加速神经网络势能面的构建:一种杂化的训练算法

Machine learning approaches have been promising in constructing high-dimensional potential energy surfaces (PESs) for molecules and materials. Neural networks (NNs) are one of the most popular such tools because of its simplicity and efficiency. The training algorithm for NNs becomes essential to achieve a fast and accurate fit with numerous data. The Levenberg-Marquardt (LM) algorithm has been recognized as one of the fastest and robust algorithms to train medium sized NNs and widely applied in recent NN based high quality PESs. However, when the number of ab initio data becomes large, the efficiency of LM is limited, making the training time consuming. Extreme learning machine (ELM) is a recently proposed algorithm which determines the weights and biases of a single hidden layer NN by a linear solution and is thus extremely fast. It, however, does not produce sufficiently small fitting error because of its random nature. Taking advantages of both algorithms, we report a generalized hybrid algorithm in training multilayer NNs. Tests on H+H₂ and CH₄+Ni(111) systems demonstrate the much higher efficiency of this hybrid algorithm (ELM-LM) over the original LM. We expect that ELM-LM will find its widespread applications in building up high-dimensional NN based PESs.     

基于退化数据和DBN算法的IGBT健康参数预测方法

绝缘栅双极型晶体管(IGBT)等电子元器件被广泛用于运输和能源部门,其健康状态对于设备安全和有效至关重要；在对IGBT的结构和损伤机制分析基础上,结合NASA艾姆斯中心开展的IGBT加速退化试验,选择集电极-发射极关断峰值电压作为失效特征参数,提出了一种基于深度信念网络的预测模型对其进行分析和预测；以Levenberg-Marquardt(LM)算法模型作为对比,实验结果显示文章提出的三隐藏层DBN模型相比于LM模型有更好的预测性能和更高的预测精度。     

PNP输入双极运算放大器在不同辐射环境下的辐射效应和退火特性

研究了PNP输入双极运算放大器LM837在1 MeV电子和⁶⁰Coγ源两种不同辐射环境中的响应特性和变化规律. 分析了不同偏置状态下其电离辐照敏感参数在辐照后三种温度 (室温, 100 ℃, 125 ℃)下随时间变化的关系, 讨论了引起电参数失效的机理. 结果表明: 1 MeV 电子辐照LM837引起的损伤主要是电离损伤, 并且在正偏情况下比⁶⁰Coγ源辐照造成的损伤大; 辐照过程中, 不同辐照源正偏条件下的偏置电流变化都比零偏时微大; 在不同的辐照源下, LM837辐照后的退火行为都与温度有较大的依赖关系, 而这种关系与辐照感生的界面态密度增长直接相关. 关键词： PNP 输入双极运算放大器 60Coγ源')" href="#">电子和⁶⁰Coγ源 偏置条件 退火     

Effect of fuel on the formation structure,transport and magnetic properties of LaMnO3+ δ nanopowders

Single-step low-temperature solution combustion (LCS) synthesis was adopted for the preparation of LaMnO_{3+ δ} (LM) nanopowders. The powders were well characterized by powder X-ray diffraction (PXRD), scanning electron microscopy (SEM), energy dispersive spectroscopy (EDS), surface area and Fourier transform infrared spectroscopy (FTIR). The PXRD of as-formed LM showed a cubic phase but, upon calcination (900°C, 6 h), it transformed into a rhombohedral phase. The effect of fuel on the formation of LM was examined, and its structure and magnetoresistance properties were investigated. Magnetoresistance (MR) measurements on LM were carried out at 0, 1, 4 and 7 T between 300 and 10 K. LM (fuel-to-oxidizer ratio; ψ = 1) showed an MR of 17% at 1 T, whereas, for 4 and 7 T, it exhibited an MR of 45 and 55%, respectively, near the T _M-I. Metallic resistivity data below T _M-I showed that the double exchange interaction played a major role in this compound. It was interesting to observe that the sample calcined at 1200°C for 3 h exhibited insulator behavior.     

The TL and room temperature OSL properties of the glow peak at 110 °C in natural milky quartz: A case study

The LM–OSL signal of quartz, while measured at room temperature, is dominated by an intermediate, broad and intense OSL component, so that its contribution and general characteristics are derived very accurately. Through a series of dose–response, bleaching and thermal decay at room temperature experiments, in conjunction with curve fitting studies, a component resolved analysis is carried out studying the correlation between this specific component, termed as LM–OSL component C₂ and the 110 °C TL glow peak in quartz. The dose–response of these two luminescence components behaves exactly similar being linear at low doses and saturating at almost 100 Gy. Both signals decay exponentially under illumination, providing identical optical detrapping cross-section values. Residual of both luminescence signals after thermal decay at room temperature follows an exponential law, yielding similar mean half-lives. All previous luminescence features provide strong evidence for the electron trap being the same for both the 110 °C TL trap and the LM–OSL component C₂. The results of the present work are very promising and clearly support the possibility of extrapolating the TL pre-dose methodology to the OSL pre-dose effect using only the LM–OSL component C₂.     

