基于PSO-SVM算法的梯级泵站管道振动响应预测

泵站管道振动响应信号实测比较困难,为实现利用较少机组数据预测管道振动状况,提出基于粒子群(particle swarm optimization,PSO)的支持向量机(support vector machine,SVM)预测方法。利用粒子群全局跟踪搜索算法优化SVM核函数和惩罚因子,弱化SVM参数优化不足导致预测精度低的问题。以景电梯级二期3泵站2号管道为研究对象,基于机组和管道的振动实测数据,首先利用频谱分析和数理统计方法确定管道振动的振源贡献率,并计算机组和管道振动相关系数,确定机组和管道之间的强耦合关系。然后建立泵站管道振动的PSO-SVM预测模型,选取机组不同时段振动实测数据作为输入因子,相应时段管道振动数据作为输出因子进行训练和振动预测,并将管道振动预测结果与BP神经网络预测结果进行对比。与BP网络神经预测结果相比,该方法预测结果与实测值吻合度高,其平均相对误差最大为6.8%,根均方误差最大为0.261,预测精度更高。能够有效实现管道的振动响应预测,从而达到管道实时在线安全运行监测的目的。     

全站仪测量立木胸径树高及材积的误差分析

立木材积表是常用的森林调查数表,胸径、树高和材积的测量精度直接影响到编制材积表的精度,该文以全站仪无损测量立木的原理和误差传播理论为基础,推导了测算胸径、树高、材积误差的数学模型,研究了各立木因子间的相关性及其误差变化规律。结果表明：树干各分段高度与分段直径间存在弱相关性,树干总材积的误差受各分段材积的方差和相邻两段材积间的协方差影响。全站仪立木因子测量理论误差材积大于树高和胸径,其中胸径、树高和材积的平均相对误差分别为0.070%、0.023%和0.235%,说明全站仪无损测量立木的理论精度均远高于不同目标的林业调查及编制材积表的精度要求,对大范围的林业生产实践有着现实意义。     

基于机器学习理论的土壤侵蚀模型构建

土壤侵蚀一直是环境问题中的重点和难点。由于影响土壤侵蚀的因素众多,传统的预测模型存在数据获取困难、适用范围小、研究周期长等不足,使得对土壤侵蚀的预测无法做到快速、便捷。支持向量机(Support Vector Machine,SVM)是机器学习中的一个重要模型,具有非线性映射、自我学习能力、全局最小值、对输入数据变化不敏感等优点,在建立土壤侵蚀量相关性预测模型方面较传统预测模型具有更强的优势。本研究应用浙江省诸暨市浦阳江水文站的降雨数据,利用ArcGIS地理信息系统确定水文站上游流域为研究区域。以降雨量、研究区域地理数据维度(包括坡度数据、坡长数据、土壤信息、土地利用类型)作为影响因子,输入支持向量机模型,进行流域内土壤侵蚀量预测。将水文站土壤侵蚀量实测数据作为对照值,用模型输出值检验,从而在取值范围内选择出模型最优的参数组。用影响因子数据和土壤侵蚀量数据对使用最优参数的模型进行检验,模型的预测准确率最高达到75%。其中,降雨量对土壤侵蚀量的影响最大,降雨量单因子预测准确率在70%以上,其余因子预测准确率在3.5%左右。最终得到一个土壤侵蚀量相关性预测模型,通过水文站降雨数据以及地理信息,即可预测当地土壤侵蚀量,准确率达到75%。     

树冠遮挡条件下全站仪测量树高及材积方法

树高、材积是森林调查中最重要的测树因子,对森林资源调查和林业科学研究有着重要的意义。为了解决林区遮挡条件下无法精确量测树高进而无法确定单木材积的问题,该研究运用全站仪进行森林样地外业数据采集,对内蒙古赤峰市旺业甸林场的188株落叶松和146株白桦进行量测。对每一株立木分别量测胸径、最高可观测处干径和该处距地面的距离以及这2处之间的中间位置的干径和该处距地面的距离,然后分别利用相似三角形法、简单生长经验方程法和孔兹干曲线方程法估算出树高及材积。试验的最高可观测的位置做了以下4种情况的假设:1/3树高、1/2树高、2/3树高和9/10树高。对这4种情况分别求得树高及材积,最后分别对其进行精度分析,得到随着最高可观测部分的增高以及观测点数的增加,测算出的树高及材积会越来越精准。其中运用生长经验方程,测算树高和材积的精度都能达到70%~80%;运用孔兹干曲线法测算树高和材积的精度都能达到70%~90%。该方法可用于林区遮挡条件下单株立木材积的测量,对实际生产有一定的参考意义。     

