Published Paper Details:

Since the study presents an innovative framework for developing precise 1D-CNN-based fault classifiers adaptable across diverse rotating machines and conditions. Utilizing Bayesian processes for optimizing architectural parameters and hyperparameters, the framework emphasizes a transferability-focused loss function. This approach yields a heterogeneous 1D-CNN ensemble for fault diagnosis, surpassing individual model performance. Comparative evaluations across various datasets highlight its superior transfer learning capabilities over traditional classifiers. With limited labeled samples from different industrial contexts, the research aspires to establish a versatile fault diagnosis ensemble for rotating machinery, ensuring easy real-world implementation and updates. Additionally, an extension integrates a Voting Classifier, combining SVM, Logistic Regression, KNN, and Decision Tree classifiers, achieving an impressive accuracy range of 99% to 100%.