3D AI AOI for Electrode Coating Porosity, -63% Missed Detections

In the critical new energy battery manufacturing process of electrode coating, WeLinkirt's 3D AI AOI equipment (featuring proprietary 3D cameras + 3D morphology reconstruction/point clouds, detecting hidden solder joints/coplanarity/micron-level morphology/porosity and other 2D optical blind spot defects, with 2D-3D fusion) precisely identifies micron-level porosity and uneven coating thickness—defects invisible to 2D optics—reducing the missed detection rate for electrode coating from 1.6% to <0.6%. This significantly enhances battery electrode quality and production line automation.

In the critical new energy battery manufacturing process of electrode coating, WeLinkirt's 3D AI AOI equipment (featuring proprietary 3D cameras + 3D morphology reconstruction/point clouds, detecting hidden solder joints/coplanarity/micron-level morphology/porosity and other 2D optical blind spot defects, with 2D-3D fusion) precisely identifies micron-level porosity and uneven coating thickness—defects invisible to 2D optics—reducing the missed detection rate for electrode coating from 1.6% to <0.6%. This significantly enhances battery electrode quality and production line automation. The new energy battery industry is experiencing rapid growth, demanding higher energy density, cycle life, and safety from batteries. This directly impacts upstream manufacturing, especially quality control in electrode coating. The electrode, as a core component of the battery, has its electrochemical performance directly determined by the uniformity, denseness, and absence of microscopic defects in its coating. A leading battery material supplier, with annual electrode production ranking among the top in the industry, urgently needed to enhance its online inspection capabilities for the electrode coating section, particularly for precise capture of micron-level defects on high-speed production lines, to meet market demands for high-quality battery materials.

Pain Points: Why This Hurdle Was Difficult to Overcome

This leading manufacturer faced several key pain points. Firstly, there was a **high missed detection rate for microscopic defects**. Traditional 2D AOI struggled to effectively identify micron-level pores, bulges, and foreign object indentations with distinct depth and morphological features in the electrode coating. These defects might appear as subtle grayscale changes or be obscured in 2D images, leading to a missed detection rate of around 1.6% for 2D AOI. Secondly, **manual re-inspection was inefficient and costly**. Due to false positives from 2D AOI, a large number of seemingly defective areas required manual re-inspection one by one, consuming significant human resources, slowing down the production line tempo, and suffering from inconsistency due to subjective human factors. Thirdly, **production line changeover and programming were time-consuming**. With traditional rule-based AOI, engineers had to manually adjust thresholds and parameters when switching between different electrode models, taking several hours or even half a day, severely impacting production line utilization. These issues collectively led to bottlenecks in product yield and persistently high production costs.

The root causes of these challenges lie in the complexity of the electrode coating process, the variability of material properties, and the stringent requirements of high-speed production lines. Electrode coating thickness typically ranges from tens to hundreds of micrometers, and any minute thickness fluctuation or surface unevenness can affect battery performance. Traditional 2D optical inspection only acquires two-dimensional grayscale information, lacking the ability to perceive defects with three-dimensional depth and morphological features (such as deep pores, protrusions). Concurrently, electrode materials (e.g., cathode/anode materials, binders) have diverse optical properties, reflecting, absorbing, and scattering light differently, making a unified 2D algorithm difficult to adapt. On high-speed production lines operating at tens or even hundreds of meters per minute, the time available for the inspection system to process images and make judgments is extremely short, demanding high computational power and real-time algorithm performance. Furthermore, the current breakthroughs in the localization of core components for intelligent manufacturing in China, particularly the reduction of high-performance 3D vision sensor costs from ten thousand yuan to thousand yuan levels, have prompted manufacturers to seek more cost-effective and technologically advanced domestic alternatives to reduce reliance on imported equipment and enhance independent control.

Technical Principles

The core advantage of WeLinkirt's 3D AI AOI equipment lies in its proprietary 3D camera and advanced 3D morphology reconstruction technology, combined with 2D-3D fused AI vision algorithms. We employ multi-line laser scanning or structured light projection to precisely acquire 3D point cloud data of the electrode surface. Through sub-pixel image registration and multi-frame fusion algorithms, we reconstruct a high-precision, high-density 3D morphological model of the electrode, achieving micron-level Z-axis accuracy. Unlike traditional 2D AOI which only analyzes grayscale images, our system can directly quantify the depth, height, volume, and other 3D geometric features of defects, such as precisely measuring the depth of pores, the height of bulges, or the size of foreign object protrusions. This enables the system to effectively distinguish true morphological defects from pseudo-defects caused by surface textures or uneven lighting, significantly reducing false positives.

Compared to traditional rule-based AOI, WeLinkirt's 3D AI AOI offers several advantages: First, **3D perception capability**. Traditional 2D AOI only acquires two-dimensional planar information and is incapable of perceiving three-dimensional information like depth and height, easily missing morphological defects such as pores, bulges, and indentations. Our 3D system, however, can directly 'see through' these 3D features. Second, **the powerful generalization and self-learning capabilities of AI models**. We integrate the DaoAI AI AOI software system, leveraging its visual foundation model's feature recognition capabilities, supporting APDT (Auto-Positive Defect Training) for positive/few-shot learning. With just 1-20 good samples, 0-code automatic programming can be completed in 5 minutes, rapidly adapting to new electrode models or defect types, whereas traditional rule-based AOI requires extensive time for parameter tuning. Third, **2D-3D fusion inspection**. The system performs pixel-level fusion of high-resolution 2D images and high-precision 3D morphological data, complementing their strengths. 2D images are used to identify planar features like color, texture, and brightness, while 3D data compensates for the shortcomings of 2D in terms of depth and height, ensuring comprehensive coverage and high-precision detection of various defects. This fusion strategy greatly enhances the robustness and accuracy of inspection, especially for defects that are not obvious in 2D images but have significant features in 3D morphology.

