Understanding AVI Reject Limits
Reject limits in automated visual inspection define the sensitivity thresholds at which the machine vision system classifies a container as defective. Setting these limits requires balancing two competing objectives: maximizing defect detection (sensitivity) while minimizing false rejects (specificity). A system that is too sensitive will reject acceptable product, reducing yield and increasing cost. A system that is too lenient will allow defective units to pass, creating patient safety risks and regulatory non-compliance.
The Reject Limit Framework
A robust reject limit framework addresses three interconnected elements:
1. Detection Capability Assessment
Before setting reject limits, you must understand what the AVI system can detect. This requires challenging the system with characterized defect test kits containing known defects of varying types and sizes. The resulting probability of detection (PoD) data establishes the system's baseline capabilities and identifies the "gray zone" — the defect size range where detection is neither guaranteed nor impossible.
2. Risk-Based Limit Setting
Reject limits should be established using a risk-based approach that considers the clinical significance of different defect types, the probability of detection at various thresholds, the acceptable quality level (AQL) for the product, and the economic impact of false rejects. This approach aligns with ICH Q9 quality risk management principles and FDA's expectation for science-based decision making.
3. Statistical Validation
Once limits are proposed, they must be validated through statistically rigorous testing. This typically involves running characterized test sets through the system multiple times to demonstrate that detection rates meet predetermined acceptance criteria with adequate statistical confidence (typically 95% confidence level).
Common Reject Limit Categories
AVI systems typically monitor multiple inspection channels, each with its own reject limit:
| Channel | Defect Types | Key Parameters |
|---|---|---|
| Particulate | Glass, metal, fiber, rubber, polymer particles | Pixel count, contrast ratio, motion detection |
| Cosmetic | Cracks, chips, scratches, seal defects | Edge detection, pattern matching, dimensional analysis |
| Fill Level | Overfill, underfill, missing product | Meniscus detection, volume estimation |
| Color/Clarity | Discoloration, turbidity, phase separation | Spectral analysis, transmission measurement |
| Container Integrity | Cap position, crimp quality, label placement | Geometric analysis, template matching |
Establishing Reject Limits: Step-by-Step
- 1Define critical quality attributes (CQAs) — Identify which defect types pose patient safety risks vs. aesthetic concerns
- 2Characterize system capabilities — Use defect test kits to map detection probability across defect types and sizes
- 3Determine acceptable quality levels — Based on product risk assessment, define the maximum allowable defect rate
- 4Set initial thresholds — Propose reject limits that achieve target detection rates while maintaining acceptable false reject rates
- 5Validate with challenge testing — Run characterized test sets to confirm performance meets acceptance criteria
- 6Document rationale — Create a reject limit justification document linking limits to risk assessment and validation data
- 7Implement monitoring — Establish ongoing trending to detect system drift and trigger revalidation
False Reject Rate Management
False reject rate (FRR) is a critical economic and operational metric. Industry benchmarks suggest that well-optimized AVI systems should achieve FRR below 1–2% for standard liquid products. Higher false reject rates indicate that limits may be set too aggressively, lighting or handling may be introducing artifacts, or the product itself has characteristics that challenge the inspection algorithm.
Important Consideration
Reject limits should never be loosened solely to reduce false reject rates without a documented risk assessment. Any limit change must be justified through formal change control with supporting validation data.
Regulatory Expectations
Regulatory agencies expect reject limits to be scientifically justified, validated, and maintained under change control. Key expectations include documented rationale linking limits to product risk assessment, validation data demonstrating detection capabilities at established limits, ongoing monitoring and trending of reject rates, formal change control for any limit modifications, and periodic revalidation to confirm continued system performance.
How YB Inspection Supports Reject Limit Development
YB Inspection provides the characterized defect test kits essential for reject limit validation. Our kits contain precisely documented defects spanning the full range of types and sizes relevant to your product, enabling you to map detection capabilities and validate proposed limits with statistical rigor. We also offer consulting support for reject limit strategy development and validation protocol design.