Rolling bearing fault diagnosis using an optimization deep belief network

TLDR

Rolling‑bearing vibration signals are highly variable due to operating conditions and background noise, making fault identification difficult. The study proposes an optimization deep belief network to diagnose rolling‑bearing faults. The network is trained by first pre‑training RBMs, then fine‑tuning with stochastic gradient descent, and its architecture is optimized via particle swarm to improve classification accuracy on simulated and experimental bearing data. Experimental results show the optimized DBN outperforms other intelligent methods in accuracy and robustness.

Abstract

The vibration signals measured from a rolling bearing are usually affected by the variable operating conditions and background noise which lead to the diversity and complexity of the vibration signal characteristics, and it is a challenge to effectively identify the rolling bearing faults from such vibration signals with no further fault information. In this paper, a novel optimization deep belief network (DBN) is proposed for rolling bearing fault diagnosis. Stochastic gradient descent is used to efficiently fine-tune all the connection weights after the pre-training of restricted Boltzmann machines (RBMs) based on the energy functions, and the classification accuracy of the DBN is improved. Particle swarm is further used to decide the optimal structure of the trained DBN, and the optimization DBN is designed. The proposed method is applied to analyze the simulation signal and experimental signal of a rolling bearing. The results confirm that the proposed method is more accurate and robust than other intelligent methods.