Feature-Based Transfer Learning for High-Value Com-ponent Recovery in Electric Vehicles: An InceptionV3 Model Evaluation

The worldwide transition to electric vehicles (EVs) has resulted in a substan-tial rise in the quantity of end-of-life (EoL) EVs that need effective recycling methods. This study examined a method of transfer learning that uses fea-tures to classify valuable electrical components from end-of-life electric ve-hicles. The study made use of a dataset consisting of high-resolution photo-graphs of different electronic control units (ECUs). The photos were pro-cessed using pre-trained InceptionV3 convolutional neural network (CNN) models to identify distinctive features. The performance of four classifiers, namely the Support Vector Machine (SVM), k-Nearest Neighbors (kNN), Random Forest (RF), and Naive Bayes (NB), was tested using the collected features. The dataset was partitioned into training, validation, and test sets using a 70:15:15 stratified split to guarantee an equitable distribution of all classes. The InceptionV3-SVM pipeline achieved the highest performance, with training, validation, and test accuracies of 100%, 97%, and 97%, re-spectively. Other classifiers also demonstrated strong performance, with val-idation and test accuracies exceeding 94%. The high accuracy and generali-zation capabilities of the InceptionV3-SVM pipeline indicate its potential for practical deployment in sustainable manufacturing processes. This study provides a foundation for further research in the automated sorting and re-covery of high-value electronic components from EVs, potentially extending to a broader range of electronic components and applications. The findings highlight the effectiveness of transfer learning techniques in enhancing the efficiency and accuracy of recycling operations in the automotive industry.