AquaAdvisor

CHEMISTRY ENGINE

A physical chemistry solver built for modeling and twins.

Most water chemistry software wraps a legacy index calculation or shells out to a batch solver. AquaAdvisor’s engine is designed for the workload modern water engineering actually needs: thousands of coupled solves per scenario, differentiable outputs, and a verifiable answer every time.

Book a technical demoAPI & embedding

WHAT IT COMPUTES

What it actually does

  • Aqueous speciation and ion activity

    Speciation and activity across temperature for the composition you measure — not a fixed spreadsheet recipe.

  • Saturation and CCPP

    Saturation indices and CCPP for the mineral set that matters in industrial water: calcite, aragonite, gypsum, silica, phosphates, magnesium silicates.

  • Precipitation and deposition kinetics

    Instantaneous rates and trajectories — deposition at the wall, not only bulk-water heuristics.

  • Corrosion and pitting

    Mixed-potential / Tafel electrochemistry with Faraday-rate and mpy outputs.

  • Gas, exchange, and surfaces

    Gas phases, exchange, and surface complexation where the program requires them.

  • Biological risk context

    Biofilm and Legionella-relevant response terms that feed program and twin layers.

SOLVER FLOW

How a value is produced

Every result carries tier, residual, database, and activity model.

Illustrative model output. Not customer data.

ARCHITECTURE

Built for modeling workloads, not a one-shot batch script

  • A full thermodynamic database registry rather than a hard-coded species set — approximately 30 elements, 235 species, and 90 phases as the full-data target.
  • Caller-specified activity model: extended Debye–Hückel, Davies/B-dot, and specific-interaction models for high-ionic-strength brines — never silently inferred.
  • Vectorized and batched solves for scenario sweeps and twin ensembles.
  • An execution ladder from closed-form to full registry; every result reports the tier used and its residual.
  • An in-process native runtime for production latency, with an independent reference path retained for audit.

Result shape (illustrative)

{
  "value": 2.14,
  "quantity": "LSI",
  "database": "aa-thermo@<version>",
  "activityModel": "b-dot",
  "tier": "full-registry",
  "residual": 1.2e-8,
  "provenance": { "T_C": 40, "envelope": "..." }
}

VALIDATION

We publish how we check ourselves.

Every release is compared against an independent reference-solver fixture set, and results are reported with their residual, the thermodynamic database used, the activity model, and the temperature and composition envelope. An accuracy number without those four things is marketing, not engineering.

DatabaseActivity modelEnvelopeFixture setAgreement
TODO — release DB id / versionTODO — activity model used for this fixture setTODO — T / composition envelopeTODO — fixture set id / versionTODO — agreement metric from release artifact
TODO — release DB id / versionTODO — activity model used for this fixture setTODO — T / composition envelopeTODO — fixture set id / versionTODO — agreement metric from release artifact
TODO — release DB id / versionTODO — activity model used for this fixture setTODO — T / composition envelopeTODO — fixture set id / versionTODO — agreement metric from release artifact

Table awaiting release artifact. No fabricated pass counts or latency targets.

Why not use a general-purpose geochemical solver?

General-purpose research solvers are rigorous, and we validate against them. They are not designed to be called ten thousand times inside an estimation loop, differentiated for sensitivity analysis, or served behind a low-latency API at production concurrency. That is the gap this engine fills.

AI LAYER

The AI layer's floor is this engine

Every insight, alarm, and recommendation in AquaAdvisor consumes solver output — speciation, saturation, kinetics, corrosion electrochemistry — with residual and provenance attached. The AI never invents chemistry; it interprets, supervises, and optimizes over what this engine computes.

How the AI layer works →

HOW IT WORKS

From input to a verifiable result

Every solve reports the tier used, the residual, the database, and the activity model.

  1. Composition and constraints

    Accept measured or design chemistry, temperature, and the activity model the caller specifies — never silently inferred.

  2. Execution ladder

    Select the solve tier on an execution ladder from closed-form to full thermodynamic registry — matched to the workload.

  3. Solve

    Run vectorized and batched solves for scenario sweeps and twin ensembles; in-process native runtime for production latency.

  4. Residual and provenance

    Attach residual, database version, activity model, and solver tier. An independent reference path is retained for audit.

  5. Downstream use

    Return values into twins, design scenarios, alarms, and API consumers with reproducible database pinning.

OUTPUTS

What you get out

  • Speciation, activity, saturation indices, and CCPP
  • Precipitation / deposition rates and trajectories
  • Corrosion and pitting rates (Faraday / mpy)
  • Solver tier, residual, database id, and activity model on every result
  • Batch and ensemble-ready outputs for twin and design workloads
  • Version-pinned chemistry databases for long-term auditability

WHO THIS IS FOR

Who this is for

  • Technical directors who need auditable chemistry they can take into a customer review
  • Service-company modeling teams running thousands of coupled solves per scenario
  • OEM and software teams embedding water chemistry via API or in-process license
  • Service companies that need identical results in cloud or on-premise — including air-gapped accounts

DEPLOYMENT

Same engine in our cloud or entirely on the customer's network

The engine runs in our cloud or entirely inside the customer's network when the account requires it. On-premise deployments ship as containers, require no outbound connectivity, and produce identical results to the hosted service against the same database version.

  • Air-gapped and DMZ-isolated operation supported
  • License-key activation
  • Versioned chemistry database bundles so results remain reproducible and auditable years later
  • Upgrade cadence under the customer's change-control process

Full deployment comparison →

Bring the chemistry at your accounts into focus.

30-minute technical walkthrough with an engineer — bring a water analysis from an account and we will model it live.

Book a technical demoTalk about API access