径向基函数神经网络在启动压力梯度预测中的应用

启动压力梯度直接影响着低渗透油田的开采量以及油藏压力的预测精度。它与流体粘度、密度、渗透率、孔隙度等影响因素呈非线性关系。而人工神经网络具有表示任意非线性关系和学习的能力，是解决复杂非线性、不确定性和时变性问题的新思想和新方法。利用径向基函数(RBF)人工神经网络对启动压力梯度进行预测，并结合岩心的启动压力梯度的实际测定结果进行研究，结果表明：RBF人工神经网络是一种较为有效的预测方法，具有较高的精度．该方法可以为低渗油田的开发提供可靠的基础数据，节省了人力、物力。     

一种应用于移动终端的语音盲分离算法

盲分离算法被广泛应用于语音信号处理领域中。为改善移动终端的通话质量，提出了一种新的语音盲分离预处理方案，应用频域复随机梯度算法，分离出移动终端待传的主语音信号，从而使其它干扰语音得到抑制。模拟和实际录音的实验结果表明，该方法对改善通话质量有明显的效果。     

12位80MHz采样率具有梯度误差补偿的CMOS电流舵D/A转换器实现

提出了一种12位80 MHz采样率具有梯度误差补偿的电流舵D/ A转换器实现电路.12位DAC采用分段式结构,其中高8位采用单位电流源温度计码DAC结构,低4位采用二进制加权电流源DAC结构,该电路中所给出的层次式对称开关序列可以较好地补偿梯度误差.该D/ A转换器采用台湾U MC 2层多晶硅、2层金属(2 P2 M) 5 V电源电压、0 .5μm CMOS工艺生产制造,其积分非线性误差小于±0 .9L SB,微分非线性误差小于±0 .6 L SB,芯片面积为1.2 7mm×0 .96 m m ,当采样率为5 0 MHz时,功耗为91.6 m W.     

物探方法在基桩质量检测中的应用

远红外和磁梯度方法作为桩身质量检测新方法,可以直观精准地提供桩身缺陷性质及程度,为设计和施工处理提供重要依据。     

Interfacial Behavior in Type IV Systems

Type IV mixtures exhibit two heteroazeotropic lines, one at low temperature and the other meeting the supercritical range, characterized by the proximity of their critical end points (CEPs). Between these CEPs, the liquid phase is homogeneous inside a narrow range of temperature. The aim of this work is to analyze interface properties and wetting transitions for Type IV mixtures. Interfacial tensions have been calculated by means of the gradient theory, applied to binary van der Waals fluids. This approach is able to predict interfacial tension and phase equilibrium using a common equation of state (EOS). Results show that interfacial properties and wetting conditions are governed by the densities and the number of phases involved in equilibrium, a scenario that changes as temperature evolves from the low- to the high-temperature heteroazeotropic line.     

Optimizing physical parameters affecting coagulation of turbid water with Morninga oleifera seeds

Moringa oleifera is a plant coagulant and coagulant aid which has potential for use on a large scale for treatment of turbid waters in tropical developing countries. This paper presents results of optimization studies on physical factors including rapid mix velocity gradient and time, slow mix velocity gradient and time, and dosage of M. oleifera for low, medium and high raw water turbidities (50–750 NTU). Regression analysis was also carried out to study the effect of multiple factor interaction in the coagulation of turbid water with M. oleifera. This was carried out using the results of experimental runs involving varying initial turbidity, M. oleifera dosage, rapid mix velocity gradient, slow mix velocity gradient, and slow mix time.     

DETERMINATION OF THE TEMPERATURE HISTORY FOR THE U THONG OILFIELD AREA (SUPHAN BURI BASIN, CENTRAL THAILAND) USING A REALISTIC SURFACE TEMPERATURE

The temperature history for the BPI-W2 well in the oil-producing Suphan Buri Basin, central Thailand, has been investigated using different surface temperatures (T_s). Two ID models using T_s values of 0°C and ∼22°C were able to fit the suppression-corrected vitrinite reflectance (VR) values and burial peak temperatures (T_peak) in the well. The geothermal gradient averaged over 3 km is ∼54°C/km for T_s= 0°C, whereas it is ∼42°C/km for T_s∼22°C. T_s= 0°C is, however, considered to be unrealistic and the ∼54°C/km gradient is therefore too high. Similarly, a previously determined geothermal gradient of 62°C/km is considered to be an overestimate. The geothermal gradient of ∼42°C/km is plausible compared to other geothermal gradients onshore and offshore Thailand, although it is at the low end. This may be due to a too low suppression correction for the measured VR values. The obtained temperature history can be used to predict measured present-day temperatures of T_s∼22°C and 77°C in the reservoir in the U Thong oilfield. The obtained temperature history associated with the geothermal gradient of ∼42°C/km seems realistic as it predicts that the onset of oil generation at 107°C will have post-dated reservoir and trap formation in Middle to Late Miocene times.     

基于BP神经网络的汽车车型识别方法

将BP神经网络应用于汽车车型的自动识别,在简述车型图像预处理和特征抽取的基础上,重点讨论学习算法的选取问题,提出一种分阶段的训练方法。对三类车型的36对样本做训练和测试实验,实验结果表明,所采用的训练算法优于传统BP算法和共轭梯度法,所设计的网络能快速有效地识别汽车车型。     

含椭球夹杂耐热梯度功能材料细观结构优化应力方程

梯度功能材料具有复杂的细观结构组合方式,对其进行细观力学的研究是当前的主要发展方向。本文作者对非均质、耐热、含椭球夹杂梯度功能材料的细观结构与性能进行了分析,研究其细观力学建模和在机械/热载荷下的损伤演化和破坏。结果可应用于多相夹杂梯度功能材料的细观应力分析和细观性能预测,并均以显式给出,以便于工程应用。     

Ti(C,N)基金属陶瓷研究的最新进展

对国内外近几年来有关Ti(C,N)基属陶瓷制备技术的研究成果进行了总结和分析;主要分析了环形结构形成机理及影响元素,介绍了超细粉体、超细晶粒合金和表面梯度结构的制备技术。