Ab initio vibrational spectra and structure of the ground state of HeScH2+

Potential energy and dipole moment surfaces of the ¹A′ ground state of HeScH²⁺ have been calculated using both the internally contracted single and double excitation multireference configuration interaction and the coupled-cluster singles and doubles augmented by a perturbative treatment of triple excitations levels of theory. Analytical functions have been fitted to the discrete surfaces employing a multidimensional least squares approach. These analytical functions have subsequently been embedded within a rectilinear normal-coordinate vibrational Hamiltonian in order to calculate vibrational states and transition intensities for low-lying states of HeScH²⁺.     

Modeling of process intensification of biodiesel production from Aegle Marmelos Correa seed oil using microreactor assisted with ultrasonic mixing

Conventionally, the batch type reactors were used for the production of biodiesel. However, in recent years, the usage of microreactors has started emerging as a significant substitute for biodiesel production due to its higher conversion rate at a short duration. These microreactors have a significantly high surface to volume ratio and high heat-mass transfer rate. The disadvantage of this type of reactors is its low mixing rate of the reagents. This can be overcome with the assistance of ultrasonic mixing. The main objective of this paper is to study the interlaced effect of a continuous flow microreactor and ultrasonic mixing on trans-esterification of Aegle Marmelos Correa seed oil using sodium methoxide catalyst. Results of microreactors with 0.3 mm and 0.8 mm diameter were compared. The effects of process parameters namely, flow rate (2–10 mL/min), reaction temperature (45–65 °C), catalyst amount (0.5–2.5 wt%), oil to methanol molar ratio (1:6–1:18) and ultrasonic mixing time (30–150 s) were studied using response surface methodology (RSM). The biodiesel yield of 98% and 91.8% were obtained for 0.3 mm and 0.8 mm microreactors, respectively. The maximum biodiesel yield observed in 0.3 mm reactor under following optimum conditions: 6.8 mL/min flow rate, 48 °C reaction temperature, 1.3 wt% catalyst, 1:9 oil to methanol molar ratio and 83 s ultrasonic mixing time. The predictive and generalization abilities of RSM and artificial neural network (ANN) models were evaluated and compared. The study showed that ANN and RSM models could predict the yield with an R² value of 0.9955 and 0.9900 respectively. However, the ANN model predicted the yield with the least mean square error value of 0.00001294, which is much lower than RSM.     

径向基神经网络模型在滇东南金矿潜力预测中的应用

采用人工网络神经法（Artificial Neural Network,ANN）有助于理解成矿系统的非线性动力学行为和对矿产资源进行预测.其中的径向基神经网络（Radial Basis Function Neural Network,RBFNN）具有优秀的逼近特性,优化过程简单,训练速度快,适合于需要大量数据综合的矿产预测.采用RBFNN方法对成矿地质条件复杂的中国滇东南地区开展金矿成矿预测.研究结果表明,该模型能快速获取成矿潜力信息.通过采用受试者工作特征（Re-ceiver Operating Characteristic,ROC）曲线进行精度验证,表明该模型具有优越的预测能力.     

半导体激光器的噪声特性神经网络仿真

通过求解含朗之万噪声项的速率方程, 给出了噪声特性解析表达式, 建立了半导体激光器噪声特性神经网络模型. 利用该模型对噪声进行了计算机仿真, 取得了与数值计算相一致的结果. 训练好的神经网络模型具有精度高、速度快等优点.     

构造性算法的神经网络集成在近红外光谱分析中的应用

针对传统的近红外数据分析方法精度较低、,应用的局限性问题,本文提出了一种基于构造性算法的神经网络集成方法,由一个构造性算法决定个体网络中隐层节点的数量以保证个体网络的精确性,运用负相关学习算法和网络个体训练次数不同保证了网络个体的多样性。这种方法在近红外光谱分析中得到了成功的应用。     

近红外光谱结合人工神经网络分析蔗汁的锤度和旋光度

应用中波近红外(NIR)光谱结合误差反传人工神经网络(BP-ANN)方法,建立蔗汁锤度、旋光度的定量分析模型。光谱范围为1 000～1 800 nm,采用2 mm光程透射方式获得蔗汁吸光度光谱。对蔗汁的吸光度光谱进行Savitzky-Golay求导和均值中心化处理,然后通过相关系数法结合样品特征吸收优化建模波长范围,再采用PLS降维获取主成分并输入BP-ANN建立校正模型,用验证样品对校正模型进行验证。结果显示,BP-ANN法建立的锤度和旋光度的预测相关系数(R2)分别为0.982,0.979,预测标准偏差(SEP)分别为0.159和0.137,均优于偏最小二乘(PLS)建模方法结果,可较好地用于蔗汁锤度、旋光度的快速测定。     

用MSC-ANN方法建立冬小麦叶片叶绿素与反射光谱的定量分析模型研究

采用多元散射校正(MSC)预处理方法对冬小麦叶片反射光谱进行预处理,有效地减小物理因素对光谱的影响,之后用非线性迭代偏最小二乘法(NIPALS)提取经MSC处理后的反射光谱的主成分,主成分个数由交叉证实法(Cross Validation)确定,将提取的主成分作为人工神经网络(ANN)的输入,建立人工神经网络分析模型(MSC-ANN),用冬小麦叶片的反射光谱来预测冬小麦叶片叶绿素含量。校准集的化学值与预测值的相关系数r达到0.960 4,预测标准偏差SD为0.187,相对标准偏差RSD为5.18%。检验集的化学值与预测值的相关系数r达到0.960 0,预测标准偏差SD为0.145,相对标准偏差RSD为4.21%。结果表明,MSC-ANN方法能在较大程度上消除了野外物理因素的影响,使用具有代表性的光谱数据点建立模型,能够建立准确的冬小麦叶绿素含量预测模型,可代替经典分析方法,满足冬小麦叶片叶绿素快速分析的需要。     

Subgrid-scale scalar flux modelling based on optimal estimation theory and machine-learning procedures

New procedures are explored for the development of models in the context of large eddy simulation (LES) of a passive scalar. They rely on the combination of the optimal estimator theory with machine-learning algorithms. The concept of optimal estimator allows to identify the most accurate set of parameters to be used when deriving a model. The model itself can then be defined by training an artificial neural network (ANN) on a database derived from the filtering of direct numerical simulation (DNS) results. This procedure leads to a subgrid scale model displaying good structural performance, which allows to perform LESs very close to the filtered DNS results. However, this first procedure does not control the functional performance so that the model can fail when the flow configuration differs from the training database. Another procedure is then proposed, where the model functional form is imposed and the ANN used only to define the model coefficients. The training step is a bi-objective optimisation in order to control both structural and functional performances. The model derived from this second procedure proves to be more robust. It also provides stable LESs for a turbulent plane jet flow configuration very far from the training database but over-estimates the mixing process in that case.