基于中心定位算子的遗传算法

针对遗传算法局部搜索能力弱,求解精度不高的缺陷提出了一种中心定位算子.在进化一定代数(T)后,选择最优的若干个(N)染色体基因来计算中心定位算子,从而确定与中心定位算子相同的基因位,并且在以后的交叉、变异操作中,都不让相同的基因位参与.随着算法的进行,染色体相同的基因位逐渐全部地被确定下来.其次,通过与小生境技术的局部搜索能力算法的结合,提高了该算子的全局优化能力;最后,通过几个非常容易陷入局部最优的测试函数测试表明几乎所有的峰值都得到了理论值.     

基于自适应小生境混合遗传算法的说话人识别

为了解决传统高斯混合模型(Gaussian Mixture Model,GMM)对初值敏感,在实际训练中极易得到局部最优参数的问题,本文提出了一种GMM参数优化的新方法.将小生境技术与最大似然估计融入到遗传训练过程,形成了一种新的混合算法,缓解了遗传算法产生的"早熟"现象,提高了算法的局部搜索能力.采用自适应策略来控制交叉和变异算子,同时在适应度评价中融入了其他用户的区分性信息,提高了模型的分类精度,增强了GMM的泛化能力.实验表明,与传统和改进的两种方法相比,本文的方法都可以得到更优的模型参数,使得系统的识别率进一步提高.     

基于性能驱动的船载稳定平台尺度参数优化设计

为提升船载稳定平台的运动学性能,针对■并联机构的尺度参数优化问题,以机构的工作空间体积和全域力传递率为综合评价指标,采用小生境遗传算法优化得到稳定平台的最佳几何构形。具体地,采用数值法与解析法相结合的方式判断支链长度、关节转角、奇异位形等约束条件的生效情况,求解出并联稳定平台的工作空间;基于力雅可比矩阵逆矩阵的最小奇异值定义机构的局部力传递性能,以工作空间内局部力传递率的平均值作为全域力传递性能评价指标;以工作空间体积和全域力传递率的加权和为优化目标,采用小生境自适应遗传算法完成优化求解,获得最优尺度参数。与初始构形的性能对比分析表明,优化构形在力传递性能方面有35%的提升,具有更好的综合运动学性能。制作试验样机并完成相关实验,验证了所提尺度参数优化方法的有效性。最后探讨了多目标优化过程中不同的权重系数取值对优化结果的作用规律,发现选用均衡的权重可获得更佳的综合性能。     

Unrestricted ramping rates and long‐term trends in the food web metrics of a boreal river

Long‐term monitoring of the food web of a regulated hydropeaking river was conducted to assess if previously documented effects of changing ramping rates (RRs) were maintained with the addition of 6 years of data. Using carbon and nitrogen stable isotope analyses, we hypothesized that: (1) macroinvertebrates and fish inhabiting areas below peaking hydrodams would be higher in δ¹⁵N and lower in δ¹³C due to increased flow velocity and the influence of light respired dissolved inorganic carbon, relative to those sampled from areas with a natural flow regime; (2) the increase in δ¹⁵N of macroinvertebrates would lead a shorter food web length in the regulated river, but δ¹³C and niche width would be similar between the restricted and unrestricted RR periods (i.e., the BACI analysis); and (3) isotopic metrics (e.g., δ¹³C, δ¹⁵N, niche width [SEA_B], and food chain length [Δ¹⁵N]) would correlate with variations in flow characteristics through time. Consistent with previous analysis conducted over a shorter time period, a shift toward higher δ¹⁵N values was observed for both fish and invertebrates, but, contrarily, only invertebrates (not fish) had a lower δ¹³C value downstream of the dam. Over the long term, the before‐after‐control‐impact analysis found no effect of RRs on any of the food web metrics, implying that the change in operation did not affect the river food web. However, analysis of the time series data indicated that flow metrics and trophic metrics were often correlated, including a negative effect of RR (invertebrates) and discharge (fish) on food chain length. This study illustrates the difficulty in detecting changes in food web structure and function under changing flow regime influenced by natural and anthropogenic effects. As such, this study highlights the need for considering large spatial and temporal scales to differentiate between confounding effects of climate, natural variability, and altered flow regimes on food webs in regulated rivers.     

