Optimizing the advance rate of an Earth Pressure Balance Machine based on geology using Machine Learning models

The estimation of the advance rate of an Earth Pressure Balance Shield (EPBS) is a crucial part in both tender and excavation phases. In both cases, there are no guarantees that the advance rate estimated for each geology is the optimum. Eiffage Génie Civil collected a large number of TBM excavation data and geology parameters especially from the Grand Paris Express project.

This study tends to analyze the collected data to optimize the geological identification and the excavation procedure. It develops a methodology combining unsupervised and supervised machine learning models (ML) with a Swarm Intelligence (SI) algorithm. The k-means clustering serves as a reconnaissance of the geology as the TBM advances. The Random Forest regression model serves to predict the data used in the optimization function. The Particle Swarm Optimization (PSO) determines the optimal advance rate that solves the optimization function.