The Challenge
A high-volume plastic bottle manufacturer in Silvassa produces over 2.4 million HDPE bottles per month across 8 production lines — supplying haircare, personal care, and edible oil brands including several national FMCG names. Their quality process relied on AQL-based random sampling: operators pulled batches every 30 minutes and visually inspected 30–50 bottles under fluorescent lighting. With line speeds of 8,000–10,000 bottles per hour, each 30-minute window produced 4,000–5,000 uninspected bottles.
Over a 12-month period, three major customer complaints resulted in formal non-conformance reports (NCRs). The first involved short-moulded bottles with insufficient wall thickness in the shoulder area — causing collapse under the customer's filling line capping pressure, resulting in a full batch recall of 48,000 bottles. The second NCR was triggered by contamination (black specks and burn marks) found at a retail distribution centre. The third involved seal surface damage that passed visual sampling but caused cap seal failures at the end-customer.
Combined, these complaints cost the manufacturer ₹28.4 lakh in returns, rework, and freight — and triggered a formal vendor performance review from their largest customer. The quality manager recognised that AQL sampling was statistically incapable of catching the defect rates occurring in reality. What was needed was 100% online inspection — every single bottle, on every shift, without slowing the line.
The Solution
Optomech deployed the BIS (Bottle Inspection System) on four production lines, rated to 10,000 bottles per hour per line — matching the existing blow-moulding output speed exactly. Each BIS unit uses multiple high-resolution industrial cameras positioned at calibrated angles to capture the full circumference, top seal surface, base, and body of every bottle in motion.
Optomech's proprietary AI-based image processing software analyses each frame in real time and classifies defects against a trained reference library built from the customer's own production samples — including short moulding, flashes, thin walls, choked necks, burn marks, black spots, material inclusions, holes, and seal surface damage.
Rejected bottles are automatically diverted to a rejection bin without stopping the line. The system stores a high-resolution image of every rejected bottle with timestamp, defect classification, and shift data — providing a complete, auditable rejection record. A 21.5-inch touchscreen console gives the line supervisor real-time defect counts, rejection rate trends, and shift productivity summaries. The 3-level password access system prevents unauthorised tolerance changes. Remote connectivity allows the Optomech service team to perform software updates and remote diagnostics without an on-site visit.
Installation and commissioning across four lines was completed in 6 days. The AI reference library was trained on the customer's parts — including normal variation in colour, gloss, and surface texture — to minimise false rejections while maintaining full defect sensitivity.
Defects Detected by the System
Results Achieved
| Metric | Result |
|---|---|
| Customer returns / NCRs | ₹28.4L annual losses → ₹0 customer returns in first 14 months post-installation |
| Defects caught per month | Avg. 3,200 defective bottles rejected per line per month — invisible to AQL sampling |
| Line speed maintained | 100% — no slowdown; BIS runs at full production speed of 10,000 bph |
| Rejection accuracy | >99.5% detection rate confirmed in commissioning validation study |
| False rejection rate | <0.3% — trained AI model distinguishes real defects from surface texture variation |
| Vendor performance review | Formal NCR status removed by largest customer after 6-month clean record |
| Audit records | Complete rejection image archive available for every shift — accepted in customer QA audits |
| Lines installed | 4 BIS units commissioned in 6 days; 2 additional lines planned in Phase 2 |
"We had three NCRs in one year and our biggest customer was reviewing us. The BIS changed everything. It runs at full line speed, catches defects our operators never even saw, and the rejection images are good enough for our customer's incoming QA team to review. Fourteen months, zero customer returns. The ROI was under six months."— General Manager – Quality, HDPE/PET Bottle Manufacturer, Silvassa
Similar defect challenges on your production line?
Share your bottle format, line speed, and quality challenge. Optomech's engineering team will recommend the right BIS configuration for your line.