane concentrate return, softening filtrate, wash water, crystallizer mother liquor, cleaning waste, off-spec product, condensate polishing reject and emergency bypass. Omitting these streams can make the pilot appear stable while the project design has no steady salt, organic or reagent balance.

Loop / residual	Pilot representation	Key measurements	Required steady or bounded condition	Design output
Concentrate recycle	Actual recycle ratio and mixing point	Flow, density, ions, organics/additives, temperature	Inventory reaches a stable or model-consistent trend	Recycle flow, buffer and maximum concentration
Filtrate / centrate return	Representative solids carryover and dissolved chemistry	TSS/turbidity, target ions, reagent residual and particle size	No uncontrolled solids or scaling load returns downstream	Clarification/polishing and return location
Mother liquor	Repeated return with defined purge	Major/minor ions, phase-formers, organics and water balance	Purity and hazard remain within limits at the selected purge	Purge rate, residual duty and product-yield impact
Cleaning waste	Full cleaning sequence and neutralization/segregation route	Volume, pH, COD/TOC, metals/ions and compatibility	Routine and peak waste have a lawful, sized destination	Tankage, treatment and operating-cost allowance
Off-spec / unsold product	Representative rejection and rework/bypass route	Mass, moisture, chemistry, storage and reprocessability	Water plant remains operable without product placement	Fallback capacity and no-product operating case

Table 33.8. The pilot must represent the recycle and residual network that controls full-scale accumulation and fallback.

Cleaning, materials, long-duration and upset tests are part of the core protocol

A stable nominal run is not enough. The programme should deliberately test the operational actions and material exposures that determine availability: start-up, planned shutdown, loss of a dosing system, feed excursion, solids breakthrough, antiscalant or polymer carryover, temperature change, cleaning, drain-down, restart and temporary loss of a product or disposal route. The selected cases should come from the project risk and operability registers rather than an arbitrary “worst case.”

A Reclamation calcium-removal pilot was framed around prevention of calcium-sulfate fouling, extended steady operation and membrane-life implications—illustrating that a pilot question can combine chemistry, operability and lifetime evidence rather than report only initial recovery.

Test family	Protocol elements	Primary measurements	Acceptance basis	Scale-up / risk output
Cleaning and recovery	Defined foul condition, chemistry, sequence, temperature, duration and rinse endpoint	Normalized performance recovery, waste volume/quality and material exposure	Required recovery without unacceptable irreversible loss or waste burden	CIP frequency, tank/dose duty, downtime and disposal
Materials and corrosion	Representative wetted materials, joints, elastomers, temperature, oxygen/redox and chemical contact	Mass loss, visual attack, coupons, corrosion indicators, swelling/embrittlement	Compatible for the demonstrated envelope with stated extrapolation limits	Materials selection, corrosion allowance and inspection plan
Long-duration operation	Multiple inventory turnovers and repeated cleaning/operating cycles	Trend, drift, variability, consumables, failures and operator interventions	Stable or bounded performance through the decision-specific exposure	Availability, replacement and maintenance basis
Upset and recovery	Credible feed, equipment, utility, control, product and residual scenarios	Alarm/control response, service deviation, time to safe state and recovery	No unsafe condition; defined service/fallback performance is restored	Buffers, redundancy, interlocks, bypass and procedures
Shutdown / restart	Hot/cold standby, drain/flush, solids suspension and restart sequence	Settling, deposition, microbial/chemical change, time and off-spec production	Restart within required service and residual constraints	Standby philosophy, storage and commissioning sequence

Table 33.9. Reliability evidence must include cleaning, material exposure, long-duration drift and recovery from credible upsets.

The protocol must make the data reproducible and auditable

EPA’s 2025 QAPP guidance treats the project plan as the specification for how environmental information will be generated and controlled. ISO/IEC 17025 provides the laboratory competence framework for valid testing and calibration, and ISO 5725 distinguishes trueness from precision. Applied to an MLD/ZLD pilot, these principles require a controlled protocol, calibrated instruments, defined analytical methods, quality-control samples, deviations, versioned data, measurement uncertainty and an explicit decision rule.

Protocol field	Minimum requirement	Typical control	Why it matters	Stop / qualify result when
Objectives and hypotheses	Named decision claim, threshold and failure condition	Approved test matrix and change control	Prevents data collection without a decision use	Objective changes after results are seen
Instrumentation	Range, accuracy, calibration, location, frequency and data acquisition	Calibration checks, redundancy for critical variables and raw-data retention	Separates process variability from measurement error	Sensor range/accuracy cannot resolve the threshold
Sampling and analysis	Location, timing, preservation, method, detection limit and chain of custody	Blanks, duplicates, spikes/standards and retained samples as relevant	Protects representativeness and comparability	Holding time, contamination or interference is uncontrolled
Replicates and controls	Number and rationale; baseline/reference where meaningful	Randomization/blocking or repeated cycles suited to mechanism	Supports precision and separates treatment effect from drift	One result is used where variability controls the decision
Mass and charge balance	Water, major salt/ion, reagent, solids/product and purge boundaries	Independent flow/analytical checks and reconciliation log	Detects hidden inventory, sampling or calculation errors	Closure error is material and unexplained
Data governance	Raw, processed and excluded data; formulas; versions; deviations; owner	Read-only raw archive, calculation review and audit trail	Allows independent reproduction and model update	Only summarized vendor charts are available

Table 33.10. Test quality is part of the evidence; a credible result must be reproducible from the protocol, raw data and balance.

