Detecting delamination in OSB before the hot press — a badly formed mat is a pallet of scrap.
Delamination in OSB is almost never visible on the final pallet — it shows up on site when the structural board fails. iLEAN Vision is an Edge terminal with a CNN over the belt, right before the hot press, that detects moisture outside range, poor resin distribution and deficient mat orientation, and pulls the stack for reforming. Raw material is recovered, and no energy is spent on pieces headed for scrap. The person signs.
The defect is not visible in the plant — it is visible on site, and by then it is too late.
Delamination in OSB is one of the costliest defects in engineered wood precisely because it almost never appears in the plant. The pallet leaves looking perfect, passes the operator's visual check, reaches the distributor, reaches the site, and there — when the board is cut for structural sheathing or gets wet during temporary exposure — the layers separate. The return is not one piece: it is a whole shipment and, with the wrong customer, a lost account.
The root cause is born three steps earlier, in the mat — the blanket of resinated strands before it enters the hot press. Three families of things can go wrong:
- Moisture outside range in the strands — unexpected rain at the drying yard or a bypass in the dryer produces strands 2-3 points of moisture above recipe and the resin does not cure properly.
- Uneven distribution of MDI or PF resin — clogged nozzles or a supplier change with a different viscosity leaves parts of the mat over-applied and others under-applied.
- Poor strand orientation in the core/face layers — a calibration change in the former makes the oriented pattern lose alignment.
None of the three is visible in the finished board. After the hot press, the energy is spent, the resin consumed and the cycle closed. The classic system works 99% of the time. The other 1% is the shipment that comes back from the site.
iLEAN Vision puts eyes where there are none — before the press.
The problem is not a lack of information: it is information living on islands that, at the critical moment (the mat moving at 2 meters per minute toward the press), does not reach the decision-maker in time. iLEAN acts as the putty between the dryer, the former, the resin applicators and the press, without asking you to change either the SCADA or the MES.
Vision sees the mat before the press. Connect reads the layer recipe and the dryer moisture. The agent cross-references them, and if the mat falls outside the window, it holds the stack. The person signs — never the other way round.
The iLEAN pieces applied to delamination detection in OSB:
- iLEAN Vision (Edge) — a terminal with machine vision (CNN) over the mat belt, trained on real OSB patterns from your plant. It picks up apparent density, resin distribution, moisture marks and strand orientation in core/face. If the mat drifts away from the recipe window, it fires the actuator (light stack, stack diversion) in milliseconds. It works without a network: as long as it has power, it infers locally and fires locally.
- Connect — it picks up the real dryer moisture (even if it lives in an old isolated PLC), the layer recipe from the MES and the resin batch changes the buyer notes in a spreadsheet. It also captures alerts from external channels (an email from the supplier about a resin viscosity change, a WhatsApp from the shift lead about last night's rain) at second zero.
- Agent — it cross-references the mat image, the real moisture, the layer recipe and the line's history. If the combination of variables predicts delamination, it holds the stack and alerts the shift lead on whatever channel they use. The person validates and signs; the former does not restart on its own.
Inspection at the final pallet vs. detection in the mat with iLEAN Vision
| Aspect | Classic inspection at the final pallet | With iLEAN Vision + Connect + Agent |
|---|---|---|
| Point of detection | Finished board or a return from the site | The mat, before the hot press |
| Raw material recovery | Zero — everything is scrap | Mat reformed or fed back in |
| Energy and resin spent | Already consumed on pieces that were going to fail | Avoided in the cycle of the held stack |
| Resin batch change | Reaction after the incident | Connect picks it up at second zero |
| Operation without a network | n/a | Vision keeps inferring on the power in the cabinet |
| Evidence for a customer claim | Rebuilt by hand | Mat photo + recipe + curve, per batch |
Impact estimate for your plant — to be validated with your numbers.
The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.
- OSB plant with one continuous press line, a mix of structural and secondary board, returns for delamination documented over recent years.
- Vision pilot on the mat belt before the press (industrial camera + Edge terminal with CNN + actuator + integration with the dryer SCADA and the MES). First value expected within a few weeks, with the first stack pulled before the press.
- Indicative payback between 4 and 9 months, depending on volume, scrap cost per pallet of structural board and the historical share of returns for delamination.
- The hard lever is every return from a site avoided: product recovered from the customer, reverse logistics, account damage. Expected reduction in returns for delamination ≥ 30% over the baseline.
And the production manager's reasonable doubt
"What if the AI gets it wrong and lets a bad mat through?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely compares an image of the mat against the layer recipe, the best models brought error below 1.5% [1]. And even so, what is critical is never decided alone: iLEAN holds the stack and the person signs. The three safety rings exist precisely for this.
[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.
What people ask about delamination detection in OSB
What causes delamination in OSB and when is it detected?
Delamination in OSB comes from three families of causes: moisture outside range in the strands, uneven distribution of MDI or PF resin in the mat before the press, and poor strand orientation in the layers. The internal defect is almost never visible to the naked eye in the finished board — it shows up when the customer cuts it or when the structural board fails in use. That is why what matters is seeing it in the mat before it enters the press, not after palletizing.
How does iLEAN Vision see delamination before the hot press?
iLEAN Vision is an Edge terminal with machine vision (CNN) trained on the real patterns of the OSB mat. An industrial camera over the belt, right before the hot press, captures the mat surface — apparent density, strand distribution, patches of over-applied or under-applied resin, moisture marks. The CNN compares the pattern against the layer recipe (core/face) and fires an actuator if the mat falls outside the window. The stack is pulled for reforming or diverted to secondary product; the pallet of structural board is not contaminated with pieces that will delaminate on site.
Why inspect in line instead of at the end of the process?
Because after the hot press you have already spent the energy, the resin, the time and the cycle. A delaminated OSB board caught at the final pallet is complete scrap — neither the mat nor the resin can be recovered. Caught in the mat, before the press, you recover the raw material: it is reformed or fed back in. The pattern is the same as in powder coating: catching the defect before the oven lets you recover both the piece and the material; catching it afterwards is pure scrap.
Does it keep working if the vision system loses its network?
Yes. iLEAN Edge is designed so that the critical cycle of detection and intervention does not depend on the internet or on the central SCADA. As long as the terminal has power, the CNN infers locally and fires the actuator. When the network comes back, it syncs the history with the central agent and traceability is rebuilt batch by batch. In a structural board plant, where the press does not wait for WiFi to return, this is not an option — it is the only defensible architecture.
How much does it cost and how long before the first value?
The order of magnitude of a Vision pilot on an OSB line is close to that of any Edge pilot in an industrial plant: an initial investment covering camera, terminal with GPU, actuator, integration with the press SCADA and MES, plus an annual license. First value expected within a few weeks, with the first stack pulled before the press. Indicative payback between 4 and 9 months, depending on volume, scrap cost per pallet of structural board and the historical share of returns for delamination. We send you the estimated ROI in 48h with your plant's data.
Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your OSB plant.
We work on your plant's real data, not ours. Diagnostic with no commitment.
Request estimated ROI in 48h See furniture and wood