应用BP神经网络模型实现内圆磨床主要零件的快速反应设计

用数字化模型代替三维实体模型，将内圆磨床主要大件的BP神经网络模型放入数据库信息管理系统，应用较简单的使用方法实现了零件结构的再修改、灵敏度分析及优化．根据所提出的方法，通过从数据库信息管理系统中调用桥板的数据文件，对其进行结构参数修改、灵敏度分析及以第1阶扭转频率最高为目标的设计变量自动搜索寻优计算，结果表明将BP神经网络模型与数据库信息管理系统结合是实现零件快速反应设计的有效途径．     

遗传算法在气体分离装置调优中的应用

提出了一种基于遗传算法和随机复合形法相结合的混合遗传算法，并以某厂气体分馏装置的生产过程的优化问题进行了测试，得出了优化操作条件。     

考虑误差影响的力学参数反分析神经网络方法研究

通过求偏导技术将反分析的参数误差分别表示为模型误差与测量误差的函数，利用这些函数研究了模型误差与测量误差在力学参数反分析神经网络方法中的传递过程以及影响反分析结果的影响因素．研究结果说明：输入变量的选择严重影响力学参数反分析的精度，选取力学参数灵敏度大的测点位移作为神经网络的输入变量，可以减小模型误差与测量误差对反分析结果的影响．这对提高力学参数反分析的实用性与可信性有重要意义．     

Inhibitory effect of the adenovirus type 5 E1A protein expressed in the yeast system

The human adenovirus type 5 E1A, a tumor- suppressor gene[1], codes for two major related proteins of 243 amino acids (12S) and 289 amino acids (13S) by al-ternative splicing in two exons[2]. Studies have been shown that E1A can regulate expression of many genes and cell cycle[3]. Both in vitro and in vivo experiments indicated that E1A could induce tumor cells differentia-tion, convert tumor cells into an epithelial phenotype, in-hibit tumor cell growth and metastasis and strongly en-ha…     

一种基于属性抽取与整合的Petri网模型

将感觉神经检测机制融合于Petri网中，提出了一种新的基于属性抽取与整合的感觉神经检测Petri网模型，使得Petri网更加适合于对神经思维的描述，提高了系统的智能行为。     

Building of hFⅨ transgenic mice by spermatogenic cells

Human FⅨ expression vector pCMVⅨ was packaged by effectene^TM reagent and injected into mice seminiferous tubules with glass pipettes.The expressional frame of pCMVⅨ was examined by PCR and Southern blotting among 41 progenies.There were 2(4%) mice being integrated with hFⅨ gene into chromosomes.4.6ng/mL of hFⅨ protein was expressed in plasma of one mouse,which was tested by ELISA.We demonstrated that building of transgenic animals by spermatogonial stem cells is an efficient method.Meanwhile,it has also been proved to be an alternative choice for mammary gland bioreactor.     

Inhibitory effect of the adenovirus type 5 E1A protein expressed in the yeast system of the human tumor cell growth

Adenovirus 5 type E1A as a tumor suppressor gene can inhibit tumor growth and enhance the censitivity of chemotherapy and radiotherapy.E1A have the ability to integrate into the host genome,resulting in long-time expres-sion that induces Rb gene inactivation and animal cells im-mortalization.This prompted us to select the E1A protein for treatment of cancer in order to overcome the limitations of E1A gene therapy.Thus,we firstly comstructed E1A eu-caryotic expression vector (pPIC9/E1A),transformated the pichia pastoris yeast cells(GS115) and screened the high-expressing recombinant strains.The positive yeast strains were cultured in the shake flask,and induced for 3d.The crude E1A protein was purified using two steps of col-umu chromatography on HiTrap Q and HiTrap SP.The pu-rified E1A protein was identified by SDS-PAGE and Western blot.E1A protein was mostly located at cellular unclear when Cheriot delivered E1A protein into cells.The analysis in vitro indicated that the E1A protein arrested LN686 cell cycle at G2/M phase,and significantly inhibited the growth of LN686 tumor cells.The current studies firstly provided an experimental basis to further develop E1A protein for tumor treatment.     

船舶初步设计阶段智能决策支持系统的应用

提出了用于辅助船舶设计者在获取船型要素信息量有限情况下，进行船型方案优选决策的支持系统理论体系框架和模型。将该系统应用于某实船船型方案优选决策，证明该方法是可行的、可靠的。此外，对未来智能决策支持系统在船舶初步设计中的应用前景进行了展望和预测。     

混沌神经网络在分离叠加模式和信息恢复中的应用

提出了一种改进的混沌神经网络模型，该模型具有两个特点：①能从叠加模型恢复出被记忆的正确信息；②能通过部份信息恢复被记亿的全部信息．设计了一系列的仿真实验证明了所提模型的有效性。     

交通噪声预测的神经网络模型

运用 Matlab语言编程 ,构造预测交通噪声的 LM算法 BP神经网络模型 ,把预测因子 (轻、重型车流量、平均车速、受声点距路肩距离、敏感点高差 )作为样本输入到网络模型 ,噪声等效声级作为样本输出 ,反复训练网络 ,通过增加隐含层节点数、改进算法 ,以降低误差 ,缩短训练时间。