[3C4] Fault diagnosis method for elevator brakes based on WTEMD and CNN-BiLSTM cross-attention
J Feng
China Special Equipment Inspection and Research Institute, China
Elevator brakes are key components ensuring the safe operation of elevators. Traditional inspections often rely on manual experience and simple signal processing methods, which struggle to handle non-linear and non-stationary failure modes. Therefore, this paper proposes a noise reduction method combining wavelet threshold denoising and empirical mode decomposition (WTEMD), and a convolutional neural network (CNN)-bidirectional long short-term memory (BiLSTM)-cross-attention hybrid model for elevator brake fault classification. Firstly, WTEMD was applied to decompose and denoise the original signals, obtaining smoother data. Subsequently, a CNN was employed to extract spatial features, a BiLSTM network was utilised to capture temporal characteristics and a cross-attention mechanism was introduced to fuse these complementary features. This enhanced feature representation enables accurate fault signal identification. Finally, the effectiveness of the proposed method was validated through experimental elevator brake failure tests, demonstrating high classification accuracy.