湿地松人工林地径与胸径树高模型的研究

应用样木资料,采用多方程分析对比法建立胸径与地径及树高与地径的相关数学模型,所建立的模型平均相对误差低于2％,预估效果理想,从而为通过地径的测定,应用二元立木材积表测定湿地松人工林被伐木材积提供了科学依据,在实践上有应用价值。     

基于自加速遗传粒子群算法的半封闭式温室能耗预测

针对半封闭式温室环境参数众多且难以测量的问题,提出了一种机理建模与系统辨识建模相结合的温室能耗建模方法。采用自加速遗传粒子群算法(self-accelerating hybrid algorithm of particle swarm optimization and genetic algorithm,SPSO-GA)对温室物理模型中难以确定的参数进行辨识,建立半封闭式温室能耗预测模型。根据上海半封闭式玻璃试验温室的气象数据和测量的能耗值,分别采用遗传算法(genetic algorithm,GA)、粒子群算法(PSO,particle swarm optimization)和SPSO-GA进行参数辨识与能耗预测比较分析。采用SPSO-GA获得的温室能耗预测结果与实测数据的相对误差为1.4%,分别比GA和PSO减少了2.9%和13.7%。根据日太阳光照辐射总量、室外日均温度2个参数及相应的变化曲线,预测的温室能耗值精确度大于86%。试验与模拟结果验证了基于SPSO-GA的温室能耗预测模型有效,可为半封闭式温室能量负载设计、管理和控制提供理论依据。     

垄沟集雨系统Laio土壤水分动态随机模型参数敏感性分析及优化

水文模型参数的敏感性分析、优化和验证对提高模型计算精度和效率具有重要意义。为探讨Laio土壤水分动态随机模型(Laio模型)各参数在垄沟集雨系统的敏感性,同时,确定参数优化和模型验证的最佳方案,本文结合多因素敏感性分析法以及改进单纯形法(ISM)、粒子群优化算法(PSO)和混合粒子群优化算法(HPSO),利用中国气象局定西干旱气象与生态环境试验基地2012—2013年垄沟集雨燕麦生长季降雨、径流和土壤水分等实测数据,对垄沟集雨系统Laio模型的13个参数进行敏感性分析、优化和验证。结果表明,平均降水量α和凋萎系数s_w对土壤水分概率密度函数p(s)最敏感,p(s)对参数α的敏感性在低土壤含水率下更明显,对参数s_w的敏感性在高土壤含水率下更明显;3种算法(ISM、PSO和HPSO)的优化参数值均能对垄沟集雨系统土壤水分概率密度函数进行较好模拟,峰值(CPV)、峰值位置(PP)和95%置信区间(CI95%)实测值与模拟值的相对误差均小于10%,CM指数均大于0.5;同时,HPSO算法优化参数的模拟效果和收敛速度均显著优于PSO算法和ISM算法,能较显著克服ISM算法和PSO算法存在的缺陷。HPSO算法可作为垄沟集雨系统土壤水分动态随机模型参数优化的待选方案。     

改进支持向量机的股票价格预测研究

为了提高股票价格预测精度,提出一种改进支持向量机的股票价格预测模型。该模型利用粒子群算法的全局寻优能力对支持向量机参数进行优化,以提高股票价格的预测精度,采用具体股票价格数据对模型性能进行测试。结果表明,改进支持向量机能够对股票价变化趋势进行预测,是一种有效、高精度的股票价格预测模型。     

基于伪氨基酸和支持向量机的蛋白质亚细胞定位预测

用电子—离子伪势能(E IIP)对蛋白质序列数字化,经离散傅立叶变换(DFT)后,取5个最高幅值对应的频率和20种氨基酸在序列中所占的百分比组成伪氨基酸。用支持向量机(SVM)方法得到分类的模型,并用几个标准的测试方法测试模型的性能。自身一致性测试和Jackkn ife测试均取得高的预测准确率,独立数据集测试的准确率超过80%。和之前报道的方法相比,本方法具有较高的预测准确率。     

