With the rapid development of the new energy battery industry, the requirements for surface quality inspection of battery PACKs are increasing day by day. DaoAI provides an efficient and accurate inspection solution for this industry with its advanced AI vision technology.
User scenario: In the battery PACK assembly line of a leading new-energy battery manufacturer, after the battery PACK assembly is completed, a comprehensive surface inspection is required. The inspection objects include defects such as scratches, pits, stains, and deformations on the battery PACK surface.
Pain points: Traditional manual inspection methods have many problems. The miss-detection rate is relatively high, about 3%, which leads to some defective products flowing into the market, affecting product quality and brand image. The false-alarm rate also reaches 10%, increasing the workload and cost of re-inspection. Moreover, the efficiency of manual inspection is low, and the labor cost is high. Five inspectors are needed for one production line. At the same time, when the product model changes, it takes up to 30min to manually readjust the inspection standards and processes, seriously affecting the production schedule. In addition, when the robotic arm loads and unloads the battery PACKs, due to the lack of accurate pose guidance, the grasping success rate is only 80%, which easily causes damage to the battery PACKs.
Technical Principles
DaoAI uses advanced AI algorithms and imaging technologies to solve these problems. In terms of imaging, the self-developed 3D camera can obtain the three-dimensional topography information of the battery PACK surface. Compared with traditional 2D imaging, it can present the tiny defects and three-dimensional features of the surface more clearly. In terms of algorithms, based on the feature recognition of the visual basic model, through APDT positive-sample/few-sample learning, only 1-20 good samples are needed to quickly learn the features of normal products. At the same time, the semantic false-alarm filtering algorithm is used to analyze and screen the detection results to reduce false alarms. In the pose guidance of the robotic arm, the 6D pose algorithm can accurately calculate the spatial position and attitude of the battery PACK to guide the robotic arm for precise grasping.
- Principle of 3D camera imaging: Using technologies such as structured light or laser, a specific pattern is projected onto the battery PACK surface, and the camera captures the reflected light. The three-dimensional coordinates of the surface are calculated by analyzing the deformation of the pattern.
- APDT positive-sample/few-sample learning: Through a small amount of positive-sample data, deep-learning algorithms are used to mine the features and laws in the data, so as to achieve accurate recognition of defects.
- Semantic false-alarm filtering algorithm: Combining the semantic information of the image, it judges and filters the detection results that may be false alarms, improving the accuracy of the inspection.
- 6D pose algorithm: Considering the three-dimensional position and three-dimensional attitude information of the object comprehensively, through the extraction and matching of feature points in the image, the precise pose of the object in space is calculated.
DaoAI Solutions and Product Introduction
DaoAI provides a series of targeted products to solve the above problems. First, the DaoAI AI AOI software system has the ability of feature recognition based on the visual basic model, which can realize zero-code automatic programming of a good product in 5min. Using APDT positive-sample/few-sample learning, only 1-20 good samples are needed to quickly build an inspection model. At the same time, the semantic false-alarm filtering function effectively reduces false alarms. Second, the DaoAI 2D / 3D AI AOI equipment uses the self-developed 3D camera and three-dimensional topography reconstruction technology to detect hidden solder joints, coplanarity, and micrometer-level topography defects, achieving a comprehensive inspection of the battery PACK surface. Finally, the DaoAI robotic vision can precisely guide the robotic arm to perform unordered bin-picking and loading/unloading operations on the battery PACKs through the 6D pose guidance function, improving the grasping success rate.
DaoAI provides a comprehensive and efficient solution for the surface inspection of new energy battery PACKs with advanced AI vision technology.
Quantitative results: After applying the DaoAI solution, the detection effect of the manufacturer has been significantly improved. The miss-detection rate has been reduced from 3% to <0.6%, effectively preventing defective products from flowing into the market. The false-alarm rate has been reduced by -63%, greatly reducing the re-inspection workload and cost. The product model change time has been shortened from 30min to 5min, improving production efficiency. The grasping success rate of the robotic arm has been increased from 80% to over 95%, reducing the damage rate of battery PACKs.