Typical Application Scenarios

• **Electrode Coating Porosity and Bubble Detection**: During electrode coating, micron-level pores and bubbles can easily form due to uneven slurry mixing or improper drying. These defects may be difficult to identify in 2D images due to lighting or background texture interference. 3D AI AOI precisely determines if they are true pores or bubbles by measuring the depth and volume of the defective areas, preventing missed detections and ensuring coating compactness.
• **Uneven Coating Thickness and Scratch Detection**: Coating thickness uniformity is a critical indicator of electrode quality. 3D morphology reconstruction technology can generate high-precision thickness maps of the electrode surface in real-time, visually displaying coating thickness fluctuations and detecting subtle scratches, indentations, or protrusions invisible to the naked eye. These minute morphological changes significantly impact battery performance.
• **Foreign Object and Contamination Detection**: Conductive dust, fibers, or metal particles from equipment wear may be introduced during production. These foreign objects might be confused with electrode particles in 2D images. 3D AI AOI distinguishes them from the normal electrode surface based on their unique morphological features (e.g., height, shape of the foreign object), achieving high-precision foreign object detection.
• **Coating Edge Defect Detection**: The neatness of the electrode coating edge, presence of burrs, overflow, or uncoated areas are crucial for subsequent processes like calendering and slitting. 3D AI AOI can precisely delineate the 3D contour of the coating edge, identifying defects such as uneven edges, missing parts, or overflow, ensuring edge quality meets process requirements.
• **Tab Welding or Riveting Coplanarity and Morphology Inspection (Extended Application)**: While this case focuses on coating, 3D AI AOI can also be applied to the electrode-tab connection. For the joint area after welding or riveting, the coplanarity of the joint surface, the morphology of the weld or rivet points, and the presence of cold solder joints or cracks are difficult to accurately assess with 2D. The 3D system can precisely measure these microscopic morphological features, ensuring connection reliability.

Implementation Case

A leading battery material supplier, with an annual electrode production exceeding tens of billions of square meters, faced severe challenges in detecting micron-level pores in the electrode coating section. Previously, their production line utilized traditional 2D AOI equipment, which, while capable of detecting some surface defects, struggled with small pores less than 50 micrometers deep and false defects caused by lighting variations. This resulted in a persistent missed detection rate of around 1.6% and a high false positive rate of 8%, necessitating extensive manual re-inspection. To address this bottleneck, the manufacturer introduced WeLinkirt's 3D AI AOI equipment. We first conducted a one-month on-site test and data collection on a high-speed coating production line. Initially, using the APDT few-shot learning function of the DaoAI AI AOI software system, a basic model was trained in 5 minutes with only 15 images of good electrode samples. Subsequently, the 2D-3D fusion model was deployed to the production line after iterative optimization with actual production line defect samples. After deployment, the system performed real-time full-surface online inspection of electrodes, simultaneously processing acquired 2D images and 3D point cloud data, and generating detailed defect reports and 3D morphological maps.

"WeLinkirt's 3D AI AOI equipment not only solved our long-standing challenge of detecting micron-level pores but, more importantly, it significantly reduced the workload of manual re-inspection, making our production line run much smoother."

WeLinkirt Solution and Products

The core solution WeLinkirt provided to this client is based on our proprietary 3D AI AOI equipment. This equipment integrates high-performance 3D cameras and image processing units, achieving micron-level 3D morphology reconstruction of the electrode surface through high-precision structured light projection or multi-line laser scanning technology. Combined with the DaoAI AI AOI software system, we first leverage its visual foundation model, which has learned from a vast amount of general industry defect features, to provide a powerful prior knowledge base for electrode inspection. In practical implementation, engineers only need to upload a small number (1-20) of good electrode images, and the system can quickly identify the normal morphological features of the electrode through the APDT positive sample learning mechanism, automatically generating a detection model. For customer-specific defects, such as pores of a particular depth or bulges of a specific height, the system supports incremental learning with a few defect samples or rapid iterative optimization through manual annotation. For deployment, our equipment supports 100% on-premise private deployment, ensuring data security and privacy by keeping data within the factory. The system provides standard interfaces for seamless integration with existing MES/SCADA systems, enabling real-time upload and traceability of defect data, and supports linkage with production line control systems for automatic sorting of defective products or alarm-triggered shutdowns. Furthermore, the DaoAI World Model, serving as a unified foundation, ensures that the AI model possesses semantic understanding and cross-scenario generalization capabilities across different electrode models and coating processes, continuously learning from production line feedback to improve detection accuracy and adaptability.

After practical operational verification, the deployment of WeLinkirt's 3D AI AOI equipment significantly enhanced the electrode coating inspection capabilities of this leading battery material supplier. The missed detection rate decreased from 1.6% to <0.6%, an improvement of 63%, which means millions of square meters of potentially defective products are prevented from entering downstream processes annually. Concurrently, the false positive rate dramatically dropped from 8% to 2.5%, reducing manual re-inspection volume by −68%, greatly freeing up human resources and improving overall production line efficiency. Thanks to the 0-code automatic programming and few-shot learning capabilities of the DaoAI AI AOI software system, production line changeover time was reduced from an average of 4 hours to 5min, significantly boosting equipment utilization. By accurately identifying and eliminating microscopic defects early, the system effectively prevented larger losses caused by defects flowing into subsequent processes, saving the manufacturer tens of millions of yuan annually in rework and scrap costs, strongly supporting its goal of producing high-quality battery materials.