One of the largest roadblocks to widespread EV adoption is the cost of the battery, which can account for 40% of a vehicle’s price tag. Materials are rare and expensive, there are numerous operations and opportunities for defects, and a few microns can separate an acceptable blemish from a legitimate flaw.
Companies like Guangdong Lyric Robot Automation are paving the way for the EV industry, addressing market challenges with automation lines, smart warehouses, and other manufacturing solutions. Its customers, including CATL, BYD, Volkswagen, Ford, and Samsung, need safe, high-performance EV batteries – and fast.
“The lithium battery and new energy vehicle industries operate at a higher level of development,” said Du Yixian, the Dean of the Research Institute at Lyric’s Dingyuan Branch. “The battery production process must be high quality and safe; these two aspects are the path to our goal.”
Lyric’s goals are twofold, Yixian said. The company aims to contribute to the global development of the lithium battery industry for a carbon-neutral future and foster a platform for engineers.
Those goals require advanced machine vision, artificial intelligence, and reliable tracking and tracing systems.
Empowering EV battery defect detection with AI
Many defect detection applications in EV battery manufacturing are too complex for rules-based machine vision alone. The variety of flaws – and the difficulty distinguishing them from acceptable anomalies – requires AI.
Lyric uses two Cognex AI technologies: edge learning and deep learning. Edge learning is an AI technology that uses pre-trained algorithms to recognize defects and gauge coordinates, requiring just 5-10 images for training. Deep learning serves a similar function but can address more complex applications and requires more training images.
“In EV battery manufacturing, there are a lot of difficulties and forms of defects,” said Feng Yingjun, head engineer of the Lyric’s vision division. “It really puts machine vision hardware and software to the test.”
One of those tests is stacking sheets of anode, cathode, and laminator material, or diaphragm, to form battery cells.
Aligning the diaphragm against sheets of anode and cathode is a straightforward operation. However, misaligning the components by just 0.2mm – about the thickness of two sheets of paper – can decrease battery performance or cause a short circuit.
Yingjun said that finding the edges of the anode, cathode, and diaphragm is one of the most challenging aspects of cell sheet stacking. It's too difficult for rule-based machine vision alone, so Lyric uses Cognex solutions powered by edge learning.
“The Cognex edge learning SmartLine tool is very widely used, mainly because it’s hard to identify the boundaries of the components,” Yixian added. “In these more complex and challenging working conditions, SmartLine really helps identify boundaries, with up to 99% accuracy.”
Yingjun said that more accurate cell sheet stacking improves the energy yield of EV batteries. If yield rates are too low, he added, the customer could reject the batteries. Edge and deep learning-based solutions help Lyric deliver solutions faster, too.
“Cognex helped reduce our deployment time by more than 30% in some applications, which is something our customers are always looking for,” Yixian added.
Shedding light on flaws and traceability
Lyric uses 3D Cognex solutions to analyze weld seams. The company combines Cognex vision tools to find weld seams and its own technology to identify defects.
While finding flaws and aligning components are essential aspects of EV battery manufacturing, it’s equally important to Lyric to monitor where components are in the production process.
Tracking and tracing EV battery parts by scanning barcodes and direct part marks is essential to quality assurance, Yingjun said. If there’s no way to track a defective product to find out the root cause, there’s a risk that faulty products will make their way to the end user. But scanning codes on reflective, curved surfaces can be challenging, especially when scanning different types of components for various customers.
Intuitive automation on a global scale
The advanced capabilities of machine vision systems and barcode scanners are essential, but an often-overlooked factor is how humans use these technologies. Lyric produces EV battery manufacturing solutions for different customers, each with its own requirements, designs, and specifications. The capability to easily reconfigure machine vision systems and barcode scanners to take on new jobs and operate new environments helps Lyric optimize throughput.
“The first thing we look for in a solution is its overall R&D capability, the product's functionality,” Yingjun said. “The second consideration is the guarantee of technical services. Those are the two factors we are most concerned about.”
Lyric realized increased production on the factory floor thanks to user-friendly tools like VisionPro and In-Sight machine vision software. Since Cognex software is programmed in a familiar drag-and-drop interface, Yingjun’s team spends less time developing source code.
“Cognex software tools are easy to operate, and their performance is still strong,” Yingjun said. The increased usability of Cognex vision tools helps improve EV battery yields and manufacturing throughput.
For example, on the hardware side, Yingjun said DataMan features like multicolor lighting and high-speed liquid lenses boost the accuracy of code reading, improving traceability. Cognex scanners are also easy to debug and reconfigure, reducing the time to deployment when addressing a new traceability challenge.
The close partnership between Lyric and Cognex extends beyond hardware and software solutions, Yixian said. Cognex helps Lyric resolve problems at customer sites and releases solutions to address new challenges in the EV industry on a global scale.
“Our customers are spread out all over the world, we have to meet different certifications and compliance requirements,” Yixian said. “Cognex has a global sales and service network that helps us meet global product and safety certifications, allowing us to provide better service domestically and internationally.”
Test driving new solutions and a mutual commitment to R&D
The EV battery market is an evolving industry, and Lyric’s customers continually require customized, integrated solutions to meet new demands. Lyric invests in research and development to keep up with new battery designs, manufacturing specifications, regulations, and customer needs.
Having a machine vision supplier with the same values helps Lyric solve technical issues quickly, Yixian said. “When our customers have issues in the field, Cognex actively cooperates with us to find the cause and provide us with timely technical services and support,” he said. “When we encounter new problems, Cognex actively introduces new products and technologies to solve them.”
That’s why Lyric continues to test and implement new solutions from Cognex. It is currently testing a new Cognex solution, the Trevista CI Dome, for EV battery defect detection.
The Trevista CI Dome uses diffused dome lighting, computational imaging, and vision software to create high-quality topographical images. The solution is especially suited for defect detection in EV battery manufacturing because it reveals subtle defects and displays depth information that can distinguish between a functional flaw and an acceptable anomaly.
“The Trevista CI Dome gives us better imaging quality, higher processing speed, and works seamlessly with other tools to carry out a systematic inspection,” Yingjun said.
Lyric is also testing the In-Sight L38, a 3D machine vision system, for post-soldering defect detection and inspecting the surfaces of EV batteries.
By investing in new solutions and technologies, Lyric and Cognex are poised to revitalize the EV industry. “I believe the cooperation between Lyric and Cognex, by virtue of the strong performance of our products, will provide better and higher-quality services to domestic and internal customers,” Yixian said.