Construction of Bearing HI Based on Digital Twin and KOA-CNN-LSTM RUL Model

traditional remaining useful life (RUL) prediction techniques for rolling bearings are affected by complex operating conditions, noise interference, and weak vibration signals, leading to poor prediction accuracy and limited generalization in practical applications. To address this issue, a digital twin-assisted KOA-CNN-LSTM method for bearing RUL prediction is proposed. Firstly, establish a dynamic simulation model of rolling bearings using digital twin technology. This model is based on a bearing dynamics model combined with measured signals to estimate the evolution of bearing defects, and uses the least squares method to match the root mean square (RMS) value of vibration signals with the corresponding relationship between defect size. Import the excitation of evolutionary defects into the bearing simulation model, create a simulation model that reflects the degradation process of the bearing, generate simulated vibration signals for the entire life cycle of the bearing, and use the generated simulation data to construct the bearing health factor. Next, a KOA-optimized CNN-LSTM model for bearing RUL prediction is proposed. The processed simulated signals are used as input to the model, where the CNN extracts local information from the vibration signals and mines deep features. The features extracted by the CNN are then fed into the LSTM layer to capture the long-term and short-term dependencies of the time series. The best hyperparameters obtained from the KOA-optimized CNN-LSTM model are used to train the model and generate prediction results. Experimental results indicate that the predicted outcomes of the proposed method exhibit a favorable monotonic trend and can closely approximate the actual life values.