Modeling and optimization of semiconductor manufacturing process with neural networks

A neural network-based process model is proposed to optimize the semiconductor manufacturing process. Being different from some works in several research groups which developed neural network-based models to predict process quality with a set of process variables of only single manufacturing step, we applied this model to wafer fabrication parameters control and wafer lot yield optimization. The original data are collected from a wafer fabrication line, including technological parameters and wafer test results. The wafer lot yield is taken as the optimization target. Learning from historical technological records and wafer test results, the model can predict the wafer yield. To eliminate the "bad" or noisy samples from the sample set, an experimental method was used to determine the number of hidden units so that both good learning ability and prediction capability can be obtained.

Generalized Viscoplastic Modeling of Debris Flow

Various concepts have been proposed or used in the development of rheological models for debris flow. The earliest model developed by Bagnold was based on the concept of the “dispersive” pressure generated by grain collisions. Bagnold’s concept appears to be theoretically sound, but his empirical model has been found to be inconsistent with most theoretical models developed from non-Newtonian fluid mechanics. Although the generality of Bagnold’s model is still at issue, debris-flow modelers in Japan have generally accepted Takahashi’s formulas derived from Bagnold’s model. Some efforts have recently been made by theoreticians in non-Newtonian fluid mechanics to modify or improve Bagnold’s concept or model. A viable rheological model should consist both of a rate-independent part and a rate-dependent part. A generalized viscoplastic fluid (GVF) model that has both parts as well as two major rheological properties (i.e., the normal stress effect and soil yield criterion) is shown to be sufficiently accurate, yet practical, for general use in debris-flow modeling. In fact, Bagnold’s model is found to be only a particular case of the GVF model. Analytical solutions for (steady) uniform debris flows in wide channels are obtained from the GVF model based on Bagnold’s simplified assumption of constant grain concentration.

The Deposition and Burning Characteristics During Slagging Co-Firing Coal and Wood: Modeling and Numerical Simulation

Numerical analysis was used to study the deposition and burning characteristics of combining co-combustion with slagging combustion technologies in this paper. The pyrolysis and burning kinetic models of different fuels were implanted into the WBSF-PCC2 (wall burning and slag flow in pulverized co-combustion) computation code, and then the slagging and co-combustion characteristicsespecially the wall burning mechanism of different solid fuels and their effects on the whole burning behavior in the cylindrical combustor at different mixing ratios under the condition of keeping the heat input samewere simulated numerically. The results showed that adding wood powder at 25% mass fraction can increase the temperature at the initial stage of combustion, which is helpful to utilize the front space of the combustor. Adding wood powder at a 25% mass fraction can increase the reaction rate at the initial combustion stage; also, the coal ignitability is improved, and the burnout efficiency is enhanced by about 5% of suspension and deposition particles, which is helpful for coal particles to burn entirely and for combustion devices to minimize their dimensions or sizes. The results also showed that adding wood powder at a proper ratio is helpful to keep the combustion stability, not only because of the enhancement for the burning characteristics, but also because the running slag layer structure can be changed more continuously, which is very important for avoiding the abnormal slag accumulation in the slagging combustor. The theoretic analysis in this paper proves that unification of co-combustion and slagging combustion technologies is feasible, though more comprehensive and rigorous research is needed.

Modeling of a coal-fired slagging combustor: Development of a slag submodel

In a slagging combustor or furnace, the high combustion temperature makes the molten slag layer cover the wall and capture the particles. If these particles contain combustible matter, they will continue to burn on the running slag. As a result, the total amount of ash deposition will be much greater than that in dry-wall combustors and the total heat flux through the deposition surface will change greatly. Considering the limitations of existing simulation methods for slagging combustion, this paper introduces a new wall burning model and slag flow model from the analysis; of particle deposition phenomena. Combined with a conventional combustion simulation program, the total computational frame is introduced. From comparisons of simulation results from several kinds of methods with experimental data, the conclusion is drawn that the conventional simulation methods are not very suitable for slagging combustion and the wall burning mechanism should be considered more thoroughly.

Modeling and optimizing of high-concentration erbium-doped fiber amplifiers with consideration of ion-clusters

Starting from the modeling of isolated ions and ion-clusters, a closed form rate and power evolution equations for high-concentration erbium-doped fiber amplifiers are constructed. Based on the equations, the effects of the fraction of ion-clusters in total ions and the number of ions per cluster on the performance of high-concentration erbium-doped fiber amplifiers are analyzed numerically. The results show that the presence of the ion-clusters deteriorates amplifier performance, such as the signal power, signal gain, the threshold pump power for zero gain, saturated signal gain, and the maximum gain efficiency, etc. The optimum fiber length or other parameters should be modified with the ion-clusters being taken into account for the amplifiers to achieve a better performance. (c) 2007 Elsevier B.V. All rights reserved.