Yarn Break Inspection System for Laminating Machines

Overview

Fiberglass / polyester scrim is the backbone of insulation facing rolls. Tension fluctuation, raw material flaws, guide wear, dust and vibration can all cause broken yarns. A missed defect travels downstream with the roll and leads to rework, downgrading or customer complaints.

Manual inspection struggles to keep up: fiberglass yarns are fine and low in contrast against the substrate, so they are easy to miss; patrols happen every few minutes, so a break is often found tens of meters late; four laminating machines on three shifts need at least 12 dedicated inspectors; manual coverage is about 60%–70%, with no images or timestamps for traceability.

The system replaces the human eye with an industrial camera and vision algorithms, inspecting every frame 24/7.

How it works

1. Capture: photoelectric trigger with synchronized strobe imaging
2. Judge: classic vision pre-screening + deep learning review
3. Respond: audible and visual alarm / automatic machine stop
4. Record: defect images, meter position and time are stored
5. Trace: query by roll and analyze with reports

Key features

• Two-stage judgment: classic vision pre-screening errs on the side of flagging, then a deep learning model reviews the full image to identify broken, merged and skipped yarns. Uncertain frames are flagged separately as “suspect”.
• Alarm and stop interlock: defects are boxed in red with an audible and visual alarm; when consecutive NG frames reach the set count the machine stops automatically. Thresholds are set per recipe, with one-click alarm reset.
• Roll management and meter positioning: each roll records inspected length, defect count and images; defects are located by meter within the roll for downstream cutting. The current roll is restored automatically after a power loss.
• Review, traceability and reports: QC staff confirm defects or mark false alarms; query by time, roll, shift, product or defect type, and export shift, daily, weekly and monthly reports to Excel.
• Imaging self-check: separate alerts when the image is too bright, too dark, blocked or blurred, so the system never runs without actually inspecting.
• Permissions and audit trail: multi-level accounts, automatic version history for recipe changes, and audit logs for key operations.

The interface is designed for the shop floor, with large touch buttons and keyboard shortcuts. Operators get started in three steps: new roll → start inspection → handle alarms.

Standard configuration (per laminating machine)

Component	Description
Industrial camera	5 MP global shutter (2448×2048), GigE, hardware trigger, IP65 housing with air purge
Lens and lighting	12 mm industrial fixed-focus lens; white LED bar strobe light (Ra≥90) with strobe controller
Trigger sensor	Photoelectric trigger, response ≥1 kHz, sync accuracy ≤1 ms
Industrial PC	17-inch touchscreen panel PC, CPU inference, no GPU required
IO module	Opto-isolated 8 DI / 8 DO for machine run / speed signals and stop / alarm / lighting control
Alarm and power	Three-color audible and visual alarm (DC 24 V); online UPS (≥10 min); control cabinet

Each machine has its own system, so a single failure does not affect other machines. Standard signal connections keep installation downtime short, and MES / ERP interfaces (REST / OPC UA) are reserved.

Implementation and service

Implementation takes about 9 weeks: requirements survey (site survey and sample collection) → design and manufacturing (delivery in about 40 working days) → factory acceptance test (FAT) → installation and commissioning (7–14 days of on-site tuning) → site acceptance test (SAT) and training.

• 12-month warranty from the date the site acceptance test is passed
• Fast response: remote response within 2 hours and a solution within 24 hours; on-site within 2 working days when needed
• Free services during warranty: 2 inspections and 1 model retraining per year, free minor software upgrades
• Long-term spare parts: spare parts supply guaranteed for ≥5 years
• Role-based training for operators, QC staff, maintenance and IT, with a Chinese user manual and operation videos, followed by assessment and certification

Detection and false detection figures depend on on-site samples; final figures are based on the site acceptance test (SAT). Specifications are subject to the technical agreement.