深水平台施工过程自然风险定量评估

在分析现有风险理论的基础上,结合海洋施工实际,提出了一种对海洋施工自然风险进行定量评估的模型。根据目标系统各风险事件与影响因素之间存在高度非线性复杂映射的特点,提出了利用BP神经网络拟合方法,对主要风险事件发生概率进行量化的模型。针对传统的梯度下降优化算法收敛速度慢的特点,采用可避免计算Hesse矩阵的Levenberg-Marquardt算法来训练神经网络。这一模型可以模拟专家评价,并准确地按照专家的评定法则进行估算,具有一定的通用性。     

交通系统智能决策的算法研究

遗传算法是近年来迅速发展起来的一种全新的随机搜索与优化算法，但由于其自身固有的缺陷，通常优化过程的收敛速度较慢，局部搜索能力不足，而且算法稳定性较差。而蚁群算法广泛应用在旅行商问题计算中，目前是较好的求解最短路由问题的算法之一。就其自身来说有很多优点，如正反馈性、鲁棒性和智能性，但是在寻优过程中容易陷入局部最佳的缺点。针对上述情况，将遗产算法与蚁群算法相结合，用于实际交通系统寻找最优路径的问题中，并定义了目标函数，以路径可靠性和路径长度为优化目标，寻找最合适的救灾路线。最后通过实际计算结果的对比验证，说明了该方法的有效性。     

三峡茅坪溪沥青混凝土心墙堆石坝位移反演分析与安全预测

本文研究了基于综合应用人工神经网络和演化算法的位移反演分析方法,并将此方法应用于三峡茅坪溪沥青混凝土心墙堆石坝的变形反演分析中。以茅坪溪一期工程原型观测成果为依据,反演出能够正确反映坝体变形特性的邓肯-张EB模型参数,从而预测了施工期末和蓄水期末该坝的变形特性,并基于总应力法研究了心墙水力劈裂破坏发生的可能性,为茅坪溪堆石坝运行期的安全生产提供依据。研究结果表明,茅坪溪心墙堆石坝不会发生水力劈裂破坏,该坝蓄水后虽然水压上升,但心墙仍是安全的。     

人员位置固定的公共场所分区疏散研究

通过将分区疏散转化为分配问题(Allocation Problem),并采用启发式的A*优化算法,对人员位置固定的公共场所分区疏散进行研究。进一步利用基于元胞自动机模型的大型公共场所人员疏散行为模拟仿真系统,在充分考虑公共场所中每个人员的状态、人员之间以及人员与周围环境的相互作用的前提下,对疏散分区的效果进行模拟;通过与未分区的模拟结果相比较,可以认为,分区疏散有助于人员快速疏散,大大缩短整体避难时间,而且基于移动路径搜索的分区更加切合实际。     

A CODED ALGORITHM FOR CAPACITY EXPANSION OF A WATER QUALITY MANAGEMENT SYSTEM1

ABSTRACT: This paper describes a mathematical model, an algorithm and a computer program that were specially developed to study the problem of a water quality management system undergoing a rapidly increasing environmental stress. The model output will determine the locations, sizes and the timing of construction of new treatment plants plus an overall treatment plant operating policy so that environmental standards are maintained at a minimum cost. The model, as formulated, is a 0-1 mixed integer programming problem which is solved by decomposing it into a capital budgeting problem (solved by Little's branch and bound algorithm) and an operational policy problem (solved by linear programming). The coded algorithm (in FORTRAN 10) has been tested with a semi-realistic example.     

人工神经网络在水环境质量评价中的应用

为了将人工神经网络应用于水环境质量评价，应用了人工神经网络Ｂ—Ｐ算法，构造了水环境质量评价模型，该模型应用于实例评价结果表明，人工神经网络用于环境质量评价具有客观性，通用性和实用性。     

基于神经网络的洪水预报研究

人工神经网络通过神经元之间的相互作用来完成整个网络的信息处理,具有自学习和自适应等一系列优点,因而用它来进行洪水预报是可行的.对洪水预报问题,初步建立了基于神经网络的洪水预报系统,给出了应用实例.     

煤炭产业国际竞争力的遗传—投影寻踪评价

根据煤炭产业发展的实际状况,提出了煤炭产业竞争力评价的指标体系。采用探索性数据分析—投影寻踪法,结合遗传算法,建立了遗传—投影寻踪综合评价模型。以全球11个主要产煤国为对象进行了实证研究,验证了该方法的科学性和实践的可行性,提出提升我国煤炭产业竞争力的对策建议。     

B—P网络逐步聚类法在环境测点优选中的应用

将Ｂ－Ｐ网络原理与逐步聚类分析思想相结合，用于环境测点聚类优选。该方法用于水清河几个监测断面的优选结果是符合客观实际的。     

Methodology of industry scale analysis under the constraint of regional resource capacity using HSY algorithm

This paper develops a methodology of proper scale analysis for regional industry development, which can be used in industry planning with the consideration of regional resource capacities. In the face of different data sources and even data scarcity, alternative methods based on linear programming and quadratic programming algorithms for calculating the resource intensity factors are designed. Based on the empirical study of industrialization, initial scenarios of industry development were set. Using HSY algorithm, sensitive industries that may cause exceeding regional resource capacity can be identified, and the risk of exceeding can be predicted and expressed in probability. Furthermore, a proper scale range can be designed for these sensitive industries according to resource capacity. Taking the case of Dalian city in China, this paper estimated the regional urban development plan, various resources capacities were studied, and land resources were estimated to be the most critical resource for the city. The land resource depletion intensities of different industries are calculated by quadratic programming algorithm. Under the constraint of 427.56 km² available industrial land resources, the electronic and power industries have the most significant impact on total land use, if the scale of power industry exceeds 1.27 billion USD, the probability of land resource capacity breakthrough will be 50%.