BP人工神经网络用于除草剂活性预测

采用AMI方法研究了10种新型的磺酰脲类除草剂的电子结构，并以原子的Mulliken净电荷和除草剂在不同浓度(100，10mg／L)下对油菜、稗草两种作物的根、茎部位的抑制率为训练样本集。构造并训练得到具有活性预测能力的BP神经网络．结果表明，该BP网络不仅能对训练样本很好拟合。亦能对未知化合物的活性作出很好的预测．     

永不消失的白螺号——访藏学大师东噶·洛桑赤列先生

东噶·洛桑赤列教授是享誉国际国内的藏学家，国家级专家，曾担任中国藏学研究中心副总干事和西藏社会科学院名誉院长等行政职务，先后任中央民族大学藏学院、西藏大学文学院教授。在他长期的坎坷人生经历中，由一名农奴成为著名的藏传佛教活佛，由一名俄让巴密宗师转变为接受马克思主义的教授，成为著述等身的大师级学者。先后经历过新旧社会的变革，新旧文化的熏陶，他的学术思想在无数磨砺中闪闪发光，他的天资在特殊的文化环境中释放出无穷的能量，他的人品、治学精神，在学术领域永无止境的追求探索的精神和他留下的20多部学术著作，已经成为中华民族一笔宝贵的文化遗产。文章根据1984年冬至1997年2月之阍陆续采访的记录（手稿），主要对东噶·洛桑赤列先生的人生历程、研究成果、学术观点、治学精神等方面的内容进行了摘编。     

Zur Synthese und Struktur von K3N

Synthesis and Structure of K₃N Two phases in the binary system K/N have been obtained via co‐deposition of potassium and nitrogen onto polished sapphire at 77 K and subsequent heating to room temperature. The powder diffraction pattern of one of these phases can be satisfactorily interpreted by assuming the composition K₃N, and the anti‐TiI₃ structure‐type, which is also adopted by Cs₃O. The resulting hexagonal lattice constants are: a = 779.8(2), c = 759.2(9) pm, Z = 2, P6₃/mcm. Comparison with possible structures of K₃N generated by computational methods and refined at Hartree‐Fock‐ and DFT level, reveals that the energetically most favoured structure has not formed (presumable Li₃P‐type), but instead one of those with very low density. In this respect, the findings for K₃N are analogous to the results on Na₃N. The thermal evolution of the deposited starting mixture has been investigated. Hexagonal K₃N transforms to another K/N phase at 233 K. Its XRD can be fully indexed resulting in an orthorhombic cell a = 1163, b = 596, c = 718 pm. Decomposition leaving elemental potassium as the only residue occurs at 263 K.     

Predicting vibration signals of automobile engine using chaos theory

Condition monitoring and life prediction of the vehicle engine is an important and urgent problem during the vehicle development process. The vibration signals that are closely associated with the engine running condition and its development trend are complex and nonlinear. The chaos theory is used to treat the nonlinear dynamical system recently. A novel chaos method in conjunction with SVD (singular value decomposition)denoising skill are used to predict the vibration time series. Two types of time series and their prediction errors are provided to illustrate the practical utility of the method.     

Fatigue Loading and Life Prediction in Three Fretting Fatigue Fixtures

Three fixtures for conducting laboratory fretting fatigue tests are described and their respective testing methods and the results of the analysis are compared. Each of these fixtures has been used to investigate the effects of various parameters of interest in fretting fatigue. These fixtures include a unique apparatus in which all load applied to the specimen is transferred to the fretting pads, an apparatus similar to many found in the literature where partial load transfer occurs across the pads, and a simplified dovetail fixture in which the clamping load, P, and the shear load, Q, are varied in phase. Select test conditions from prior experiments performed on identical material and resulting in similar lives ranging from one to ten million cycles from these fixtures are identified. The various testing conditions were used to compute the unique stress field for each case. The resulting contact stresses were used to calculate crack initiation based criteria, and to calculate stress intensity factors. The three fixtures were shown to be able to accommodate a range of loads, fretting pad contours, and specimen geometries that produced a variety of stress fields. A crack-initiation-based criterion was shown to predict the failure lives of thinner specimens accurately. The stress intensity factor calculations showed the possibility of a crack arresting for a stress field that decays rapidly and the possibility of a local minimum for K as a function of depth. The fixtures are shown to be complementary in generating data for development of robust fretting fatigue models that use these criteria.     

Long-Term Prediction of Creep Strain of Expanded Polystyrene

Data obtained in investigating the creep of expanded polystyrene in compression are discussed. An analysis of the results of creep tests 65–608 days long showed that the creep can be predicted for 50 years based on experiments of less duration than the 605 days indicated in EN 13163. For this purpose, it is suggested to employ a 95-fold extrapolation in time if the creep curves are described by a power function (as recommended by EN 13163) or a 50-fold extrapolation if the exponential equation proposed in the present paper is used. __________ Translated from Mekhanika Kompozitnykh Materialov, Vol. 41, No. 6, pp. 795–802, November–December, 2005.     

保险公司离散型崩溃模型研究

本文提出并讨论了含投资因素、红利分配因素的崩溃模型，得出了关于崩溃概率、保险公司的期望寿命的结论，这些结论推广了没有考虑投资因素的崩溃模型的相关结论．     

具有离散参数齐次随机场的线性预测(V)

本文提出并研究下列四类线性预测问题：2（n0），6（n0），5（NO，n0）及3（N0，n0）．不仅求出了它们的预测误差，而且也求出了它们的预测值．因而不仅推广了而且也改进了[2－5]中的结果．本文还给出了场｛X（m，n）｝的一个新形式的表示式．     

煤与瓦斯突出预报的模糊聚类关联分析法

首先采用模糊聚类分析对煤与瓦斯突出的样本集合进行分类，建立不同突出程度的模糊模式。然后用关联分析确定待预报样本与模式的关联程度，以此预测预报样本的煤与瓦斯突出危险程度。这种预报方法相对于模糊聚类分类后，将模式与待预报样本组成新的样本集合，再进行聚类分析，以此分类结果进行预报法。不仅可靠程度高，而且能定量描述待报样本与模式的亲和程度。     

指数型寿命数据中异常值的检验

本文讨论了在指数型寿命数据中,对同时存在的异常大数据和异常小数据的检验方法,给出了一个明确的判别标准,并以一例说明其应用。