Collision damage assessment in lithium-ion battery cells via sensor monitoring and ensemble learning

Abstract

The safety of lithium ion batteries (LIBs) is an important issue in electric vehicle industry. Collision damage characterization is an essential aspect of the overall safety assessment of electric vehicle LIBs. Although immediate consequences may not appear evident, battery cells long-term safety and performance can be seriously affected by damages resulting from collisions, leading to dangerous failures. In this paper, a framework and associated methodology for battery cells collision damage assessment is proposed. An experimental rig was designed and built for the realization of a collision tests campaign. During such tests a number of sensor signals were collected and processed to extract significant features. The collision damages were then characterized in terms of physical inspection and electrical performances. An ensemble learning based pattern recognition decision making support system was setup by inputting environmental conditions parameters and sensor signal features to assess the collision damage class and the charge/discharge ability. Classification results are discussed and hints for future developments are proposed.