BP人工神经网络预测邻苯二甲酸酯光化学降解

将反向传播人工神经网络（BP－ANN）用于邻苯二甲酸酯类化合物的光化学降解的预测中，选取正交设计的试验点作为反向传播人工神经网络的训练集，实现对全实验域试验点的预测，并与实测的试验数据比较。结果表明反向传播人工神经网络（BP－ANN）具有良好的预测能力。     

人工神经元网络对水处理系统建模适应性的研究

针对水处理系统的特点和研究中的难点，引入人工神经元网络的理论和思想，提出了水处理系统的神经网络分析方法．通过建立基于ＢＰ人工神经元网络的臭氧生物活性炭系统的模型，具体地考察了人工神经元网络对水处理系统建模的适应性，使水处理系统的研究迈向智能化和控制化．     

壳聚糖衍生物-镁盐复配去除中性兰模拟废水色度的研究

研究了有机高分子絮凝剂NCTS-M对中性兰染料的絮凝性能。通过测量絮体的Zeta电位以及对絮体进行彩色电视显微扫描初步探讨了该絮凝过程的絮凝机理。结果表明，当投加量为30 mg/L，pH=6时，絮凝效果最明显，脱色率可达93.1%。其絮凝机理主要是压缩双电层以及高分子吸附架桥作用，以压缩双电层为基础。絮凝剂中镁离子在中性条件下对絮凝过程也起到了较强的助凝作用。     

Using lysosomal membrane stability of haemocytes in Ruditapes philippinarum as a biomarker of cellular stress to assess contamination by caffeine, ibuprofen, carbamazepine and novobiocin

Although pharmaceuticals have been detected in the environment only in the range from ng/L to g/L, it has been demonstrated that they can adversely affect the health status of aquatic organisms. Lysosomal membrane stability (LMS) has previously been applied as an indicator of cellular well-being to determine health status in bivalve mussels. The objective of this study is to evaluate LMS in Ruditapes philippinarum haemolymph using the neutral red retention assay (NRRA). Clams were exposed in laboratory conditions to caffeine (0.1, 5, 15, 50 μg/L), ibuprofen (0.1, 5, 10, 50 μg/L), carbamazepine and novobiocin (both at 0.1, 1, 10, 50 μg/L) for 35 days. Results show a dose-dependent effect of the pharmaceuticals. The neutral red retention time measured at the end of the bioassay was significantly reduced by 50% after exposure to environmental concentrations (p < 0.05) (caffeine = 15 μg/L; ibuprofen = 10 μg/L; carbamazepine = 1 μg/L and novobiocin = 1 μg/L), compared to controls. Clams exposed to these pharmaceuticals were considered to present a diminished health status (retention time < 45 min), significantly worse than controls (96 min) (p < 0.05). The predicted no environmental effect concentration (PNEC) results showed that these pharmaceuticals are very toxic at the environmental concentrations tested. Measurement of the alteration of LMS has been found to be a sensitive technique that enables evaluation of the health status of clams after exposure to pharmaceuticals under laboratory conditions, thus representing a robust Tier-1 screening biomarker.     

Oxidative stress and DNA damages induced by cadmium accumulation

Experimental evidence shows that cadmium （Cd） could induce oxidative stress and then causes DNA damage in animal cells, however, whether such effect exists in plants is still unclear. In the present study, Vicia faba plants was exposed to 5 and 10 mg/L Cd for 4 d to investigate the distribution of Cd in plant, the metal effects on the cell lipids, antioxidative enzymes and DNA damages in leaves. Cd induced an increase in Cd concentrations in plants. An enhanced level of lipid peroxidation in leaves and an enhanced concentration of H2O2 in root tissues suggested that Cd caused oxidative stress in Vicia faba. Compared with control, Cd-induced enhancement in superoxide dismutase activity was significant at 5 mg/L than at 10 mg/kg in leaves, by contrast, catalase and peroxidaseactivities were significantly suppressed by Cd addition. DNA damage was detected by neutral/neutral, alkaline/neutral and alkaline/alkaline Comet assay. Increased levels of DNA damages induced by Cd occurred with reference to oxidative stress in leaves, therefore, oxidative stress induced by Cd accumulation in plants contributed to DNA damages and was possibly an important mechanism of Cd-phytotoxicity in Vicia faba plants.     

