Single Plane Temperature Sensors Placement Optimization for Aligned Air-Cooled Cylindrical LiB Module

In lithium-ion batteries (LiBs), thermal runaway (TR) poses a significant risk of fire that requires early identification to control its propagation throughout the battery module and possible module-wide fires. In an attempt to improve the TR cell position prediction performance by Machine Learning (ML) models in the aligned arrangement of cells and single-mid plane uniform thermal sensor distribution scenario that under-performed in the previous work, the present study incipiently implements the randomization approach of temperature sensor distribution on the single-mid plane of the battery computational domain. In addition to the regular sensor distribution pattern discussed in the previous study for the aligned cell arrangement and single-mid plane regular thermal sensor distribution scenario, two more random patterns of distribution of temperature sensors were considered in the present study, maintaining the initial number of temperature sensors distributed to be 100 in all three cases. The work aims to optimize the temperature sensors for all the chosen sensor distribution patterns, which could be further used to develop an ML model and test its accuracy in predicting the cell position undergoing TR in the battery domain under different operation conditions. To identify the optimized sensors, the Pearson Correlation Coefficient (heat map) optimization approach is implemented by analyzing the correlations between temperature sensors and potential fault positions based on the coefficient limiting threshold of 0.85. Utilizing the optimized set of temperature sensors obtained from this study for each of the chosen distribution patterns of temperature sensors, the effects of randomization of temperature sensor distribution on a single-mid plane and associated variation in the TR cell position prediction accuracy from the ML can be accounted for.