Research On Topological Manifold-coordinated Quality Evaluation Methods For Complex Structural Engineering Rock Masses

The quality evaluation of complex structural engineering rock masses represents a pivotal step in rock mass engineering.Accurate characterization of fracture parameters is essential,as these parameters influence the geometric,mechanical and hydraulic properties,as well as the deformation and stability of rock masses.The spatial pattern of fracture networks serves as a reflection of the topological properties of rock masses,often interconnected with various attributes.However,existing methods for characterizing fracture parameters and network features are deemed unsatisfactory,and the correlation between the topology of fracture networks and rock mass quality remains inadequately established.Conducted within the context of a significant road engineering project in the southwestern mountainous region of China,this research leverages abundant fracture data collected through field investigation and multi-angle nap-of-the-object photogrammetry facilitated by unmanned aerial vehicles.Adhering to the“rock mass structure control theory”,the study optimizes the three-dimensional（3D）fracture network modeling approach under the collaboration of topological manifolds,identifies the appropriate representative elementary volume and devises a method for evaluating rock mass quality integrating non-persistent fracture network attributes,effectively overcoming challenges in the field of engineering rock mass evaluation.The primary research contents and contributions are outlined as follows:（1）This research proposes a novel means for characterizing rock mass fractures,leveraging topological manifolds,manifold learning and fractal models.This comes after a comprehensive exploration and reproduction of several visualization techniques within the traditional stereonet-based system.The new means delves into the manifold structure of orientation data,capturing diverse fracture properties and quantifying both monofractal and multifractal behaviors inherent in the manifold structure.Additionally,an equivalent manifold structure is proposed to enhance the engineering applicability of the new means.The results indicate that the manifold structure of orientation data exhibits pronounced multifractal features,signifying a dispersed and heterogeneous distribution of orientations.A comparative analysis between the new means and traditional systems is conducted,considering data representation and fractal behavior,validating the effectiveness of the proposed system.（2）This research proposes a novel technique that integrates initial units and fracture-concentration domains to identify statistically homogeneous domains within fractured rock masses.The approach is based on a high-resolution 3D realistic slope model.The boundaries of initial units are established using lithology and the distribution of long fissures.Local shear zones are employed to partition each initial unit into multiple sub-fracture-concentration domains,facilitating the identification of fracture-concentration boundaries.Additionally,the topological manifolds incorporate deep learning to merge similar sub-domains and determine statistically homogeneous domains.The results demonstrate that the study area can be effectively divided into 15homogeneous domains when considering fracture orientations,and into 17homogeneous domains when considering orientations,trace lengths and roughness.（3）This research proposes a novel unsupervised learning algorithm that integrates the topological manifolds with clustering algorithms to identify fracture sets based on orientations or multiple parameters.The algorithm eliminates the need for selecting initial clustering centers and achieves global optimization throughout the iteration process.A key feature is the incorporation of the Silhouette index,facilitating an automatic search for optimal grouping results.The algorithm undergoes testing using a complex artificial dataset and Shanley and Mahtab’s dataset,demonstrating its effectiveness.Furthermore,it is applied to the fracture dataset within the study area and compared with other effective methods.The results show that the new algorithm identifies three fracture sets when considering only orientations and four fracture sets when considering orientations,roughness and apertures.（4）This research proposes an inference framework and an inversion method designed to deduce each fracture’s size and spatial distribution pattern from trace samples contained within rectangular sampling windows.The framework comprises four steps:data preparation,estimation of true trace length distribution,estimation of fracture diameter distribution and the solution of each fracture size.The non-unique inverse problem is addressed using theories such as stereology and high-order moments to acquire all fracture sizes of contained traces within the set.Using spatial geometric relationships based on the orientation,size,endpoints and modulus of each contained trace sample,the true location of each fracture in the rock mass space is determined.This information allows the lower limits of volumetric fracture density and intensity of the rock mass to be obtained.Consequently,the true spatial distribution pattern of each fracture can be inverted based on its location,orientation and size.The proposed framework and inversion method are validated through Monte Carlo simulation,demonstrating successful inversion of the in-situ spatial distribution pattern of each fracture within a rectangular sampling window on the slope of the study area.（5）This research proposes an innovative computational framework and a harmony dimension method for the estimation and prediction of the representative elementary volume（REV）within homogeneous domain#7.A high-fidelity 3D discrete fracture network（DFN）is generated,leveraging geometric information from 939 fractures and probability statistical theory.The framework incorporates extended computer graphics algorithms,extended Stokes’theorem and an equivalent porous medium model for secondary development.Key functions include fracture intersection analysis,calculation of volumetric fracture intensity P₃₂,seepage simulation and determination of the equivalent permeability coefficient within the DFN.The framework identifies the optimal REV size as 32.5 m×32.5 m×32.5 m.Employing the harmony dimension method,supported by a nonlinear optimization algorithm,four prediction equations are established.Using two-dimensional parameters,the REV size is predicted to be 35 m×35 m×35 m.The prediction method’s applicability is verified through 10 additional Poisson modeling processes and 570 seepage simulations.（6）This research proposes a novel approach for assessing rock mass quality and establishing quantitative conversion relationships between assessment systems.A thorough evaluation of rock mass quality is conducted in homogeneous domain#7within the study area.The rock mass samples are enhanced based on optimal REV.Using four widely recognized rock mass quality assessment systems,the quality of the rock mass for all samples is determined,and explicit expressions for these assessment systems are derived.Expanding on the core principles of these systems,a new method for rock mass quality assessment is introduced.Twelve quantitative conversion relationships,applicable to different rock strength ranges,are systematically deduced between the new method and traditional systems based on correlations among evaluation indicators.Through on-site investigations and comparisons with multiple studies,the efficacy of the new method is validated,establishing Grade III as the optimum rock mass quality.The results demonstrate that,for specific fractured rock masses,the new method exhibits higher conversion accuracy when using the same assessment system,underscoring its robust applicability.Applying the identical methods to additional homogeneous domains on the left bank slope of Chada can ultimately yield an overall rock mass quality rating of II～IV for the entire slope and obtain the main controlling factors of rock mass quality,including the number of moderately dipping fracture sets,the size of the representative elementary volume and the mean fracture size.Notably,the rock masses in homogeneous domains#1,#4,#6,#8,#12 and#14 exhibit subpar quality,categorized as Grade IV,indicating a heightened risk of rock mass instability and failure.Consequently,focused attention is imperative for protective measures in these areas.