基于支持向量机的多类分类研究

现今流行的分类方法的重要基础是传统的统计学，前提是要有足够的样本，当样本数目有限时容易出现过学习的问题，导致分类效果不理想。引入支持向量机方法，它基于统计学习理论，采用了结构风险最小化原则代替经验风险最小化原则，较好的解决了小样本学习的问题；又由于采用了核函数思想，把非线性空间的问题转换到线性空间，降低了算法的复杂度。对其相关内容包括优化算法及多类分类问题的解决进行了研究，最后用一个实例说明了该方法的可行性和有效性。     

The Influence and Interpretation of Surface Parameters for Wetting Transitions in Ternary Mixtures

Starting from a microscopic Hamiltonian defined on a semi-infinite cubic lattice, and employing a mean-field approximation, the surface parameters relevant for wetting in confined ternary mixtures are derived. These are found in terms of the microscopic coupling constants, and yield a physical interpretation of their origins. In comparison with the standard expression for the surface free-energy density, several new terms arising from the derivation are identified. The influence of the surface parameters on a predicted unbinding transition in a mixture of oil, water, and amphiphile demonstrate that existing results are robust to the addition of the extra surface terms.     

芴酮在合成功能高分子方面的应用

介绍了由芴酮合成的双酚芴在高分子材料方面的应用，尤其是在环氧树脂、聚碳酸酯、有机玻璃方面，与双酚A相比，用双酚芴改性的高分子材料具更高的耐热性及透明性。     

Topology optimization of trusses using genetic algorithm,force method and graph theory

In this article size/topology optimization of trusses is performed using a genetic algorithm (GA), the force method and some concepts of graph theory. One of the main difficulties with optimization with a GA is that the parameters involved are not completely known and the number of operations needed is often quite high. Application of some concepts of the force method, together with theory of graphs, make the generation of a suitable initial population well‐matched with critical paths for the transformation of internal forces feasible. In the process of optimization generated topologically unstable trusses are identified without any matrix manipulation and highly penalized. Identifying a suitable range for the cross‐section of each member for the ground structure in the list of profiles, the length of the substrings representing the cross‐sectional design variables are reduced. Using a contraction algorithm, the length of the strings is further reduced and a GA is performed in a smaller domain of design space. The above process is accompanied by efficient methods for selection, and by using a suitable penalty function in order to reduce the number of numerical operations and to increase the speed of the optimization toward a global optimum. The efficiency of the present method is illustrated using some examples, and compared to those of previous studies. Copyright © 2003 John Wiley & Sons, Ltd.     

Requirements engineering in health care: the example of chemotherapy planning in paediatric oncology

Health care is characterized by highly complex processes of patient care that require unusual amount of communication between different health care professionals of different institutions. Sub-optimal processes can significantly impact on the patient’s health, increase the consumption of services and resources and in severe cases can lead to the patient death. For these reasons, requirements engineering for the development of information technology in health care is a complex process as well: without constant and rigorous evaluation, the impact of new systems on the quality of care is unknown and it is possible that badly designed systems significantly harm patients. To overcome these limitations, we present and discuss an approach to requirements engineering that we applied for the development of applications for chemotherapy planning in paediatric oncology. Chemotherapy planning in paediatric oncology is complex and time-consuming and errors must be avoided by all means. In the multi-hospital/multi-trial-centre environment of paediatric oncology, it is especially difficult and time-consuming to analyse requirements. Our approach combines a grounded theory approach with evolutionary prototyping based on the constant development and refinement of a generic domain model, in this case a domain model for chemotherapy planning in paediatric oncology. The prototypes were introduced in medical centres and final results show that the developed generic domain model is adequate.     

宝力格油田保护油层钻井液技术

钻井液作为与油层首先接触的外来流体极易对油层造成损害,做好钻井液保护油层工作在整个油层保护系统工程中非常重要.针对华北油田宝力格油田的储层物性,室内从岩心的敏感性评价入手,开展了钻井液保护油层技术的研究.通过对油层封堵技术的研究,研制出了适合于该油田储层特征的油层保护剂,结合现场钻井液施工的特点,制定了便于现场操作的保护油层施工措施.该技术自2002年开始,在二连地区宝力格油田共实施162口井.采取油层保护措施的井,完善井为86%,说明该钻井液技术在宝力格油田应用效果较好.     

双转向垂向控缝技术实验研究及现场应用

针对压裂过程中缝高较难控制的问题，研制开发出新型的油溶性垂向双转向剂，并开展了室内实验研究。首次验证了转向剂对裂缝延伸的控制机理，确定了转向剂的适用浓度及控缝能力。实验表明，转向剂可以增加隔层应力3～4MPa左右。在此基础上研究了控制裂缝高度施工方案的设计方法，并利用全三维软件结合垂向应力剖面对控缝技术进行了优化。现场推广应用30井次，取得了较好的控缝效果。     

活化疏松剂用于过磷酸钙生产的性能研究

介绍了在过磷酸钙中添加活化疏松剂的试验情况及运用效果。试验结果表明 :添加活化疏松剂后鲜肥转化率提高 4.0 % ,产品物性得到改善 ,结块倾向明显降低。     

低温低渗透砂岩油藏窜流大孔道深部封堵技术研究

所用堵剂为高强度的淀粉接枝聚丙烯酰胺交联凝胶SAMG-1,由<6%淀粉、4.5%-5.5%丙烯酰胺、0.003%-0.006%交联剂组成,35℃成胶时间受淀粉和交联剂用量控制,为18-20小时以上,成胶前黏度-100 mPa.s。该堵剂具有长期稳定性,在储层岩心中注入深度15 cm的堵剂,在35℃候凝48小时后及老化90天后,封堵强度分别为0.61和0.59 MPa/cm,封堵率分别为98.6%和98.1%。该堵剂优先进入高渗层,注入0.5 PV并成胶后,2组双填砂管组成的模型低、高渗管渗透率保留率分别为68.4%、0.7%和69.4%、0.0%。吉林扶余油田西一区+15-8.2区块有水井6口,油井13口,含水率达91.5%,注入水最快在5天内到达油井。报道了该区块整体深部调剖封堵窜流通道的情况,详细叙述了+15-9.2井施工中通过注入压力和井底回压控制注入流量,使堵剂陆续进入原生和次生孔道的工艺作业,该井设计注入堵剂92 m3。6口水井整体调剖后,油井产液量差别减小,产油量增加,有效期已超过了9个月,共增油843 t,含水平均下降3.87%。图7表1参8。     

羧酸系列集油剂的水面集油性能

研究了十二烷酸、十四烷酸、十六烷酸和十八烷酸在淡水中的集油能力。结果表明，这4种羧酸在淡水表面的表面膜压随其分子中碳数的增加而增大。其中后两者的表面膜压高达45～50mN／m，显著高于各类油膜膜压，满足集油剂对表面活性物质膜压的要求。4种脂肪羧酸在淡水中使柴油油膜的收缩率在1min时达到最大值，分别为80．0％、81．8％、83．8％和88．2％。它们的己醇、正庚醇、正辛醇、仲辛醇溶液的集油持续性，总体上随羧酸分子碳数增加而增大。其中十六烷酸和十八烷酸集油能力非常接近，它们的正庚醇、正辛醇和仲辛醇溶液使柴油膜面积收缩率在120h后仍保持在66．0％～70．2％。该类集油剂毒性低，易为生物降解，有实际应用前景。