Real-Time Dynamic Earth-Pressure Regulation Model for Shield Tunneling by Integrating GRU Deep Learning Method with GA Optimization

This paper proposes an intelligent framework to predict and automatically regulate earth pressure using a deep learning technique during earth pressure balance shield tunneling. A prediction model was proposed by integrating a new cost function (relative mean square error) with a gated recurrent unit (GRU). The moving average smoothing method was also incorporated into the GRU model to reduce the noise of the dataset and improve the accuracy of the proposed model. A real-time dynamic regulation model for adjusting the operational parameters was proposed by integrating the GRU model into a genetic algorithm-based optimizer. By adjusting the operational parameters, the dynamic regulation model regulates the excessive predicted earth pressure within a suggested range. The proposed prediction and regulation models were applied to a metro tunnel construction in Luoyang, China. The results show that the proposed models provide good guidance for automated tunnel construction.