基于小波网络与模糊神经网络的非线性系统辨识方法的研究

非线性系统辨识是现代控制理论的一个重要分支，是人们认识非线性系统的理论基础。今年来，利用神经网络辨一识非线性系统，是非线性系统辨识理论i}jf究的热点，现己形成了比较完善的理论体系。本文以某种反坦克导弹武器系统的弹道轨迹辨识为背景，在研究神经网络的辨识理论基础上，更深入地研究了纂于小波网络与模糊神经网络的非线性系统辨识的方法。一方面完善了现有的理沦，另佩方面提出了几种新的算法，主要内容如下：(1)提出了一种基于输入变量重要性指标的神经网络输入变量选择算法，对候选输入变量的重要性进行降序排序，引入一种正规化准则函数作为性能指标，最终确定同时具有良好的逼近能力和泛化能力的输入组合。该算法同基于宽度优先的搜索树算法相比，效率明佩Ul-Ul-提高。(2)基于小波多分辨率分析理论，引入非正交的小波基，从低尺度到高尺度逐渐逼近，L2(Rd)上的非线性连续函数，并且引用递推思想对辨识的偏差再进行多分辨率辨识，这样既降低了对正交小波基的要求，又能以较高的精度来辨识。（3）针对高维非线性系统，提出一种基于正交化思想的小波网络的参数和结构辨识算法。提出递推正交最小二乘法RGLS实现对小波向量矩阵的正交分 解，同传统的CGS, MGS算法相比，ROLE算法计算量小，可实现权值参数的在线辨识。利用ROLS算法的结果，采用从复杂到简单及从简单到复杂两种思想进行结构辨识，结构辨识算法是在上三角矩阵上进行的，而不是在小波向量矩阵上进行的，且在上述两种结构辨识算法中都不需要 重新训练网络，计碑算量小，运行效率高。本文所提的小波网络辨识方法，改变以往小波网络受限于低维非线性系统的状况。（4）针对基于聚类的模糊神经网络对维数不敏感，国内外很多研究集中到聚类算法上。本文完善了Ching－Chang Wong等1999年，提出的一种新聚类算一法，并根据Brouwer不动点定理给出了算法收敛的理论证明，该聚类算法同其他聚类算法相比不需预先知道聚类数目及初始中心点的设置，且．有较好的聚类效果。在新聚类算法的基础上，建立一阶TSK模糊神经网络模型，进行非线性系统辨识，收到很好的辨识效果。（5）对Ching－Hung Lee等2000年，提出的一种递归型模糊神经网络，引入遗传算法来完善该网络的学习算法。利用新聚类算法的结果，对初始种群的个体进行初始化，避免了通常采用遗传算法由于初始值选择不当而产生的早熟问题。对种群中的个体采用实数编码，实现网络参数的全局寻优。

New progress in R&D of lower olefin synthesis

This paper gives a brief review of R&D researches for light olefin synthesis directly and indirectly from synthesis gas in the Dalian Institute of Chemical Physics (DICP). The first pilot plant test was on methanol to olefin (MTO) reaction and was finished in 1993, which was based on ZSM-5-type catalyst and fixed bed reaction. In the meantime, a new indirect method designated as SDTO (syngas via dimethylether to olefin) was proposed. In this process, metal-acid bifunctional catalyst was applied for synthesis gas to dimethylether(DME) reaction, and modified SAPO-34 catalyst that was synthesized by a new low-cost method with optimal crystal size was used to convert DME to light olefin on a fluidized bed reactor. The pilot plant test on SDTO was performed and finished in 1995. Evaluation of the pilot plant data showed that 190-200 g of DME were yielded by single-pass for each standard cubic meter of synthesis gas. For the second reaction, 1.880 tons of DME or 2.615 tons of methanol produced 1 ton of light olefins, which constitutes of 0.533 ton of ethylene, 0.349 ton of propylene and 0.118 ton of butene. DICP also paid some attention on direct conversion of synthesis gas to light olefins. A semi-pilot plant test (catalyst 1.8 1) was finished in 1995 with a CO conversion > 70% and a C(2)(=)-C(4)(=) olefin selectivity 71-74% in 1000 h. (C) 2000 Published by Elsevier Science B.V. All rights reserved.

