- ホーム
- 機能創成セミナー
機能創成セミナー Seminar on Mechanical Science and Bioengineering
第190回 | 2023年12月11日(月) 13:00-14:00 [Dec 11th 2023 (Mon) 13:00-14:00] 基礎工学研究科 C419(C棟共用セミナー室) |
|---|---|
| Predicting the phase and strength of multi-principal element alloys: a mechanistic data-driven and machine learning approach | |
| Refractory multi-principal element alloys (RMPEAs) comprise five or more constituent elements and potentially offer superior mechanical performance. A key challenge lies in effectively navigating vast compositional design spaces to identify optimal microstructures and properties. Although RMPEAs have shown promise, the path towards holistic property optimization remains elusive. We have developed a probabilistic machine learning approach integrating experiments and simulations to elucidate connections between processing, complex microstructures, and resulting mechanical response in RMPEAs. The methodology first constructs a dataset sampling millions of microstructures with corresponding mechanical properties. Mixture density networks combined with neural networks subsequently learn from this data to predict property distributions and rank microstructural feature importance. We apply this technique to modeling yield strength in pure Ni, Al, and Cu metals, as well as NiCoCr, NiCoV and CrMnFeCoNi RPMHEAs, and we show excellent agreement with experiments. The model quantifies the statistical influence of morphological features like grain size in line with established theories while revealing additional impactful microstructural contributors. This work overcomes hindrances imposed by experimental data constraints through integrated data-driven and mechanistic modeling within a probabilistic learning paradigm. The technique represents a versatile materials-by-design tool for optimizing microstructures in RPMHEAs and beyond while providing fundamental insights into property-structure interconnections. | |
| Jaafar A. El-Awady (Department of Mechanical Engineering, Johns Hopkins University) | |
| 世話人：尾方成信 | |
第189回 | 2023年9月29日(金) 10:30-11:30 [Sep 29th 2023 (Fri) 10:30-11:30] 基礎工学研究科 D404(D棟共用セミナー室) |
| Structured Illumination for Resolving Microchannel Flows | |
| Measurements in microchannel flows using optical diagnostic techniques are challenging in many cases due to spatial resolution limits and a lack of optical access. This talk will discuss using nonuniform, or structured, illumination for optical sectioning, i.e., to resolve a thin slice of a flow containing a fluorescent tracer and illuminated over its diameter, by reconstructing a single image of the flow from multiple (typically two) images. Results for particle tracking velocimetry (PTV) in Poiseuille flow through 100μm deep channels demonstrate that structured (sinusoidally modulated) illumination (SI) halves the slice thickness compared with uniform illumination. Optical diagnostics in liquid-vapor flows is even more challenging because of the scattering and distortion of light due to the refractive-index difference between the phases. Nonuniform laser light sheets, or structured laser illumination planar imaging (SLIPI), have been used to visualize dense sprays of fluorescently dyed liquids. Here, SI is used instead to visualize liquid-vapor flows where the vapor is the discrete phase. Images of vapor bubbles in flow boiling of the fluorinated dielectric fluid HFE-7200 (containing a fluorinated rhodamine) in a 1mm deep channel obtained with SI have much greater contrast, quantified by their signal-to-noise ratio, than those obtained with uniform illumination. | |
| Mimami Yoda (George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology) | |
| 世話人：河原源太 | |
第188回 | 2023年9月14日(木) 13:30-14:45 [Sep 14th 2023 (Thu) 13:30-14:45] 基礎工学研究科 B303講義室 |
| Modeling hydrodynamically mediated collective responses of active matter – active turbulence and cilia synchronisation | |
| I shall present our recent theoretical works on how hydrodynamics can induce collective dynamics of active matter systems. (1) Both active nematics (microtubule-motor system) and bacteria suspensions can exhibit turbulent behaviours, but with different scaling in the energy spectra. We construct an analytical model to understand such scaling behaviours in the two systems, where we reach good agreements between analytic calculations and experimental measurements. (2) Cilia can coordinate with each other by hydrodynamic interactions, beating in the form of a metachronal wave. With the dynamic model, we deal with the coordinated dynamics of a cilia array which contains an infinite number of model cilia. With the theory, we can provide the dispersion relation and predict the stable wave patterns in the model cilia arrays, which may guide future fabrications of model cilia for industrial applications. | |
