Presenting the best prediction model of water inflow into drill and blast tunnels among several machine learning techniques

During the construction of a tunnel, water inflow is one of the most common and complex geological disasters and has a large impact on the construction schedule and safety. When serious water inflows occur in tunnel construction, huge economic losses and casualties can occur. Therefore, this phenomenon's prediction is an important task to ensure the safety and schedule during the underground construction process. In this article, water inflow into tunnels was predicted using six machine learning techniques of long short-term memory (LSTM), deep neural networks (DNN), K-nearest neighbors (KNN), Gaussian process regression (GPR), support vector regression (SVR), and decision trees (DT) by applying 600 datasets. The key features of the models mentioned above were discussed. Finally, in terms of accuracy, the models were ordered as LSTM, DNN, GPR, SVR, KNN, and DT with the route mean squared errors of 4.07486, 4.66526, 5.77216, 12.95589, 16.63670, and 17.99058, respectively.

Automated summary

This study predicted water inflow into tunnels using six machine learning techniques and ranked the models based on their accuracy, with LSTM being the most accurate and DT being the least accurate. This provides valuable insights for ensuring safety and progress in tunnel construction.