Intelligent fault diagnosis technique for rotating machinery based on parameter-optimised variational mode decomposition and improved bidirectional gated recurrent unit

The vibration signals in rotating machinery are characterised by their non-stationary, non-linear nature and are often mixed with noise. These complexities make it difficult to extract effective fault features, which reduces the accuracy of fault diagnosis. In this paper, a fault diagnosis method for rotating machinery based on adaptive variational mode decomposition (VMD) and an improved bidirectional gated recurrent unit (BiGRU) is proposed. They are both optimised using an improved marine predator algorithm (IMPA). Firstly, a revised differential variational strategy and variable step size Lévy flight strategy are introduced into a marine predator algorithm (MPA). The search phase of the algorithm is adjusted to balance the exploration and exploitation capabilities. The IMPA is used to optimise the optimal number of modal layers and penalty factor in order to decompose the vibration signals of the rotating machinery into a number of intrinsic mode function (IMF) components and the energy entropy of each component is calculated to form the eigenvectors. Meanwhile, the five initial parameters of the BiGRU are optimised by the IMPA to ensure the accuracy of diagnosis and the feature vectors are trained by this IMPA-BiGRU network model for pattern recognition and classification. Finally, the model is validated on the Southeast University (SEU) gearbox dataset and the Jiangnan University (JNU) bearing dataset. The results show that the proposed method is effective in identifying fault types under different operating conditions and more accurate diagnostic results can also be obtained in a noisy background.