几种液体金属自由表面的MHD不稳定性实验研究

介绍了在核工业西南物理研究院的液态金属实验回路(LMEL)上获得的几种可供液态偏滤器-限制器系统选用的液体自由表面的磁流体动力学(MHD)效应不稳定性的实验结果。实验发现：自由表面射流在穿越梯度横磁场时射程变短、截面变扁平，但MHD效应稳定，调节射流与磁场的夹角可以控制射流的流动特性；自由表面膜流MHD效应存在三种现象，即层流、溪状流和湍流。层流是由多束射流打到固体表面产生的(简称“射-膜流”)，从MHD效应角度考虑，“射-膜流”将是四种可选液态偏滤器-限制器系统的液体自由表面形式中最佳的选择。同时，探讨了从物理的角度来理解四种自由表面形式的MHD效应的现象。     

多纵模短腔染料激光放大前后光谱的比较

实验观测了短腔染料激光器输出的多纵模激光及其经一级染料放大的激光光谱，比较了光谱特性。在一定情况下，短腔染料激光器的多纵模激光经放大器放大可产生一个至几个新纵模，新纵模与短腔染料激光器输出的纵模有相似的频率间隔、线宽及频率牵引等特征。新纵模产生属于三阶四光子混频（即四波混频）为主导的非线性光学效应。存在频率牵引表明，增益介质中的多波混频存在频率失配，在这一四波混频中光子能量并不守恒     

一种基于多准则决策和PSO-LM混合优化算法的多峰Brillouin散射谱的特征提取方法

采用传统方法对多峰Brillouin散射谱进行拟合的过程中,通常是以谱线最大功率点为基准的,却忽略了其他比该点小但却是极值的功率点。这样获得的拟合曲线通常只有一个峰值,相当于把除最高峰之外还有多个小峰的多峰Brillouin散射谱进行了简化,导致大量有用信息的丢失。为了提高Brillouin散射谱的特征提取精度,提出了一种基于MCDM和PSO-LM混合优化算法的多峰Brillouin散射谱特征提取方法(MCDM-PSO-LM)。MCDM可以识别和准确定位多峰Brillouin散射谱的各个波峰和波谷;PSO-LM混合优化算法可以实现分别对各个波峰和波谷的曲线进行拟合并找到每一个波峰的中心频率,该算法既克服了PSO算法过早收敛于局部极值和LM算法依赖初值的问题,又可以将PSO算法的全局搜索能力和LM算法的局部收敛能力结合在一起。较传统算法而言,MCDM--PSO—LM算法保证了对最优值求解的速度和精度,提高了运算能力,使解析解最大限度地接近最优值。分别在不同信噪比和不同线宽条件下进行仿真验证,频移和温度误差分析结果表明,MCDM--PSO—LM方法可以对多峰Brillouin散射谱的各个波峰与波谷进行准确定位,可用于多峰Brillouin散射谱的特征提取,识别效果明显强于传统算法,提高了信息分析的准确性。     

Thermoelastic properties of MgSiO(3)-perovskite: insights on the nature of the Earth's lower mantle

We have determined by means of first principles quasiharmonic calculations the elastic constants and acoustic velocities of MgSiO3 perovskite, the most abundant mineral of the Earth's lower mantle (LM), at pertinent pressures and temperatures. Using these results, along with the effects of low concentration iron alloying and the thermoelasticity of the most important secondary LM phase, MgO, we predict the isotropic elastic moduli of likely LM aggregates. Comparison with seismic values extracted from the preliminary reference Earth model indicates that the top of the LM behaves as a typical aggregate of pyrolitic composition, likewise the upper mantle. But systematic deviations that cannot be accounted for by alterations in the geotherm alone develop toward the deep LM. This result could be viewed as evidence in support of radially inhomogeneous LM models.     

Comparative study on interactions between laser and arc plasma during laser-GTA welding and laser-GMA welding

This paper describes an investigation on differences in interactions between laser and arc plasma during laser-gas tungsten arc (LT) welding and laser-gas metal arc (LM) welding. The characteristics of LT heat source and LM heat source, such as plasma behavior, heat penetration ability and spectral information were comparably studied. Based on the plasma discharge theory, the interactions during plasma discharge were modeled and analyzed. Results show that in both LT and LM welding, coupling discharge between the laser keyhole plasma and arc happens, which strongly enhance the arc. But, the enhancing effect in LT welding is much more sensitive than that in LM welding when parameters are adjusted.