D���人格与冠心病患者身心健康的关系

D���人格（指经历消极情感和社交压抑的综合行为模式）作为冠心病的心理危险因素，得到了广泛的关注。本研究在中国人群中首次探索D���人格与冠心病的关系。作者建构了心理、行为、生理、医学相结合的综合研究框架，采用问卷调查法、临床实验法以及访谈法，考察D���人格对冠心病患者身心健康的影响，以及可能的行为和生理机制。结果表明： 1．D���人格量表中文版具有良好的信度和效度。根据消极情感≧10且社交压抑≧10的划分标准，D���人格在我国冠心病人群和健康人群中的分布比例均为31％。 2．A型人格与D���人格的共同点在于消极情感，区别在于D���人格具有特殊的社交压抑结构。我国冠心病患者中A型人格比例（58％）高于健康人群（43％）。 3．A型人格与冠心病的病情（即冠状动脉狭窄程度）、患者的自评身心健康均无关。D���人格与冠心病的病情无关，但D���人格能够独立地预测自评身心健康的下降，并且不依赖于疾病状况（患病或健康）。 4．D���人格患者在应对疾病的过程中，更多地采用屈服，更少地采用面对。疾病应对方式对D���人格与患者自评身心健康的关系起中介作用。 5．在冠状动脉造影的应激条件下，冠心病患者的D���人格总分与应激下心血管反应无关，但男性患者的消极情感、社交压抑能够预测应激下舒张压反应。 此外，本研究的结果具有重要的临床价值，比如采用D���人格量表中文版鉴别高危人群以提高他们的身心健康水平，在进行D���人格干预时要关注行为方式，在对冠心病进行预防和治疗时要注意心理生理反应的性别差异等。

A型、D���人格影响冠心病预后的心理生理机制

Coronary heart disease （CHD���is a common cardiovascular disease in the elderly, is also a typical psychosomatic disease. Personality factors are very important in many psychological factors impacting on the prognosis of patients with CHD. The most influential personality factors to CHD are Type A and Type D personality. The previous research has shown that although Type A personality increased the prevalence of CHD, it cannot predict the development and prognosis after diagnosis. In contradict, Type D personality can predict prognosis. There is still no clinic-based or theory-based answer to the question: Why Type A personality cannot predict the outcome while Type D personality could predict the prognosis independently. The current research conducted a systematic investigation to the above question, which included one comparison study between CHD patients and control group, and four studies on reaction experiment and answered the question: why Type A personality cannot predict whereas Type D personality could effectively predict prognosis of CHD. The findings of the current research were: Type A and Type D personality influence CHD prognosis through different psychological mechanisms: both dimensions of Type D personality have direct influence on social support, whereas neither dimensions of Type A personality related to social support, directly of indirectly. Negative affection component of Type D personality significantly related to anxiety and depression, Social repression significantly related to anxiety but not depression. Both dimensions of Type A personality significantly related to anxiety but not depression. Neither under rest or diaphragmatic breathing conditions, Type A personality had no significant influence on vestibular autonomic reaction among healthy young males. Neither Type A nor Type D personality had significant influence on vestibular autonomic reaction among old CHD patients under rest condition. Type D personality predicted lower sympathetic excitation under rest condition, and lower cardiac vagal tone under diaphragmatic breathing condition among healthy young males. When actively reacted to stimuli (math calculation) under rest condition, Type A personality increased sympathetic excitation among healthy young males. When actively reacted to stimuli (math calculation) under diaphragmatic breathing condition, Type A personality increased cardiac vagal tone among the same group of subjects. When actively reacted to stimuli under neither condition, Type D personality showed no significant influence on vestibular autonomic reaction among young males. When passively reacted to stimuli under neither condition, Type A personality showed no significant influence on vestibular autonomic reaction among young males. When passively stimulated followed rest, Type D personality increased sympathetic excitation and decreased cardiac vagal tone among young males. When passively stimulated followed diaphragmatic breathing, Typed showed no significant influence on vestibular autonomic reaction among young males. The above results indicated that Type A and Type D personalities had different psychological mechanisms to the outcome of CHD treatment: neither dimensions of Type A personality had direct or indirect effects on social support; both dimensions of Type D personality had direct and indirect effects on social support. Negative affection component of Type D personality significantly related to anxiety and depression, Social repression significantly related to anxiety but not depression. Both dimensions of Type A personality significantly related to anxiety but not depression. Social support positively related to the outcome after CHD treatment. The biological mechanisms of Type A and Type B personality to CHD prognosis differed in the following ways: Type A personality increased sympathetic excitation when actively stimulated, but had no influence when passively stimulated among young male subjects. When passively stimulated after rest, Type D personality predicted high sympathetic excitation and low cardiac vagal tone among young males, but not vestibular autonomic reaction among young males. Key words: Type A personality, Type D personality, Coronary Heart Disease (CHD), Prognosis, Psychobiological Mechanisms