A dynamic learning method based on the Gaussian process for tunnel boring machine intelligent driving

Introduction: The application of intelligent learning methods to the mining characteristics and rules time-series data has gained increasing attention with rapid development deep learning. One critical such is assistant driving tunnel boring machines (TBMs), for which optimization parameters essential improve construction efficiency. However, existing prediction models TBM are “static” cannot dynamically capture parameter evolution during real-time cycles. Methods: In this study, we propose a novel dynamic model by introducing Gaussian process address problem. can learn decision-making experiences from historical cycles, update based on small sample current simultaneously achieve prediction. We focused in project western China. Results: results show that average relative errors predicted total thrust torque values were 1.9% 2.7%, respectively, accuracy was higher than conventional as random forest long short-term memory. fully exploited updating samples parameters, reducing time cost 29.7 s, satisfies requirements efficient application. Discussion: strategy adopted study provides reference other similar engineering applications. proposed thus facilitating enhancing efficiency tunnels. Conclusion: summary, establishes accurate contribute TBMs