Camera station or dedicated empty bottle inspector?
Dedicated empty bottle inspectors combine cameras with sensors a camera station does not have. HEUFT, for example, describes optical base inspection for low-density foreign objects such as film, cigarette ends and insects; finish and thread inspection for defects, cracks and rust rings; a 360° sidewall inspection with up to eight views and the bottle rotated between stations; and residual liquid detection that finds lye by high frequency and oil or varnish by infrared, at up to 84,000 bottles per hour (HEUFT InLine II IS). Those are the vendor’s statements. The table shows which checks a camera station can take on.
Empty bottle checks: what needs a camera, what needs another sensor| Check | How it is done | Possible with a camera station? | Note |
|---|
| Base: opaque foreign objects, glass, dirt | Camera with light from below | Yes | HEUFT names film, cigarette ends and insects as low-density objects its optical base inspection finds |
|---|
| Finish and thread: chips, cracks, rust rings | Camera with ring or angled light from above | Yes, where in view | Dedicated machines add separate finish and thread optics |
|---|
| Sidewall: inclusions, cracks, contamination | Cameras against a backlight, several views | Partly: the views you can afford | HEUFT describes up to eight views covering more than 360°, with the bottle rotated by servo |
|---|
| Residual caustic lye | High-frequency sensor | No | A non-camera sensor; part of dedicated inspectors |
|---|
| Residual oil or varnish | Infrared sensor | No | A non-camera sensor; part of dedicated inspectors |
|---|
The practical consequence for returnable glass in South Africa: if residual caustic or liquid is a real risk after the washer, a camera station alone is not the answer. Where the risk is foreign objects, chips and damaged finishes on a slower line or on non-returnable containers, a camera station is often the proportionate choice.
Worked example: time per container and tracking to the reject
At 36,000 bottles per hour, ten bottles pass each second, so each has 100 ms for imaging and the decision. With bottles touching at a 100 mm pitch, the conveyor runs at about 1 m/s. If the reject sits 1.5 m after the cameras, about 15 bottles are between camera and reject at any time, which is why each result is tracked by an encoder and shift register rather than by a timer. Figures are arithmetic on the stated assumptions.
Example: line rate, time budget and tracking distance| Bottles per hour | Bottles per second | Time per bottle | Conveyor speed at 100 mm pitch | Bottles between camera and a reject 1.5 m away |
|---|
| 18,000 | 5 | 200 ms | 0.5 m/s | 15 bottles |
|---|
| 36,000 | 10 | 100 ms | 1.0 m/s | 15 bottles |
|---|
| 60,000 | about 16.7 | 60 ms | about 1.5 m/s | about 17 bottles (90 mm pitch) |
|---|
Preforms: inline or sample inspection?
Preform inspection is offered both as 100 % inline inspection behind the moulding machine and as laboratory sample inspection of detailed quality features; INTRAVIS, for example, lists an inline system at 72,000 preforms per hour and a laboratory system at up to 1,200 per hour (INTRAVIS brochure). A line station from addaScan sits in the first group: gate, specks, bubbles and short shots on every preform; dimensions such as neck ovality are a measurement task and stay with the laboratory or a dedicated gauge.
Light and acceptance
Base and sidewall checks use transmitted light: a backlight shows the container as a silhouette in which opaque objects block the light, and shadowless light evens out the bright and dark patches of moulded glass (HIKROBOT light catalogue, glossary). Acceptance uses your own containers with real faults placed in them. A run of 100 seeded containers with no misses bounds the miss rate for that fault at about 2.6 % with 95 % confidence, using the Wilson method in NIST’s handbook (NIST e-Handbook, 7.2.4.1); the claim tightens as challenge passes accumulate in production.