| Fanlong Meng (Institute of Theoretical Physics, Chinese Academy of Sciences・中国科学院) | |
| 世話人：松永大樹 | |
第187回 | 2023年8月29日(火) 11:00-12:00 [Aug 29th 2023 (Tue) 11:00-12:00] 基礎工学研究科 J棟3階生物工学コースセミナー室 |
| Nanoscale molecular architecture determines the biophysics of calcium regulation and ATP dynamics in dendritic spines | |
| Post-synaptic transmission of neuronal synapses can occur on a dendritic spine – a microdomain with a bulbous head connected to the dendrite through a thin neck. Calcium in spines is known to regulate synaptic plasticity, a process that underlies learning and memory in the hippocampus. Upon strong calcium influxes similar to synaptic inputs, we observed very fast signal transmission occurring from the spine's head to its base in a timescale of a few milliseconds. We developed a framework of stochastic modeling and simulations that explained these fast responses based on the extreme statistics of the times taken by diffusing calcium ions to arrive at a distant, autocatalytic reaction site. Thereby we predicted the nanoscale geometric organization of calcium regulators that are closely associated with the internal calcium stores in spines: ORAI channels are located towards the spine head and are colocalized with SERCA pumps, while Ryanodine receptors are located at the base of the spine neck. These predictions were verified microscopically using in-vivo slices of mouse hippocampal neurons. Moreover, we extended our model and combined with ATP imaging data to show that the ATP production sites of dendritic mitochondria have to be located strictly below the spine neck to guarantee an efficient ATP delivery into the spine head. The timescales of ATP delivery into the spine head (few tens of milliseconds) shown by the simulations also matched the experimental measurements. These insights will be useful also to understand the consequences of signaling and metabolic disruptions in the brain during detrimental conditions such as neurodegenerative diseases. | |
| Kanishka Basnayake (University of California, San Francisco) | |
| 世話人：野村泰伸 | |
第186回 | 2023年8月24日(木) 13:30-14:30 [Aug 24th 2023 (Thu) 13:30-14:30] 基礎工学研究科 D404(D棟共用セミナー室) |
| Interactive active matter | |
| Active matter encompasses a wide range of systems that consist of individuals, which extract energy from their surroundings and convert it into mechanical work. Such individuals, including bacteria and eukaryotic cells, exist in extremely diverse environments, which, in turn, lead to a complex network of inter- and intra-population interactions. Mechanical forces, exerted and experienced by cells, can act as messengers, regulating individual behaviour, however, how they lead to emergent collective behaviours and the emergence of liquid crystalline features in such active systems have not been elucidated yet. Here, by utilising a suite of theoretical models at different scales, such as continuum liquid crystal theory and cell-based phase-field formulation, we investigate how mechanical interactions with surroundings and other cells can lead to collective cell migration. | |
| Aleksandra Ardaseva (Niels-Bohr Institute; Denmark) | |
| 世話人：松永大樹 | |
第185回 | 2023年8月18日(金) 16:00-17:00 [Aug 18th 2023 (Fri) 16:00-17:00] 基礎工学研究科 D404(D棟共用セミナー室) |
| Particles in the ABC flow, and morphology of droplets in turbulent flows | |
| I will present the seminar in two parts; firstly, I will investigate the effect of spherical particles on the turbulent Arnold-Beltrami-Childress (ABC) flow. Using direct numerical simulations, we see that dispersions of heavy particles significantly modulate the largest scales of the flow towards an anisotropic, more energetic state. Secondly, I will look at the morphology of surfactant-laden droplets in homogeneous isotropic turbulence. Particularly in relation to the Hinze-scale; the length-scale at which turbulent fluctuations balance surface tension forces. We find that droplets smaller than the Hinze scale have ellipsoid-like shapes, while larger droplets are long and filamentous. To extend our analysis of the filamentous droplets, we measure the Euler characteristic of their surfaces. | |
| Ianto Cannon (Okinawa Institute of Science and Technology) | |
| 世話人：本告遊太郎 | |
