Operating Condition Recognition Methods of Mechanical System Based on CEEMDAN and GA-DBN

This paper presents a novel method for identifying the operating condition of a mechanical system by combining the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and the Genetic Algorithm-optimized Deep Belief Network (GA-DBN). The study focuses on the extraction of condition features and the optimization of DBN parameters. Firstly, the CEEMDAN algorithm is employed to decompose the signal and extract several intrinsic mode functions (IMFs), which are used for signal reconstruction. Subsequently, the GA is employed to optimize the parameters of the DBN, including learning rate, number of hidden layers and number of hidden layer nodes, etc. Then the power spectrum of the reconstructed signals are used as feature vectors, serving as input to the GA-DBN model for identifying the mechanical system operating condition. To evaluate the effectiveness of the proposed method, a case study is conducted using a sliding bearing friction test data. The results demonstrate the high accuracy of the method, which achieving 96.00%. Moreover, compared to the DBN model, the GA-DBN model has a significant improvement of 7.67% in the operating condition recognition.