Prediction of extreme geophysical, industrial and biophysical flows using particle methods

The flow of particulates and fluids is critical to many applications. Examples include human processes such as eating and locomotion, extreme geophysical flows such as tsunamis and dam-collapse inundation, biomedical applications such as transport of material in the gastro-intestinal tract and heart assist pumps, and industrial processes such as crushing and grinding (comminution). Meshless particle based methods such DEM (Discrete Element Method) and SPH (Smoothed Particle Hydrodynamics) have a range of strong advantages over traditional grid based continuum methods for these types of applications. The advantages include: • the ability to include particle level information in collision dominated particle flows • ability to resolve complex fluid free surfaces including splashing and fragmentation • ability to predict very large deformations, including fracture • interaction with moving and deforming solid bodies or objects • ability to track material history and use this in the flow modelling such as the rheology. These advantages are demonstrated in a series of computationally demanding examples and case studies including crushers and mills, biomedical pumps and transport of faecal matter in the bowel, geophysical extreme flow events (risk/disaster modelling), eating of food by humans and elite water based sports. In all cases, there are significant sources of variability that need to be taken into account in the modelling: • For human based models the major sources of variability are the nature of the motions of the human (both internal and external). Such variations can be short term (cycle variation between repetitions), medium term such as changes in response to fatigue or the relative timing and strengths of repetitions over time or longer term (such as in response to injury or deliberate planned changes). • For the geophysical/disaster modelling, variability is mainly reflected in the range and nature of the scenarios that need to be considered in building risk frameworks in which these models can predict specific consequences. For examples, the range of possible failure scenarios of a dam or the range of heights, speeds and water volumes for tsunamis generated by poorly understood initiation events. • Finally, in all such applications there is uncertainty about material properties and the details of the initial conditions whose effects need to be understood. For example, in comminution (either in an industrial crusher or mill or in the human mouth during chewing) the performance of the process is heavily influenced by the specific nature of the material being comminuted. This can vary strongly between materials and between different instances of what is nominally the same material. In the human eating case, this is further complicated by feedback from the material nature into the detailed nature of the chewing motion. Such variability influences both the inputs to specific instances of the deterministic models and the usage scenarios that need to be considered to understand the range of behaviours that exist for these processes. Optimisation and decision making based on such models cannot be reasonably performed without considering these issues. The ultimate usefulness of the model outcomes is heavily dependent on the quality of the management of these uncertainties.