Hierarchical growing cell structures: TreeGCS

We propose a hierarchical clustering algorithm (TreeGCS) based upon the Growing Cell Structure (GCS) neural network of B. Fritzke (1993). Our algorithm refines and builds upon the GCS base, overcoming an inconsistency in the original GCS algorithm, where the network topology is susceptible to the ordering of the input vectors. Our algorithm is unsupervised, flexible, and dynamic and we have imposed no additional parameters on the underlying GCS algorithm. Our ultimate aim is a hierarchical clustering neural network that is both consistent and stable and identifies the innate hierarchical structure present in vector-based data. We demonstrate improved stability of the GCS foundation and evaluate our algorithm against the hierarchy generated by an ascendant hierarchical clustering dendogram. Our approach emulates the hierarchical clustering of the dendogram. It demonstrates the importance of the parameter settings for GCS and how they affect the stability of the clustering     

Deep reinforcement learning for drone navigation using sensor data

Neural Computing and Applications - Mobile robots such as unmanned aerial vehicles (drones) can be used for surveillance, monitoring and data collection in buildings, infrastructure and...     

Biological effects of embedded depleted uranium (DU): summary of armed forces radiobiology research institute research.

The Persian Gulf War resulted in injuries of US Coalition personnel by fragments of depleted uranium (DU). Fragments not immediately threatening the health of the individuals were allowed to remain in place, based on long-standing treatment protocols designed for other kinds of metal shrapnel injuries. However, questions were soon raised as to whether this approach is appropriate for a metal with the unique radiological and toxicological properties of DU. The Armed Forces Radiobiology Research Institute (AFRRI) is investigating health effects of embedded fragments of DU to determine whether current surgical fragment removal policies remain appropriate for this metal. These studies employ rodents implanted with DU pellets as well as cultured human cells exposed to DU compounds. Results indicate uranium from implanted DU fragments distributed to tissues far-removed from implantation sites, including bone, kidney, muscle, and liver. Despite levels of uranium in the kidney that were nephrotoxic after acute exposure, no histological or functional kidney toxicity was observed. However, results suggest the need for further studies of long-term health impact, since DU was found to be mutagenic, and it transformed human osteoblast cells to a tumorigenic phenotype. It also altered neurophysiological parameters in rat hippocampus, crossed the placental barrier, and entered fetal tissue. This report summarizes AFRRI's depleted uranium research to date.     

Phenotypic and genotypic characterization of human and nonhuman Escherichia coli

Estuarine waters receive fecal pollution from a variety of sources, including humans and wildlife. Escherichia coli is one of several fecal coliform bacteria that inhabit the intestines of many warm-blooded animals that sometimes contaminate water. Its presence does not specifically implicate human fecal input, therefore it is necessary to differentiate contamination sources to accurately assess health risks. E. coli were isolated from human sources (HS) and nonhuman sources (NHS) in the Apalachicola National Estuarine Research Reserve and analyzed for fatty acid methyl ester (FAME), O-serogroup, and pulsed-field gel electrophoresis (PFGE) profiles. For FAME and PFGE analyses, there was no relationship between profile and isolate source. Human source PFGE profiles were less diverse than NHS isolates, and conversely for FAME. In contrast, O-serogrouping showed less diversity for HS vs. NHS isolates, and the predominant HS O-serogroups differed significantly (P < 0.01) from those of NHS isolates.     

Active mission success estimation through functional modeling

Through the application of statistical models, the active mission success estimation (AMSE) introduced in this paper can be performed during a rapidly developing unanticipated failure scenario to support decision making. AMSE allows for system operators to make informed management and control decisions by performing analyses on a nested system of functional models that requires low time and computational cost. Existing methods for analyses of mission success such as probabilistic risk assessment or worst case analysis have been applied in the analysis and planning of space missions since the mid-twentieth century. While these methods are effective in analyzing anticipated failure scenarios, they are built on computational models, logical structures, and statistical models that often are difficult and time-intensive to modify, and are computationally inefficient leading to very long calculation times and making their ability to respond to unanticipated or rapidly developing scenarios limited. To demonstrate AMSE, we present a case study of a generalized crewed Martian surface station mission. A crew of four astronauts must perform activities to achieve scientific objectives while surviving for 1070 Martian sols before returning to Earth. A second crew arrives at the same site to add to the settlement midway through the mission. AMSE uses functional models to represent all of the major environments, infrastructure, equipment, consumables, and critical systems of interest (astronauts in the case study presented) in a nested super system framework that is capable of providing rapidly reconfigurable and calculable analysis. This allows for AMSE to be used to make informed mission control decisions when facing rapidly developing or unanticipated scenarios. Additionally, AMSE provides a framework for the inclusion of humans into functional analysis through a systems approach. Application of AMSE is expected to produce informed decision making benefits in a variety of situations where humans and machines work together toward mission goals in uncertain and unpredictable conditions.     

Solid‐state structure of thermally crystallized syndiotactic polystyrene

The solid‐state structure of syndiotactic polystyrene (s‐PS) after crystallization from the melt and the glassy state was examined by differential scanning calorimetry (DSC), density, and X‐ray diffraction analysis. It was possible to prepare semicrystalline s‐PS containing either the pure α‐ or the pure β‐crystalline form by melt crystallizing s‐PS from 280 or 330°C. The measurements confirmed the low density of both crystalline forms, which in the case of α‐crystalline form was smaller and in the case of β‐crystalline form was only slightly larger than the density of the glassy amorphous s‐PS. An endeavor to introduce the crystalline phase in s‐PS through cold crystallization at constant temperature above the glass transition resulted in a complex ordered phase. This ordered phase, depending on the crystallization temperature, contained the planar chain mesomorphic phase and the α‐crystalline phase with a low degree of perfection (cold crystallization in the range 120–175°C) or a mixture of the α‐ and β‐crystalline forms with a high degree of perfection (cold crystallization in the range 210–260°C). The combination of DSC and X‐ray measurements enabled us to resolve the complex ordered structure in semicrystalline s‐PS after cold crystallization. © 2002 Wiley Periodicals, Inc. J Appl Polym Sci 83: 2705–2715, 2002     

A piecewise linear theory of plasticity for an initially isotropic material in plane stress

A 20-faced polyhedron is chosen as a reasonable piecewise linear approximation to either the Mises or Tresca yield criterion. Strain hardening and motion of the faces during hardening are assumed to be linear functions. Subject to the above assumptions, to initial isotropy and to some reasonable symmetry requirements it is shown that the most general possible theory contains 11 material constants. Some simple experiments with a thin-walled tube are suggested for determining these constants.