Precursor-guided soft recurrence and topological proxy representation learning for early warning of coal and gas outbursts

Chengfei Liu, Zhonghui Li, Enyuan Wang, Jiankun Xu and Yubao Zhao

Chaos, Solitons & Fractals, 2026, vol. 210, issue P2

Abstract: Reliable and accurate early warning of coal and gas outbursts depends on the identification of effective precursors from continuous monitoring data that can characterize the process of risk incubation. Existing early-warning methods for coal and gas outbursts primarily include traditional approaches based on statistical features and empirical criteria, as well as data-driven approaches using deep learning. The former are often unable to adapt to the nonlinear evolution of multi-indicator signals under complex operating conditions. Although the latter exhibit strong nonlinear feature-learning capability, most of them still focus primarily on feature extraction and class discrimination within time-series windows, while insufficient use is made of the state associations and structural characteristics among different temporal positions within each window. To address this limitation, a PKRT-Net model was developed for early warning of coal and gas outbursts using continuous acoustic emission (AE), electromagnetic radiation (EMR), and gas concentration monitoring sequences. Precursor analysis was first conducted to identify how normal and risk windows differ in sustained trend variation, energy organization, cross-channel coupling, and intra-window state associations. Based on these observations, each multi-indicator monitoring window was transformed into a soft recurrence structure, in which trend and energy-related precursor patterns were used to guide the representation of state associations among different temporal positions. Compact structural descriptors were then introduced to summarize the connection strength, local closure, and overall complexity of the recurrence structure, while latent dynamical constraints were used to preserve the continuity of risk evolution across adjacent windows. Experimental results showed that PKRT-Net outperformed multiple baseline models in training stability, threshold consistency, F1 score (94.8%), AUPRC (95.0%), and AUC (95.1%). Analyses of representative samples and real-time field cases further confirmed its capability to track risk stages and support graded early warning. These findings indicate that, for coal and gas outburst warning, modeling the process of risk approach from the perspective of multi-indicator relational structural evolution offers greater stability and practical potential.

Keywords: Coal and gas outburst; Early warning; Soft recurrence; Topological proxy features; Structural evolution (search for similar items in EconPapers)
Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0960077926007745
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:chsofr:v:210:y:2026:i:p2:s0960077926007745

DOI: 10.1016/j.chaos.2026.118633

Access Statistics for this article

Chaos, Solitons & Fractals is currently edited by Stefano Boccaletti and Stelios Bekiros

More articles in Chaos, Solitons & Fractals from Elsevier
Bibliographic data for series maintained by Thayer, Thomas R. ().