Glass manufacturing
Catch bubbles, inclusions, scratches, and surface defects before bottles, jars, and containers leave the cold end.
Automated quality inspection for glass manufacturing, running on a refurbished iPhone alongside the IS machine, the cold end, and the palletiser.
What is automated quality inspection for glass manufacturing?
AI defect detection for glass manufacturing uses a camera and an AI model to watch every container or sheet as it leaves the IS machine, the lehr, or the cold-end inspection station, and to flag non-conforming product before it reaches the palletiser. Instead of relying on rigid rule-based vision tuned for a single bottle shape, the AI learns the specific container geometry, glass colour, and surface signature of your portfolio, and applies a consistent visual checkpoint across shifts, mould changes, and shade variations.
Container glass is particularly hard to inspect at line speed because the optical refraction of the glass itself complicates every camera read: a bubble inside the wall and a reflection on the surface look almost identical from a fixed angle, and a stone in the heel can be invisible from a side view but obvious from below. Rule-based vision built around a single bottle shape breaks the moment you swap to a different SKU, a different colour, or a different finish. AI-led inspection handles those variations because the model learns from real production frames rather than from a fixed threshold.
The result is an automated visual checkpoint that complements your end-of-line cold-end station and gives you a container-by-container image record. When a customer query comes back six weeks later, you can pull the frames from the exact production window and either confirm the defect or push back with evidence.
Defects we catch on container-glass lines
Bubbles, seeds, and inclusions
Bubbles and seeds are gas pockets trapped in the glass wall, caused by under-fining of the melt, refractory degradation in the furnace, or excess water in the batch. Inclusions are foreign material caught in the glass, ranging from refractory chips to unmelted batch particles, and they show up as dark or coloured spots inside the wall. Cold-end inspectors catch the obvious bubbles but miss the sub-millimetre seeds that ruin a clear-glass cosmetic bottle and the dark inclusions that look like surface dust under cold-end lighting. The AI model learns the in-spec wall texture and flags every seed and inclusion above your acceptance threshold.
Stones and unmelted batch
Stones are crystalline inclusions in the glass, caused by refractory degradation, devitrification of the melt, or unmelted batch particles that survive the fining stage. They show up as opaque spots in the wall, often with a halo of stress around them, and they can crack the container weeks after delivery when temperature cycles open the stress field. Manual inspectors catch the larger stones but miss the small ones that pass the side-view station. The AI model picks up both the opaque spot and the surrounding stress pattern from a single frame and flags the container before the lehr exit.
Surface scratches and abrasions
Surface scratches and abrasions are produced when containers contact each other or contact line equipment between the IS machine and the palletiser, and they show up as fine longitudinal lines on the bottle body. Severe abrasion patterns weaken the surface and trigger pressure failures on the customer's filling line. Manual inspectors catch the worst cases under cold-end lighting but miss the borderline scratches that pass and fail at the customer. The AI model learns the in-spec surface finish for each container shape and flags scratches above your tolerance, with the frames available so the operator can adjust line guides or palletiser tooling.
Stress cracks and checks
Stress cracks and checks are fine cracks in the glass wall, often invisible to the naked eye but lethal under fill or transit pressure, caused by uneven cooling in the lehr, sudden temperature changes, or poor mould design. Polariscope inspection catches some checks but misses the surface ones that develop after the lehr. The AI model holds the in-spec polarised reflectance signature and flags stress patterns at the lehr exit, giving the line a chance to correct upstream conditions before a pallet of cracked containers reaches the customer.
Dimensional defects and out-of-round
Dimensional defects are deviations in container height, body diameter, neck diameter, finish dimensions, and capacity, caused by mould wear, uneven gob delivery, or temperature drift in the IS machine. Out-of-round bottles fail the customer's capping line and trigger full-pallet rejections. Calliper sampling at break catches the trend but misses the windows in between. The AI model picks up the silhouette deviation across multiple angles and flags the containers that fall outside your acceptance band before they reach the palletiser.
Surface stains and lubricant marks
Surface stains include hot-end coating drips, cold-end lubricant marks, and oil or grease contamination from line equipment, and they show up as cloudy patches or streaks on the container surface. Severe stains trigger label-adhesion failures at the customer's bottling line. Manual inspectors catch the obvious cases but miss the gradual drift that builds up after a coating-spray miscalibration. The AI model holds the in-spec surface clarity for each colour and flags stains as soon as the local clarity delta exceeds your spec.
The lighting setup that makes this work on a container-glass line is a diffuse backlight at the lehr exit to read bubbles, seeds, and stones, plus a polarised filter for stress patterns and a low-angle ring light at the cold end to read surface scratches and stains. An iPhone Pro with macro and wide-angle lenses handles the seven defect families from a single inspection station per critical control point. We synchronise the rig with the line encoder so flagged containers drive a downstream divert decision before the palletiser. We spec the optics with you during onboarding.
How Enao runs on a container-glass line
The full hardware rig costs less than €1,000 and consists of a refurbished iPhone Pro, a diffuse backlight with optional polarised filter and low-angle ring light for surface inspection, a USB-C cable, and a mount that clamps over the lehr exit, the cold-end station, or the palletiser infeed. PLC integration is not required for the first deployment, the rig fits in a flight case, and the line keeps running while you set it up.
Onboarding is self-serve. Your line team mounts the rig, opens the Enao app, and starts collecting reference frames at the next changeover. Day one returns 80% accuracy without any prior labelling, and by day fourteen the model is operating above the manual inspector on the defect families it has seen, improving with every flagged container that the line confirms or rejects.
Each line teaches its own model what its glass colour, container geometry, and surface finish look like. When you swap to a new SKU on the same machine, the model adapts in a single shift. When you bring a sister line online with a similar product family, the second model starts from the first model's experience and the marginal effort drops sharply.
Out-of-spec containers stop reaching the palletiser, scrap is logged at the inspection point rather than at the QC office, and your operators get back the hours of attention they need for the parts of the job that still need a human, including IS machine setup, mould wear management, and customer complaints.
How Enao compares to manual inspection and traditional machine vision
For container-glass producers the comparison sharpens around five dimensions.
Setup time on a container-glass line. — Manual cold-end sorting misses bubble and inclusion defects at speed. Traditional machine vision (Cognex, Heraeus, Robovision, averroes, industrialmind) requires three to nine months of integration and a six-figure budget. Enao is deployed in a week by your own team on a refurbished iPhone, day one at 80% accuracy.
Hardware cost per line. — Manual visual inspection: none upfront, ongoing labour cost. Traditional cold-end vision: €150,000 to €500,000 per line for industrial cameras, multiple inspection heads, and integration. Enao: under €1,000 per line with a refurbished iPhone Pro, lamp, and mount.
Handling new colours, shapes, and finishes. — Manual visual inspection: re-train inspectors for every new SKU. Traditional cold-end vision: rewrite the recipe per SKU, often outsourced to the integrator. Enao: re-teach the model on new colours and shapes in a single shift, no code to touch.
Detection accuracy on subtle seeds and stress patterns. — Manual visual inspection: high at shift start, drops measurably after three hours. Traditional cold-end vision: strong on dimensional checks, weak on subtle seed drift and surface-stain progression. Enao: learns wall and surface signatures from reference frames and holds accuracy across shifts and runs.
Who runs it. — Manual visual inspection: trained cold-end inspector. Traditional cold-end vision: system integrator or a specialised vision engineer. Enao: your line team, no external specialist required.
FMCG and pharma customers change vendors over the cost of a glass-fragment recall, and the cost of a chargeback or a quiet category-manager phone call sits well above the cost of an iPhone-based inspection rig. Enao is built for that gap.
