A deformation-based approach to the SoH estimation of collided lithium-ion batteries

Abstract

Extensive research work has been carried out mainly focusing on the assessment and prediction of battery cell State of Health (SoH) under operating conditions, however limited contributions focus on SoH following collision impacts. This paper proposes a method for estimating the battery cell SoH from collision deformation features. Experimental tests of collision impact were designed and realized on brand new battery cells to investigate deformation features. Deformed battery cells were subject to a 3D scanning procedure to retrieve the contour data, subsequently a number of geometrical features were extracted from the 3D image instances. The battery cells damage characterization was carried out by characterizing both physical and electrical performances following the collision impact tests. An intelligent assessment was carried out by adopting a neural network-based supervised machine learning paradigm for classification of deformed battery cells into safe, latent danger and unsafe cells respectively. Training and testing results show a clear pattern between geometrical deformation features and battery cells SoH, with classification accuracy up to 96.7% demonstrating the suitability of the proposed method for an effective assessment. Within electric vehicles applications, such method can provide a basis for safety design enhancement of lithium-ion battery system via finite element simulation of collisions impacts.