Define success before operation starts

Success criteria should combine service, operability, quality, residual, safety/environmental and scale-up requirements. Averages alone are insufficient: the protocol should state allowable variability, event-specific limits, minimum exposure or cycles, how missing data are treated, and whether a failed hard criterion can be repeated after a documented cause and corrective action. Weighted scoring must not override a failed safety, lawful-residual, permit or minimum-service condition.

Success domain	Example criterion structure	Evidence basis	Hard gate?	Action if not met
Water service	Net recovered-water quantity and quality within defined normal/design/upset windows	Closed water balance and accepted analytical methods	Yes when required for operations or permit	Redesign duty, pretreatment, controls or endpoint
Process performance	Target removal/concentration/yield with defined normalized conditions and variability	Repeated runs/cycles and representative feed	Usually	Change mechanism, operating window or scale-up factor
Operability and availability	Start, stop, clean and recover within defined service and staffing constraints	Logged events, downtime, interventions and root causes	Yes where residual/water service depends on it	Add buffer, redundancy, automation or simplify train
Product / residual quality	Defined specification, on-spec fraction, off-spec route and fallback capacity	Representative lots/streams and qualification methods	Yes for legal route and claimed revenue	Remove credit, add conditioning or return to residual route
Materials and safety	No unacceptable corrosion, incompatibility, exposure or uncontrolled reaction	Coupons/inspection, alarms, events and safety review	Yes	Change materials, chemistry, containment or operating envelope
Data quality and balances	Instrument/method uncertainty supports threshold; material balances close to agreed tolerance	QA/QC, calibration, reconciliation and independent review	Yes for load-bearing claims	Repeat, qualify or reject result
Scale-up action	Named model/design/guarantee inputs are replaced and residual uncertainty bounded	Approved scale-up report and model update	Yes for next gate	Additional test or conditional/recycle decision

Table 33.11. Pilot success criteria must be defined before testing and linked to a project action.

Figure 33.3. Pilot success criteria. All eight domains must be addressed; hard service, safety, legal and residual failures cannot be averaged away.

Convert results into scale-up parameters—without hiding residual uncertainty

A pilot report should not stop at plots and removal percentages. It should translate evidence into the controlled project model: design flows, loads, residence times, fluxes, transfer coefficients, mixing or heat duties, solids separation rates, wash and purge requirements, consumables, cleaning intervals, material limits, control set points, availability allowances and product/residual specifications. Each parameter should identify the demonstrated range, normalization, scale-up method, uncertainty and any factor retained for untested effects.

Short-term element evaluation can be valuable for inspection and performance change, but ASTM D7601 explicitly limits its design use. The same discipline applies more broadly: a result should be used only for the claim and scale it actually supports.

Pilot result	Translation required	Scale effect / uncertainty	Project register updated	Do not do
Performance curve	Normalize to defined feed, temperature, pressure/recovery or thermal basis	Commercial geometry, control range and fouling/scale drift	Process model and equipment datasheet	Copy peak performance as design duty
Kinetic / residence data	Fit the mechanism and operating window used for sizing	Mixing, nucleation, surface/volume and RTD differences	Reactor/crystallizer/softener basis	Use one batch time as full-scale residence time
Solids separation	Area/flux, cake resistance, moisture, wash and cycle basis	Feed distribution, equipment geometry and solids variability	Solids balance, equipment duty and OPEX	Size only from dry solids mass
Cleaning and lifetime	Loss/recovery per cycle, chemical/water use, downtime and residual waste	Longer-term degradation and supplier warranty limitations	Availability, replacement, CIP and risk register	Extrapolate a few cycles directly to asset life
Product/residual quality	Distribution, on-spec fraction, rejection and fallback under cases	Scale, storage, transport and buyer/permit qualification	Product/residual specification and commercial/legal register	Credit every recovered kilogram as saleable
Upset response	Trigger, safe state, service deviation, recovery time and operator action	Multiple failures, full-plant buffers and utilities	Control philosophy, HAZOP/operability and contingency	Treat successful manual recovery as proof of automatic plant resilience

Table 33.12. Scale-up is an explicit translation from demonstrated evidence to model, design and risk registers.

Laboratory and pilot acceptance gate

The programme may advance only when the tests are representative enough to support the next decision, the data can be audited, the integrated residual and fallback duties are operable, and every scale-up claim is bounded by the evidence actually generated. A technically impressive result is not sufficient when the protocol, feed, duration, balance, materials or residual route do not represent the project.

