For wood and building products lines (OSB, particleboard, MDF) and industrial manufacturing more broadly. Start with a single tag, or a full process evaluation if you're not yet sure which tag matters most, then scale to a line or a plant once it's proven. Every engine below runs against your existing PLC and OPC UA data, and nothing writes back to the PLC until it has earned that trust.
Not sure which tag matters most, or whether your data can support live monitoring at all? These are scoped, stand-alone engagements, on wood and building products lines or any continuous industrial process, and none of them require signing up to a tier first.
We walk the line, map how material flows through it and where it backs up, and review your PLC and OPC UA tag list against your P&IDs to see what's measured against what actually drives quality and output, whether that's a press, a refiner, a blow line or a single unit operation.
Before any dashboard is built, we check whether the data underneath it can be trusted: sample rates, autocorrelation, sensor drift and tag naming, and whether the historian will hold up going from a single tag to fifty.
A point-in-time capability and stability snapshot on the tags you already have: Cp/Cpk, Pp/Ppk and where the process is drifting, delivered as a report rather than a subscription.
Your historian likely already holds months or years of readings. We mine it for patterns, correlations and recurring upsets, so you know what's worth monitoring live before committing to it.
Start with a single tag on the line that's costing you the most, prove it, then scale to the rest of the line and the plant. Each tier adds capability on the same live data: nothing is rebuilt or replatformed to move up a tier, and nothing writes to your PLC until you're ready for it to.
Live control charts and process capability, checked before they're shown, not after.
Everything in Visibility, plus the modeling layer that tells you why a number moved.
Everything above, plus validated output acting on your process: alarms, interlocks, discrete control.
Tier 1 runs SPC, validation and the dashboard. Tier 2 adds anomaly detection and influence scoring. Tier 3 turns validated output into action. All of it shares one live data pipeline.
Xbar-MR charts, Cp/Cpk, Pp/Ppk, sigma drift and out-of-spec counts, computed from sample statistics rather than the population-variance default that quietly inflates confidence.
EWMA and CUSUM with control limits that tighten correctly over the first samples of a run, and a clean state reset on every recalculation, so early points aren't judged against the wrong limits.
Four methods, normalized to a common scale and combined with configurable weights. Where they agree, confidence goes up. Where they disagree, you see it rather than a blended number that hides it.
Lag-1 autocorrelation, Durbin-Watson and effective sample size, with time-ordered cross-validation instead of methods that assume independent samples and overstate how good a model is.
A fast poller filling a bounded buffer and a slower scorer reading snapshots from it, so live scoring never blocks live data collection, and memory stays stable on a long-running service.
Direct integration with Ignition Perspective and similar SCADA layers, including dynamic tag binding so one view scales across many assets, and a health tile for every engine.
A simplified version of the TAG: INFLUENCE_SCORE engine above, running in your browser: Pearson, Spearman and a standardized regression coefficient, combined into one ranking. The real engine also uses PLS and Ridge; those need a proper statistics library and aren't reproduced here.
MATLAB scripts and Simulink models, fed with your live tags or CSV exports. Find where the process backs up, compare what it should do with what it actually does, and try changes on the model before they touch the line.
Dynamic models of tanks, loops, dryers and conveyors, with your real control logic in the loop. Tune a PID or test a sequence change on the model, not the line.
Scripts that pull your historian or CSV data, clean it and run the analysis the question needs: capability, regression, frequency analysis. Handed over documented, so your team can rerun them.
How material actually moves from infeed to stacker, with rates, buffers and downtime at each step. The bottleneck shows up as the step everything else waits on, backed by your own data rather than a guess.
A theoretical model of how the process should behave runs alongside one driven by your real data. Where the two diverge, and by how much, points straight at drift, losses or assumptions that no longer hold.
Example: tank fill and consumption modeled over time and checked against real levels, so a leak or a drifting flow meter shows up before it becomes a shortage or overflow.
Compare rated capacity with what the line actually delivers, and see where the difference goes. Then test the questions that are expensive to try for real: a faster speed, a bigger buffer, a different demand pattern.
Hands-on exercises on a model of your own process. Inject a pump trip, a sensor fault or a demand surge and practise the response, or tune a loop, with nothing at risk on the real plant.
Projects that use the same live tags and validation discipline, scoped to your process.
If it can be triggered from your PLC or HMI and needs custom logic, whether inspection, measurement or verification, we scope it, build it and wire it in over OPC UA as its own dedicated service.
A PLC signal captures an image, measures the part against a trained profile and returns pass/fail. Runs on-site in Python, with no cloud dependency.
Purpose-built logic for each tag: the exact thresholds, calculations and oversight your process needs, not a generic template stretched to fit.
Most teams start with Visibility. Tell us what you're monitoring today and we'll tell you what's realistic to add.
Request a demo