【作者】 缪鹍；

【导师】 李亮；

【作者基本信息】 中南大学 ， 道路与铁道工程， 2011， 博士

【摘要】 公(铁)路建设项目在我国处于蓬勃发展期,国家每年大量投资用于此类建设,而设计的前期工作——选线设计是首先要解决的问题,是勘测设计中决定全局的重要工作。路线空间位置的好坏直接影响项目的投资、运营、社会和环境保护等方面。传统方法一般是在人工反复比选后才能确定路线走向、线位和线形的几何参数,而现代的公路、铁路设计不仅要提高选线的质量,还要提高选线的效率,这对现有的路线设计理论和方法提出了进一步的要求。群智能优化算法是一种近年来新兴的优化方法,是受到关注最多的优化研究领域之一,其主要通过社会性动物的各种群体行为的模拟,以达到群体中的个体之间的信息交互和合作来实现寻优的目的。与其它类型的优化方法相比,其实现较为简单、效率较高。尽管对群智能优化的研究已经取得了一定的成果,但是从整体上来说,这一新兴的领域仍然处于开放状态,如何进一步提高寻优效率,如何将具体问题与算法有机的结合等尚有待进一步研究。本文以自动选线及相关问题为背景,开展群智能优化算法中的蚁群优化及粒子群优化方法在交通选线中的应用研究。主要开展了以下方面的研究：基于蚁群算法的纵断面优化研究、智能计算方法改进的研究、空间选线方法的研究和土石方调配方法的研究。主要研究成果及创新点如下：在蚁群算法的纵断面优化研究方面：主要针对当前大多纵断面模型不能直接实现自动确定变坡点的问题,建立了离散的纵断面优化算法模型,并采用基本的蚁群算法进行求解,解决了合理的坡段数、坡长和变坡点标高的自动确定问题。在智能计算方法改进的研究方面,主要进行了两种算法的改进研究：(1)为解决较大空间的离散域优化问题,通过将局部更新规则、最大最小蚂蚁和精英蚂蚁策略综合,并将确定性选择和随机性选择相结合对蚁群算法进行改进,研究了算法涉及的信息素更新机制、选择机制及候选集构造等相关问题。(2)针对选线走向优化设计中的连续域优化问题,研究了粒子群算法。为提高粒子群算法在多维变量、多约束条件的复杂条件下的全局搜索能力,提出了在算法中嵌入局部探测和转轴机制的基于Rosenbrock思想的改进粒子群算法,通过大量实验,验证了算法性能得到改善。在计算机空间选线方法的研究方面,为了提高计算效率和改善搜索的全局性能,将路线线形及线位设计分为三维空间走廊线搜索和三维空间线形定位两个阶段。前者着重于发现路线的概略位置,后者着重于局部线形及参数的计算。第一阶段的研究中：(1)提出了基于三维空间网格的轴层模型,使搜索空间与选线空间一致,改进了传统的基于平面网格搜索走廊线的计算模型；(2)提出了基于蚁群算法的三维轴层结构的路线走向(走廊线)构建方法,通过候选集策略实现了大规模离散空间的搜索,并实验验证了基于此网格空间搜索的可行性。(3)为解决搜索中的相关费用计算问题,提出了将数字地价模型和三维轴层结构相结合的策略。第二阶段的研究中：(1)提出了在三维连续空间中三维线形的平纵面同时优化模型,实现了线形参数和线位的优化。(2)实验验证了改进的RPSO粒子群算法在多维、多约束的复杂三维空间中搜索的有效性,并显示了基于数字地面模型的优化计算效率。最后,为将土石方调配中的非线性因素纳入目标,利用土石方累计曲线的同层调配思想,提出了土石方调配的离散模型,并作为改进的蚁群算法的应用。更多还原

