Combining neural networks and genetic algorithms for optimising the parameter design of the inter-metal dielectric process

An inter-metal dielectric (IMD) is deposited between metal layers to provide isolation capability to a device and separate the different metal layers that are unnecessary in the conduction of electricity. Owing to the complicated input/response relationships of the IMD process, the void problem results in electric leakage and causes wafer scraping. In the current study, we combined neural networks, genetic algorithms (GAs) and the desirability function in order to optimise the parameter settings of the IMD process. Initially, a backpropagation (BP) neural network was developed to map the complex non-linear relationship between the process parameters and the corresponding responses. Moreover, the desirability function and GAs were employed to obtain the optimum operating parameters in respect to each response. The implementation of the proposed approach was carried out in a semiconductor manufacturing company in Taiwan, and the results illustrate the practicability of the proposed approach.     

Interface characterizations of diamond-coated tools by scratch testing and simulations

In this study, micro-scratch tests were conducted on diamond-coated tungsten-carbide substrates to investigate coating adhesion. During the scratch testing, high intensity acoustic emission (AE) signals can be clearly detected when the coating delamination occurs. It is also found that the tangential force increase gradually with the normal force, but fluctuates significantly when the critical load of coating delamination is reached. A three-dimensional (3D) finite element (FE) model with a cohesive-zone interface was developed to simulate the scratch process, and by comparing with the delamination critical load from the experiment, the interface characteristic length, the maximum strength and the fracture energy can be obtained. The preliminary results indicate that it is feasible to use the FE simulation combined with scratch tests to evaluate the coating interface behaviors.     

Estimating the Ungauged Natural Flow Regimes for Environmental Flow Management

Regime-based approach recently becomes an important strategy while considering aquatic ecosystems in environmental flow management. The key element for supporting this strategy is long streamflow data which is usually not available for determining natural flow regimes. This study uses a back-propagation network to estimate ungauged natural flow regimes. A set of the upper reaches of Taiwan’s 42 flow stations with non-human control streamflow and at least 20 years daily flow data is used to quantify the natural flow regimes using 31 Indicators of Hydrologic Alteration (IHA). Watershed geomorphologic characteristic factors and rainfall parameters are used to classify homogeneous flow regime areas. The results show that there are three types of flow regimes from the flow stations, and each group of indicators in the IHA has different correlations with different geomorphologic characteristic factors and rainfall parameters. The results of using an artificial neural network model to estimate IHA show that the group average percent error fell from 21 % to 8 % and the average correlation coefficient was over 0.7, indicating that the model presented in this study is able to accurately estimate the natural flow regime in ungauged stations. Instead of predicting daily streamflow, this study estimates indicator values for ease of ecological water resources management.     

A label‐based information flow control model for object‐oriented systems

Abstract

This paper proposes a label‐based information flow control model to prevent information leakage within object‐oriented systems. It offers the features of: (a) adapting to dynamic object state change, (b) adapting to dynamic role change, (c) preventing indirect information leakage, (d) detailing the control granularity to variables, (e) allowing purpose‐oriented method invocation, (f) controlling method invocation through argument sensitivity, (g) allowing declassification, and (h) allowing only trusted sources to write a variable.     

On a transistor-type hydrogen gas sensor prepared by an electrophoretic deposition (EPD) approach

