Probabilistic quantification of tsunami current hazard using statistical emulation

In this paper, statistical emulation is shown to be an essential tool for the end-to-end physical and numerical modelling of local tsunami impact, i.e. from the earthquake source to tsunami velocities and heights. In order to surmount the prohibitive computational cost of running a large number of simulations, the emulator, constructed using 300 training simulations from a validated tsunami code, yields 1 million predictions. This constitutes a record for any realistic tsunami code to date, and is a leap in tsunami science since high risk but low probability hazard thresholds can be quantified. For illustrating the efficacy of emulation, we map probabilistic representations of maximum tsunami velocities and heights at around 200 locations about Karachi port. The 1 million predictions comprehensively sweep through a range of possible future tsunamis originating from the Makran Subduction Zone (MSZ). We rigorously model each step in the tsunami life cycle: first use of the three-dimensional subduction geometry Slab2 in MSZ, most refined fault segmentation in MSZ, first sediment enhancements of seabed deformation (up to 60% locally) and bespoke unstructured meshing algorithm. Owing to the synthesis of emulation and meticulous numerical modelling, we also discover substantial local variations of currents and heights.

Optimization-Based Unstructured Meshing Algorithms for Simulation of Hydraulically and Naturally Fractured Reservoirs With Variable Distribution of Fracture Aperture, Spacing, Length, and Strike

Summary Multistage hydraulically fractured wells are applied widely to produce unconventional resource plays. In naturally fractured reservoirs, hydraulic-fracture treatments may induce complex-fracture geometries that one cannot model accurately and efficiently with Cartesian and corner-point grid systems or standard dual-porosity approaches. The interaction of hydraulic and naturally occurring fractures almost certainly plays a role in ultimate well and reservoir performance. Current simulation models are unable to capture the complexity of this interaction. Generally speaking, our ability to detect and characterize fracture systems is far beyond our capability of modeling complex natural-fracture systems. To evaluate production performance in these complex settings with numerical simulation, fracture networks require advanced meshing and domain-discretization techniques. This paper investigates these issues by developing natural-fracture networks with fractal-based techniques. After a fracture network is developed, we demonstrate the feasibility of gridding complex natural-fracture behavior with optimization-based unstructured meshing algorithms. Then we can demonstrate that one can simulate natural-fracture complexities such as variable aperture, spacing, length, and strike. This new approach is a significant step beyond the current method of dual-porosity simulation that essentially negates the sophisticated level of fracture characterization pursued by many operators. We use currently established code for fractal discrete-fracture-network (FDFN) models to build realizations of naturally fractured reservoirs in terms of stochastic fracture networks. From outcrop, image-log, and core analysis, it is possible to extract fracture fractal parameters pertaining to aperture, spacing, and length distribution, including center distribution as well as a fracture strike. Then these parameters are used as input variables for the FDFN code to generate multiple realizations of fracture networks mimicking fracture clustering and randomly distributed natural fractures. After incorporating hydraulic fractures, complex-fracture networks are obtained for further reservoir-domain discretization. To discretize the complex-fracture networks, a new mesh-generation approach is developed to conform to nonorthogonal and low-angle intersections of extensively clustered discrete-fracture networks with nonuniform aperture distribution. Optimization algorithms are adopted to reduce highly skewed cells, and to ensure good mesh quality around fracture tips, intersections, and regions of extensive fracture clustering. Moreover, local grid refinement is implemented with a predefined distance function to control cell sizes and shapes around and far away from fractures. Natural-fracture spacing, length, strike, and aperture distribution are explicitly gridded, thus introducing a new simulation approach that is far superior to dual-porosity simulation. Finally, initial sensitivity studies are performed to demonstrate both the capability of the optimization-based unstructured meshing algorithms, and the effect of aforementioned natural-fracture parameters on well performance. This study demonstrates how to incorporate a fractal-based characterization approach into the current work flow for simulating unconventional reservoirs, and most importantly solves several issues such as nonorthogonal intersections, extensive clustering, and nonuniform aperture distribution associated with domain discretization with unstructured grids for complex-fracture networks. The proposed meshing techniques for complex fracture networks can be easily implemented in existing preprocessing, unstructured mesh generators. The sensitivity study and the simulation runs demonstrate the importance of fracture characterization as well as uncertainties associated with naturally fractured reservoirs on well-production performance.

Computational Study of a Transonic Turbine Cascade: Validation and Comparison of Griding Approaches and Challenges

This paper presents a computational study, with some experimental validation, of a low-turning transonic turbine cascade. A comparison is presented between the time-consuming and difficult to generate hexa-structured meshing approach, and the mostly automated tetra-unstructured meshing approach. The paper compares the predicted flow physics and losses, with discussion of the challenges in griding and convergence between both approaches. Computations were carried out using a commercial RANS solver (ANSYS CFX 12) using the Shear Stress Transport turbulence model, and the Gamma-Theta transition model. The computational domain encompassed a half blade span, and one blade pitch with periodic boundary conditions; griding for both approaches was done using ANSYS ICEM CFD. Computational results from both griding approaches were compared to corresponding experimental data. The outlet Mach number was 0.90. The experiment was carried out using a linear cascade in a blow-down type wind tunnel. Downstream seven-hole pressure probe measurements at 1.8 axial chord lengths from the leading edge provided loss, streamwise vorticity, and secondary kinetic energy distributions and integrated coefficient values. It was found that both griding approaches predicted similar downstream endwall flow structures to those observed in the experiment. The tetra-unstructured mesh solution predicted higher losses, but both predicted lower losses than the experiment. Overall results suggest that for capturing of the basic flow physics, both approaches suffice, with the tetra-unstructured being the much easier approach, but with limitations on the level of grid refinement. For more accurate capturing of the flow physics, the time-consuming and difficult to generate hexa-structured meshing approach can be justified.