Non-destructive residual moisture measurement of lithium-ion battery electrodes using near-infrared spectroscopy

The production environment accounts for a significant portion of the battery manufacturing costs. Most manufacturing processes in battery production require a dry environment. Otherwise, reactions between moisture and battery materials, especially active materials and electrolytes, can adversely affect battery life. Current research focuses on smaller dry rooms to reduce costs. However, with fully automated, digitized production processes, more inline measurement tools are needed to ensure quality and efficiency. Residual moisture measurement during the battery production process is, therefore, crucial. In this work, a contactless, non-destructive method for determining residual moisture was developed. Numerous experiments were conducted to examine potential influencing factors of electrode moisture. Near-infrared (NIR) spectroscopy was used to detect the residual moisture of electrodes. Artificial intelligence (AI) models were trained to predict residual moisture using a non-destructive, inline-capable hyperspectral imaging (HSI) camera. The true labels of the residual moisture content were determined based on Karl Fischer titration (KFT) results. For data preprocessing, multivariate image analysis techniques were required to handle the large data sets generated by the HSI camera. In the test data set, 84 % of the residual moisture predictions fell within the 1-σ uncertainty band. This scan-level evaluation was subsequently complemented by a stricter, sample-independent validation in which entire electrode samples (rather than individual scans) were withheld from training; under this stricter split, the coefficient of determination decreased from R 2 = 0.66 to R 2 = 0.39, while the model still outperformed a partial least squares (PLS) regression baseline evaluated under the same conditions, indicating both the promise and the current generalization limits of the approach. The robustness of the predictive AI model was further tested by moving samples to align with the battery manufacturing environment. The presented methodology enables inline residual moisture measurements integrated into the battery production process to monitor electrode moisture during cell assembly.