第184回 | 2023年8月18日(金) 14:00-15:30 [Aug 18th 2023 (Fri) 14:00-15:30] 基礎工学研究科 D404(D棟共用セミナー室) |
| Geostrophic turbulence and the formation of large scale structure | |
| Rotating convection is studied using an asymptotically reduced system of equations valid in the limit of strong rotation. The equations describe four regimes as the Rayleigh number Ra increases: a disordered cellular regime near threshold, a regime of weakly interacting convective Taylor columns at larger Ra, followed for yet larger Ra by a breakdown of the convective Taylor columns into a disordered plume regime characterized by reduced heat transport efficiency, and finally by a new type of turbulence called geostrophic turbulence. Properties of this state will be described and illustrated using direct numerical simulations of the reduced equations. These simulations reveal that geostrophic turbulence is unstable to the formation of large scale barotropic vortices or jets, via a process known as spectral condensation. The details of this process will be quantified and its implications explored. The results are corroborated via direct numerical simulations of the Navier-Stokes equations; in the presence of boundaries, robust boundary zonal flows resembling topologically protected edge states in chiral systems are present. A simple procedure for eliminating these zonal flows from experimental studies of geostrophic turbulence is suggested. | |
| Edgar Knobloch (Department of Physics, University of California, Berkeley) | |
| 世話人：河原源太 | |
第183回 | 2023年6月23日(金) 12:10-13:10 [Jun 23rd 2023 (Fri) 12:10-13:10] 基礎工学研究科 G215セミナー室 |
| Local ordering in multi-principal element alloys and its impact on diffusion and dislocation patterning | |
| Multi-principal element alloys (MPEAs), commonly termed medium- or high-entropy alloys containing three or more components in high concentrations, are presumed to be random solid solutions corresponding to maximum configurational entropy. The ideal random solid solutions in MPEAs, however, may only be possible at temperatures close to the melting point. As the temperature decreases in material processing or during service, solute-solute interaction and mixing enthalpy (enthalpic contribution) predominate the Gibbs free energy and induce local ordering of the chemistry. When this chemical short-range order (SRO) appears in the materials, fundamental questions arise as to (i) how does the SRO influence diffusion, and (ii) what are the deformation mechanisms and dislocation patterning under impact loading? In this presentation, I will first review the background of MPEAs and chemical SRO. In the second part, I will discuss the specific role of SRO on diffusion and deformation mechanism. Through computational modeling and theoretical analysis, we found that SRO considerately reduces and localizes diffusion. Regarding fundamental deformation physics, leveraging on large-scale deformation simulations of a model fcc NiCoCr alloy at the atomistic level, we uncover strikingly different characteristics of dislocation patterning and deformation microstructure evolution under a high loading rate. At the end of the talk, I will present our deep learning framework using convolutional neural networks (CNNs) that can promisingly predict diffusion kinetics in vast compositional spaces of MPEAs and the associated timescale of chemical SRO formation. | |
| Penghui Cao (The University of California, Irvine) | |
| 世話人：尾方成信 | |
第182回 | 2023年6月22日(木) 15:30-16:30 [Jun 22nd 2023 (Thu) 15:30-16:30] 基礎工学研究科 国際棟セミナー室 |
| Resolvent Analysis and Control of Turbulent Flows | |
| Resolvent analysis reveals the harmonic relations between the forcing inputs and response outputs with respect to a base flow. By combining our understanding of the baseline flow physics and the resolvent mode characteristics, effective and efficient flow control strategies can be developed to alter the time-averaged flows. This approach provides us with a physics-driven flow control strategy, without having to resort to expensive parametric computational and experimental studies. We also discuss our recent efforts in extending the traditional resolvent analysis to (1) incorporate randomized numerical linear algebra to enable examining higher-Reynolds number turbulent flows; (2) establish a sparsity-promoting formulation to determine the optimal actuator location and the forcing variable; and (3) integrate network science into the resolvent formulation to study time-varying base flows. In this talk, we demonstrate the use of these resolvent analysis techniques for examples of actively controlling turbulent separated flow over a NACA 0012 airfoil, high-speed open cavity flow, and isotropic turbulence. | |
