Research on the application of improved density peak clustering algorithm in structural plane grouping identification under multiple noise interference

The distribution characteristics of structural planes are one of the important controlling factors affecting the stability of rock mass slopes. Addressing the issues of strong noise interference and difficulty in identifying local high-density regions in structural plane occurrence data, this paper proposes an improved fast density peak search algorithm (PE-CFSFDP) that integrates potential energy density and entropy regulation mechanisms to enhance clustering accuracy and stability. This method introduces a potential density weighting strategy based on the traditional CFSFDP algorithm, and constructs a potential entropy change model to enhance the adaptability of local density estimation and the sensitivity of cluster center identification. This article systematically compares and analyzes the performance differences between PE-CFSFDP and traditional CFSFDP methods in terms of density evolution, entropy distribution characteristics, and cluster boundary identification by constructing density heatmaps and entropy change curves of structural plane occurrence data. Comparative experiments conducted under various noise levels demonstrate that the PE-CFSFDP algorithm exhibits significantly superior clustering accuracy in high-density distribution areas compared to the traditional CFSFDP method and demonstrates stronger robustness in high-noise backgrounds. Among them, both the CH index and the Dunn index were significantly improved, and the fluctuation range of density entropy was significantly reduced. The results of engineering application verification show that PE-CFSFDP can effectively extract high-density blocks of local structural planes, enhance the identification ability of dominant structural planes, and possess good adaptability and engineering promotion value. Propose a clustering algorithm that enhances density recognition ability, which can improve the accuracy of structural plane recognition. Demonstrate the robustness and separation capability of the method under high noise conditions. Provide a scalable method for automatic grouping and analysis of structural planes to facilitate rock mass stability research.