Stenter Fabric Defect Inspection System

Overview

Mosquito nets and mesh fabrics go through warping, weaving, dyeing and stenter finishing. Raw material variation, machine condition, process parameters and operator handling can all cause defects such as crescent marks, scars, holes, pulled threads, snowflake spots, color spots and poor joints. When quality relies mainly on downstream or post-roll sampling, problems often surface only after a large batch is produced, leading to rework, downgrading and customer complaints.

Manual inspection has clear limits on fast, continuous lines: staring at fine mesh and faint flaws for long periods causes fatigue, which leads to missed and false detections; different shifts and inspectors judge “minor, can pass” and “serious, must handle” differently; and inspection posts require a lot of labor.

The Stenter Fabric Defect Inspection System is installed at the stenter or other finishing stations. It images the tensioned, moving fabric inline and uses machine vision and deep learning to detect, grade and record defects, moving quality control from after-the-fact sampling to inline control.

How it works

1. Capture: time-sliced imaging of the moving fabric, with exposure, lighting and trigger timing coordinated to reduce motion blur
2. Pre-process: image enhancement and region-of-interest division, with adaptive enhancement for mesh texture and weak areas
3. Detect and grade: classic vision features + deep learning to classify and grade defects
4. Alarm and mark: on-screen highlighting, audible and visual alarm, with synchronized marking or signal output
5. Record and trace: defect snapshots, position, time and batch are stored and can be queried by shift and roll number

Compared with manual inspection

Item	Manual offline / sampling	This system (inline vision + AI)
Coverage	Sampling by batch and time window	Continuous inspection along the line
Consistency	Affected by fatigue and subjective judgment	Fixed rules with version control
Records	Paper or scattered electronic records	Structured logs and defect snapshots
Fine defects	Easy to miss over long shifts	Stable imaging + algorithms for low-contrast, small targets
Marking	Manual cutting or marking after a defect is found; new defects can be missed while marking	Inspection and marking run together, with traceability
Staffing	Many repetitive inspection posts	Shifts to spot checks and sample maintenance

Key features

• Live monitoring: multi-window, split-screen view of the fabric with defect areas highlighted.
• Defect classification and grading: detects and grades multiple defect types; the defect list can be refined per project.
• Audible and visual alarms: alarm strategies configurable by defect grade.
• Position: relative coordinates of each defect along and across the fabric, mapped to the roll.
• Records and statistics: defect snapshots, time and batch are stored in a structured way; filter by shift, roll number and type, with report export extendable per project.
• Packing prompts: prompts operators to pack based on the current length, and can be linked to roll diameter, work orders or takt time.
• Permissions and logs: operators and parameter changes are traceable.
• Model maintenance: incremental samples and threshold releases, with small-sample adaptation at changeover.

Standard configuration

Component	Description
Industrial camera	Global shutter area-scan, sharp imaging of moving fabric
Industrial lens	Matched to field of view and working distance
LED lighting	Raises contrast of yarns, mesh structure and defects; strobe optional
Vision controller / edge computer	Inference, communication, recording and result output
Mounting and protection	Anti-vibration, temperature control and dust protection for the workshop
HMI	Live monitoring, reports and parameter maintenance; touchscreen optional

Exact models are selected after an optical calculation based on the machine, line speed, width and site conditions.

Deployment and benefits

10 systems are running on site at the stenter stage of mosquito net / mesh fabric production, operating stably for more than a year and a half.

• Automatic detection and classification: defects are detected inline and filed by type, with fixed judgment criteria that rely less on individual experience.
• Marking linked to the process: marks or signals are output per the process rules, so downstream steps can cut, avoid or re-check.
• Lighter workload: the system takes over the intensive fabric watching, and staff move to spot checks, reviews and sample maintenance.
• Less downstream loss: fewer defective sections reach cutting and sewing, reducing waste, rework and lost hours.
• Earlier quality control: defects are intercepted and graded at the stenter / inspection stage, instead of being found in finished products.

Notes

Miss rate, false detection rate and minimum detectable size depend on line speed, depth of field, oil and dust, and fabric sheen. They are verified during trial operation using defect samples provided by the customer, and final figures are based on the technical agreement and site acceptance. Send us your line speed, width, fabric type and defect list, and we will provide an optical calculation and implementation proposal.