A Supervised Learning-based Framework for Failure Detection in Toy Cars Using Acoustic Signal Analysis

Failure detection in toy cars has emerged as a critical area of study in both academic and industrial contexts. Traditional techniques are often inefficient, time-consuming, and error-prone. Previous studies in the field of failure detection using acoustic signal analysis also struggle with detecting subtle or intermittent failures. To address these challenges, this paper presents a novel framework for toy car failure detection using supervised machine learning and acoustic signal analysis. The proposed methodology integrates three essential steps: data augmentation, preprocessing, and supervised learning. Data augmentation techniques, such as noise injection and pitch shifting, simulate anomalous conditions. Preprocessing ensures the extraction of meaningful features, noise reduction, and standardization of audio signals. Finally, the framework employs multiple supervised learning algorithms: Random Forest (RF), Gradient Boosting Machines (XGBoost), Support Vector Machines (SVM), Convolutional Neural Networks (CNNs), and Multilayer Perceptron (MLPs) for accurate anomaly detection. Simulation results show the superiority of the XGBoost algorithm, which achieves the best performance with an F-Score of $44 \%$ and an accuracy of $91.6 \%$. Compared to existing approaches, the proposed framework addresses limitations related to imbalanced detection results while demonstrating competitive accuracy. Consequently, the research contributes to advancing automated, accurate, and reliable toy car failure detection systems in both educational and industrial applications.