基于均分法的小生境遗传算法

为了避免遗传算法种群中个体过早陷入局部最小,在以往随机初始种群的基础上提出一种均分法,使得初始种群随机平均地分为若干个子种群,形成小生境,这样既维持了种群的多样性,也使得种群中的个体不会过早出现早熟现象,更提高了算法的收敛速度.同时采用了自适应技术控制交叉和变异的概率,使得算法能更快速地找到最优解.仿真结果表明,与传统的遗传算法优化RBF网络相比较,新算法的迭代次数更少,精度更高,大大提高了收敛速度.     

一种改进的人工蜂群算法

从经典人工蜂群算法机制出发,针对原始算法在初始种群构造、子种群分组、步长更新和种群淘汰方面的不足进行了改进.新算法运用均匀设计理论构造初始种群,提出了一种种群交叉的Z型分组方法,设计了一种对数函数自适应步长代替原来的随机步长,引入了小生境技术及时淘汰陷入局部最优的个体.实验结果表明,改进后的算法有效地解决了人工蜂群算法早熟收敛、搜索速度较慢等问题,并提高了解的精度.     

Cruise missile multiple routes planning based on hybrid particle swarm optimization

In order to solve cruise missile route planning problem for low-altitude penetration,a hybrid particle swarm optimization( HPSO) algorithm is proposed. Firstly,K-means clustering algorithm is applied to divide the particle swarm into multiple isolated sub-populations,then niche algorithm is adopted to make all particles independently search for optimal values in their own sub-populations. Finally simulated annealing( SA) algorithm is introduced to avoid the weakness of PSO algorithm,which can easily be trapped into the local optimum in the search process. The optimal value obtained by every sub-population search corresponds to an optimal route,multiple different optimal routes are provided for cruise missile. Simulation results show that the HPSO algorithm has a fast convergence rate,and the planned routes have flat ballisticpaths and short ranges which meet the lowaltitude penetration requirements.     

基于生态位理论的黄河流域用水结构分析

为分析和探究黄河流域用水结构的变化特征,科学支撑流域水资源合理配置与调度,保障流域可持续发展,以黄河流域为研究对象,基于生态位理论构建了用水结构生态位及其熵值模型,分析了流域及流域内主要省区用水结构演变趋势和特征,并与发达国家用水特征作比较。在此基础上提出了相关建议。结果表明:黄河流域的农业用水生态位(0.7～0.78)最大但呈现波动减少趋势,工业用水生态位(0.12～0.15)次之且变化趋势为波动且有所下降,生活用水生态位(0.06～0.08)与城镇公共及生态用水生态位(0.03～0.09)相对较少但皆有不同程度的增长趋势。各省区的生态位熵值差异较大,表明流域及各省区间用水结构差异较大。统计年间宁夏、内蒙古及山东的农业用水占有率高于全流域,陕西、山西、河南及甘肃的工业用水占有率高于全流域,各省区城镇公共及生态用水的占有率变化剧烈,陕西、山西及河南的生活用水占有率高于全流域。相较于黄河流域,美国的工业用水(0.42～0.52)在用水结构中的优势更加突出,农业用水(0.31～0.44)相对处于劣势地位;两个区域生活用水在用水结构中的地位相对其他的用水类型更为接近。     

基于优育子群迁徙策略的多模态遗传算法研究

通过分析小生境遗传算法(NGA)的不足,提出了用于优化计算时多模态遗传算法的完全收敛性问题.结合传统梯度进化技术,设计了基于优育子群迁徙策略的遗传算法(MBGA),并分析了其运行机理和完全收敛特性.采用新算法与NGA进行了对比仿真实验,不仅验证了所提算法的完全收敛性结论,同时验证了其求解多模态问题的有效性、完全收敛能力及其收敛的稳定性.     

丁酸型产氢-产酸发酵细菌pH生态位探讨

本试验结果认为,细菌在3 86,或4 5>PH>5 3生态位理论。分析原因在于,试验过程中的环境因素———C/N比的降低,氮源物质浓度的提高,相应提高了微生物的合成代谢水平,并且使得细菌发酵过程在pH值较低的环境中,向合成代谢水平较高的丁酸型发酵转变。细菌发生丁酸型发酵是在环境内多种环境因子协同作用下进行的。该试验结果拓宽了前期理论研究中得到的丁酸型发酵生态位范围,为今后相关的理论研究及实际生产提供了理论依据。