Hazard assessment based on the combination of DAN3D and machine learning method for planning closed-type barriers against debris-flow

Abstract

If a slope located near a densely populated region is susceptible to debris-flow hazards, barriers are used as a mitigation method by placing them in flow channels; i.e., flowpaths. Selecting the location and the design of a barrier requires hazard assessment to determine the width, volume, and impact pressure of debris-flow at the moment of collision. DAN3D (Three-Dimensional Dynamic Analysis), a 3D numerical model for simulating debris-flow, has been widely used to perform hazard assessment; however, solely using DAN3D would be both insufficient and inefficient in finding the optimal barrier location. Therefore, the present study developed a framework that interprets the results from DAN3D simulation without considering any barriers. Then, the framework generates hazard assessment maps showing the impact parameters of debris-flow along the flowpath by various algorithms and machine learning methods, such as the k-means clustering algorithm, and also computes the width of the debris-flow, which is not explicitly calculated in DAN3D. A case study of the debris-flow at Umyeon mountain, Korea, in 2011, was used to generate hazard assessment maps. The maps were demonstrated to be a tool to quickly compute the impact parameters for conceptual barrier design with the aim of finding potential barrier locations.