Predicting the geological condition beyond the tunnel excavation face using MSP monitoring data and LSTM algorithm

Abstract The ground conditions beyond an excavation face, especially discontinuities in rock masses, have a significant influence on tunnel construction. However, the actual observed during construction are often different from predicted geotechnical site explorations carried out design stage. Changes may require alterations design, leading to substantial disruptions schedule and budget. In this regard, accurate evaluation prior stages essential for successful projects. Machine learning models been developed order evaluate condition of within 50 m face. data (rotational pressure, feed drilling (advance) speed) obtained large boring hole machine, called MSP, at NATM granite formation located South Korea were logged, Discontinuity Score (DS) was appraised by analysing internal bore images taken after drilling. Then, LSTM algorithm applied develop machine model determine DS based logged data. most accurately when speed included input data, whereas those cases using only rotational pressure showed low prediction accuracy. Therefore, seems higher correlation than hydraulic with regard conditions, including discontinuities. Once additional is collected various sites, could be further enhanced become more robust provide solutions engineering problems.