A CNN-BiGRU Predictive Maintenance Method for Industrial Equipment Based on Attention Mechanism

To improve RUL prediction under complex operating conditions, this work employs the NASA C-MAPSS aero-engine dataset as a benchmark and presents a hybrid deep model combining a convolutional neural network (CNN), bidirectional gated recurrent units (BiGRU), and a self-attention mechanism. The CNN extracts local patterns from multivariate sensor inputs, BiGRU captures long-term temporal dependencies, and self-attention highlights critical time steps to enhance accuracy. Training uses mean squared error loss with the Adam optimizer. Experiments demonstrate that the proposed model outperforms common baselines across varied scenarios, offering a practical framework for RUL estimation on high-dimensional noisy data and significant engineering value for predictive maintenance.