Online process monitoring with near-zero misdetection for ultrasonic welding of lithium-ion batteries: An integration of univariate and multivariate methods

Ultrasonic metal welding is used for joining lithium-ion batteries of electric vehicles. The monitoring of battery joining processes requires near-zero misdetection in order to prevent any battery joints with a low quality connection going into the downstream assembly. The conventional control chart techniques widely used in many process monitoring systems were designed based on a pre-specified false alarm rate. To ensure weld quality and reduce manual inspection at the same time, a near-zero misdetection rate is desired foremost while achieving a low false alarm rate. A monitoring algorithm targeting near-zero misdetection is developed in this article by integrating univariate control charts and the Mahalanobis distance approach. The proposed algorithm is capable of monitoring non-normal multivariate observations with flexible control limits to achieve a near-zero misdetection rate while keeping a low false alarm rate. By implementing this algorithm on the ultrasonic welding process of battery manufacturing, the developed algorithm proves to be effective in achieving near-zero misdetection in process monitoring to ensure battery weld quality. The developed algorithm also shows great potential for monitoring other processes that target at near-zero misdetection.