Abstract: The performance of tunnel pavements is one of the key factors affecting traffic safety in tunnels. With the continuous expansion of the scale of tunnels in China, the issue of pavement performance prediction has received widespread attention. This paper focuses on the prediction of tunnel pavement performance by analyzing key factors that influence it and designs a deep learning model based on ConvLSTM (Convolutional Long Short-Term Memory). The model effectively captures both temporal and spatial features from operational data, demonstrating superior performance in predicting tunnel pavement smoothness. The Mean Squared Error (MSE) of the predicted value is around 0.7, with a Mean Absolute Error (MAE) as low as 0.2, and the Coefficient of Determination (R²) is over 0.90, confirming the effectiveness and accuracy of the model. This study provides robust support for tunnel pavement maintenance decision-making and lays the foundation for further optimizing tunnel pavement management and service quality.