Case studyFood & beverageImprove quality & yield

First-pass yield up 2.1pp on the dryer line.

A dairy producer replaces lab-delayed moisture checks with a real-time soft sensor, catching off-spec batches before they're packed.

Talk to an expertDownload 2-page PDF
Soft sensorForecastingOperator workflow
BATCH · dryer_Bsoft-sensor v2
Predicted lab quality vs target — in band
At a glancetimeframes and caveats below
+2.1pp
first-pass yield
6 months
Audited line; other lines staged
−38%
off-spec rework
Same period
Internal QA records
8 weeks
pilot to production
Single line
Historian access in week 1
4
operators trained
Before go-live
Coverage on every shift
Challenge

What was happening before.

Off-spec batches were discovered too late to correct. Rework was costly and rejection rates depended heavily on which shift was running.

Constraints
  • Soft sensor must be explainable to a QA auditor.
  • Cannot run closed-loop on a regulated product line — operators stay in control.
  • Recommendation must use existing PLC tags only — no new instrumentation budget.
Solution

What we built, on what data.

Data
  • › Historian tags: inlet/outlet air temp, feed rate, exhaust humidity, atomiser pressure
  • › Lab measurements: moisture, particle size
  • › Batch records: SKU, start/end, operator
  1. 01
    Time-align lab samples with PLC tags
    Used Data Explorer to map historical lab moisture to the right point in the batch trajectory.
  2. 02
    Train a per-SKU soft sensor
    Predicts outlet moisture every 30 seconds with a calibrated confidence band.
  3. 03
    Push prediction to HMI tile
    Operator sees live moisture estimate, deviation from target, and a 15-minute forecast.
  4. 04
    Weekly recalibration with QA
    QA team re-labels drift days; model refits weekly with audit trail.
Adoption

Who uses it, and where it fits in the day.

Every dryer operator on every shift uses the HMI tile. The QA lead owns the weekly recalibration. Engineering reviews drift monthly.

Modules in use
Soft sensorForecastingOperator workflow
Results

Quantified, with timeframe and caveats.

  • R1First-pass yield up 2.1 percentage points across the audited line.
  • R2Off-spec rework reduced by 38%.
  • R3Operator-to-operator variability on moisture dropped meaningfully — measured by inter-shift sigma.
BATCH · dryer_Bsoft-sensor v2
Predicted lab quality vs target — in band

Operational outcome (proof rule): First-pass yield improved by 2.1 percentage points over six months on the audited line, with a 38% drop in off-spec rework.

Voice from the floor
"The moisture estimate is in front of the operator before the batch is half-dry. We adjust feed rate ourselves; we don't wait on the lab to tell us we already lost the batch."
Senior Process Operator · Dryer line, B shift · Nordic Foods
Replication

Relevant for teams trying to…

F&B teams running batch dryers, evaporators, or any unit where the spec is confirmed off-line.

Solution: Improve quality & yield →Industry: Food & beverage →
Lessons
  • A soft sensor only earns trust if its confidence band is visible. Hide the uncertainty and operators learn not to act on it.
  • Per-SKU models beat one big model — small data, high variance between products.
  • Pair model rollout with a one-hour shift training. Don't ship to an HMI without it.
Get started

Talk to an expert about a similar use case.

Tell us your unit and your KPI. We'll come back with what data you'd need and a realistic timeline based on Nordic Foods's rollout.

Book a demoDownload 2-page PDF