Engineering topics

Durian, Douglas J.

Publications and source records attributed to Durian, Douglas J..

7 records

DOE OSTI2024

Machine learning without a processor: Emergent learning in a nonlinear analog network

Standard deep learning algorithms require differentiating large nonlinear networks, a process that is slow and power-hungry. Electronic contrastive local learning networks (CLLNs) offer potentially fast, efficient, and fault-tolerant hardware for analog machine learning, but existing implementations are linear, severely limiting their capabilities. These systems differ significantly from artificial neural networks as well as the brain, so the feasibility and utility of incorporating nonlinear elements have not been explored. Here, we introduce a nonlinear CLLN—an analog electronic network made of self-adjusting nonlinear resistive elements based on transistors. We demonstrate that the system learns tasks unachievable in linear systems, including XOR (exclusive or) and nonlinear regression, without a computer. We find our decentralized system reduces modes of training error in order (mean, slope, curvature), similar to spectral bias in artificial neural networks. The circuitry is robust to damage, retrainable in seconds, and performs learned tasks in microseconds while dissipating only picojoules of energy across each transistor. This suggests enormous potential for fast, low-power computing in edge systems like sensors, robotic controllers, and medical devices, as well as manufacturability at scale for performing and studying emergent learning.

Science & Technology - Other Topics↗

NASA NTRS2002

Foam Optics and Mechanics

The Foam Optics and Mechanics (FOAM) project will exploit the microgravity environment to more accurately measure the rheological and optical characteristics of wet aqueous foams. Using both rheology and laser light scattering diagnostics, the goal is to quantify the unusual elastic character of foams in terms of their underlying microscopic structure and dynamics. Of particular interest is determining how the elastic character vanishes, i.e., how the foam 'melts' into a simple viscous liquid, as a function of both increasing liquid content and increasing shear strain rate. The unusual elastic character of foams will be quantified macroscopically by measurement of the shear stress as a function of shear strain rate and of time following a step strain. Such data will be analyzed in terms of a yield stress, shear moduli, and dynamical time scales. Microscopic information about bubble packing and rearrangement dynamics, from which the macroscopic non-Newtonian properties ultimately arise, will be obtained non-invasively by multiple-light scattering: diffuse transmission spectroscopy (DTS) and diffusing wave spectroscopy (DWS). Quantitative trends with materials parameters, most importantly average bubble size and liquid content, will be sought in order to elucidate the fundamental connection between the microscopic structure and dynamics and the macroscopic rheology.

Durian, Douglas J.↗

NASA NTRS2000

Diffusing Wave Spectroscopy Used to Study Foams

The white appearance of familiar objects such as clouds, snow, milk, or foam is due to the random scattering of light by the sample. As we all know, pure water is clear and easily passes a beam of light. However, tiny water droplets, such as those in a cloud, scatter light because the air and water droplet have different indexes of refraction. When many droplets, or scattering sites, are present, the incident light is scattered in random directions and the sample takes on a milky white appearance. In a glass of milk, the scattering is due to small colloidal particles. The white appearance of shaving cream, or foam, is due to the scattering of light at the water-bubble interface. Diffusing wave spectroscopy (DWS) is a laser light-scattering technique used to noninvasively probe the particle dynamics in systems that strongly scatter light. The technique takes advantage of the diffuse nature of light, which is reflected or transmitted from samples such as foams, dense colloidal suspensions (such as paint and milk), emulsions, liquid crystals, sandpiles, and even biological tissues.

Zimmerli, Gregory A.↗

NASA NTRS1997

Microgravity Foam Structure and Rheology

To exploit rheological and multiple-light scattering techniques, and ultimately microgravity conditions, in order to quantify and elucidate the unusual elastic character of foams in terms of their underlying microscopic structure and dynamics. Special interest is in determining how this elastic character vanishes, i.e. how the foam melts into a simple viscous liquid, as a function of both increasing liquid content and shear strain rate. The unusual elastic character of foams will be quantified macroscopically by measurement of the shear stress as a function of static shear strain, shear strain rate, and time following a step strain; such data will be analyzed in terms of a yield stress, a static shear modulus, and dynamical time scales. Microscopic information about bubble packing and rearrangement dynamics, from which these macroscopic non-Newtonian properties presumably arise, will be obtained non-invasively by novel multiple-light scattering diagnostics such as Diffusing-Wave Spectroscopy (DWS). Quantitative trends with materials parameters, such as average bubble size, and liquid content, will be sought in order to elucidate the fundamental connection between the microscopic structure and dynamics and the macroscopic rheology.

Durian, Douglas J.↗

NASA NTRS1996

The Melting of Aqueous Foams

Diffusing-wave spectroscopy measurements show that ordinarily solid aqueous foams flow by a series of stick-slip avalanche-like rearrangements of neighboring bubbles from one tight packing configuration to another. Contrary to a recent prediction, the distribution of avalanche sizes do not obey a power-law distribution characteristic of self-organized criticality. This can be understood from a simple model of foam mechanics based on bubble-bubble interactions.

Durian, Douglas J.↗

NASA NTRS1996

Microgravity Foam Structure and Rheology

The objective of this research was to exploit rheological and multiple-light scattering techniques, and ultimately microgravity conditions, in order to quantify and elucidate the unusual elastic character of foams in terms of their underlying microscopic structure and dynamics. Special interest was in determining how this elastic character vanishes, i.e. how the foam melts into a simple viscous liquid, as a function of both increasing liquid content and shear strain rate.

Durian, Douglas J.↗

NASA NTRS1994

Microgravity foam structure and rheology

Our long-range objective is to establish the fundamental interrelationship between the microscopic structure and dynamics of foams and their macroscopic stability and rheology. Foam structure and dynamics are to be measured directly and noninvasively through the use and development of novel multiple light scattering techniques such as diffusing-wave spectroscopy (DWS). Foam rheology is to be measured in a custom rheometer which allows simultaneous optical access for multiple light drainage of liquid from in between gas bubbles as the liquid:gas volume fraction in increased towards the rigidity-loss transition.

Durian, Douglas J.↗