Separator, a critical component in lithium batteries, is highly sensitive. Even minor errors or contaminants during manufacturing can create defects like tears and other flaws, resulting in a relatively low yield rate. Recently, LUSTER has delivered an integrated AI-powered inspection solution for a leading battery separator manufacturer, significantly enhancing the precision and efficiency of separator inspection.

01| A Major Challenge for Separator Yield Improvement

Excessively large pores on separators can cause electrode contact, potentially triggering cell explosion. Industry experts emphasize the need for clearly defined defect types to ensure accurate detection, classification, and grading for high-quality separator production. Leveraging years of experience in the new energy sector, LUSTER has developed an integrated AI-powered inspection solution which precisely identifies dozens of separator defect types, slashing grading time from 2 minutes manually to 30 seconds automatically. By overcoming traditional machine vision limitations with AI inspection and auto-grading, our solution saves more than 10 human inspectors. Crucially, it boosts yield and cuts costs by shifting from passive defect acceptance to proactive analysis and optimization.

02| Cut from 2 Minutes Manually to 30 Seconds Automatically

To break through these bottlenecks limiting yield improvement, our separator project team conducted in-depth on-site research. They developed a customized system integrating Data, AI, and Vision for real-time defect classification, automated grading, and root cause analysis.

Data: An expert knowledge base was built, collecting defect data and process expertise. Over 30 specific defect characteristics were defined from an initial 10+ types, based on severity, frequency, and impact, providing a solid data foundation.

AI: Leveraging our LusterLVM general industrial vision model, our solution was trained specifically for separator defects. The advanced F.Brain algorithm enabled efficient annotation, training, and optimization, achieving highly precise defect identification and classification. Additionally, Grand Master of Quality Management (GMQM) module was integrated for online process quality monitoring.

Vision: A data acquisition platform was developed to interface with all machines, capturing defect images and critical features (shape, size, area, location coordinates). Image consistency was also optimized.

03| Closed-Loop System Integration for Smarter Quality Control

LUSTER's project team identified a key pain point: the client was dedicating significant labor power to defect data collection and import, easily causing production delays. To reduce labor dependency and boost efficiency, we focused on the client's core goals of minimizing waste and enhancing quality management. A tailored intelligent Quality Management System (QMS) was developed and some integration modules were created to deeply connect this system with the client's existing QMS and other IT systems. This established a closed-loop connecting hundreds of shop-floor machines with the GMQM systems, enabling cross-device data sharing and delivering functionalities such as quality traceability, defect warnings, and root cause analysis.

With over 20 years of machine vision expertise, LUSTER has accumulated 10 million+ finely annotated defect datasets, enabling efficient defect classification from separator inspection to broader industrial quality scenarios. As a driver of smart factory transformation, LUSTER revolutionizes lithium battery separator production and empowers the rise of "Dream Factories".