Alcyone
The dose, argued from chemistry.
Every morning a treatment operator decides how much of four chemicals to feed, against raw water that changed overnight and targets that don’t move. Alcyone reads the day’s raw water, predicts what the settled water will do, and recommends the cheapest dose set that still clears the plant’s limits — then keeps the ledger of what it recommended, what was fed, and what the plant actually produced.
Visit alcyone.us.comAsk about your plant
A report, not a control: no SCADA connection, no setpoints written, nothing that goes on a state form. The certified operator decides.
What it answers
The numbers the plant is judged on
Settled turbidity, settled dissolved aluminum, and settled dissolved manganese — plus settled UV254 and chlorite — predicted from the day’s raw water and the doses of coagulant, permanganate, chlorine dioxide, and powdered activated carbon.
The cheapest dose that still clears
A constrained optimizer searches the dose space against the plant’s own hard limits and targets, using the delivered cost of each chemical as it’s actually fed — so the recommendation is the least expensive feed that meets the spec, not the safest-looking one.
Chemistry first, statistics second
A mechanistic core does the reasoning — aluminum solubility, oxidant demand allocation for manganese, chlorite yield, carbonate chemistry, Edwards DOC sorption — with a residual model and honest prediction intervals layered on top, rather than a black box fit to a spreadsheet.
Calibrated to your plant, not a vendor default
Coefficients are fit to the plant’s own lab record and jar tests, and every fitted number carries its provenance. Until that fit has run, the model says out loud that it’s standing on literature placeholders.
A savings ledger you can audit
Recommended against fed against produced, day by day, priced with the plant’s own invoiced chemical costs — and labeled “illustrative” whenever those costs are still estimates.
Runs as a page, not a project
A browser URL on a small always-on host, behind authentication, with a dead-man’s switch watching it. No plant-floor install, no agent on the control network.
Where the line is
A tool that advises a licensed operator has to be exact about what it is — so Alcyone is built with the boundary written into it, not bolted on as a disclaimer.
- It never touches the processNo SCADA integration, no setpoint writes, no CT computed for reporting, nothing produced for a state form. It is a reading aid for the person who holds the license.
- It targets a number that is safe to advise onSettled water quality is an operational judgment, not a reported compliance value — which is exactly why it is the right place for a model to help.
- Uncertainty is shown, not smoothedPredictions come with intervals, and each one ships with a model card saying what it is and is not good for. Cards stay empty until the plant's own calibration has run.
- Plant two is configurationThe treatment train, feed points, chemicals, costs, limits, and targets are a config file; reading another utility’s data export is a mapping, not a rewrite. Every plant still gets its own fit.
Where it stands
Alcyone is running in pilot on a Front Range conventional surface-water plant, with calibration against that plant’s own multi-year lab record underway. It’s built to take a second plant as configuration rather than a rebuild — so if you run a conventional plant and your operators are making this call on feel and a jar test, it’s worth a conversation.
Start a conversationalcyone.us.com →
Built and operated by True Ascent Labs. Advisory output only — every dosing decision stays with the plant’s certified operator.
