Real-Time Prediction of TBM Driving Parameters Using In Situ Geological and Operation Data

The real-time prediction of tunnel boring machine (TBM) driving parameters is a critical step during the process of tunnel excavation. To obtain the dynamic characteristics and the approximate range of TBM driving parameters, a hybrid in situ prediction dataset is collected from the water conveyance channel project, which consists of operation data and geological data. The TBM driving parameter prediction task is transformed into a multiperiod and multivariate time series prediction task according to the technical specifications for construction. To fulfill the task, a TransTP network is designed in this article. The network consists of two novel modules, which are the temporal pattern attention detectionmodule and the temporal pattern attention mechanism module. The former is designed to learn the multiperiod feature representation with a convolutional component, and the latter is designed to assist the deep network to extract the multivariate features of inputs. The experimental results for TransTP promise superiority, which achieves a state-of-the-art performance for the TBM driving parameter prediction dataset.

Automated summary

This article proposes a TransTP network for real-time prediction of tunnel boring machine (TBM) driving parameters, achieving superior performance.