Skip to content

Portfolio · Industrial water treatment

A plant that reports itself,and says what needs doing.

Industrial water treatment is measured constantly and read rarely: the data exists, the interpretation does not, and the gap is filled by an expert visiting the site. The platform closes it — a wireless sensor estate across the vital systems, analytics at the edge and in the cloud, predictive models on top and monitoring around the clock — and it ends where the work is, in a process visualisation an operator reads rather than a table somebody exports.

The system is measured continuously and understood occasionally. That gap is the whole engagement.
01Sector
Industrial water treatment — process monitoring and management
02Estate
Wireless sensor network across the vital systems
03Analytics
At the edge and in the cloud, on one pipeline
04Delivery
Remote monitoring and predictive modelling, around the clock
Act 01/ 033 figures

The estate

Measurement first, and it has to survive the plant.

A sensor network in an industrial setting is not a data problem to begin with; it is a reliability problem. Wireless sensors sit on the vital systems, an edge runtime does the first pass on site so a lost link is a delay rather than a hole, and what reaches the cloud is already a reading rather than a raw stream. The chemistry the plant is actually managing — dosing, fouling, corrosion, microbiological control — is what decides which measurements are worth taking at all.

Wireless sensors across the vital systems, reporting continuously rather than on a visit.
What to measure is a chemistry question before it is a data question.
The first pass happens on site, so a dropped link costs a delay and not a gap in the record.
  1. 01.01

    Continuous, not periodic

    The alternative is a specialist on site with a test kit, which is one sample on one date.

  2. 01.02

    The edge does the first pass

    On-site processing means a connectivity failure delays the record rather than losing it.

  3. 01.03

    Chemistry chooses the sensors

    The estate is designed around what the plant is managing, not around what is easy to instrument.

Act 02/ 033 figures

From the plant to the platform

One pipeline, and everything downstream reads from it.

Readings land in cloud storage, are processed by scheduled functions, and settle into the stores the rest of the platform reads — a time-series history for trend, a document store for state, a relational store for the things that have to join. Building it as one pipeline rather than one per feature is what makes a new sensor type, a new site or a new model an addition instead of a project.

One path in, however many things read from it — which is what makes the next sensor cheap.
History is the asset. A prediction is only ever as good as the record it was fitted to.
Storage, scheduled functions, machine learning and the pipelines between them, all managed.
  1. 02.01

    One path in

    Every feature reads the same ingested record, so two screens cannot disagree about a reading.

  2. 02.02

    History is the asset

    Prediction is fitted to the record, which means the record has to be complete before it is useful.

  3. 02.03

    A new site is configuration

    Adding an estate is registering it, not extending the pipeline that carries it.

Act 03/ 033 figures

The reading

Not a data dump — the two or three things to do today.

The failure mode of industrial monitoring is a dashboard nobody opens, because it answers every question except the one an operator has. Predictive models turn the series into an expectation, expert monitoring watches it around the clock, and the output is a process visualisation: the state of the system, what is drifting, and what needs a person. The number of readings collected is not the measure of the platform; the number acted on is.

The state of the system, at a glance, in the terms the operator already thinks in.
What the series is about to do, so an intervention is scheduled rather than reactive.
Watched around the clock, because a process does not keep office hours.
  1. 03.01

    Insight, not the data dump

    The platform is judged on the readings acted on, not on the readings collected.

  2. 03.02

    Predicted, then scheduled

    A drift caught early is a planned intervention; caught late it is an outage.

  3. 03.03

    Visual, and for the operator

    The last mile is a graphic somebody reads on shift, not a report somebody exports on Monday.

The last word

This shows how Famysys builds for a physical process: instrument what the chemistry says matters, process at the edge so the record survives the site, fit the prediction to a history worth having, and end on the two or three things somebody has to do today. If your plant produces more data than decisions, the last mile is where we would start.

Built in

  • IoT sensing
  • Edge analytics
  • Telemetry pipeline
  • Time-series history
  • Predictive models
  • Remote monitoring
  • Process visualisation
  • Cloud data platform

All projects

Let's build what's next

Technology alone doesn't transform businesses.The right partnership does.

Modernizing systems, building a new product, or exploring AI-driven transformation — Famysys helps you move forward with confidence.

Schedule Executive SessionEmail Us Directly
  • SOC2 Type II Compliant
  • Strict Commercial NDA
  • Zero Lock-In Guarantee

Famysys

We Engineer Clarity

Scan to Connect

Instant digital business card & WhatsApp link