Effects of air permeability and perimeter area on air losses in compressed air tunnelling

Field measurement data obtained from a tunnel in Munich was used to develop a model to predict air losses in the process of compressed air tunnelling. In the implemented case study, compressed air was used to control the groundwater followed by placing a shotcrete lining as temporary support. Evolutionary polynomial regression was used as a data-driven method based on evolutionary computing aimed to search for polynomial structured model representing air losses. EPR uses a combination of the genetic algorithm (GA) and least square method to search for the most suitable structures. Data from different tunnel lengths were obtained to train/develop the model and also validate it with lines of data not used in the training phase. Comparisons made by the actual measurements represents robustness of the suggested models in learning and predicting the behaviour of the system. The sensitivity analysis conducted on the developed model further revealed the reliability of the models by presenting the expected effects from air permeability of the soil deposits in non-homogeneous and layered nature of the geomaterials involved in this particular study