基于PSO-BP的甘蔗施肥监控系统设计与试验

针对甘蔗横向种植机的施肥机构由于肥料潮湿结块易堵塞等问题,该研究对施肥机构进行电液传动与控制改造,构建了一套基于粒子群（Particle swarm optimization,PSO）-前反馈（Back propagation,BP）神经网络预测的施肥监控系统。以施肥马达的压力、转速及肥料箱中肥料量为输入参数,将施肥机构的工作状态（空载状态、正常状态、重载状态、堵塞状态）作为输出,通过BP算法建立输入与输出之间的映射关系,并用PSO算法优化BP算法的权值与阈值,相比未优化BP算法,优化后的工作状态预测准确率由97%提高到99%。以识别施肥机构工作状态响应准确率以及重载状态下堵塞预防概率为试验指标进行车间试验,结果表明：工作状态响应识别准确率为89%;重载状态下,控制系统控制施肥马达正反转并消除堵塞的概率为87.5%。在田间试验中,监控系统能准确预测施肥机构的重载状态并自动执行防堵控制指令,没有出现堵塞故障。该施肥防堵塞监控系统无需上位机,能够满足复杂多变工况下施肥机构的工况预测及防堵控制要求,可为其他施肥机构的自动化改造提供参考。     

Pedologic and geomorphic impacts of a tornado blowdown event in a mixed pine-hardwood forest

Biomechanical effects of trees on soils and surface processes may be extensive in forest environments. Two blowdown sites caused by a November 2005 tornado in the Ouachita National Forest, Arkansas allowed a case study examination of bioturbation associated with a specific forest blowdown event, as well as detailed examination of relationships between tree root systems, soils, and underlying bedrock. The sites occur within mixed shortleaf pine and hardwood forests. More than 95% of trees in the severe blowdown areas were either uprooted or suffered trunk break, with uprooting more common than breakage. Within the most heavily damaged areas all uprooted trees were pines, while all trees left standing were hardwoods. Root wads of uprooted trees had a mean surface area of about 3 m² and volume of about 2 m³, though individual sizes were quite variable. Nearly 4% of the ground surface area was affected by uprootings, with a soil volume equivalent to a disturbance of the entire surface area to a depth of 2.4 cm. Tree size (as measured by diameter at breast height) was significantly related to the area and volume of root wads (R² = 0.55, 0.71, respectively), with volume of uprooted soil varying as diameter to the ~ 3 power, suggesting that the timing of blowdown events relative to tree age or growth stage significantly influences the area of disturbance and the mass and volume of material involved. In 93% of cases the roots of the uprooted trees contacted or penetrated the underlying bedrock, and in all those cases bedrock was quarried by uprooting. Only 11% of the tree throws showed evidence of general lateral root turning at the soil–bedrock interface; in most cases roots penetrated bedrock along joints. The propensity for tree roots to penetrate bedrock joints, facilitate weathering, and excavate bedrock during uprooting supports the idea that tree roots play a predominant role in locally deepening soils.     

Pedologic and geomorphic impacts of a tornado blowdown event in a mixed pine-hardwood forest

Biomechanical effects of trees on soils and surface processes may be extensive in forest environments. Two blowdown sites caused by a November 2005 tornado in the Ouachita National Forest, Arkansas allowed a case study examination of bioturbation associated with a specific forest blowdown event, as well as detailed examination of relationships between tree root systems, soils, and underlying bedrock. The sites occur within mixed shortleaf pine and hardwood forests. More than 95% of trees in the severe blowdown areas were either uprooted or suffered trunk break, with uprooting more common than breakage. Within the most heavily damaged areas all uprooted trees were pines, while all trees left standing were hardwoods. Root wads of uprooted trees had a mean surface area of about 3 m² and volume of about 2 m³, though individual sizes were quite variable. Nearly 4% of the ground surface area was affected by uprootings, with a soil volume equivalent to a disturbance of the entire surface area to a depth of 2.4 cm. Tree size (as measured by diameter at breast height) was significantly related to the area and volume of root wads (R² = 0.55, 0.71, respectively), with volume of uprooted soil varying as diameter to the ~ 3 power, suggesting that the timing of blowdown events relative to tree age or growth stage significantly influences the area of disturbance and the mass and volume of material involved. In 93% of cases the roots of the uprooted trees contacted or penetrated the underlying bedrock, and in all those cases bedrock was quarried by uprooting. Only 11% of the tree throws showed evidence of general lateral root turning at the soil–bedrock interface; in most cases roots penetrated bedrock along joints. The propensity for tree roots to penetrate bedrock joints, facilitate weathering, and excavate bedrock during uprooting supports the idea that tree roots play a predominant role in locally deepening soils.     

