崩塌運移之連體與非連體模型數值模擬比較 - 以RAMMS與PFC3D為例

A Comparison of Continuum and Discontinuum Models in Landslide Run-out Simulation – Examples Using RAMMS and PFC3D

蕭靖平(Ching-Ping Hsiao)

指導教授 ： 潘以文

國立交通大學/工學院/土木工程系所/碩士(2019年)

摘要

邊坡破壞之致災程度取決於影響範圍、運移速度與土石厚度等致災性參數。評估致災性參數的方法有經驗法、解析法與數值模型，其中數值模型是可以取得較詳細預測結果的方法，本研究比較兩種數值模型，其一是屬於連續體模型的RAMMS::AVALANCHE(後面簡稱RAMMS)，其二是屬於非連續體模型的PFC3D。RAMMS相較於PFC3D更容易使用、運算快速，且需要的控制參數很少(摩擦係數與紊流係數)，然因用於大規模崩塌運移模擬之案例相對於PFC3D仍少，本研究因此分別使用此二工具模擬同一崩塌地前後二期崩塌事件，藉以比較PFC3D與RAMMS的分析過程、模擬結果以及對應參數。
PFC3D的參數率定係藉由模擬力學試驗，將巨觀參數轉換成對應之微觀參數。RAMMS的參數率定則是藉由前一期的崩塌事件，藉由九宮格率定法，定性比較運移模擬結果與實地堆積，以得到最佳參數組合。以第一期崩塌事件率定出之參數再用於模擬次一期的崩塌事件。山區地形粗糙度影響運移行為頗大；就RAMMS之模擬，本研究比較是否依地形分區屬性設定不同參數之模擬結果；不考慮地形分區差異的案例稱為係數均勻設定，反之稱為係數分區設定。此案例之結果顯示：係數分區設定雖然需要更多時間完成參數率定，但是較係數均勻設定能夠模擬得更佳結果。
比較PFC3D與RAMMS之模擬結果，兩者皆可得到接近兩期崩塌事件堆積的結果。於本案例中PFC3D更能接近現地實際堆積狀況，但RAMMS模擬的結果和實際堆積狀況差距也不大。若對應相似之運移及堆積分析結果，RAMMS模擬所率定出的摩擦係數顯著高於PFC3D所率定出的摩擦係數，其原因是RAMMS僅以摩擦係數和紊流係數兩項考量運移的阻抗。另由比較五處莫拉克颱風期間不同破壞機制的崩塌案例，可見率定出之RAMMS參數會受破壞機制的影響。
RAMMS雖然不如PFC3D準確，但也能得到與實際崩塌接近的結果，再加上易於使用、計算時間快、電腦效能需求較小及控制參數少，為其用於致災性參數評估的優點。

關鍵字

崩塌 ； 運移模擬 ； 連續體模型 ； 非連續體模型 ； RAMMS ； PFC3D ； 參數率定

並列摘要

The hazard from slope failure is highly dependent of the runout behavior of the failure slope. Various approaches may be used to estimate the runout behavior of a slope failure; these methods include empirical-statistical methods, analytical methods and numerical models. Among the possible approaches, numerical models are able to obtain more details of the runout characteristics. This study aims to compare the modeling of the continuum model RAMMS::AVALANCHE (referred as RAMMS hereafter) and the discontinuum model PFC3D. Compared to PFC3D, RAMMS is easier to use, less computation intensive and with requiring much less input parameters (friction coefficient and turbulence coefficient only). However, compared to PFC3D, much less case studies using RAMMS in large-scale landslide simulation can be retrieved in literatures. This study made use of these two software to simulate a same large landslide occurred in two stages. The simulated results were compared to examine their simulation capability, computation efficiency and proper input parameters for each software.
In this study, the input parameter for PFC3D was calibrated by correlating the microscopic parameters with the macroscopic mechanical parameters based on the triaxial test simulations. The input parameters for RAMMS were calibrated through back calculation of the actual run-out terrain change in the first-stage failure in the landslide site. These back-calculated parameters were then used to simulate the landslide in the subsequent stage. The terrain roughness may affect the runout behavior of landslide mass; thus, this study considered two cases in RAMMS simulation. In case 1, different input parameters were assigned for different partitions of terrain (to be referred as “partition-coefficient setting”); in case 2, terrain-partition was not considered (i.e., parameters were uniform over the whole simulated region). Although it took longer time to complete the calibration for the case 1, simulation using partition-coefficient setting could generally simulate much better results than using case-1 setting.
Comparing the simulated results from PFC3D and RAMMS, both software were able to simulate the deposition area in two consecutive stages of the simulated landslide. However, the simulated results from PFC3D was relatively closer to the actual situation occurred in the field. Yet, the difference between the RAMMS simulated result and the actual situation was still acceptable. For the same landslide case, the friction coefficient calibrated for RAMMS was significantly higher than the friction coefficient calibrated for PFC3D. It is likely because there are more sources of energy dissipation in PFC3D than in that in RAMMS. This study also used RAMMS to simulate five landslide cases in Typhoon Marakot, 1999; these landslide sites failed owing to various failure mechanisms. The compared results show that the calibrated parameters were dependent of the failure mechanism.
Although the simulation by RAMMS may not be as accurate as that by PFC3D, RAMMS may still acceptably simulate the run out of a landslide as long as the input parameters have been properly calibrated. RAMMS also has the merits including ease of use, short computation time and less input parameters.

並列關鍵字

Landslide ； Run-out simulation ； Continuum model ； Discontinuum model ； RAMMS ； PFC3D ； Parameter calibration

參考文獻

1. Buhler, R., Argue, C., Jamieson, B. & Jones, A. “Sensitivity analysis of the RAMMS avalanche dynamics model in a Canadian transitional snow climate”. Proceedings of the International Snow Science Workshop, Innsbruck, Austria, pp. 767-771, October 2018.

2. Cannon, S. H. “Empirical model for the volume-change behavior of debris flows”, 1993.

3. Cesca, M., & d’Agostino, V. “Comparison between FLO-2D and RAMMS in debris-flow modelling: a case study in the Dolomites”, WIT Transactions on Engineering Sciences, 60, 197-206, 2008.

4. Christen, M., Kowalski, J. & Bartelt, P. “RAMMS: Numerical simulation of dense snow avalanches in three-dimensional terrain”, Cold Regions Science and Technology, 63, pp. 1-14, 2010.

5. Crominas, J. “The angle of reach as a mobility index for small and large landslides”, Canadian Geotechnical Journal, 33, pp. 260-271, May 1996.