[2O5-GS-13] AI application: Civil Engineering

[2O5-GS-13-02] Application of deep learning and ensemble learning on the rock mass evaluation of mountain tunnenl

〇Hirokazu Hasegawa¹, Ryo Hemmi¹, Satoshi Torii¹, Kosuke Tanimura¹, Dohta Awaji¹ (1. Shimizu Corporation)

Keywords:Mountain tunnel, Rock mass evaluation, Tunnel face, Deep learning, Classification algorithm

During the excavation of the mountain tunnel, the characteristics of the rock mass have been evaluated based on the rock strengths and on observing the tunnel face. However, these evaluations may differ depending on the variability of observers’ experience. Therefore, a system is demanded that objectively makes these evaluations. We are developing an AI system that evaluates the rock mass using the photos of tunnel face, a digital elevation model, and the measured pressure values derived from the drilling machine. We created and verified a model that predicts the evaluations of professional engineers. This prediction is calculated based on a deep-learning algorithm, and on its pattern classification by ensemble learning method. As a result of applying this system to the rock mass of mudstone, the predicted results indicated the same tendency as the professional observers.