Ensemble AI Fault Diagnosis Model Using Ultrasonic Microphone

the earliest method for predicting industrial faults is through ultrasonic diagnostics. This study focuses on analyzing a dataset comprising 10 main faults, from which 29 different features are extracted in both time and frequency domains. To reduce dimensions, the t-SNE method is utilized, which effectively distinguishes between the classes. Four predictive models, namely k-nearest Neighbor (kNN), Logistic Regression (LR), Gaussian Naive Bayes (GNB), and Support Vector Machine (SVM), are then applied. To tackle the classification problem, a stacking classifier based on a meta-classifier, which combines multiple classification models, is introduced. The proposed method achieves an accuracy approximately 6% higher with lower variations over 5 cross folds, exhibiting superior performance compared to other approaches.