A 6 × 6 Cells Interconnection-Oriented Programmable Chip for CNN

The implementation of a versatile VLSI chip certainly represents an important step to improve the research on Cellular Neural Networks. In this paper a VLSI realization of the multi-chip oriented, the 6 × 6 Digitally Programmable Cellular Neural Network (6 × 6 DPCNN) chip, will be presented. This chip covers most of the available one-neighbourhood templates for image processing applications. Moreover, it can be easily interconnected to others to carry out very large CNN arrays. The designs and some measured results of a single chip and a multi-chip board (the 720 DPCNN System) will be shown.     

CNN cell for computing disparity map

The real-time estimation of the distance of objects from an observer is a critical issue in several application fields. A new cellular neural network circuit that uses a stereo vision algorithm to compute the disparity map is presented     

A 3 × 3 digitally programmable CNN chip

In this paper a VLSI implementation of a 3 × 3 cell digitally programmable cellular neural networks (CNN) with discrete templates is presented. This chip is the first successfully tested, multichip-oriented CNN chip. In fact, this chip has been designed to be easily interconnected to others to carry out very large CNN arrays. This feature has been verified in a two-chip prototype board. The fully programmable capability covers most of the available one-neighbourhood fixed templates.     

A Dedicated Multi-Chip Programmable System for Cellular Neural Networks

Cellular Neural Networks (CNN's) represent a remarkable improvement in the hardware implementation of Artificial Neural Networks (ANN's). In fact, their regular structure and their local connectivity feature contribute to render this class of neural networks especially appealing for VLSI implementations. CNNs are widely applied in several fields, including image processing and pattern recognition. In this research, the authors already presented two fully digitally programmable CNN chips with 3×3 (3×3DPCNN chip) and 6×6 cells (6×6DPCNN chip) respectively. In this paper, a system with twenty of the latter chips will be presented. The main features of this electronic system consist of the full digital programmability of the templates, the digital input/output for logic operations, the analog outputs for dynamic analysis and the implementation of space-variant as well as space-invariant CNNs.     

Advanced neutronics tools for BWR design calculations

This paper summarizes the developments implemented in the new APOLLO2.8 neutronics tool to meet the required target accuracy in LWR applications, particularly void effects and pin-by-pin power map in BWRs. The Method of Characteristics was developed to allow efficient LWR assembly calculations in 2D-exact heterogeneous geometry; resonant reaction calculation was improved by the optimized SHEM-281 group mesh, which avoids resonance self-shielding approximation below 23 eV, and the new space-dependent method for resonant mixture that accounts for resonance overlapping. Furthermore, a new library CEA2005, processed from JEFF3.1 evaluations involving feedback from Critical Experiments and LWR P.I.E, is used. The specific “2005-2007 BWR Plan” settled to demonstrate the validation/qualification of this neutronics tool is described. Some results from the validation process are presented: the comparison of APOLLO2.8 results to reference Monte Carlo TRIPOLI4 results on specific BWR benchmarks emphasizes the ability of the deterministic tool to calculate BWR assembly multiplication factor within 200 pcm accuracy for void fraction varying from 0 to 100%. The qualification process against the BASALA mock-up experiment stresses APOLLO2.8/CEA2005 performances: pin-by-pin power is always predicted within 2% accuracy, reactivity worth of B4C or Hf cruciform control blade, as well as Gd pins, is predicted within 1.2% accuracy.     

An extension of the Generalized Perturbation Theory for spatial and energy collapsing

In this paper an extension of the Classical and Generalized Perturbation Theory is proposed which enables evaluating, at the 1st order, the contribution of any single spatial mesh to the variation of reactivity and reaction-rate ratios when an energy, spatial or energy/spatial collapsing is performed. An important achievement of the method is that it allows evaluating the individual asymptotic contribution of each spatial mesh to the variation, which enables investigating its spatial origin, very easily. Together with 1st order formulation, a simplified version is provided, which allows saving computation time while keeping an acceptable quality of the results. Some simple examples are provided, aimed at highlighting that the results obtained adopting the proposed formulation have a physical meaning and could be used to analyze the usual spatial/energy collapsing process in a convenient way.This formulation enlarges to the energy and spatial meshing the Classical and Generalized Perturbation Theory approaches to evaluate reactivity and reaction-rate ratio sensitivity to parameter variations.     

Digitally programmable nonlinear function generator for neural networks

The design of a new digitally programmable analogue circuit well suited for the implementation of several sets of nonlinear functions by approximating them by using a linear combination of sigmoidal terms is presented. The proposed circuit, allowing the building of several collections of nonlinear functions, would be useful in modelling artificial neural networks, fuzzy as well as partial differential equations based circuits