Parallel, self-organizing, hierarchical neural networks. II

For pt.I see IEEE Trans. Neural Networks, vol.1, p.167-78 (1990). Parallel, self-organizing, hierarchical neural networks (PSHNNs) involve a number of stages with error detection at the end of each stage, i.e., rejection of error-causing vectors, which are then fed into the next stage after a nonlinear transformation. The stages operate in parallel during testing. Statistical properties and the mechanisms of vector rejection of the PSHNN are discussed in comparison to the maximum likelihood method and the backpropagation network. The PSHNN is highly fault tolerant and robust against errors in the weight values due to the adjustment of the error detection bounds to compensate errors in the weight values. These properties are exploited to develop architectures for programmable implementations in which the programmable parts are reduced to on-off or bipolar switching operations for bulk computations and attenuators for pointwise operations     

A time-domain approach to transmission network equivalents viaProny analysis for electromagnetic transients analysis

This paper presents a method of obtaining transmission network equivalents from the network's response to a pulse excitation signal. The proposed method is based on modal decomposition representation for the large-scale interconnected system. In this framework we use Prony analysis to identify the network function of the system and to decompose the large system into a parallel combination of simple first-order systems. As a result the network function of the transmission network can be identified easily, and a Thevenin-type of discrete-time filter model can be generated. It can reproduce the driving-point impedance characteristic of the network. Furthermore, the proposed model can be implemented into the EMTP in a direct manner. The simulation results with the full system representation and the developed equivalent system showed a good agreement     

Strain analysis of moving objects using a Fourier transform grid method

There is no good method to measure the shape and the strain distributions of a structure changing with time. We have previously proposed the Fourier transform grid method (FTGM) to measure the three-dimensional shape and surface strain distributions of stationary objects by analysing the two-dimensional grating images recorded with two cameras. In the stereoscopic method, it is very important to determine the accurate geometric parameters of the camera system. In this paper, the positions and the directions of cameras are accurately determined using the FTGM applied to images of a reference object on which a two-dimensional grating is drawn. Applications for analysing shape and strain distributions of vibrating rubber plates and a moving human skin are shown.     

Electrical properties and positron annihilation study of (Ba<Subscript>1−<Emphasis Type="Italic">x</Emphasis></Subscript>Ho<Subscript><Emphasis Type="Italic">x</Emphasis></Subscript>)TiO<Subscript>3</Subscript> ceramics

DC resistivity, dielectric constant, dielectric loss and positron annihilation spectra of (Ba_1−x Ho _x )TiO₃ ceramics have been measured as a function of holmium concentration x. It has been found that the DC resistivity of (Ba_1−x Ho _x )TiO₃ is strongly dependent on the Ho content: it decreases three orders of magnitude and reaches a minimum at x = 0.4%. Doping with 0.6% holmium increases the permittivity of BaTiO₃ by approximately three times (from ∼1,300 to ∼4,000), with only a slight increase in the corresponding dielectric loss. The local electron density and defect concentration estimated using positron annihilation technique conforms well to the features found in the dielectric and resistivity measurements. The results have been discussed in terms of a mixed compensation model.     

Preliminary<Emphasis Type="Italic">in vitro</Emphasis> and<Emphasis Type="Italic">in vivo</Emphasis> characterizations of a sol-gel derived bioactive glass-ceramic system

This study investigates quantitatively and qualitatively the sol-gel derived bioactive glass-ceramic system (BGS)—apatite-wollastonite (AW) type granules in the size range of 0.5–1 mm, as an effective graft material for bone augmentation and restoration. Scanning electron micrographs (SEM) of the sintered granules revealed the rough material surface with micropores in the range 10–30 μm. X-ray diffraction (XRD) pattern of the granules revealed the presence of crystalline phases of the hydroxyapatite and wollastonite, and the functional groups of the silicate and phosphates were identified by Fourier transform infrared spectroscopy (FT-IR). Thein vitro cell culture studies with L929 mouse fibroblast cell line showed very few cells adhered on the BGS disc after 24 h. This could be due to the highly reactive surface of the disc concomitant with the crystallization but not due to the cytotoxicity of the material, since the cellular viability (MTT assay) with the material was 80‰ Cytotoxicity and cytocompatibility studies proved that the material was non-toxic and biocompatible. After 12 weeks of implantation of the BGS granules in the tibia bone of New Zealand white rabbits, the granules were found to be well osteointegrated, as observed in the radiographs. Angiogram with barium sulphate and Indian ink after 12 weeks showed the presence of microcapillaries in the vicinity of the implant site implicating high vascularity. Gross observation of the implant site did not show any inflammation or necrosis. SEM of the implanted site after 24 weeks revealed good osteointegration of the material with the newly formed bone and host bone. New bone was also observed within the material, which was degrading. Histological evaluation of the bone healing with the BGS granules in the tibial defect at all time intervals was without inflammation or fibrous tissue encapsulation. After 2 weeks the new bone was observed as a trabeculae network around the granules, and by 6 weeks the defect was completely closed with immature woven bone. By 12 weeks mature woven bone was observed, and new immature woven bone was seen within the cracks of the granules. After 24 weeks the defect was completely healed with lamellar bone and the size of the granules decreased. Histomorphometrically the area percentage of new bone formed was 67.77% after 12 weeks and 63.37% after 24 weeks. Less bone formation after 24 weeks was due to an increased implant surface area contributed by the material degradation and active bone remodeling. The osteostimulative and osteoconductive potential of the BGS granules was established by tetracycline labelling of the mineralizing areas by 2 and 6 weeks. This sol-gel derived BGS granules proved to be bioactive and resorbable which in turn encouraged active bone formation.     

Residual Stress Simulation of Injection Molding

In this study, distribution and history of residual stresses in plaque-like geometries are simulated based on linear thermoviscoelastic model, which helps to understand the mechanics and evolution of the residual stresses in the injection molding process. The numerical calculation of direction, combined with the specified boundary conditions. Results show that the stress variation across the thickness exhibits a high surface tensile value changing to a compressive peak value close to the surface, with the core region experiencing a parabolic tensile peak. Residual stress distribution throughout the thickness is almost same along the flowpath and the final residual stresses value near the gate is lower than the value near the end of flowpath.