Center for Mathematical Science and Advanced Technology (MAT)

Seminar Schedule

[MAT Seminar]

Date:

2023/12/19(Tuesday) 11:00 - 12:00

Language:

English

Speaker:

Ettore Barbieri

Title:

Algorithms for unsupervised digraph learning from centralities - applications to food web reconstruction

[MAT Seminar]

Date:

2023/11/07(Tuesday) 14:00 - 15:00

Language:

English

Speaker:

Pinaki Chakraborty (Okinawa Institute of Science and Technology)

Title:

Surprising thermodynamics of landfalling typhoons

[Dr. Daniel Weihs invited to give a special MAT seminar]

Date:

2023/09/21(Thursday) 11:00 - 12:00

Language:

English

Location:

Miyoshi Memorial Auditorium, Yokohama Research Institute

ZoomURL:

https://us02web.zoom.us/j/84410582844?pwd=YjhMYlJyRzdiaXcxUUtPWEE1THJRdz09

Speaker:

Daniel Weihs (PhD 1971)

Title:

Hydrodynamic Interactions in Animal Groups: A 50-Year Research Review

Abstract:

Research spanning 50 years describing and analyzing fluid mechanical interactions between members of animal groupings is presented. These range from fish schools and invertebrate groupings in water, through bird flocks in air.
The different modes of hydrodynamical interactions will be discussed, starting from large fish schools, where wake vortex effects through lateral interactions for neutrally and negatively buoyant animals are calculated. The past controversy on the importance of hydrodynamics in schooling patterns will be resolved next. Large schools, as well as smaller groups, such as the tuna soldier formation and mother-calf assistance in cetaceans will be discussed.
The dolphin mother calf interaction study was a result of the problem of dolphin bycatch by tuna fishermen, helping to understand and reduce the problem. Because of the difficulty in measuring the forces on the cetacean pairs, wind tunnel models were built and tested, showing the significant savings in energy for the calf or weaker individual.
The hydrodynamic aspects of Salp chains are shown next, solving an apparent paradox of how individual periodic pulsatile motion can be reconciled with group motion, especially as these chains are physically attached to each other.
Some developments in understanding the V- formation in migrating birds are shown to be closely related to the previous work, pointing out some of the control problems arising from the fluid dynamics of adjacent bodies.

Speaker Biographies:

Daniel Weihs (PhD 1971) is Distinguished Professor of Aerospace Engineering Emeritus. Holder of the Richmond Chair of Life Sciences. At present he heads the Technion Autonomous Systems Program.
He has been Dean of Aerospace Engineering , the Technion Graduate School, and Provost at Technion , Head of the Neaman institute, Chief Scientist of the Ministry of Science, Chair of the Interuniversity Marine Institute IUI in Eilat, and Chair of the Israel National Committee for Space Research.
He is a member of the Israel Academy of Sciences, the US National Academy of Engineering and holds honorary doctorates from Ben Gurion University and the Hong Kong Univ. of Science and Technology . He is Fellow of the American Physical Society, the International Society for Bionic Engineering , and the Marine Biological Society. He visited Japan as Fellow of JSPS in 2012.
He has been on the Board of several companies and institutes, including the Israel Oceanographic and Limnology Research Corp.
He has published over 180 archival papers including several groundbreaking studies on fish locomotion.

[MAT Seminar]

Date:

2023/09/13(Wednesday) 13:00 - 14:00

Language:

English

Speaker:

Jian Chen

Title:

JKR theory-based force-displacement relation for DEM simulations of adhesive particles

Abstract:

There is a lot of scientific and commercial interest in adhesive particles. Discrete Element Method (DEM) simulations can be used to study their collective behavior but require a well-grounded force model, namely a force-displacement relation, in order to evaluate the contact force from the relative particle displacement. The Johnson-Kenneth-Robberts (JKR) theory reliably predicts the contact radius under loading for adhesive contact,but deriving a force-displacement relation from it has remained challenging. We derived a force-displacement relation from the JKR theory as a composite function of displacement via an intermediate function which is the effective adhesive contact radius. In addition, we analyzed contact geometry errors in the Hertz and JKR models and we determined the exact contact centroid for calculating contact torque and relative tangential velocity. Our analysis shows that the exact contact centroid depends on the stiffness quotient. Neglecting this leads to incorrect torque and velocity. Contact geometry errors could be significant for some nano-sized particles due to their large relative contact size. We show that using a coefficient of restitution derived from the Hertz model results in higher damping and a lower actual coefficient of restitution in the JKR models. Our study provides insight into contact mechanics-based models to simulate adhesive particles.

[MAT Seminar]

Date:

2023/08/09(Wednesday) 13:00 - 15:00

Language:

English

Speaker:

Takahiro Arai

Title:

Application of Data-driven Approach Based on the Phase Reduction Theory: Example of Analysis of Human Gait Motion and Further Application to Earth Science

Abstract:

Synchronization phenomena are widely observed. They often play important roles in many fields, such as biological and engineering systems, circadian rhythms, and the neuronal synchronization of heartbeat. Of course, synchronization also appears in earth sciences.
Recently, some synchronization phenomena connecting the opposite sides of the globe have been reported: Arctic Oscillation and Antarctic oscillation, Kuroshio Current and Gulf Stream. The synchronization appears in a simple phenomenon where the rhythms adjust with each other, although the background of the system is too complex to elucidate the details of the causality between two or more elements. The modeling of causality is difficult for such a complex system. However, the application of the phase reduction theory to a data-driven approach may give a decisive solution to such a problem. Without the details of the governing equation underlying the complex system, the method can elucidate the intrinsic causality of synchronization and rhythmic relationship. This approach is now widely used, mainly in neuroscience.

