Deep Ensemble-Based Classifier for Transfer Learning in Rotating Machinery Fault Diagnosis

Nowadays, intelligent models can correctly detect faults by analysing signals from rotating machinery. However, most of the studies are run in controlled environments and performance industrial real-world is not yet fully validated. Hence, a suitable tool to implement fault diagnosers transfer learning, this topic under development challenges persist. This paper proposes framework for creating accurate 1D-CNN based classifiers that be transferred between different machines working conditions. Multiple Bayesian processes select architecture parameters hyperparameters, which minimize loss function related their transferability other same machine operating conditions (such as load engine speed). The resulting model fitted heterogeneous diagnosis data ensemble improves unitary model. Subsequently, learning capability analyzed on two source sets using parameter transfer. results compared with classical classifiers. Finally, additional operations five target domain datasets validate our limited labeled samples allow interpretation results. ultimate goal find an generalize machinery easy implementation update settings.