A Pd/GaN/AlGaN heterostructure field-effect transistor (HFET)-type hydrogen gas sensor, based on an electrophoretic deposition (EPD) approach, is fabricated and studied. Due to the formation of good Schottky gate contact by an EPD approach, the studied HFET shows improved DC performance including the suppressed gate current and better thermal stabilities on current–voltage (I–V) characteristics. This is mainly attributed to the reduction of interface trap density and improved Pd morphology. The EPD-based Pd morphologies are examined by X-ray diffraction, energy dispersive spectroscopy, Auger electron spectroscopy, scanning electron microscopy, and atomic force microscopy. For the used gate-dimension of 1 μm × 100 μm, an EPD-based HFET shows low gate current of 2.9 nA, maximum drain saturation current of 490 mA/mm, and maximum extrinsic transconductance of 78.9 mS/mm at room temperature. Also, solid thermal stabilities on maximum drain saturation current (−0.46 mA/mm K) and maximum extrinsic transconductance (−0.08 mS/mm K) are found as the temperature is increased from 300 to 600 K. For hydrogen gas sensing application, at 370 K, the maximum hydrogen sensitivity of 600.1 μA/mm ppm H₂/air under a 5 ppm H₂/air ambiance and fast response time (30 s) and recovery time (47 s) under a 10,000 ppm H₂/air ambiance are obtained. The EPD approach also demonstrates advantages of low cost, simple apparatus, easy process, little restriction on the shaped substrate, composited deposition, and adjustable alloy grain size. Therefore, the proposed EPD approach gives the promise for fabricating high-performance HFET devices and hydrogen gas sensors.     

Probabilistic simulation for developing likelihood distribution of engineering project cost

Inaccurate early project cost estimates can eliminate investment benefits. This study focuses on assisting estimators who are attempting to enhance the accuracy and reliability of engineering project cost in the pre-conceptual stage. This aim has recently garnered the attention of the transportation communities. Data from the Texas Department of Transportation (TxDOT) were utilized to develop an alternative approach that aids decision makers in terms of probability and confidence level. The proposed procedure comprises heuristic and practical simulation models that can be employed to calculate the probabilistic costs of highway bridge replacement projects. The simulation models utilize independent, correlated, and Latin Hypercube sampling approaches that incorporate major work items, roll-up work items, and project-level engineering contingencies. Cumulative distribution functions (CDFs) are then developed as a user-friendly chart for decision makers and these CDFs can be used to assess project risks during the pre-conceptual stage. Trial runs using these estimating procedures generate reliable pre-conceptual estimates. Additionally, these procedures can be extended to other project types along with programming techniques for developing an engineering project cost decision support system.     

Infinite single-walled boron-nitride nanotubes studies by LGTO-PBC-DFT method

Localized Gaussian type orbital-periodic boundary condition-density functional theory (LGTO-PBC-DFT) method was used to determine the electronic and detailed geometrical structures of (n, 0) zigzag type for n = 6 – 33 and (n, n) armchair type for n = 3 – 15 single-walled boron-nitride (BN) nanotubes with infinite tubular lengths. The calculations reveal that the calculated E_g (band gap between HOCO and LUCO) increases with increasing tubular diameter and eventually converge to 5.03 eV for BN nanotubes of larger tubular diameter. According to the calculated E_gs, the BN nanotubes are semi-conductor and their conductivities are not sensitive to the tubular diameter. Theoretically, the calculated bond length decreases with increasing tubular diameter. Based on our calculations, the bond length and angle do converge to 1.45 Å and 120 degree, respectively. Thus, the structures of BN nanotubes with the infinite tubular length approach the perfect hexagonal network when the tubular diameter increases. The calculated results also indicate that zigzag BN nanotubes with the tubular diameter larger than 18 Å display 3n properties in the calculated E_g, which is also obtained for zigzag carbon nanotubes.     

Ripening temperature affects the content and distribution of isoflavones in sufu,a fermented soybean curd

Sufu, a fermented soybean curd, was prepared by ripening the salted tofu cubes in the Aspergillus oryzae‐fermented rice–soybean koji mash at 25, 35 or 45 °C for a period of 16 days. It showed that the 16‐day ripened sufu contains less total isoflavone content (629.29–739.68 μg g^?1 dried matter) than the salted tofu before ripening (942.59 μg g^?1 dried matter). Regardless of ripening temperature, ripening causes a major reduction in the content of β‐glucoside and malonylglucoside isoflavones along with a significant increase of aglycone isoflavone content. These changes were enhanced as the ripening period extended. Among the various treatments examined, sufu ripened at 45 °C showed the greatest increase in aglycone content coupled with the greatest decrease in malonylglucosides. The distribution of malonylglucosides decreased from an initial 40.32% in the tofu cube to 9.78% after 16 days of ripening at 45 °C. Meanwhile, the distribution of aglycone increased from 13.17% to 39.88%.