Fast Classification of Black Mass by Handheld LIBS Based on
Machine Learning
CHEN Nan1, 2*, ZOU Zhao-hua1, LUO Zi-xun1, SHEN Xin-jian1, LIU Yan-de1, 2
1. School of Mechatronics & Vehicle Engineering, East China Jiaotong University, Nanchang 330000，China
2. National and Local Joint Engineering Research Center of Fruit Intelligent Photoelectric Detection Technology and Equipment, East China Jiaotong University, Nanchang 330000，China
Abstract：With the rapid development of new energy vehicles and energy storage devices, the number of waste lithium batteries has surged. Black mass, as the most critical material in the battery recycling process, has a complex and diverse composition, which is very likely to cause resource waste and environmental pollution if it cannot be effectively identified and categorized. Traditional detection methods are time-consuming and costly, making it difficult to meet the demand for real-time classification of black mass in industrialized scenarios. Laser-induced breakdown spectroscopy (LIBS) offers a new approach for rapid identification of black mass, leveraging its advantages of simultaneous multi-element detection, rapidity, and high efficiency. In this study, a handheld LIBS spectrometer is combined with machine learning algorithms to achieve accurate identification and efficient classification of black mass from used lithium batteries. The experiment firstly purchased nine common lithium battery black mass samples from Ganzhou Haohai New Material Co., Ltd. and collected the spectra of the black mass samples by a handheld LIBS instrument; In order to improve the quality of spectral data and the accuracy of the subsequent modeling, maximum and minimum normalization (MMN) and Savutzky-Golay smoothing filter (SG) were used to optimize the preprocessing of LIBS spectral data; In the feature extraction stage, the pre-processed spectral data were subjected to dimensionality reduction by introducing two data dimensionality reduction methods, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), respectively; Finally, three types of classification models, namely, Random Forest (RF), Partial Least Squares Discriminant Analysis (PLS-DA) and Back Propagation Neural Network (BPNN), were established based on the dimensionality-reduced spectral data; The optimal black mass classification model is selected by comparing four aspects: classification accuracy, precision, recall and F1 score of the test set. The experimental results show that the classification model constructed using a combination of Linear Discriminant Analysis (LDA) and a Backpropagation Neural Network (BPNN) achieves the best recognition performance, with an overall accuracy of 99.70% on the test set. The results validate the feasibility and effectiveness of LIBS technology combined with machine learning methods for identifying lithium battery black mass, providing a theoretical basis and practical value for the efficient classification and reuse of waste lithium battery black mass.
陈 楠，邹招华，罗兹循，沈新建，刘燕德. 基于机器学习的手持式LIBS对黑粉的快速分类[J]. 光谱学与光谱分析, 2025, 45(12): 3342-3348.
CHEN Nan, ZOU Zhao-hua, LUO Zi-xun, SHEN Xin-jian, LIU Yan-de. Fast Classification of Black Mass by Handheld LIBS Based on
Machine Learning. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(12): 3342-3348.
