The table of content
Introduction
The purpose of this use case is to put numbers and clear rules around something that operators and quality teams often feel only “by experience”: how veneer thickness and cutting quality along the FC + CVX line affect scrap, rework and customer complaints. Instead of looking only at end‑of‑line defects, the solution combines thickness, cutting and detection signals into one view of how stable the process really is and where problems start. The goal is to catch issues as early as possible—ideally already at peeling, drying or first cutting—so that defective sheets can be corrected, downgraded or rejected before they turn into expensive scrap after pressing or, worse, problems at the customer.
Veneer thickness and cut quality are critical because they directly determine how well joints survive pressing, handling and final use. If certain zones are too thin, if thickness varies too much across the width, or if edges are frayed or misaligned, joints are more likely to crack, delaminate, or break at the edge even when the press program looks correct. On the FC + CVX line this is especially important, because sheets from different zones and species are combined into complex lay‑ups; a local weak spot in one zone can compromise an entire panel. By systematically linking thickness profiles and cut events to actual OK / rework / scrap outcomes, the use case makes these hidden relationships visible and gives engineers concrete levers to stabilise quality.
The key objectives are to detect thickness and cutting issues early, learn which patterns lead to which defects, and translate these insights into practical recommendations: optimal thickness windows per species and moisture level, limits on under‑tolerance per zone, and clear guidelines for peeling, drying and cutting parameters. The intended users are quality engineers, process engineers and production supervisors who need a shared, data‑driven picture of where quality is lost and how to adjust the process. Operators benefit as well, because alerts and dashboards give them simple signals—such as “Zone 10 under‑tolerance rising” or “edge quality deteriorating”—that explain why a change in settings or extra inspection is being requested.
Data Used
Thickness measurement sensors and moisture‑related information
The solution starts from the thickness sensors already installed along the FC line, typically arranged as several measurement points (for example zones 1..12) across the veneer width. These sensors provide continuous or sampled thickness values as sheets pass, allowing the system to reconstruct how thick the veneer is at the edges, in the centre, and in any critical zones. For each veneer species and grade, calibrated nominal thickness and tolerance bands are loaded from existing recipes or quality standards, and, where available, moisture information from drying or NIR/moisture sensors is added. Together, this makes it possible to say not just “this sheet is thin,” but “for this species at this moisture level, this zone is outside the safe window.”
Cutting and composition signals (cutoff, composition shear, miscuts)
Cut‑related signals describe how well sheets are being cut and composed into joints. Sensors at cutoff stations track when cuts happen, how long each cutting cycle takes, and whether cycles are regular or show outliers such as double cuts, missed cuts or aborted cycles. Composition shear sensors provide similar information for the joint‑forming step, including start/end of each cycle and its timing relative to sheet movement. By combining these signals with thickness data, the system can see, for example, whether miscuts cluster when thickness is low in certain zones or when cycle timing drifts from its normal pattern.
Workpiece detection, edge quality and thread‑guard signals
Workpiece detection sensors before and after cutting and gluing indicate whether sheets are present, correctly aligned, or missing altogether at key points in the process. Luminescent or edge‑detection sensors provide a proxy for edge quality, flagging situations where edges look fuzzy, chipped or misaligned rather than clean. Thread guards add another layer by detecting loose threads, overlaps or stray material that can interfere with cutting or create local defects in joints. These signals help the system distinguish between thickness‑driven problems and issues caused by poor cutting, overlaps or contamination along the edges.
Quality outcomes, defect codes and production/material context
Finally, all of this process data is linked to what actually happened to each panel or joint: whether it passed as OK, needed rework, or ended up as scrap, and which defect codes were assigned (joint cracks, delamination, thickness out of tolerance, edge breaks, and so on). Additional context—veneer species and grade, moisture range, product type and lay‑up structure, press recipe, shift, line speed and key upstream settings—provides the background needed to compare like with like. This combination of detailed process signals with clear quality labels is what allows the models to learn which patterns of thickness and cutting behaviour genuinely matter for quality, and which are harmless variation.
