FMCG — Snacks Manufacturing
Data latency achieved, down from once-daily PLC pulls
Of emerging bottlenecks flagged before they cost output
About
The manufacturer needed a safe flight simulator to test line speeds and shift patterns before touching the shop floor across four production lines. PLC data arrived once a day, hiding micro-stoppages that cost 6% of OEE and produced costly overtime. We built a WebGL digital twin streaming live MQTT data into a discrete-event engine.
Industry
FMCG — Snacks Manufacturing
Company size
500 – 1,000 employees
Founded
1998
The Company
Four lines, one blind spot
The manufacturer runs four production lines producing a fast-moving snacks portfolio. Line-level PLC data existed, but it was pulled into reporting once a day — enough for a retrospective view of yesterday's output, useless for catching a problem while it was still happening.
Engineers wanted to test line-speed and shift-pattern changes before committing to them on the actual shop floor, but had no simulation environment to do that safely.
The challenge
Micro-stoppages that never showed up until the next morning
Because PLC data refreshed only once a day, short stoppages — a jam, a changeover running long, a minor mechanical fault — were invisible in real time. By the time they appeared in a report, the shift that suffered from them was already over.
These micro-stoppages were quietly costing about 6% of OEE and driving overtime as supervisors tried to make up lost output after the fact rather than catching the cause while it was still fixable.
The Solution
A WebGL twin fed by live MQTT streams
We built a WebGL-based digital twin of all four lines, streaming live MQTT sensor data into a discrete-event simulation engine so engineers could trial any scenario — a shift pattern, a line-speed change — and watch bottlenecks surface in real time rather than discover them after the fact.
The same live stream feeds the operational floor directly, so a developing stoppage shows up as it starts rather than in tomorrow's report, giving supervisors the lead time to intervene before it costs the shift.
The Results
Latency down to 3 seconds, five stoppages averted in month one
Data latency dropped from a once-daily refresh to an average of 3 seconds, and the twin now flags roughly 90% of emerging bottlenecks early enough to act on them. In the first month alone, five stoppages were averted before they became a lost shift, saving an estimated Rs. 25 lakh.
Engineers now use the twin to test shift-pattern and line-speed changes before ever touching the physical line, removing the risk that used to come with any operational change.
KEEP READING
Explore more studies on AI, analytics, and enterprise intelligence.
