AI-Powered X-Ray Inspection for BGA & Fine-Pitch Assemblies
Boolean & Beyond delivers AI-driven X-ray and 3D inspection for Ball Grid Array (BGA) and fine-pitch components, solving the visibility and speed limitations of conventional optical and manual X-ray review.
Why Traditional Methods Struggle
- Hidden joints: BGA solder balls are fully obscured under the package; top-down optical cameras cannot see critical joints.
- Ultra-fine pitch: Pitches down to 0.3 mm and hundreds of I/Os make manual inspection slow and error-prone.
- Complex X-ray images: AXI images require expert interpretation; partial or subtle defects are easily missed.
What Our AI Vision System Does
Our deep learning models analyse X-ray and 3D data to detect:
- Head-in-pillow
- Cold joints
- Bridging
- Voiding (with quantitative void %)
- Insufficient solder
- Missing balls
Models are trained on thousands of labelled X-ray images, enabling detection of subtle, low-contrast and partial defects that even experienced operators struggle to classify consistently.
For fine-pitch devices (QFN, CSP, micro-BGA) in dense assemblies across Bangalore and Pune, the system fuses:
- X-ray cross-sections
- Oblique-angle X-ray views
to reconstruct a complete view of solder joint quality beneath each package.
Implementation Approach
Phase 1 (Weeks 1–4): X-Ray Data Capture & Annotation
- Capture comprehensive X-ray datasets from your existing inline or offline systems.
- Focus on your actual:
- BGA package families
- PCB stack-ups and layouts
- Reflow profiles and process windows
- Annotation team labels defects to IPC-A-610 standards, creating a ground-truth dataset aligned with your quality criteria.
Phase 2 (Weeks 5–8): Deep Learning Model Development
- Train specialised detectors per defect category (voiding, head-in-pillow, bridging, etc.).
- Pay particular attention to:
- Voiding: Distinguish acceptable vs. excessive voids per IPC class and your internal limits.
- Head-in-pillow: Detect subtle grayscale and texture differences.
- Bridging: Robust detection in dense, fine-pitch arrays.
- Validate models against:
- Your reject/accept criteria
- IPC Class 2 or Class 3 requirements
Phase 3 (Weeks 9–11): Production Deployment & Optimisation
- Deploy models alongside your existing X-ray equipment (e.g. Nikon, Nordson DAGE, Viscom; inline AXI or offline stations).
- Optimise inference speed to match or exceed line takt time.
- Stream real-time results into your MES for:
- Traceability
- SPC (Statistical Process Control)
- Closed-loop process feedback
Expected Results & ROI
Typical outcomes for electronics manufacturers using Boolean & Beyond for BGA inspection:
- Defect detection: 90–97% detection of hidden BGA defects (vs. 70–80% with manual X-ray review).
- Inspection speed: 3–5× faster X-ray analysis, increasing throughput without new hardware.
- Voiding accuracy: Automated void percentage within ±2%, replacing subjective visual estimates.
- Operator dependency: ~80% reduction in reliance on scarce X-ray experts (especially critical in Bangalore and Chennai).
- Field reliability: 40–60% fewer warranty claims tied to BGA solder failures.
- Payback: Typical ROI within 8–12 months for high-mix, high-density assembly operations.
Integration with Existing X-Ray & Quality Workflows
Boolean & Beyond augments your current X-ray infrastructure instead of replacing it:
- Hardware-agnostic AI layer:
- Inline AXI systems