【Abstract】 Linear engineering structures such as highway or railway in China are in booming development, and the country has been pouring money into these projects each year. Route location, which is a kind of overall work in road alignment survey and design, is the preliminary work of a project. It is very important when dealing with a linear engineering. The location of the route has great influence on construction costs, operating expenses, and environmental impacts on the study area. A traditional method solves the complicated problem by repeated comparison of the corridor, location and geometric parameter for the linear engineering structures, while a modern one demands not only a good alignment in all affairs but also the efficiency to design the structure. Requirements for the further improving the method are put forward for the route location.The swarm intelligence optimization algorithm is a kind of modern optimization method, and more and more attention has been paid to the research field. Some animals have exhibited complex social behaviors. A most surprising behavioral patterns exhibited by ants is the ability of certain ant species to find what computer scientists call shortest paths, and another one is a population-based optimization technique inspired by the motion of a bird flock, or fish schooling. It is this behavioral pattern that inspired computer scientists to develop algorithms for the solution of optimization problems. The optimization methodology of the swarm intelligence algorithm is the interaction of information and the cooperation between the individuals. The methods are simple and efficient compared to other traditional methods. Although the research for the intelligence optimization has attains plenty of important achievements, the new research field is still open and further research works on how to raise the calculation efficiency, and how to integrate the method with the realistic problem should be given.This dissertation focuses the attention on finding a realistic three-dimensional route alignment. Around this issue, ant colony optimization (ACO) and particle swarm optimization (PSO) algorithm and their applications in the route location automatically are researched.This dissertation focuses on the following works:(1) Ant colony optimization algorithm for the vertical alignment of route. (2) Improving performance about swarm intelligence optimization algorithm. (3) Method of route location in 3D space (4) Earthwork allocation model for nonlinear factors.The main works and contributions of this dissertation are as follows:Researches on the vertical profile:an optimization method to produce an optimum vertical highway or railway profile for a pre-selected horizontal alignment is developed based on discrete theory. The aim of the program was to establish an initial vertical alignment according to discreet ground elevation of station. Considering the discreet characteristic of the ground elevation and the intersection point of grade line, a discrete model is presented. The automatic design problem is set to select the number, location and elevation of the intersection point of the grade line after considering several designing constraints.The two swarm intelligence optimization algorithms (ACO and PSO) are impoved:(1)A combination approach with local pheromone update ruler, elitist ants and MAX-MIN ant system (MMAS) is designed for not only developing the ant search scope, but also strengthening the ability of the ants to pass the complex space. The method combines probabilistic selection and deterministic selection to design transition probability. Some key factors for pheromone update, selection mechanism and allowed set strategy also are researched.(2) A two-stage probing method (RPSO) is proposed to improve PSO method. The first stage guarantees the particle to get away from feasible region as little probability as possible, and the second stage probes further to overcome local minima by Rosenbrock method. The proposed method is implemented and tested for several functions. The results show that the combining method demonstrates a quite good performance in finding global minima reliably in dealing with multidimensional variables and multiple constraints.As for as route location in 3D space, the problem is broken into two parts:one is a corridor finding, another is 3D alignment location. The former aims at a coarse route location while the latter aims at a local alignment and parameter calculation.The first stage of the research on route location in 3D space:(1) A space model consisted in axes and layers are proposed in accordance with route location space, which is an improvement on the traditional plane grids model to seek a corridor.(2) A corridor alignment construction approach on 3D axis and layer model, which is based on an ACO algorithm, is proposed. It can search a good solution on a large discrete space by allowed sets strategy, and a digital example proves its feasibility on the 3D grids elevation model.(3) Right-of-way costs including those associated with land and environmental impacts as well as impacts to stream and other water conduits. A strategy on combining the 3D right-of-way cost model with the 3D grids elevation model is proposed for the cost calculation in the grid point search process.The second stage of the research on route location in 3D space:(1) A PSO model for simultaneously optimizing three-dimensional highway or railway alignments is proposed to get the alignments parameter and location.(2) The experiment about route location has proved that RPSO algorithm has high computational efficiency in calculating earthwork quantity on digital elevation models (DEM) with multidimensional variables and multiple constraints.In the end, a new optimized highway earthwork allocation model from mass-haul diagram idea is built for nonlinear nature. The model aims at generating the optimal earthmoving plan automatically. With it, the earth moving operations can be represented as discrete events systems, and an ant colony optimization algorithm is developed to be equipped with the model.更多还原