Soft Sensing

ML-based virtual sensors that estimate hard-to-measure variables in real time — product quality, composition, emissions and efficiency — from instrumentation you already have.

Free AI Consultation

All solutions

INSTRUMENTS IN · ESTIMATE OUT

// What it is

How Soft Sensing works for your plant

A soft sensor is a machine-learning model that estimates a variable that is difficult, slow or expensive to measure directly. Trained on your historian and lab data, it reads the live signals your plant already records and publishes a continuous estimate — like any other instrument, but without new hardware.

SOFT SENSOR // SIGNAL PATH

TI-204 · 84.6 °CPI-101 · 6.3 bargFI-330 · 416 m³/h

ML MODEL
GBM · v2.3 · 42 MS

QC-EST · 98.2 % ± 0.3OPC UA · EVERY 5 S

EXISTING INSTRUMENTS IN — LAB-GRADE ESTIMATE OUT

// How it works

From data to running system

Collect

Historian and lab data assembled and cleaned.

Train and validate

Model benchmarked against your lab reference.

Deploy plant-side

Runs on-site, reading live tags.

Publish

Estimate published via OPC UA every few seconds.

// Industries

Where it applies

// Lab vs soft sensor

What changes when the estimate is continuous

	LAB SAMPLE	SOFT SENSOR
Latency	Hours — sample, transport, analyze	Seconds — published continuously
Coverage	A few samples per shift	Every cycle, every unit
Cost per value	Technician time, reagents, lab load	Existing instruments, no consumables
Use in control	Too late to act on	Direct input to MPC and RTO
Validation	Is the reference	Benchmarked against the lab before go-live

// Benefits

What it changes

SS-01 // IMPACT BOARD

Lab-quality visibility, continuouslyOK Fewer manual samples and lab delaysOK Quality deviations caught earlyOK Ready-made inputs for MPC and RTOOK

FROM THE FIELD

Stop waiting hours for the lab. A validated soft sensor publishes lab-grade estimates every few seconds.

Book a free consultation

// FAQ

Common questions

How accurate is a soft sensor compared to the lab?+

Accuracy is benchmarked against your lab reference before go-live and reported in your units, with a confidence band published beside every estimate.

What data do we need to start?+

Most engagements start with 6–24 months of historian, lab or document data. We audit coverage in the first week and tell you plainly if it is not enough — before you commit.

How long until we see results?+

A proof of concept on your historical data typically lands in 4–8 weeks. Production deployment follows once the KPI target is met and your team signs off.

Where does it run, and who sees our data?+

On your infrastructure — plant-side servers, your Azure or AWS tenancy, or an air-gapped network. Data never leaves your boundary, and nothing is used to train third-party models.

Who owns the result?+

You do. Handover includes source code, documentation, training sessions and a retraining plan.