Learning Machines Applied to Potential Forest Distribution

The clearing of forests to obtain land for pasture and agriculture and the replacement of autochthonous species by other faster-growing varieties of trees for timber have both led to the loss of vast areas of forest worldwide. At present, many developed countries are attempting to reverse these effects, establishing policies for the restoration of older woodland systems. Reforestation is a complex matter, planned and carried out by experts who need objective information regarding the type of forest that can be sustained in each area. This information is obtained by drawing up feasibility models constructed using statistical methods that make use of the information provided by morphological and environmental variables (height, gradient, rainfall, etc.) that partially condition the presence or absence of a specific kind of forestation in an area. The aim of this work is to construct a set of feasibility models for woodland located in the basin of the River Liébana (NW Spain), to serve as a support tool for the experts entrusted with carrying out the reforestation project. The techniques used are multilayer perceptron neural networks and support vector machines. Their results will be compared to the results obtained by traditional techniques (such as discriminant analysis and logistic regression) by measuring the degree of fit between each model and the existing distribution of woodlands. The interpretation and problems of the feasibility models are commented on in the Discussion section.     

DEVELOPMENT OF LARGE SCALE GRIDDED RIVER NETWORKS FROM VECTOR STREAM DATA1

ABSTRACT: The Network Tracing Method (NTM) has been developed to determine gridded coarse river networks for modeling large hydrologic systems. For a coarse resolution grid, the NTM determines the downstream cell of each cell and the distance along the actual meandering flow paths between them. Unlike previously developed methods, the NTM uses fine resolution vector river networks as the source of information of the flow patterns rather than digital elevation models. The main advantage of using vector river networks as input is that they capture the hydrologic terrain features better than topographic data do, particularly in areas of low topographic relief. The NTM was applied to South America with a grid resolution of 1 degree by 1 degree and to the globe with a resolution of 2.815 degrees by 2.8125 degrees. Overall, the method captured the flow patterns well. Generated digital river networks and drainage divides showed minor disagreement with those obtained from existing maps, and most of them were consistent with the resolution of the coarse river network. The majority of estimated basin areas were also close to documented values. River lengths calculated with the NTM, however, were consistently underpredicted.     

人工神经元网络辅助酚类化合物构效关系研究

分子连接性指数（^tXv）与在正辛醇和水之间的分配系数logP是反映化合物性能的重要结构参数。本文计算了酚类化合物的分子连接性指数及分配系数,运用新型的模式识别方法──人工神经元网络对酚类化合物的构效关系进行了研究。所得结果优于逐步判别法,对于预测未知化合物的毒性,具有重要意义。     

潮汐河网调水工程的环境与经济效益分析——以北江、西江调水工程为例

本文针对潮汐河网的基本特点,提出了运用非恒定流水环境实时数学模型和“影子工程法"对一系列引水方案的环境与经济效益进行定量的预测、分析和比较。该方法对潮汐河网地区水务工程的环境与费用效益分析研究具有普遍的参考意义。文中的算例为广州市市政洪水与污水整治战略决策提供了科学的依据。      

潮汐河网可降解有机物降解系数研究

潮汐河网具有其独特的水物理及水化学特性。本文采用珠江三角洲河网区域20个水样的BOD降解实验数据,并用BOD反应一级动力学方程进行数据拟合,结果表明,潮汐河网水体BOD降解过程的延滞期消失,远离污染源河段河水的BOD降解过程两阶段区分不明显,降解过程符合BOD反应一级动力学方程。此外,潮汐河网河水的降解系数随时间的变化过程是随机的。      

BP及RBF人工神经元网络对臭氧生物活性炭水处理系统建模的比较

运用BP和RBF人工神经元网络建立臭氧生物活性炭系统模型,考察了两个网络对水处理系统建模的适应性。研究表明,BP和RBF人工神经元网络的臭氧生物活性炭系统模型准确地描述了系统影响因素的关系,可以求出系统中臭氧的经济投量;用BP人工神经元网络建立水处理系统模型,泛化能力好,但逼近速度较慢;运用RBF人工神经元网络建模,泛化能力较差,但逼近速度快。该项研究克服了运用传统方法建模的不足,为实现水处理系统的优化设计提供了可行的途径。     

稠油热采井管柱失效动态风险分析研究

为保证稠油热采井筒管柱的安全运行,采用故障树与贝叶斯网络相结合的方法开展稠油热采井管柱失效风险分析,应用故障树分析方法定性识别管柱失效诱因,建立稠油热采井管柱失效场景演化模型,并将故障树转化为贝叶斯网络模型,应用贝叶斯网络对管柱失效风险进行定量分析,得出热采井管柱失效的关键致因和管柱失效的动态失效概率,以新疆油田某区块热采井的管柱失效为对象进行了案例分析。研究结果表明,该方法可为稠油热采井管柱失效风险分析及井筒完整性管理提供理论支撑和工程设计参考。     

Scenario deduction of Natech accident based on dynamic Bayesian network: A case study of landslide accident in a liquor storage tank area in Guizhou Province,China

Industrial technical accidents caused by natural disasters are defined as Natech accidents, such as earthquakes and landslides, which can cause tremendous damage to industrial storage tanks, and lead to accidental leakage and even serious fire and explosion accidents. In this study, a landslide-induced storage tank accident model under earthquake disasters was proposed, and the relationship between landslide mass impact and target impact resistance was taken into account. Also, tank failure and the formation of the pool fire were considered to be the consequences of the Natech accident. Through scenario deduction, the dynamic process of landslide Natech was transformed qualitatively into a disaster chain network diagram composed of a scenario state, a disaster-causing factor and emergency management. The Bayesian network was used to learn and deduce the parameters of the network diagram, and in this process, the prior probability and conditional probability of nodes were obtained primarily by Monte Carlo simulation, and by an improved expert scoring method based on the fuzzy set theory. Through visualization software, the sensitivity analysis of landslide Natech was achieved. Finally, a case study of a liquor storage tank area in Guizhou Province, China was carried out, and the results show that a large amount of hazardous material leakage caused by buckling is key to the formation of pool fire accidents, and several prevention measures for earthquake-induced landslide Natech was proposed according to the sensitivity analysis.     

Reliability evaluation of fire alarm systems using dynamic Bayesian networks and fuzzy fault tree analysis

In almost all industries, fire alarm systems play a vital role in the reducing the risks associated with fires and damages. Therefore, it is necessary to evaluate their reliability and performance in emergency situations. The present study aimed to use fault tree analysis (FTA) to determine the root causes involved in the failure of fire alarm systems, to use Fuzzy set theory and expert elicitation to determine relative probabilities, and finally, to evaluate the reliability of a fire alarm system using dynamic Bayesian networks (BNs) during a thirty-six months period. A total of 29 basic events were detected from the FT. The reliability of the fire alarm system was estimated at 0.954 according to the FT and 0.957 according to conventional BNs. The reliability of the fire alarm system after 36 months was estimated at 0.375 according to dynamic BNs. All the events involved in the failure of fire alarm systems were drawn in the fault tree diagram. The results indicate that remodeling of these systems and simultaneous construction activities are the most important factors in the failure of the fire alarm system. System reliability can also be increased to 0.965 by providing preventive and control measures to reduce the probability of critical events.