導入事例

One Station, Every Model: AI Inspection for Battery Top Shell Defects

ケーススタディのダウンロード

CONTENTS

シェアする

Introduction:Where Cosmetic and Functional Quality Meet 

The battery top shell serves a dual purpose: it is the outward face of the product and a vital structural boundary. Because it is an exposed component, any surface defect instantly degrades customer confidence. However, the engineering stakes are much higher—flawed top shells lead to poor assembly alignment, failed seals, and compromised battery cells.

To stop these issues at the source, manufacturers need a definitive quality gate. UnitX provides automated, early-stage detection that stops defective shells from advancing to subsequent processes. This proactive approach eliminates wasted downstream processing, safeguards the integrity of the final cell, and ensures every battery meets the highest standards.

The Challenge: High-Mix Production Meets Complex Geometry 

The production line processes top shells across multiple models, creating a high-mix environment with a vast, unpredictable defect spectrum. Furthermore, the component’s circular geometry and glass-bead press-fit structure make imaging inherently difficult.

In this environment, manual inspection and traditional, rule-based vision systems create dangerous coverage gaps. They simply lack the flexibility to maintain accuracy across shifting product variants. To eliminate escape rates, cut production costs, and secure traceable data, the manufacturer needed a unified, automated system capable of detecting:

  • Glass Defects: Cracked, broken, or missing glass — direct safety and sealing failures.
  • Mechanical Defects: Burrs, damaged posts, loose electrodes, stringers, and tooling crash marks—upstream errors that compound aggressively during final assembly. 

Battery top shell (top view): the circular geometry and glass-bead press-fit structure create imaging challenges across multiple product models.

Battery top shell (top view): the circular geometry and glass-bead press-fit structure create imaging challenges across multiple product models.

The Solution: Rotary Dial Inspection Covering All Variants

To overcome the high-mix bottleneck, UnitX deployed a rotary dial inspection station that evaluates every top shell variant with zero changeover tooling.

Positioned directly after the lid-glass bead press-fit operation, the automated cell utilizes a rotary dial to continuously index parts. As each shell moves into position, a UnitX OptiX vision system captures the complex circular geometry, engineered to expose the entire spectrum of both glass and mechanical defects. The high-resolution visual data is immediately fed into CorteX - UnitX’s dedicated AI training and inference engine. Executing real-time, multi-model AI inference, CorteX instantly identifies the product variant on the dial and applies the correct defect criteria, delivering a flawless single-station quality gate without ever slowing down production. 

Rotary dial inspection cell: A OptiX camera positioned over the part as it indexes through the dial, with the UnitX controller managing image acquisition and automated pass/fail output.

Rotary dial inspection cell: A OptiX camera positioned over the part as it indexes through the dial, with the UnitX controller managing image acquisition and automated pass/fail output.

ビジョンシステムコンポーネント:

  • AI Industrial PC: CorteX × 1 — real-time multi-model training & inference.
  • イメージングシステム: OptiX × 1 — single camera covering all defect types.
  • 検査セル: integrated cell with rotary dial for automated part handling.

Results: Industry-Leading Accuracy Across All Models

Operating in a demanding high-mix environment, the UnitX system successfully established a definitive quality gate, delivering the following validated production metrics: 

  • Near-Zero Escapes (False Acceptance Rate ≤ 0.09%): The AI reliably intercepts 99.91% of all defects across every model variant, protecting downstream assembly.
  • Minimal Scrap (False Rejection Rate = 0.19%): The system accurately distinguishes critical flaws from benign variations, ensuring perfectly good parts are not wasted by false alarms.
  • High-Speed Throughput (Cycle Time = 2 s): The system evaluates all defect types on a given part in a single station pass without bottlenecking the line.

Defect Visualization

The images below show raw captures (top row) alongside AI-labeled detection results (bottom row) for the four primary defect classes. The AI draws precise segmentation masks, enabling accurate defect classification and location reporting.

  • “Stringer and burr”: Foreign material and sharp protrusion on the glass surface.
  • “Tooling crash”: Visible tool impact mark — a process alarm requiring upstream investigation.
  • “Loose electrode”: Electrode not seated correctly — a structural assembly failure.
  • “Damaged post”: Post deformation that affects electrical connectivity.

UnitX AI identifies and segments each defect class with precise boundary annotations, logging results directly to the production MES.

UnitX AI identifies and segments each defect class with precise boundary annotations, logging results directly to the production MES.

結論

Battery top shell quality is visible quality. UnitX’s single-station rotary dial inspection delivers consistent, objective standards across every product model — with near-zero escapes and minimal scrap — eliminating the variability and coverage gaps of manual inspection.

Standardize quality across all your top shell variants.
接触 UnitX to discuss multi-model surface defect inspection.

関連ケーススタディ

Keeping Motors Turning: AI Inspection for Slip Ring Surface Defects
Signal-Perfect: AI Inspection for Automotive Connector Surface Defects
UnitX AI-Powered 2.5D Inspection for Zinc Die-Casting Surface Defects
Defect-Controlled Delivery: AI Inspection for Automotive Sleeve Surfaces
Micro-Defects, Maximum Stakes: AI Inspection for Three-Way Valve Surfaces
Full-Surface at Speed: AI Inspection for EV Transmission Motor Shafts
Beyond the Surface: AI Inspection for Rubber Bushing Adhesion and Cracks
10 Million Cells, Zero Escapes: BiCell Internal Defect Inspection with CT-Scan AI
Automating the X-ray Line: AI Detection for Cell Positive and Negative Tab Inspection
Gear spring assembly (D65mm, H15mm): the AI evaluates two spring zones on both the front and back sides of the gear in a single 9-second inspection cycle.
上へスクロール