A construction framework of hybrid model for mechanical vibration fault diagnosis based on multiple classifiers in few samples

Fault diagnosis is critical to predictive maintenance and health management of mechanical systems. While sufficient data enables high-precision fault diagnosis models through machine learning, diagnostic accuracy tends to decline with limited samples, rendering single-classifier approaches inadequate. To address this, a hybrid fault diagnosis model incorporating multiple machine learning algorithms is proposed. This model integrates two discriminators and one optimizer: the discriminators select appropriate models and samples, while the optimizer fine-tunes model parameters to achieve high diagnostic accuracy. Comparative analysis shows that the proposed framework effectively enhances the processing capability for fault samples. Moreover, it exhibits extensibility and is not restricted to specific classifier types.