South KoreaManufacturing2024 – ongoing

LG Innotek

AI inspection of incoming semiconductor substrate materials

LG Innotek said incoming-material AI cut defect-cause analysis time by up to 90%

WorkingCompany's own claimOfficial ↗

Problem

Visual inspection could not reliably identify material defects affecting increasingly fine semiconductor substrates.

Approach

LG Innotek trained an image-processing model on material composition data. It deployed the system in RF-SiP production and expanded it to FC-BGA substrates, using visualized lot differences to support material adjustments.

Results

up to 90%reduction in defect-cause analysis time, reported by LG InnotekCompany's own claim[1]

1 minutetime to analyze material composition and defect regionsCompany's own claim[1]

at least 90%reported analysis accuracyCompany's own claim[1]

Takeaway

Inspection moved upstream to incoming materials, before those materials entered production.

Sources

업계 최초로 AI 활용해 불량 원자재 투입 ‘원천 차단’
LG Innotek25 Sep 2024Company publication

Source summaries

[1]LG Innotek25 Sep 2024

LG Innotek described deployed AI inspection for substrate raw materials. It reported analysis within 1 minute at at least 90% accuracy and up to 90% less time spent identifying defect causes.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026