<p>As the substantial increase in sales in the electric vehicle market, lithium-ion batteries. Lithium-ion batteries are widely used in various devices. However, they will inevitably have some damages due to external mechanical abuse of loads. Those damages may lead to different consequences, such as defects without triggering a hard short circuit or thermal runaway. If the safety risk level of a battery is unknown, we will fail to make an early decision. Here, we develop a machine learning (ML) classifier to successfully realize online safety risk level classification. Random Forest (RF) algorithm is selected due to its low time cost and high accuracy for this problem. The battery safety risk level is classified into four groups: 1. Normal; 2, Latent risk (shorted but works as normal); 3, Low risk (shorted but the temperature is lower than the TR triggering temperature); 4, High risk (shorted and the temperature will be higher than the TR triggering temperature). The dataset will be obtained and generated by an experimental and FE modeling combined manner. At first, a series of internal short circuit (ISC) triggering tests (including different loading conditions and SOCs) and cycling tests (including normal batteries and defective batteries) were conducted. Secondly, A multiphysics modeling that consists of several sub-modes, such as battery model, thermal model, short circuit model, and thermal runaway model, was developed and validated by the experimental data. The multiphysics model was used to conduct a large number of parametric calculations. Different SOCs, charging/discharging rates, and short circuit resistance are considered. The COMSOL Multiphysics was used to develop the model and conduct the non-linear solving. Third, a dataset that has 40,000 samples was generated. The dataset consists of the battery voltage, current, and temperature signal at different working conditions and different scenarios which have risk levels. For each data sample, input X is a vector that combines the current and a sequence of voltage or voltage slope, output Y is an integer number from 1 to 4 that indicates the safety level and is labeled manually. Finally, a Random Forest (RF) classifier is developed and trained based on this dataset. The classifier can realize online safety risk level classification via inputting the current state of the battery. We also find that the prediction accuracy is highly related to the battery capacity and ISC current (or ISC resistance). This method will provide an insight into the identification, monitoring, and warning of battery safety issues.</p>

Data-Driven Safety Risk Classification of Lithium-Ion Batteries

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.