Tree influence on soil microbial community structure

Biological communities differ over time and in space, and in the forest these communities often vary according to trees and tree gaps, mediated by mechanisms that are likely to change over time and as a tree are removed. In this paper we ask the questions: What is the influence of individual trees on soil microbial community structure? Does the soil microbial community change in the short-term when a tree is removed, and does this change depend on the initial influence of the tree? We use phospholipid fatty acid (PLFA) analysis and a geostatistical approach to study effects of trees and tree removal (thinning) on soil microbial community structure in a young boreal Norway spruce (Picea abies) forest. An experiment was setup where half (four) of the included trees were cut and soil was collected prior to (t0) and one month after (t1) tree felling. The samples were collected along two perpendicular transects originating from each of the eight study trees. A tree influence index was calculated for each sample point from the distances to neighbouring trees, weighted by tree diameter. We found that individual trees are important in structuring the soil microbial community as microbial community structure responded to the gradient in tree influence. Also strong spatial structure was found corresponding to the patch structure induced by trees. Changes in microbial community structure before and after tree felling (t0 and t1) was found to differ significantly between felled and non-felled trees: samples from felled trees came to resemble samples with a low value of tree influence and samples from below non-felled trees came to resemble samples with a high value for tree influence. We thus found that soil microbial community structure in a boreal forest is spatially structured by the distribution of single trees, and that soil microbial community structure varies seasonally and is affected by tree removal, in an intricate manner that reflects the initial influence of trees.     

基于太赫兹光谱和支持向量机快速鉴别咖啡豆产地

结合太赫兹时域光谱技术和支持向量机对3种典型产地的咖啡豆进行了鉴别。选取埃塞俄比亚(Ethiopia)、哥斯达黎加(Costa Rica)以及印度尼西亚(Indonesia)3个产地咖啡豆样品进行压片处理,采用太赫兹透射模式获取样品的时域和频域光谱信号,并用主成分分析法对太赫兹频域光谱信号进行分析;构造了基于粒子群(partical swarm optimization,PSO)参数寻优的支持向量机(support vector machine,SVM)鉴别模型,模型对不同产地咖啡豆样品的综合识别正确率达到95%。试验结果表明,太赫兹作为新型的检测手段结合模式识别方法可用于咖啡豆的产地鉴别。该文为一类在太赫兹波段下没有明显特征吸收峰的农产品/食品安全检测和产地追溯研究提供了一种快速、准确的方法。     

牛姆林米槠天然林主要树种直径分布结构特征的研究

运用5种分布函数模型对福建永春县牛姆林米槠天然林主要树种的胸径和地径进行拟合,以筛选出拟合效果较好的分布函数模型,探索其林分直径结构特征。结果表明:1米槠林分直径分布结构应用正态分布模型拟合效果最好;2峭度、偏度和变动系数等特征值可用于描述天然林的生长特点;3正态分布参数可通过较易获取的林分直径实测值来预估,进而预测林分各径阶林木株数、蓄积量及其动态变化,为米槠天然林合理的经营管理提供科学的参考依据。     

Comparison of breeding bird and vegetation communities in primary and secondary forests of Great Smoky Mountains National Park

We compared breeding bird communities and vegetation characteristics at paired point locations in primary (undisturbed) and mature secondary forest (70-100 years old) sites in Great Smoky Mountains National Park, USA to understand how sites logged prior to creation of the park compare to undisturbed sites following 70 years of protection from human disturbance. We found that bird and vegetation communities are currently similar, but retain some differences in species composition. Rank abundance curves for primary and secondary forest bird communities showed very similar patterns of species dominance. Species composition was also similar on the two sites which shared 24 of the 25 most frequently recorded species. Nonetheless, comparisons of density estimates derived from distance sampling showed three bird species were more abundant on primary forest sites and that one bird species was significantly more abundant on secondary forest sites. Notably, comparisons based on raw counts (unadjusted for potential differences in detectability) produced somewhat different results. Analyses of vegetation samples for the paired sites also showed relative similarity, but with some differences between primary and secondary forests. Primary forest sites had more large trees (trees greater than 50 cm diameter at breast height) and late successional species. Primary forest sites had a denser tall shrub layer while secondary forest sites had a denser canopy layer. Nonetheless, tree species richness, basal area of live trees and number of standing snags did not differ between primary and secondary forest sites. Results indicate that breeding bird communities on sites within the park that were logged commercially 70 years ago are currently quite similar to bird communities on sites with no history of human disturbance. Similarities between the bird communities on previously disturbed and undisturbed sites in Great Smoky Mountains National Park may exceed those on more fragmented landscapes because large patches of primary forest, adjacent to commercially logged sites, remained in the park when it was established in 1935. These patches of primary forest may have served as source areas for commercially logged sites.