基于BP神经网络的上市公司CEO报酬实证研究

近年来的研究表明,CEO报酬不仅受公司业绩的影响,还受到其他诸多因素的影响,而且,CEO报酬与其决定因素之间往往存在着非线性关系。本文以2003-2005年沪深股市的A股上市公司为样本,采用BP神经网络对CEO总报酬、CEO年薪、CEO持股价值及其决定因素分别进行训练和学习,结果表明：（1）网络训练输出值与实际值的拟合度分别达到91.09%、97.23%和78.44%;（2）网络的预测能力相对于传统的线性回归模型分别提高了92.72%、92.08%和53.89%。因此,本文认为在分析和确定CEO报酬水平时引入神经网络模型是可行的。     

The structure of the fire fighting foam surfactant Forafac®1157 and its biological and photolytic transformation products

For several decades, perfluorooctane sulfonate (PFOS) has widely been used as a fluorinated surfactant in aqueous film forming foams used as hydrocarbon fuel fire extinguishers. Due to concerns regarding its environmental persistence and toxicological effects, PFOS has recently been replaced by novel fluorinated surfactants such as Forafac®1157, developed by the DuPont company. The major component of Forafac®1157 is a 6:2 fluorotelomer sulfonamide alkylbetaine (6:2 FTAB), and a link between the trade name and the exact chemical structure is presented here to the scientific community for the first time. In the present work, the structure of the 6:2 FTAB was elucidated by ¹H, ¹³C and ¹⁹F nuclear magnetic resonance spectroscopy and high-resolution mass spectrometry. Moreover, its major metabolites from blue mussel (Mytilus edulis) and turbot (Scophthalmus maximus) and its photolytic transformation products were identified. Contrary to what has earlier been observed for PFOS, the 6:2 FTAB was extensively metabolized by blue mussel and turbot exposed to Forafac®1157. The major metabolite was a deacetylated betaine species, from which mono- and di-demethylated metabolites also were formed. Another abundant metabolite was the 6:2 fluorotelomer sulfonamide. In another experiment, Forafac®1157 was subjected to UV-light induced photolysis. The experimental conditions aimed to simulate Arctic conditions and the deacetylated species was again the primary transformation product of 6:2 FTAB. A 6:2 fluorotelomer sulfonamide was also formed along with a non-identified transformation product. The environmental presence of most of the metabolites and transformation products was qualitatively demonstrated by analysis of soil samples taken in close proximity to an airport fire training facility.     

藻类水华对太湖梅梁湾不同粒径有机组分结构的影响

运用过滤和切向流超滤系统,分离得到2009年8月至2010年8月每月的太湖梅梁湾不同粒径水体有机物,测定了水体中颗粒态有机碳,高分子量和低分子量可溶解性有机碳的含量,并同步分析了叶绿素浓度,水体中各种有机物的碳氮比值和中性单糖含量.对不同形态有机碳浓度的比较发现颗粒态有机碳是太湖碳存在的主要形式.统计分析结果表明叶绿素浓度与颗粒态有机碳之间具有显著的相关性,说明浮游植物是颗粒态有机物的主要来源.此外高分子量可溶性有机物的碳氮比值和中性单糖含量均相对较高,这表明该有机碳的生物可利用性比其它形态的水体有机碳高.     

基于数据分解的AQI的CEEMD-Elman神经网络预测研究

针对Elman神经网络在预测空气质量指数（AQI）时易受到数据非平稳性的影响导致预测趋势良好但准确度较低的问题,提出以互补集合经验模态分解（Complementary Ensemble Empirical Mode Decomposition,CEEMD）为基础的CEEMD-Elman模型.应用CEEMD对AQI序列分解成不同时间尺度上的本征模态函数分量和剩余分量,进而首次将对非平稳的AQI序列的预测研究转化为对多个平稳的本征模态函数分量的研究.分别与Elman单一模型、EMD-Elman模型、BP单一模型及CEEMD-BP模型进行实验对比.结果表明：应用该方法建立的模型的均方误差、平均绝对误差和平均绝对百分比误差分别为4.80、0.71、1.84%,均小于其他模型结果;对应空气质量等级预报正确天数的频率为94.12%.该模型能有效的降低非平稳性对实验预测结果的影响,实现对空气质量等级的准确预报;该研究为进一步预测AQI的走向提供了有效依据,也为政府决策和管理部门制定空气污染控制提供了更充分的参考.     

煤矿水害风险量化评价方法

为提高煤矿防治水管理水平，预防和消除矿井水害，在层次分析法的基础上建立中性值作为参照对象对矿井水害风险进行实时评判的方法。根据《煤矿防治水细则》建立以矿井水文地质类型、矿井涌水量标准分数、突水预兆、采掘面位置、探水结果为准则层的层次结构模型，并对各评价指标赋权。依据制定的水害风险评价指标的评分细则和监测监控数据并结合其权重得到水害评价总得分。通过总得分与中性参照分数比较得出预测结果:水害评价总得分大于中性参照分数，证明水害的威胁小，分数越高越安全；反之则水害的威胁较大，分数越低越危险，这时需要加强防治水的力度，令评价分数管控大于中性参照分数。这种方法依赖于井下监测监控数据进行量化评价，能实时、客观、全面且准确地反映煤矿水害的风险情况。