Development and application of a GPGPU-parallelized hybrid finite-discrete element method for modelling geo-structure collapse and resultant debris flow
FDEM is rarely implemented to model non-cohesive soils due to the computationally intensive costs required for contact detections and interactions of irregularly shaped non-cohesive soil particles. This study first reviews a series of authors' recent developments for speeding up the contact detections and interactions for FDEM including GPGPU-parallelization, efficient contact activation approach, mass scaling, hyperplane separation theorem, as well as the adaptive and semi-adaptive contact activation scheme. With their implementation, our GPGPU-parallelized HFDEM is about 8,000 to 61,000 times faster than sequential FDEM code, which paves the way for investigating the instability and collapse of geo-structures and resultant debris fragmentation and flow involving in a large numbers of irregular-shaped non-cohesive debris. The GPGPU-parallelized HFDEM is then implemented to investigate the collapse process of 3D irregular-shaped and non-cohesive soil heaps under gravity, and the excavation-induced slope instability as well as the resultant complex debris fragment action and flow process.