| 平邦彦 (University of California, Los Angeles) | |
| 世話人：後藤晋 | |
第181回 | 2023年6月5日(月) 15:10-17:10 [Jun 5th 2023 (Mon) 15:10-17:10] 基礎工学研究科 国際棟 セミナー室 |
| Exploring Spinal Locomotor Circuitry through Computational Modeling | |
| To effectively navigate complex and changing environments, animals must control locomotor speed and gait, while precisely coordinating and adapting limb movements to the terrain. The underlying neuronal control is facilitated by circuits in the spinal cord that integrate supraspinal commands and afferent feedback signals to produce coordinated rhythmic muscle activations necessary for stable locomotion. I will present a series of computational models investigating dynamics of central neuronal interactions as well as a neuromechanical model that integrates neuronal circuits with a model of the musculoskeletal system. These models closely reproduce speed-dependent gait expression and experimentally observed changes following manipulation of multiple classes of genetically- or anatomically-identified neuronal populations. I will discuss the utility of these models in providing testable predictions for future studies. | |
| Simon M. Danner (Drexel University College of Medicine) | |
| Real-to-Sim: Methods, tools, and experiences from computational modelling of animal locomotion | |
| In the field of robotics, computer-aided simulations have become a crucial tool for rapid prototyping of robot designs, testing, and training of controllers. However, the critical challenge is effectively transferring the simulations to the physical hardware. This Real-to-Sim transfer is the opposite process of transferring experimental observations into neuromechanical simulation models for animal locomotion studies. This presentation will focus on Real-to-Sim and its role in developing neuromechanical models of animals to study animal movement, particularly locomotion. The creation of neuromechanical models involves complex processes such as modelling rigid-body dynamics, muscle-models, and neural networks. The talk will highlight the development of neuromechanical models of Mouse, Drosophila Melanogaster, and Rhesus Macaque. These case studies will demonstrate how neuromechanical simulations are used to investigate various aspects of animal motor control and complement animal experiments. Finally, the presentation will cover the use of open-source tools and frameworks, along with personal experiences, in developing neuromechanical simulations. Attendees will gain insights into the techniques and applications of Real-to-Sim transfer for neuromechanical simulations. | |
| Shravan Tata Ramalingasetty (Drexel University College of Medicine) | |
| Using robotics to investigate neuronal control of turning. | |
| Locomotion is a critical behavior that allows animals to move in the external world. The underlying neuronal circuitry combines information from the spinal cord with afferent feedback to execute descending commands. Most of what we know about the neuronal control of locomotion pertains to straightforward locomotion; how turning or the change in direction is controlled, remains poorly understood. I will present a quadrupedal robot controlled by a model of spinal locomotor circuits to study various mechanisms of turning. Using a robot allows us to study the impact of these turning strategies on stability and balance. The quadrupedal robot has 13 degrees of freedom: three joints per limb and one joint within the torso that allows for lateral bending of the body. The neuronal model controlling the robot includes four coupled rhythm generators, each controlling one limb. These rhythm generators receive sensory feedback that characterizes loading and extension of the corresponding limb. The robot model was able to exhibit speed-dependent changes of stance and swing phase durations. We studied the effectiveness of left-right asymmetric changes to induce turning at different speeds and gaits. Our model suggests that a combination of strategies is needed to effectively turn while maintaining stability, and that the optimal strategy depends on the locomotor gait and speed. Thus, control of turning likely involves task- and speed-dependent modulation of the spinal neuronal circuits at multiple levels. | |
| Andrew B. Lockhart (Drexel University College of Medicine) | |
| 世話人：青井伸也 |