In this seminar, I introduce the study of the application of the data-driven approach based on the phase reduction theory to the analysis of human gait motion. The human gait motion is mainly controlled by the cerebellum and spinal cord, but its mechanism is still uncertain. We use the method mentioned above for the analysis of the control of interleg coordination. As a result, we found that interleg coordination is not strictly controlled until the deviation from the antiphase condition between two legs exceeds a certain threshold.

At the end of the presentation, I briefly explain my research plan in JAMSTEC as a postdoctoral researcher. My research aims to extend the scope of the application of the data-driven approach based on the phase reduction theory to a variety of data in the field of earth science.

[MAT Seminar]

Date:

2023/05/31(Wednesday) 13:00 - 15:00

Language:

English

Speaker:

Vedasri Godavarthi (University of California, Los Angeles)

Title:

Optimal waveform for fast entrainment of airfoil wakes

Abstract:

Unsteady fluid flows over bluff bodies and airfoils form a large class of periodic flows and their flow control has several engineering applications in the context of fluid-structure interactions and enhanced aerodynamic performance. The phase synchronization properties of such periodic flows encode insights needed to modify the flow behavior. Recently, the characterization of the asymptotic synchronization (entrainment) properties of periodic wake flows in the presence of external forcing has been done using phase reduction analysis. However, for flow control applications, the time it takes to synchronize to external input is important to quickly modify the flow behavior. In this work, we obtain the actuation waveform that maximizes the entrainment speed of periodic airfoil wakes analytically. We investigate the influence of the angle of attack on the synchronizability and we observe that phase synchronization becomes harder to achieve for high angle of attack and the corresponding optimal actuation becomes increasingly non-sinusoidal. We numerically validate the performance of fast-entrainment analysis by comparing it to sinusoidal forcing. We obtain entrainment in as low as two shedding cycles using optimal waveform in comparison to the O(10) cycles using sinusoidal waveform. The optimal actuation waveform identifies the "when" and "how" to efficiently modify the vortex formation process of airfoil wakes. Since, the vortex formation process over airfoils is directly related to the lift production, we can achieve a transient increase in lift for these flows. This has wide applicability for unsteady active flow control to promote aerodynamic performance of wake flows.

[MAT Seminar]

Date:

2023/05/10(Wednesday) 13:00 - 15:00

Language:

English

Speaker:

Satoshi Noguchi

Title:

Extraction of process-structure-property relationship in material design by deep learning methodology

Abstract:

I would like to begin this presentation with a brief self-introduction, followed by an overview of my previous research activities. My research can be broadly categorized into two parts: (i) the geometric formulation of discrete systems of physical phenomena, and (ii) the development of a general deep learning methodology for material design. In this seminar, I will primarily focus on the latter, which was carried out during my doctoral program.

To start, I will introduce the fundamental concept of material design and then explain the deep learning framework, which consists of a vector quantized variational autoencoder (VQVAE) and a pixel convolutional neural network (PixelCNN). Subsequently, I will discuss the framework's application through several examples in the context of material design. The seminar will cover the following three topics: (i) extracting the process-structure relationship of structural materials using our deep learning framework, (ii) identifying critical microstructures influencing target properties without relying on background physics, and (iii) optimizing molecular structures using PixelCNN.

Lastly, I will conclude the presentation with a brief comment on my research plan.

[MAT Seminar]

Date:

2023/04/26 (Wednesday) 13:00 - 15:00

Language:

English

Speaker:

Furuichi Mikito

Title:

Virtual earthquakes in the numerical granular rock box experiment

Abstract:

A new numerical simulation of virtual earthquakes reproduced by the 3D DEM (Discrete Element Method) is presented. The numerical granular rock box experiment was developed based on the shortening sandbox simulation which is known as an analog model of the accretionary prism formation. An adhesive contact force is applied to simulate the rock failure envelope. Calibration of the adhesive contact model to the failure envelope of the laboratory triaxial test requires a high young modulus, such as 2.3 GPa as a specimen, which has not been used in the previous granular models at geologically relevant scales. Thus, the scope of the new rock box model can be not only the quasistatic deformation as in previous studies but also the seismic events. The simulation of the shortening granular layer shows the sequential formation of the reverse and normal thrusts with building tectonic structures. At the same time, the element-wise frictional contact force induces the stick-slip event as a virtual seismic event. Especially at slow convergence speed, the clear seismic cycle, various quasistatic and rapid shear deformations with rebound and rotation processes are confirmed. The slip displacements at each stage of the seismic event, such as frictional slip failure and shear fracture, are analyzed to understand the granular process behind them. The observable precursors of the seismic event from such as borehole monitoring are also discussed. The rock box simulation reproduces from the large-scale tectonic deformation such as mountain building to the brittle elastic seismic event in a unified manner. Although the required computational cost is high, it is achievable with our parallel DEM simulation code DEPTH and ES4.

©JAMSTEC, www.jamstec.go.jp