Intelligent Optimization and Machine Learning Algorithms for Structural Anomaly Detection using Seismic Signals

TL;DR

The paper tackles structural anomaly detection in mechanized tunneling by marrying forward seismic simulations with parameter-estimation inversion and machine-learning surrogates. It introduces the unscented hybrid simulated annealing (UHSA) framework, which combines global search with local UKF refinement, and a deep Gaussian covariance network (DGCN) surrogate to accelerate evaluations. On a small-scale aluminium-block experiment, UHSA achieves near-PSO accuracy with a fraction of the forward runs (e.g., 293 or 122 vs. 600), while multi-output DGCN surrogates further reduce computation and can outperform single-output variants. The findings point to a practical route for rapid, robust localization of structural anomalies from seismic data, with potential for real-time guidance in tunneling operations.

Abstract

The lack of anomaly detection methods during mechanized tunnelling can cause financial loss and deficits in drilling time. On-site excavation requires hard obstacles to be recognized prior to drilling in order to avoid damaging the tunnel boring machine and to adjust the propagation velocity. The efficiency of the structural anomaly detection can be increased with intelligent optimization techniques and machine learning. In this research, the anomaly in a simple structure is detected by comparing the experimental measurements of the structural vibrations with numerical simulations using parameter estimation methods.

Figures (11)

• Figure 1: Principle of the UHSA algorithm Nguyen2016
• Figure 2: Gaussian process approximation with noisy data
• Figure 3: Schematic overview of a DGCN. The input dimension is not related to the presented application.
• Figure 4: Visualisation of a multi-output SML problem
• Figure 5: Small-scale laboratoy experiment