Ductile Fracture Modeling And Parallel Framework Based On Material Point Method

Ductile fracture is a common phenomenon in our daily life,and it is of important practical significance to use computer to simulate it accurately and efficiently.The physics theory behind the ductile fracture is complex,usually including three basic processes,such as elasticity,plastic extension,and fracture.In physics simulation,a unified mechanics model for these processes is necessary,especially when calculating the elastoplastic behavior,yield extension,and fracture damage of solid materials.On the other hand,there are many repetitive reduction and summing behaviors in large-scene simulation systems,and it is often time-consuming when using CPU for simulation.Therefore,GPU-based parallel schemes are very important for physics simulation.In this paper,the plastic phasefield(PPF)is established for ductile fracture phenomena,and the GPU-based material point method(MPM)is improved,in which novel optimization schemes are proposed to fit the physical model to simulate ductile fracture phenomena and other phenomena with high efficiency and fidelity.In terms of physics model,this paper proposes PPF for ductile fracture simulation,which couples the phase-field and yield criterion.First of all,in the phase-field fracture method,the solid material carries a damage field,the phase-field is used to represent the damage state of the material,and the plastic strain of the material is introduced into the phase-field evolution,so that the damage change and plasticity are related.Secondly,in the plastic yielding of materials,this paper uses evolution of the phase-field to modify the yield surface accordingly,so that the basic yield surface follows the material damage.Through the above two improvements,the ductile fracture process can be expressed as:the solid material is stretched until it exceeds the elastic deformation threshold,the plastic strain continues to accumulate,and then the material damage is triggered.When the material is damaged to a certain extent,fracture occurs.This method models the entire process of ductile fracture in real life.In terms of parallel computing,this paper improves on the GPU-MPM algorithm,and proposes a new reduction algorithm and an efficient prefix sum algorithm.The reduction algorithm makes full use of the hierarchical structure of the GPGPU memory,which make threads and shared memory work well.Its computational efficiency is 2.77 to 15.2 times that of the traditional algorithm,and it can be applied to MPM and the phase-field subsystem.In addition,this article also proposes an efficient prefix summation algorithm,which changes the thread mapping of traditional algorithms,enables threads correspond to continuous computing tasks.This method can save GPU memory resources and improves computing parallelism.This paper also provides multiple sets of experiments to compare the above methods with classic algorithms,which proves the efficiency and universality of the our framework.Integrating the plastic phase-field method and GPU acceleration strategy,the PPFMPM framework can efficiently simulate general phenomena in life.Aiming at ductile fracture,including damage,extrusion and tearing,PPF can well show the characteristics of ductile fracture,including necking,elongation,damage and fracture.Moreover,the GPU-based framework allows PPF-MPM to achieve 56%acceleration as fast as the basic GPU-MPM algorithm.