水力破裂試驗三維顆粒流模擬技術精進

Abstract

本研究旨在提升三維水力破裂模擬技術之準確性與數值效率，建構一套整合流體滲流與裂隙破壞行為的顆粒流模擬模型。本研究以離散元素法（Discrete Element Method, DEM）為基礎，配合 Delaunay 四面體剖分演算法與幾何拓樸概念，實現動態網格更新與斷鍵判定機制，並透過 Python 程式語言開發自動化流程，以有效處理三維裂隙生成與流體導通過程中的網格重建與導水性計算。整體模擬流程考量了岩石裂縫內部的流場變化、鍵結破壞行為以及顆粒位移分布，使模型能夠同時呈現壓力變化與裂縫幾何成長之特徵。 為驗證模型之準確性與可行性，本研究亦同步進行室內物理水力破裂試驗。試體選用來自印尼之凝灰岩，具有完整結構、無明顯天然裂隙，適合用於破裂機制觀察。試體尺寸為 30×30×30 公分，並進行單軸應力條件施加約1.5MPa之軸向壓力，再施以定流量注水進行破裂試驗。注水系統由低流量注水機控制，約 0.1 ml/s，水壓可達 3 MPa 以上，壓力感測與數據紀錄則由監測系統即時完成。 數值模擬結果顯示，在等向圍壓與三軸應力條件下皆可觀察到典型的壓力–時間曲線，並能清楚辨識破裂壓力（P_b）、再開壓力（P_r）與閉合壓力（P_s）等重要特徵壓力點。在均向應力情境下，雖各主應力相等，仍可發現隨著圍壓提升，破裂壓力 P_b 呈現上升趨勢，顯示模型能合理反映流體破裂臨界壓力對應力環境的敏感性。此外，滲流模型中壓力傳遞方向與球顆粒的位移變化皆符合基本物理現象；例如裂縫擴展大多沿垂直於最小主應力方向發展，並伴隨孔隙水壓從裂縫起裂點向外傳導之行為，驗證本模擬在耦合力學與滲流場之描述上具備高度一致性與合理性。案例分析亦指出，當主應力差異不足時，裂縫成長方向將顯現隨機性，此結果與理論預期相符，突顯主應力差對裂縫導向的重要性。 然而，本研究在物理實驗過程中仍遭遇若干限制。例如 PU 膠封水效果不佳導致壓力累積與破裂時機錯判，部分試體產生非預期洩壓現象，可能造成水流外洩與壓力曲線異常。此外，模擬模型中部分參數尚未完成率定，如鍵結強度、水力導通參數等，影響壓力反演精度與裂隙形態準確度，未來仍須進一步改進與驗證。

This study aims to enhance the accuracy and computational efficiency of three-dimensional hydraulic fracturing simulations by developing an integrated particle flow model that captures both fluid seepage and fracture behavior. Based on the Discrete Element Method (DEM), the model incorporates Delaunay tetrahedral meshing and geometric-topological concepts to enable dynamic mesh updates and bond failure detection. A Python-based automated workflow was developed to efficiently handle 3D fracture propagation and flow path reconstruction, including transmissivity calculations. To validate the model’s accuracy and feasibility, a series of laboratory-scale hydraulic fracturing experiments were conducted. The tested specimens were intact tuff samples from Indonesia, free of visible natural fractures, making them suitable for fracture mechanism observation. The specimens measured 30×30×30 cm and were subjected to an axial stress of approximately 1.5 MPa under uniaxial loading conditions. Constant-rate fluid injection was performed using a low-flow-rate pump (approximately 0.1 ml/s), with pressures exceeding 3 MPa, and the entire injection process was monitored in real time using a pressure-sensing and data acquisition system. Numerical simulation results revealed characteristic pressure–time curves under both isotropic and triaxial stress conditions, clearly identifying key pressure points including breakdown pressure (P_b), reopening pressure (P_r), and shut-in pressure (P_s). Under isotropic stress, where all principal stresses are equal, the simulations still showed that P_b increased with confining pressure, indicating that the model appropriately captures the sensitivity of fracture initiation to the surrounding stress environment. Furthermore, the simulated fluid transmission paths and particle displacements aligned well with fundamental physical principles. For instance, fractures generally propagated perpendicular to the minimum principal stress direction, with pore pressure diffusing outward from the initiation site—demonstrating strong consistency between the coupled mechanical and hydraulic fields. Case analyses also indicated that when the stress difference between principal directions is minimal, the resulting fracture propagation becomes more random in orientation, consistent with theoretical expectations and emphasizing the role of principal stress anisotropy in guiding fracture direction. Nevertheless, several experimental limitations were encountered. Incomplete sealing by polyurethane (PU) led to premature pressure buildup and misinterpretation of fracture timing, resulting in fluid leakage and abnormal pressure responses in some specimens. Additionally, some key simulation parameters—such as bond strength and hydraulic conductivity—have not yet been fully calibrated, affecting both the accuracy of pressure inversion and the fidelity of simulated fracture geometries. These issues highlight areas for future refinement and model validation.