Comparative study of swarm intelligence and simplex shape algorithm for acoustic emission localization in thin-layered rocks

Thin-layered rocks, a common rock type in deep underground engineering, exhibit fracturing and instability under high stress conditions at great depths. This phenomenon is a key factor triggering major engineering hazards such as large rock mass deformation and rockbursts, posing severe threats to construction and operational safety. Research into the fracture instability mechanism of thin-layered rocks requires precise spatial localization of microfracture initiation and propagation within them. Taking four typical thin-layered rocks — carbonaceous slate, quartz-mica schist, quartz schist, and black shale — as study subjects, an anisotropic wave velocity model was established using function fitting methods based on the anisotropy dominated by bedding planes in thin-layered rocks. Numerical experiments systematically compared the localization accuracy of anisotropic wave velocity models versus uniform wave velocity models. Based on this anisotropic wave velocity model, acoustic emission localization methods were developed using different swarm intelligence algorithms and the simplex method. The performance differences of these algorithms for acoustic emission localization in rocks with varying wave velocity conditions were investigated. The optimized algorithms were validated through true triaxial compression tests, leading to comprehensive recommendations for algorithm selection under different rock conditions. The results show that: In acoustic emission localization wave velocity models, the anisotropic wave velocity model demonstrates significantly superior localization accuracy compared to the uniform wave velocity model. In numerical comparisons of acoustic emission localization algorithms, swarm intelligence algorithms demonstrated less sensitivity to overall sample wave velocity values and higher stability, whereas the simplex algorithm’s performance was closely tied to wave velocity — exhibiting higher accuracy in high-velocity rocks but relatively larger errors in low-velocity rocks. Using the macroscopic fracture morphology from triaxial tests as a reference, both the simplicial algorithm and adaptive particle swarm algorithm demonstrate high localization accuracy and good applicability for locating fracture sources in thin-layered rocks.