Solving a Nonlinear Problem in Magneto-Rheological Fluids Using the Immersed Interface Method

Many applications lead to a nonlinear elliptic interface problem in which the discontinuous coefficient depends on the solution and the material properties. A finite difference method based on Cartesian grids and the maximum principle preserving immersed interface method is proposed for the nonlinear elliptic interface problems discussed in this paper. Numerical experiments against the exact solutions reveal that our method is nearly second order accurate in the infinity norm. The method is applied to study the magneto-rheological field-responsive fluids that contain iron particles. Numerical experiments are performed against the results from the literature.     

Avoiding Routing Loops on the Internet

Suppose that some particular link in the Internet is currently congested. A natural solution is to try to make packets bypass that link. This can be done by increasing the cost of that link intentionally, say from a ₁ to a ₂, since the Internet uses shortest-path routing. Unfortunately, however, this often causes temporary loops for packet traveling, called routing loops. In this paper we show that routing loops can be avoided by increasing the cost of the link not directly from a ₁ to a ₂ but through an intermediate value, a ₃, i.e., from a ₁ to a ₃ and then to a ₂. We may need several intermediate values. We show that in this case the greedy strategy, namely, raising the cost as much as possible in each step, is optimal.     

Parallel processing circuit of line image using neuristor

A simultaneous detecting circuit system of centers and lengths of some lines in one dimension is presented applying an electronic circuit model of the nervous axon with respect to signal transmission. This model consists of series of one basic section, and has functions of Neuristor. The circuit system has a paralell processing function. It is constructed by three layers. In the first layer each visual signal of lines is changed to electric pulse signal. In the second layer edges of each line are detected, and in the third layer these detected signals are converted into voltages which correspond to lengths of the lines. These functions were confirmed experimentally using its hardware.     

Reduction of NOx from combustion flue gases bysuperimposed barrier discharge plasma reactors

NO_x reduction from combustion flue gases by superimposed barrier discharge plasma reactors is experimentally investigated. The experiments are conducted for applied voltages from 0 to 28 kV, flue gas rates from 0.5 to 2 L/min, ammonia mixture concentrations from 0.7 to 2.65 stoichiometry, and applied voltage phase differences from 0° to 180°, where two 60-Hz AC power supplies are used. The results show the following: (1) NO_x reduction rate decreases with increasing discharge power for surface discharge operations, however, NO_x reduction rate increases with increasing discharge power for silent and superimposing discharge operating modes; (2) NO_x reduction rate increases with increasing discharge power, gas flow rate and ammonia stoichiometry under in-phase operations; (3) NO_x reduction rate for out-of-phase operations is much higher compared with in-phase operations, however, NO_x reduction rate has an optimum condition on ammonia stoichiometry, discharge power, and gas flow rate; and (4) energy efficiency of NO_x reduction increases with increasing ammonia mixture and gas flow rate and decreases with increasing discharge power     

Improving the Performance of Feedforward Neural Networks by Noise Injection into Hidden Neurons

The generalization ability of feedforward neural networks (NNs) depends on the size of training set and the feature of the training patterns. Theoretically the best classification property is obtained if all possible patterns are used to train the network, which is practically impossible. In this paper a new noise injection technique is proposed, that is noise injection into the hidden neurons at the summation level. Assuming that the test patterns are drawn from the same population used to generate the training set, we show that noise injection into hidden neurons is equivalent to training with noisy input patterns (i.e., larger training set). The simulation results indicate that the networks trained with the proposed technique and the networks trained with noisy input patterns have almost the same generalization and fault tolerance abilities. The learning time required by the proposed method is considerably less than that required by the training with noisy input patterns, and it is almost the same as that required by the standard backpropagation using normal input patterns.     

Role of number of consecutive voltage zero-crossings in propagation of water trees in polyethylene

Effects of superposition of a low-frequency (0.1 to 50 Hz) voltage and a high-frequency 2 kHz voltage on the growth of water trees in polyethylene are examined. It has become clear that both the number of total voltage zero-crossings during the whole voltage application period and the number of consecutive voltage zero-crossings in the instant of the polarity reversal of the low-frequency voltage play important roles in the length and shape of water trees. Namely, the water tree length becomes longer as the number of total zero-crossings increases. Furthermore, among the voltages with the same number of total zero-crossings, the length becomes even longer with a prolonged shape in the direction of electric field as the number of consecutive zero-crossings increases. Effects of superposition of a high-frequency voltage onto a dc voltage are also examined. In this case, the dc voltage seems not to have any influence to the tree growth. By assuming the presence of space charge around the water-tree tip, the effective zero-crossings become only sensitive to the high-frequency component. Therefore, this result also indicates that the number of voltage reversals at the tree tip plays a crucial role in the growth of water trees.     

Linear bipolar OTAs employing exponential‐law circuits

This paper presents design of linear bipolar OTAs, which are composed of two function blocks; one is an exponential‐law circuit and the other is a core cell. Multi‐tanh cells are employed as the core cell. This kind of OTA has lower power dissipation relatively to the conventional multi‐tanh cell. According as the order of the multi‐tanh core cell becomes higher, the number of circuit realization for the core cell increases. For example, we have two OTAs for the core circuit of an emitter‐coupled pair and four OTAs for the doublet core cell. Thus, we consider the generalized OTAs for an arbitrary order n of the core cell and obtain a formula to give the realization number of the linear OTAs for n. According to the formula, there must be eight OTAs in the case of n=3. All of the eight OTAs are examined. Analysis and simulation results show that the OTAs have advantage in their characteristics, such as linear input range, power dissipation, noise, and frequency response. Copyright 2004 John Wiley & Sons, Ltd.     

Growth of Yb-doped YLiF4 laser crystal by the Czochralski method. Attempt of Yb energy level assignment and estimation of the laser potentiality

YLiF₄ (YLF) single crystals undoped and Yb³⁺-doped with different concentrations were grown by the Czochralski technique under CF₄ atmosphere. Detailed analysis of Yb³⁺-doped YLF spectroscopy were made to contribute to the determination of energy levels in this host. We are dealing with temperature and concentration dependences of both π and σ polarizations of the infrared (IR) absorption and emission spectra. Raman spectra were also used to give an attempt of interpretation of electronic and vibronic levels. The radiative energy transfer (self-trapping) and strong phonon–electron coupling make the assignment of Yb³⁺ energy levels difficult. Evaluation of the laser potentiality of this fluoride host is also presented.