Classifications of Lithium-Ion Battery Electrode Property Based on Support Vector Machine with Various Kernels

Manufacturing chain of lithium-ion batteries belongs to a significantly complex process with many coupled product parameters and intermediate products. To well monitor and optimize battery manufacturing process, it is vital to design a data-driven approach for effectively modelling and classifying the product properties within this complicated production chain. In this paper, a support vector machine (SVM)-based framework, through using four various and powerful kernels including linear kernel, quadratic kernel, cubic kernel and Gaussian kernel, is proposed to well classify the electrode mass loading property of battery. The effects of four crucial variables including three product features from mixing step and one product parameter from coating step on the electrode property classification are also investigated. Comparative results illustrate that electrode mass loading can be effectively classified by the designed SVM framework while Gaussian kernel-based SVM achieves the best classification for all labelled classes. This is the first time to systematically evaluate and compare the performance of different kernel-based SVMs on the battery electrode property classification. Due to data-driven nature, the proposed SVM-based framework can be easily extended to classify other product properties and analyze other variables in battery production domain.