NN Approach Research Articles

Use of Neural Network for Automatic Trajectory Plotting

Automatic trajectory plotting refers to the target recognition by computer processing of a sequence of images and plotting of the trajectory using the match locations. There are various approaches for achieving the target detection. This paper presents the ATP using NN approach for target recognition. The effect of the template size, the search area and the training set on the performance of the system is also discussed.

Neural-network-based decision making in diagnostic applications.

In this article the NN approach for medical decision making was applied for three specific examples. The first example was decision making with single-valued data for IUGR detection. The second example was decision making with double-valued data in prediction of ovulation. The third example was the use of independent NN modules and consensus theory for prediction of ovulation time. The NN approach has superiority over classical statistical approaches for decision making with medical data for the following reasons: 1. It is distribution-free. 2. It captures correlative features and does not need any specific consideration for mutual test dependence. 3. It provides weighted reliability of various tests. 4. It produces fast, accurate results. The statistical decision approach will probably outperform the NN approach in making decisions when an accurate distribution model is provided. However, the NN is proposed as a useful tool to help physicians in decision making and diagnosis of certain symptoms. The capability and performance of this tool has generally been proven in combining mutually dependent medical tests.

NEURAL NETWORK APPROACH TO LOCATING ACOUSTIC EMISSION SOURCES IN NON-DESTRUCTIVE EVALUATION

We expand upon the idea of using AE to monitor crack in a system or structure, give further motivation for the NN approach, and describe some details of the NN implementation, training, and application. Also included are experimental results on a steel I beam. We conclude with a summary of the many open issues remaining in traversing from the concept stage of this paper to a practical large scale test or implementation

A neural network model of parametric nonlinear hysteretic inductors

A neural network model of nonlinear hysteretic inductors is presented. The proposed neural network model is shown to reproduce the dynamic scalar hysteretic behavior of the current-flux relationship. Moreover, the intrinsic characteristics of the neural network approach yield a model particularly suitable when dependencies on different parameters are present, e.g., influence of mobile part positions, of temperature etc. Both theoretical comparisons (e.g., Chua-Stronsmoe model and Jiles-Atherton model) and experimental measurements (e.g., variable reluctance linear motor) are considered. Results point out the numerical accuracy and computational efficiency of the proposed NN approach, that results in a general framework useful in the field of electric machines, of electronic power circuits, of control and identification.

Potential Energy Surface of the O(1D) + N2O → 2NO, O2 + N2 Reactions

Using the Møller−Plesset method the ab initio N2O2 potential energy surface (PES) is calculated in order to study the exothermic singlet state reaction channels O(1D) + N2O → 2NO, O2(a 1Δg) + N2. According to the results of the calculation, the O(1D) + N2O → 2NO channel PES demonstrates the main energy release in the entrance valley. In this channel a wide interval of ∠O(1D)NN approach angles around the collinear configuration is easily accessible, but the minimum energy approach is not collinear. The O(1D) + N2O → O2(a 1Δg) + N2 channel is strongly noncollinear, as the minimum energy ∠O(1D)ON approach angle is close to 90°. The main energy release of this channel takes place in the exit valley. The results of the ab initio calculations have been used to construct an analytical model potential that describes roughly the main features of the N2O2 PES and can be used in the dynamical studies of both O(1D) + N2O reaction channels.

Postbuckling Analysis of Composite Laminated Cylindrical Panels Under Axial Compression

Concluding Remarks Two different back-propagation neural networks were developed to represent the nonlinear stress-strain behavior of (±0) graphiteepoxy laminates under monotonic and cyclic loadings. The NN predictions for both monotonic and cyclic loadings are in good agreement with the experimental data obtained from the literature. These preliminary results support the use of a NN approach to composite material modeling. The network developed in this study aids in identifying important aspects of the stress-strain behavior, such as breaking stress, fracture stress, etc. This approach can also be used to predict failure mechanisms.

Discrimination of $$p\bar p \to t\bar t$$ events by a neural network classifier

Neural network and conventional statistical techniques are compared in the problem of discriminating\(p\bar p \to t\bar t\) events, with top quarks decaying into anything, from the associated hadronic background at the energy of the Fermilab collider. The NN we develop for this sake is an improved version of Kohonen's learning vector quantization scheme. Performance of the NN as a\(t\bar t\) event classifier is found to be less satisfactory than that achievable by statistical methods. We conclude that the probable reasons for the are: i) the NN approach presents advantages only when dealing with event distributions in the feature space which substantially differ from Gaussians; ii) NN's require much larger training sets of events than statistical discrimination in order to give comparable results.