Plenary lectures

PL01 Title: Trans-scale simulation of aeolian sand/dust transport and landform evolution process
Lecturer: Prof. Xiaojing Zheng, Xidian University, Xi’an, China
Aeolian disaster is a key environmental problem and great threat for human beings, and deep understanding of it can help for accurate prediction and efficient protection. Aeolian sand and dust transport are typical but also special particle-laden high-Reynolds number wall turbulent flows, which has well-known scientific significance. Coherent structures in the atmospheric surface layer (ASL) turbulence and aeolian landform evolution processes are both with inherent multi-scale nature, i.e. over ten orders of magnitude in length scale, posing great challenges for modeling and simulation. In this talk, we will introduce our trans-scale simulation efforts and new results on the problems of aeolian sand/dust transport in high-Reynolds number ASL flow, as well as aeolian landform evolution processes (sand ripples and dunes). Outlook for future studies in this field is also discussed.
PL02 Title: Development of Integrated Earthquake Simulation Enhanced with High Performance Computing
Lecturer: Prof. Muneo Hori, The University of Tokyo, Japan
Integrated Earthquake Simulation (IES) is aimed at combining earth science simulation, earthquake engineering simulation and social science simulation for more rational earthquake hazard and disaster assessment with higher resolution and reliability [1, 2]. For a given earthquake scenario, it computes a distribution of ground motions, structural damages, and social reaction against disasters in a target urban area. Assessment of earthquake hazard and disaster made by the sequential numerical analysis is more rational than conventional assessment that uses empirical equations.
Since required computation is large, it is essentially important to enhance IES with High Performance Computing (HPC), to increase the resolution and accuracy as well as the number of earthquake scenarios. Capability computing and capacity computing (namely, numerical analysis of large scale model and numerical analysis of numerous cases, respectively) are used for all or some components of the numerical analysis that are implemented in IES.
As for the earth science simulation, we are developing analysis models of the largest scale for the Japanese Islands. This model is of highest fidelity and highest spatial resolution; observed data of the crust structures are used to make a finite element model whose elements are of a few 100 meters. Earthquake cycle simulations and earthquake wave propagation simulation are being conducted in considering various uncertainties. These simulations are regarded as typical capacity computing. In particular, the earthquake wave propagation simulation can accurately evaluate topographical effects of the non-uniform crust structure upon the concentration of seismic waves.
The earthquake engineering simulation is divided into ground motion amplification simulation and seismic structural response simulation. As for the former, a massive model whose degree-of-freedom exceeds 1 trillion is used [3]. As for the latter, needed is the automated modeling for numerous structures, which include residential buildings, infrastructures, and road networks, since third party programs of seismic structural analysis are implemented in IES. We are developing a smart data conversion system which combines several data resources and generates suitable analysis models. The fidelity of the automated model depends on the quality of available data, and uncertaintiy of the models could be accounted by capacity computing in which a few thousand models are analyzed for one structure.
At this moment, the social science simulation consists of the following three: 1) mass evacuation simulation; 2) traffic demand and flow simulation; and 3) economic activity simulation. Multi Agent System (MAS) is developed for the first two simulations, which makes finest grain simulation in this field and takes advantage of capacity computing in which numerous cases of possible earthquake damages are studied.
IES is readily extended to other natural disasters such as tsunami and flooding. The basic procedures of the natural disaster are common, as they consist of automated modeling and automated execution of plugged-in numerical analysis methods. As for tsunami, IES can combine the earthquake engineering simulation and the mass evacuation simulation. The results of the combined simulations are visualized so that stakeholders can share a possible situation of a large earthquake and tsunami.
The quality of IES for the earthquake hazard and disaster assessment depends on data resources from which various urban are models are constructed. There are various digital data resources which are not open to the public, and the use of them surely increase the quality of IES. Non-digital data such as blue print drawings are transformed to digital data, and can contribute the increase of the quality of IES. Moreover, the use of measured or sensed data is desirable to accurately evaluate the actual state of the urban area which is computed by the earthquake engineering simulation. The data assimilation of human activity monitoring data is needed for the social science simulation.
PL03 Title: Recent Advances in Accelerated and Stabilized Meshfree Method for Modeling Man-made and Natural Disasters
Lecturer: Prof. J. S. Chen, University of California, San Diego, United States
Meshfree methods such as the Reproducing Kernel Particle Method (RKPM) are well suited for modeling materials and solids undergoing fracture and damage processes, and nodal integration is a natural choice for this class of problems. However, nodal integration suffers from spatial instability, and the excessive material deformation and damage process could also lead to kernel instability in RKPM. This presentation reviews the recent advances in nodal integration for meshfree methods that are stable, accurate, and with optimal convergence. A variationally consistent integration (VCI) is introduced to allow correction of many low order quadrature rules to achieve optimal convergence, and several stabilization techniques will be discussed. The application of the new RKPM formulation for fracture to damage multiscale mechanics and materials modeling, and their applications to the modeling of extreme events, will be demonstrated. These include the modeling of man-made disasters such as fragment-impact processes, penetration, shock and blast events, as well as natural disasters such as landslide will be presented to demonstrate the effectiveness of the new developments.
PL04 Title: Petascale Finite Element Simulation Based Investigation on Structural Integrity of Nuclear Power Plant Attacked by Strong Earthquake
Lecturer: Prof. Shinobu Yoshimura, The University of Tokyo, Japan
Leading supercomputers offer the computing power of petascale, and exascale systems are expected to appear in a few years. Supercomputers with more than tens of thousands of computing nodes, each of which has many cores, cause serious problems in developing practical finite element software. Not only the time consuming hot spots of the algorithms but also the entire algorithms must be well tuned and parallelized. We have been developing an open source parallel finite element software system known as ADVENTURE, which enables very precise analyses of practical structures and machines using over 100 million to billions DOFs mesh. The core parallel solution algorithm employed is the hierarchical domain decomposition method (HDDM) with balancing domain decomposition (BDD) as preconditioner. As one of practical target applications, we have been studying seismic response and structural integrity evaluation of existing nuclear power plants attacked by a strong earthquake exceeding design limit.
In this talk, after introducing some key technologies for practical petascale structural simulations, we describe finite element seismic response analyses of the high-fidelity full scale integrated model of Boiling Water Reactor and Reactor Building of Unit 1 at Fukushima Daiichi Nuclear Power Plants subjected to the Tohoku Off-Pacific Coast Earthquake of 9.0 Mw occurred on March 11th, 2011. Here, we precisely modeled pressure vessel, containment vessel, suppression chamber, vent pipes, a number of supports and reactor building with two hundred million DOFs finite element mesh. Its dynamic response during 65 seconds (from 80 sec to 145 sec) (Δt=0.01sec) was successfully solved using ADVENTURE_Solid Ver.2 on the K computer of 10 Petaflops peak performance. We discuss calculation performance, postprocess of Petabytes results, V&V and three dimensional behaviors of the plant. Then we conclude key roles of petascale finite element structural simulations for such practical and socially-important problems.
PL05 Title: Stability Analysis Based on Inequality Equations
Lecturer: Prof. Gen-hua Shi, University of Chinese Academy of Sciences, China
Stability analyses such as dam, slope or tunnel stability analysis play a critical role on the safety of related personals and engineering projects. Discontinuous computation is essential for stability analysis. The general equation for discontinuous computations is the second order programming which is minimization of total potential energy ∏ with the constrain inequality equations

