Press Releases

OPT > News > Press Releases > Content

OPT Inspection of Lithium Batteries for Tab Defects

Share to:
Release time:2022-09-08 Source :

Defect types in lithium battery tabs are complex and diverse. Apart from random locations, some subtle defects are hardly distinguishable from the background color of the tab, making it difficult to accurately determine the defect characteristics. Therefore, their visual inspection is a great challenge.

Based on the two main technology platforms, i.e. visual imaging and image analysis, OPT improves the quality and speed of image capturing hardware. In addition, integration with internally developed AI algorithms enables accurate classification and assessment of the smallest or most complex defects on battery tabs. The inspection covers processes such as cutting, winding, welding, etc., providing a comprehensive solution for battery tab inspection.


High Resolution Optical Design for Clearer Images

With the expansion and efficiency improvement of the power battery industry, battery manufacturers have a high demand for optical inspection of tab defects and correspondingly very high demands on the quality and speed of visual imaging.

Through a continuous, in-depth understanding of the difficulties of tab defect inspection in any manufacturing process, OPT has introduced a complete set of optical machine vision solutions and developed a range of hardware products optimized for tab defect inspection. The advantages of fast image acquisition and accurate imaging enable them to meet the inspection requirements of any battery tab process.

Application of an OPT Machine Vision Solution in a Machine for Cutting
and Stacking Battery Tabs

For example, in the tab cutting and winding machine, OPT uses front and back lights to inspect the burrs or folded shape defects of the winding tabs that may result from the high-speed tab cutting process.

The machine vision line scan camera lens used in the OPT solution consists of a high-resolution optical design and uses technologies such as achromatic, vignette free, tolerance sensitivity optimization, and automatic detection of lens resolution to ensure that clear images of the inspection object can be obtained. It also provides a good basis for subsequent image analysis.

Optimized Tolerance Sensitivity

Furthermore, OPT fully utilized FPGA edge technology to accelerate the image acquisition of line scan cameras, which are additionally capable of processing binarizations, morphological operations, and blob analysis in real time. Subsequently, the processing results are transferred to the host computer together with the images. This significantly reduces the CPU load on the industrial computer. The requirements for the inspection of high-speed strips during cutting and winding are thus fully met.

Application of an OPT Machine Vision Solution in Tab Winder

OPT uses dome and coaxial lights for uniform illumination during the welding process of the tabs. In combination with an area scan camera lens, high resolution, clear images and sharp image contours are ensured. This allows very accurate detection or localization of weld marks, welding piece leakage, yellow tape position, yellow tape absence and so on at any inspection station of tabs.

OPT Machine Vision Solution for Tab Welding Defect Inspection


Deep Learning for Precise Extraction of Defect Features

Besides high-precision optical image acquisition, machine vision software is also key to dealing with complex defects in the tab area. OPT used the combination of Deep Learning and traditional algorithms to accurately extract and classify the defect characteristics of the battery tabs. This helped to avoid misjudgments and overlooked defects, and solves the problem of multi-stage defect inspection of battery tabs.

Process Visualization of the OPT Deep Learning Algorithm

OPT Deep Learning technology for tab error checking uses training and learns from a manageable number of different defect samples to build an AI model. 

There are three innovative advantages. First, the Deep Learning-based small sample inspection framework can reduce the number of faulty samples to single digits, solving the problem of difficulty in detecting faulty samples and high labeling costs in the past. This improves testing accuracy and robustness. Secondly, the self-adaptive data augmentation techniques using the data samples helps to recommend the most representative samples for manual labeling, which shortens the training time of the model and increases the testing accuracy by 10%, so that no false-negative results are obtained when testing battery tabs.

Application of OPT Deep Learning in the Inspection of Tab Defects

Lastly, in order to better meet the inspection needs of frequent battery product replacement, OPT has introduced self-adaptive transfer learning technology to shorten the training cycle of the AI model and transfer the inspection of tabs of similar size and process with a single click. For the inspection of battery tabs of different sizes, it is only necessary to provide a small amount of training data after transmission to fine-tune the AI model.

Self-Adaptive Transfer Technology Based on Deep Learning

Currently, the OPT Deep Learning algorithm is used in multi-process optical inspection of lithium batteries, including electrode sheet coating defect inspection, laser tab cutting defect inspection, tab winding defect inspection, and aluminum plastic film packaging defect inspection.