The research of the resource optimization in engineering network

Abstract

Resource optimization is considered as a highly important content to the scheduling of the project management. A suitable resource arrangement not only brings a huge facility for organizing management of the construction, but also achieves much benefit economically for the builder during the construction. The research of resource optimization in the early stage is mostly based on the analysis and the heuristics method. However, these methods pose a great many problems such as the low calculation efficiency, the poor transplantation, and the monotonous results; meanwhile the methods have many requirements for the goal questions. As for these problems resulted from the analysis and the heuristics method, another two methods (namely particle swarm algorithm and the random search algorithm) are applied for the resource optimization in this paper. Particle swarm algorithm is a new optimization approach that developed recently. Compared with the analysis and the heuristics method, the particle swarm algorithm has a quicker rate of convergence, higher accuracy and the better global convergence ability. Furthermore it has little requirement to the goal questions with better transplantation. The author defines a variable to denote the estate of the activities, establishes the evaluation function whose independent variable was the actual start time of the activities, makes the further improvement to the optimization foundation, and presents a new optimization based on the dynamic float of the noncritical activities in the paper. The particle swarm algorithm is firstly applied to the unlimited leveling optimization problem which obtains satisfactory results. Like the particle swarm algorithm does, the random search algorithm is another approach that needs little requirements to goal question with good commonality. It builds the sub-aggregate of the feasible solution of the goal problem by the homogeneous random numbers and searches the best solution in this finitude sub-aggregate as the approximate optimization solution for the goal problem. The usage of this method to the unlimited leveling optimization problem achieves a better outcome too. For the resource-constraned project scheduling problem, the author presents a new conception called the random priority to differentiate the precedence relationship of the activities in parallel relationship, and constructs the activity sequences of the network by combining the random priority with the topology sort, than uses the random search algorithm to find the optimization solution in this paper. It indicates that this method is better than the heuristics method through the results, and makes up the monotonous and the poor transplantation characters for the heuristics method. By the comparison of the results, it indicates that the two new methods are feasible and contain many advantages over other methods. The two methods applied in this paper are applicable to the solution of the resource optimization problems and reveal the practical value for the project management.

Alternate abstract:

资源优化是工程项目管理中进度控制的一项重要的内容，在工程建设的施工过程中，合理的安排资源，可以给施工组织管理带来很大的方便，同时还可以给施工单位带来良好的经济效益。长期以来，人们对资源优化问题的研究主要是基于数学解析法和启发式算法，这些方法优化效率比较低、可移植性差、求解结果也比较单一，并且对目标问题要求很高。本文在前人的研究基础上，针对传统解析法和启发式算法的若干问题，对资源优化问题采用了两种新的方法（微粒群算法和随机搜索算法）进行了应用研究。 微粒群算法是近来发展起来的一种新的优化方法，它相对解析法和启发式算法具有收敛速度快，求解精度高，全局搜索能力强等特点；尤其是对目标问题要求不高，具有很好的可移植性。本文针对“工期固定—资源均衡”问题，定义了开关变量 来表示活动的工作状态；建立了以活动的实际开始时间为自变量的评价函数；对问题优化基础做了进一步改进，提出了基于非关键活动的动态时差的优化方法；将微粒群算法首次应用到资源均衡优化问题，取得了较好的优化效果。 随机搜索算法和微粒群算法一样对目标问题不需要特定的要求，是另一种通用性比较好的算法：它通过均匀地产生目标问题可行解空间中的子集空间，在有限的子集空间上搜索最优个体作为目标问题的近似最优解。将该方法运用到资源均衡问题，也取得了较好的优化效果。 对于“资源有限—工期最短”问题，本文提出了随机优先度的概念，用于区分处于平行位置的活动的先后关系；结合拓扑排序，生成即定网络结构的多种顺序序列；然后用随机搜索算法进行寻优，优化结果显示该算法很好的解决了目前启发式算法中序列法求解该问题时出现的结果单一和通用性差的缺点。 最后通过两种规模算例的计算分析，将结果和现行采用的解法做比较，验证了这两种算法在求解资源优化问题的可行性和优越性，对工程实际应用具有一定的参考价值。