Technical Realisation
Deriving thickness profiles and tolerance checks per sheet and zone
From the raw sensor streams, the system reconstructs a thickness profile for each sheet across all measurement zones. For every pass it calculates simple but powerful statistics: average thickness, minimum and maximum values, variation across the width, and how often readings fall below or above the tolerance band for that veneer species and moisture range. This makes it possible to say, for example, that “Zone 10 on the right edge is below tolerance in 12% of readings this shift,” or that “edge zones are consistently thinner than the centre,” which is far more actionable than a single overall average.
Cut‑quality and edge‑quality indicators from cycle and sensor data
In parallel, the system analyses cut cycles and edge‑related signals. It measures cycle times for cutoff and composition shear, detects irregular cycles (double cuts, missed cuts, aborted cuts), and checks their timing against workpiece‑detection signals so that events like “cut without sheet” or “sheet without confirmed cut” are flagged. Edge‑quality proxies from luminescent sensors and thread guards are summarised as simple flags—such as “weak edge signal,” “repeated thread‑guard activations,” or “overlap suspected”—so they can be compared across shifts and products. Together, these indicators quantify how “clean” and stable cutting and edge formation really are, instead of relying only on spot checks or visual impressions.
Linking process signals to OK / rework / scrap outcomes with models
The next step is to link these thickness and cutting indicators to what ultimately happened to the panel or joint. For each finished piece, the system looks back at the thickness profile, cut cycles, and edge signals recorded while its veneer was processed, and assembles them into a feature set alongside species, moisture, product type and press settings. Supervised learning models—regression and classification—are then trained to predict whether a panel is likely to be OK, need rework or become scrap, or to estimate a continuous defect‑risk score. Once trained, these models can score new production in near‑real time and highlight panels or conditions that resemble past problem cases.
Discovering patterns that drive joint cracks, delamination and edge breaks
Because the aim is to support engineers, not just produce scores, the models are analysed for interpretability. Feature‑importance and rule‑extraction techniques are used to highlight patterns such as “a high share of under‑tolerance readings in zone 10 strongly increases the chance of joint cracks,” or “large thickness variation across the width, combined with certain moisture ranges, is often present before delamination defects.” Similar patterns can link weak edge sensor signals and frequent thread‑guard hits to edge breaks after trimming. These relationships are presented as clear rules and examples in the dashboards, so quality and process engineers can see which combinations of thickness, cutting and edge signals are most harmful and where to focus corrective actions first.
Recommendations and Online Monitoring
Optimal thickness windows per veneer species and moisture level
Once the models have learned how thickness and cutting behaviour relate to quality outcomes, the system can propose optimal operating windows rather than just flagging problems. For each veneer species and moisture range, it identifies the thickness interval where the probability of an OK panel is highest and scrap is lowest, along with limits on how many under‑tolerance readings are acceptable in each zone. The result looks like a simple recipe: “For species A at 6–8% moisture, target 0.60–0.65 mm; under‑tolerance share in any edge zone should stay below 5%.” These windows give quality and process engineers a concrete target to aim for when tuning the line.
Upstream peeling and drying adjustments derived from the models
Deviations from these optimal windows are then translated into upstream actions that operators and technologists can understand. If the system sees that certain zones are regularly too thin for a given species and moisture, it can suggest increasing the peeling thickness setpoint by a small amount or adjusting knife settings. If thickness shrinks too much after drying in specific conditions, it can propose changes to dryer temperature or line speed to reduce variation across the sheet. In this way, abstract model outputs become clear “turn these knobs” recommendations for peeling and drying, helping to stabilise veneer before it even reaches FC and CVX.
