Industry Solution · INDUSTRY

Electronics / PCBA

SMT / DIP and semiconductor back-end — the benchmark for full-board PCBA inspection.

Full-board inspection of SMT placement, DIP through-hole, solder joints and component mounting.

Feedback from real lines and real scenes can't be bought.
Eight years on the floor, line by line — hard-won engineering experience.
Full-stack in-house — the whole chain from sensor to world model is ours.
100M+ real industrial data points continuously refine one world model.

Key checks · KEY CHECKS

Solder: cold joints / bridging / insufficientComponents: missing / wrong / reversed / misalignedPolarity & silkscreen errorsBGA / QFN voids & bridgingGold-finger scratches / oxidationConformal-coating coverage & contaminants

Recommended systems: 2D AI AOI · 3D AI AOI · AOI software

Book a Demo Explore Products

Case studies · CASE STUDIES

Electronics / PCBA · proven on the line

From SMT solder joints to hidden BGA balls and micro-cracks in passives — AI-AOI drives false calls down while still catching the hard-to-see defects.

Scene · post-reflow SMT

A leading PCB plant · taming post-reflow AOI false calls

ChallengeConventional AOI flags many good boards as defects; operators re-judge image by image and burn review capacity.

SolutionAI-AOI as a second gate plus APDT positive-sample learning separates true from false defects by learning only good units — 5-minute, no-code changeover.

−80%False calls

<1%Escapes

−80%Re-review

Scene · hidden BGA/QFN joints

An automotive-electronics PCBA plant · joints beneath the package

ChallengeBGA/QFN joints hide under the package where optical AOI cannot see; voids and bridges are hard to judge reliably.

Solution3D AI-AOI hardware plus the world model spots voids, bridging and non-wetting in low-contrast imagery.

99%+Detection

3DHidden joints

100%Inline 100%

Scene · MLCC micro-cracks

A passive-component maker · MLCC surface & micro-cracks

ChallengeMLCC micro-cracks and electrode misalignment are micron-scale; manual reading is fatiguing and inconsistent, and defect samples are scarce.

SolutionAPDT positive-sample learning goes live on 1–20 good units, with micron-grade precision and instance segmentation for exact localisation.

1–20Good samples

μmMicron-grade

msPer part

Cases are anonymised industry scenarios; the figures are typical ranges achievable with WeLinkirt DaoAI solutions.

In-depth cases · IN-DEPTH

Electronics / PCBA · in-depth cases

Each case breaks down the path to deployment in a real scenario — industry context, pain points, the DaoAI solution and the results.

Electronics · 2026-06-26

Taming Post-Reflow AOI False Calls: An AI Second Gate That Cuts Re-Review

A leading consumer-electronics PCBA plant fought chronic AOI false calls that buried its re-review stations. DaoAI AI-AOI software, paired with APDT positive-sample learning, sat as a second gate after the legacy AOI: false calls fell about 80%, manual re-review dropped 80%, and escapes stayed under 1%.

Electronics · 2026-06-24

3D AI Inspection of Hidden BGA/QFN Joints: Voids, Bridges, Non-Wetting in One Pass

At a communications-module contract manufacturer, BGA/QFN balls hide beneath the package—2D optics can't see them and X-ray sampling is slow and costly. DaoAI's 3D AI-AOI hardware images the hidden joints in three dimensions, reaching 99%+ combined detection of voids, bridges, and non-wetting.

Robotic bin picking of PCB components with 3D vision

Electronics · 2026-06-22

MLCC Micro-Cracks: Live on 1–20 Good Parts, Micron-Scale, Milliseconds per Part

At a passive-component maker, MLCC end-face micro-cracks are only microns wide—hard for classic vision to catch—and defect samples are nearly impossible to collect. DaoAI APDT positive-sample learning goes live on just 1–20 good parts, resolving at micron scale and clearing each part in milliseconds.

DaoAI 2D AI-AOI automated optical inspection system

Electronics · 2026-06-20

Component Polarity False Calls: 164 Images, <1s Inference, 99% Accuracy

At an automotive-electronics PCBA plant, diode and electrolytic-cap polarity calls were plagued by false positives—legacy AOI read silk glare and body color shift as reversed. DaoAI used just 164 historical false-call images for few-shot learning: under 1 second inference, ~99% polarity accuracy.

Electronics and PCBA production inspection

Electronics · 2026-06-18

PCBA Missing Parts & Connector Misloads: Full Inspection at High Takt

At an industrial-control-board PCBA plant, connectors, jumpers, and heatsinks were prone to omission, misload, and wrong orientation, and manual final inspection couldn't cover it all. DaoAI AI-AOI runs inline full inspection of assembly completeness at high takt, reaching 97%+ combined detection of missing parts and misloads.

FAQ · FAQ

Electronics / PCBA · AI Visual Inspection FAQ

What defects can AI visual inspection detect in electronics and PCBA?

It covers post-SMT solder defects (insufficient solder, bridging, shift, tombstoning), missing and wrong components, reversed polarity, hidden BGA / QFN joints and MLCC micro-cracks. DaoAI uses a vision foundation model to catch these subtle, low-contrast defects while keeping false calls low.

How does DaoAI's AI-AOI reduce false calls on SMT lines?

It adds an AI second-pass gate after traditional AOI, re-judging suspect points with deep learning. This cuts false-call rework volume by over 70 percent without missing true defects, freeing up inspectors.

What is needed to inspect hidden joints under BGA and QFN packages?

DaoAI 3D AI-AOI reconstructs 3D topography to inspect hidden joints and coplanarity under BGA / QFN devices, covering blind spots that 2D optics cannot see through.

With frequent product changeovers, is AI inspection slow to set up?

No. DaoAI software auto-programs in five minutes from a single good sample, with no CAD and no defect library needed, and reuses quickly across changeovers, ideal for high-mix, low-volume electronics.