Applying Machine Learning to the Smart Welding of Cylindrical Lithium-ion Battery Cells within Battery Modules

Battery packs are used in transportation and energy storage applications to provide sufficient power for machine operation. There are various types of battery cell structures as well as numerous methods of interconnecting the battery cells. This paper will focus on cylindrical battery cells, the 21700 type, and Smart Welding as the interconnect technology. Smart Welding is a hybrid version of an ultrasonic welder and an ultrasonic heavy wire bonder that was developed by Hesse GmbH. The problem addressed in this study was that previous research pointed to a need for improvement in determining the optimally smart welding parameters for cylindrical lithium-ion battery cells within a module. By using Taguchi techniques and a Naïve Bayes Classifier, a machine learning algorithm was developed to help determine optimum Smart Weld parameters that produced acceptable weld shear values and nuggets.