Online quality‑risk indicator and early warnings based on live signals
In day‑to‑day operation, the same models power an online quality‑risk indicator that runs on live thickness and cutting signals. As sheets pass the sensors, the system estimates the current risk of defects for the material being produced and compares it against thresholds. When risk climbs—because, for example, under‑tolerance in a zone spikes or cut cycles become irregular—it raises straightforward early warnings such as “Zone 10 under‑tolerance 3× baseline in the last 30 minutes; expect higher joint‑crack risk, check peeling and dryer settings.” Depending on the plant’s preferences, it can also recommend temporary actions like slowing the line slightly, routing borderline material to less critical products, or triggering extra inspections until conditions are back in the safe window.
User Interface, Alerts and Workflows
Quality analytics dashboards for shifts, zones and defect types
The user interface brings all the pieces together in dashboards that are easy to read during daily meetings and on the shopfloor. One view compares scrap and rework rates by shift, alongside a model‑based quality‑risk curve, so teams can see whether recent actions are actually reducing problems. Other views focus on thickness zones, showing which zones have the highest share of under‑tolerance readings and how strongly they are linked to specific defect types such as joint cracks or edge breaks. This makes it straightforward for quality and process engineers to move from “we have too much scrap” to “these three zones, in this species, are driving most of our complaints.”
Alerts on under‑tolerance zones, cut irregularities and edge‑quality drift
On top of the dashboards, the system provides targeted alerts that watch for early signs of trouble. If a zone’s under‑tolerance frequency rises above its limit, if cut cycles become irregular, or if edge‑quality indicators from luminescent sensors and thread guards deteriorate, an alert is triggered with a plain‑language description and suggested checks. For example, “Zone 12 edge under‑tolerance increasing; inspect peeling knife and dryer profile,” or “Cut cycle outliers at composition shear; check timing and sensor alignment.” These alerts can appear on quality dashboards, HMIs, or in email/messaging tools, depending on how the plant prefers to work.
Support for quality and process decisions (recipes, reports, improvement projects)
To support longer‑term decisions, the system also generates information that can be used in recipes and improvement projects. Recommended thickness windows and process parameters per product are summarised so that standard operating recipes can be updated and kept under control. Periodic reports highlight where upstream drift—such as a change in peeling behaviour or dryer performance—is causing higher scrap, and estimate the potential gain if specific issues are fixed. This helps managers and engineers choose which projects to tackle first and provides a data‑based story when discussing quality with operators or customers.
Business Value and Conclusions
Reduced scrap, rework and complaints through earlier detection
By linking thickness, cutting and edge‑quality signals to downstream outcomes, the use case makes it possible to catch problematic patterns well before panels reach the press or the customer. Instead of discovering issues only when joints crack in finished boards or when complaints arrive, the plant can spot rising under‑tolerance in a zone, unstable cutting cycles or edge‑quality drift in near real time and react with targeted adjustments. This typically reduces scrap and rework rates, cuts the amount of labour spent on repairs, and lowers the risk of costly customer claims.
More stable product quality and better use of raw veneer
Clear thickness windows per species and moisture level, combined with concrete recommendations for peeling, drying and cutting, help keep veneer much closer to its “sweet spot” in everyday production. That leads to more consistent joint strength and edge quality across orders and shifts, so quality becomes less dependent on individual operator experience. At the same time, better control over thickness reduces over‑thick veneer that wastes raw material and under‑thick veneer that turns into scrap, improving yield from each log and making the most of available veneer resources.
Foundation for predictive quality analytics across the FC + CVX line
Perhaps the biggest long‑term benefit is that this use case creates a reusable data and model foundation for predictive quality analytics on the entire FC + CVX line. The same infrastructure that now connects thickness and cutting to joint defects can be extended to include additional sensors, vision systems, press data or customer feedback, gradually building a richer picture of how process choices influence quality. With each new dataset and each confirmed pattern, the models become more accurate and more helpful, enabling the plant to move from reactive quality control toward a predictive regime where risks are anticipated and prevented, and high, stable quality is simply how the line normally runs.
Our success stories
.avif)
Tell us about your project needs
.png)
.png)