The inequality equation system represents all discontinuous constrains: no penetrations, no tensions. Using open-close iteration, the inequality equation system can be solved. The following figures are the bolting computation of rock slope and the time-depending forces of blots.
To build the inequality equations is based contact theory. Contact theory is to ensure the block contacts and block movements are exactly same as real cases.
A new concept of an entrance block E(A, B) of two general A and B is introduced. Choosing a reference point a0 in block A, the contact between blocks A and B transferred to the contact between a point a0 and entrance block E(A, B).

Following figures are convex and complex entrance blocks E(A, B), where B is the solid block, A is the frame block, E(A, B) is the transparent block and a0 is the black dot which is the center of gravity of block A.
PL06 Title: Microstructural material database for self-consistent clustering analysis of strain softening materials
Lecturer: Prof. Wing Kam Liu, Northwestern University, United States
The advent of advanced processing and manufacturing techniques provides unparalleled freedom to design new material classes with complex microstructures across scales from nanometers to meters. As these new materials become more prevalent in design and manufacturing, so does the need to understand the failure limits of these materials. To design using these materials it is necessary to develop high fidelity multiscale material models that capture the bulk structural performance while accounting for the microscopic aspects, which are extremely important in advanced materials. In this lecture a new data-driven computational framework for the analysis and design of these complex material systems will be presented. A mechanistic concurrent multiscale method called self-consistent clustering analysis (SCA) is developed for general inelastic heterogeneous material systems. The efficiency of SCA is achieved via data compression algorithms which group local microstructures into clusters during the off-line training stage, thereby reducing required computational expense. Its accuracy is guaranteed by introducing a self-consistent method for solving the Lippmann-Schwinger integral equation in the on-line predicting stage. The integration of microstructure reconstruction and subsequent high-fidelity multiscale predictions of the materials behavior leads to the generation of vast amounts of reliable data. This structure-property feedback loop enables the design of new material systems with new capabilities. In mathematical physics, the “structure” and “property” can be interpreted as the nonlocal interaction of the microstructure clusters and the virtual work at the corresponding material point, respectively. Based on the computational design of experiments, data mining techniques offer the ability to discover the influence of the microstructure on the macroscopic materials behavior. The proposed framework will be illustrated for advanced composites and the integrated design of various advanced material systems.