Gate field	Proceed	Conditional / recycle	Stop	Required record
Decision linkage	Each test closes a named threshold and changes a project input or decision	Exploratory test is segregated from decision evidence	Testing has no defined decision use	Claim-to-test register
Representative basis	Feed, ageing, pretreatment, recycle, temperature and cases match the stated claim	One secondary condition is bounded by sensitivity or follow-up work	Synthetic/aged/partial basis is presented as representative	Feed and condition register
Integration and duration	Controlling hydraulics, solids, recycle, cleaning and time-dependent mechanisms are demonstrated	One scale effect remains with justified factor and owner	Isolated short test is used to claim integrated long-term operation	Configuration and exposure matrix
Data quality and balance	Methods, calibration, uncertainty, raw data and balances support thresholds	A non-critical gap is documented and does not change the decision	Load-bearing result cannot be reproduced or materially fails closure	QAPP/protocol, QA/QC and reconciliation record
Success / failure criteria	Predefined criteria are met across required cases and cycles	A defined repeat or redesign condition is authorized	Criteria are redefined after failure or hard gate is averaged away	Acceptance report and deviation log
Scale-up and project impact	Design/model/guarantee/cost inputs are replaced with bounded evidence	Next stage carries explicit conditional factor and test plan	Peak result is copied to design without translation	Scale-up report and updated registers
Residual and no-product operation	All routine/upset/off-spec streams have sized, lawful routes and the water system survives without sales	One route awaits confirmation with full temporary capacity	Pilot success depends on absent disposal or uninterrupted buyer	Residual/fallback and no-product case

Table 33.13. Laboratory and pilot evidence advances only when representation, data quality, integration and scale-up use are auditable.

Stop rule Stop, redesign or re-stage when the programme demonstrates only isolated performance; when feed ageing or synthetic reconstruction controls the result; when a critical time-dependent mechanism has not been exposed; when balances or data quality cannot support the threshold; or when the process cannot operate through off-spec product, cleaning waste or residual fallback. Do not extend a short successful test into an unsupported lifetime or guarantee claim.

Handover to Chapter 34 — procurement and technology evaluation

Chapter 34 evaluates technology offers and owner procurement decisions. It should receive the complete evidence package rather than vendor headline performance: approved protocol, representative feed basis, raw and reconciled data, normalized results, demonstrated operating envelope, failure modes, cleaning and materials evidence, residual/fallback duties, scale-up factors, unresolved uncertainty and the exact project-model inputs replaced by testing.

Procurement should distinguish what has been independently demonstrated, what is vendor reference evidence, what remains a guarantee or exclusion, and what the owner must still validate. A bidder should not receive credit for matching the pilot peak while excluding the pretreatment, recycle, solids, product qualification or residual systems that made the result possible.

Handover field	Required basis for Chapter 34	Procurement risk if missing
Tested configuration and boundary	Complete equipment, pretreatment, controls, recycle, purge, residual and utility boundary	Bids exclude the systems that created the demonstrated result
Feed and operating envelope	Normal/design/upset distributions, ageing controls, temperature and additives	Guarantee applies only to a cleaner or narrower feed
Normalized performance	Raw and normalized water, energy, chemical, solids/product and quality data by case	Vendors compare incomparable peak or short-term values
Cleaning, materials and lifetime basis	Demonstrated cycles/exposure, recovery, waste, compatible materials and residual uncertainty	Replacement, availability and warranty risk is hidden
Success, deviations and failures	Predefined criteria, results, anomalies, root causes, repeats and unresolved conditions	Only successful runs enter the tender package
Scale-up factors and model impact	Parameters replaced, retained factors, equipment duty, cost/risk range and independent review	Bidder or EPC must rediscover the process basis during procurement
Data rights and assurance	Owner access to protocol, raw data, calculations, samples, software/version and audit rights	Owner cannot verify, compare or enforce the evidence

Table 33.14. Chapter 34 requires a procurement-ready evidence package that preserves the tested boundary, limitations and unresolved risks.

Chapter conclusion

Laboratory and pilot work are not smaller versions of the final plant and they are not marketing demonstrations. They are controlled instruments for converting specific uncertainty into evidence. The programme begins with a project decision, selects the lowest scale that reproduces the controlling risk, protects the representative feed, integrates the relevant recycle and residual network, exposes the required duration and cycles, and defines success before operation begins.

The strongest result is not the highest recovery or cleanest graph. It is an auditable package that shows what operated, on which feed and boundary, for which exposure, with what variability and failures, and exactly how the evidence changes the process design, materials, controls, cleaning, product/residual route, guarantees, cost and next gate. Chapter 34 now uses that evidence to compare technologies and procurement offers on a common basis.

Chapter 33 in one sentence Pilot the integrated process against predefined decision claims—because isolated performance is not scale-up evidence.