A Transformer-Based Framework for Misfire Detection From Blasting-Induced Ground Vibration Signal

Abstract

The growing popularity of electronic detonators in tunnel construction facilitates the application of short-delay millisecond blasting. However, the occurrence of misfire endangers personnel safety, which the current blasting initiation system cannot identify. This article proposes a transformer-based framework for misfire detection from blasting-induced ground vibration signals and compares its capability with the signal processing method variational mode decomposition (VMD). The framework is established on a transformer structure cutting with locality self-attention vision transformer (CtL-ViT), which includes a cut embedding process and a locality self-attention calculation mechanism. Field experiments were conducted at a shallow tunnel construction site, and ground vibration signals were recorded. The proposed framework surpasses the VMD method in terms of higher misfire detection accuracy, precision, recall, and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${F}_{1}$ </tex-math></inline-formula> score, which are 97.5%, 91.9%, 99.7%, and 95.8%, respectively. The current results demonstrate the feasibility of CtL-ViT and the inapplicability of VMD in handling a short-delay hole-by-hole vibration signal and reveal the potential of the CtL-ViT framework as a second layer of protection.