Machine learning-based fault-tolerant inertial navigation system for micro tunnel boring machines

Abstract

Underground tunneling poses a navigation challenge as Micro Tunnel Boring Machines (mTBMs) operate in GPS denied environments. In such settings, Inertial Measurement Units (IMUs) offer an alternative navigation solution. However, faults in mTBM operation caused by various environmental and operational factors lead to inaccuracies in IMU data. This paper presents a machine learning-based fault tolerance system for IMU based mTBM navigation. IMU data corresponding to four specific mTBM faults: vibration, high temperature, magnetic interference and mTBM precession were collected under rigorous testing in a lab environment. Support Vector Machine (SVM) was applied to classify these faults and predict the mTBM's operational state based on the IMU readings. The results demonstrate the effectiveness of the SVM model in identifying and classifying the faults, thereby enhancing the reliability and safety of mTBM operations in challenging environments.