The battery cell in a new energy vehicle is constructed by stacking dozens of layers of electrodes. The metal strips extending from the edges of the electrodes are called "tabs", serving as the conductive pathways for current flow. If a tab on any layer folds over unnoticed, it could pose a risk of short circuits and fires, while appearing completely normal from the outside. So what is the solution? "Let AI count the number of electrode layers", said Gao Hongchao, R&D Director at OPT Machine Vision. "If AI registers 24 or 26 layers in a 25-layer cell, it indicates a folded tab."

Dongguan is home to over 220,000 industrial enterprises, with the electronic information manufacturing sector achieving a gross output value of 1.13957 trillion yuan for industrial enterprises above designated size. The larger the industrial scale, the more critical quality inspection becomes. The algorithm powered by AI has reduced the miss rate to one in 100,000 and cut the number of quality inspectors on the relevant production lines by over 30%. This breakthrough marks the transition of quality inspection in the Dongguan manufacturing sector from manual visual inspection to AI visual inspection.

Empowered by AI, quality control in manufacturing has shifted from experience-driven to data-driven. "Traditional manufacturing relies heavily on engineers' experience. With the advancement of AI, we have evolved from an experience-driven approach to one that is fueled by data and intelligence," said Cao Ling, Senior R&D Director at OPT Machine Vision.

Dongguan can truly transform its isolated cases of AI visual inspection into a core capability in the manufacturing sector, and ultimately into a new foundation for quality by turning manufacturing experience into data, making algorithmic models work in tandem across the industry, and fully leveraging its supply chain ecosystem.

东莞制造装上“AI之眼”，可将漏检率降至十万分之一

在新能源汽车电池内部，电芯由几十层极片叠成。极片边缘伸出的金属片叫“极耳”，是电流进出的通道。如果某一层极耳在内部悄悄翻折了，从外面看毫无异常，却可能埋下短路、起火的隐患。怎么办？“让AI去数极片的层数。”奥普特研发总监高红超给出答案，“25层的电芯，数出来24层或26层，就说明有一层出了问题。”

东莞全市工业企业超过22万家，电子信息制造业完成规上总产值11395.7亿元。产业体量越大，质检这道“眼睛”就越关键。这套“聪明算法”把漏检率降到了十万分之一，相关产线的质检人员减少了30%以上。这一突破，正是东莞制造业质检从人工目检走向AI全检的一个缩影。

AI让制造业质量控制从“经验驱动”走向“数据驱动”。“传统制造业更多依赖工程师的经验。随着AI技术发展，我们从经验驱动演变成了数据驱动、智能驱动。”奥普特研发高级总监曹玲说。

当制造经验被沉淀为数据，当算法模型在产业内协同，当一座城的供应链生态共同参与，东莞制造才可能把“AI之眼”从案例变成能力，从能力变成新的质量底座。

策划、统筹 | 唐波
文 | 记者 彭钦
图 | 记者 王俊伟
译 | 黄嘉嘉
审 | 曾敏