Crack detection in carbide anvil using acoustic signal and deep learning with particle swarm optimisation

Abstract

The carbide anvil plays a significant role in producing synthetic diamond. However, it suffers from complex alternating stresses and consequently results in fatigue damage such as cracks. Accurate crack detection of the carbide anvil still faces a significant challenge. This paper develops an acoustical crack detection method of the carbide anvil using the deep learning. In the method, an online sound impulse extraction strategy is designed to construct an anvil dataset. Subsequently, the stacked autoencoder model is designed to learn a robust feature representation of the anvil states from the measured sound impulse signals. Besides, an improved particle swarm optimisation method based on classification probability is proposed for the hyper-parameter optimisation. Finally, the performance of the proposed method is evaluated using experimental data. This research can provide a potential tool for the engineers to automatically detect the crack of the carbide anvils in the diamond industry.