Optimizing time and cost by genetic algorithm method for construction projects

Dissertation for Master's degree (M.A)

Trend: "Strategic"

Introduction

Today, the need for proper planning in order to correctly estimate the time and cost of the project and the amount of resources required in a project that have a direct impact on the implementation, administration and proper exploitation of projects such as dam and building construction is clear. One of them is the correct modeling and forecasting of the project cost and time. Achieving this goal significantly contributes to the optimal management of the project and decision-making in certain situations. The issue of planning and then controlling the project schedule becomes more important every day. It is very important to be based on the facts and to be able to contain all the economic facts in the model of a project. A comprehensive program has the ability to take into account the necessary changes in the cost and time of resources in a project and to put the appropriate solutions in front of the users so that they can have a proper estimate of the time and cost of implementation and the amount of resources needed in the project. The project is defined. Sometimes the project management decides to reduce the project time, which will have a direct impact on the finished cost.

Time reduction is achieved with special measures, including the use of executive resources, which increases project implementation costs.

In order to optimize time-cost, various methods are used in the two areas of time-cost balance analysis. For this purpose, various mathematical and research (exploration) methods are used.

Among the exploration models are: Fondal method, Oprager structure model, Moslehi modeling, and Siemens model. Mathematical methods include the linear programming method, the major integer programming model, the dynamic model, and the combined model of linear programming and the integer number.

Genetic algorithm is a search and optimization method based on natural evolutionary principles. This algorithm stipulates that a population consisting of a large number of individuals selected under special rules will optimize the fit function during an evolutionary process. The genetic algorithm has advantages over other optimization methods, such as the optimization of continuous or discrete variables with much more complex objective functions, the use of probabilistic transfer rules instead of deterministic rules, and the ability to work with a large number of variables. For the first time, Feng and Lui (1997) used the genetic algorithm method in solving the time-cost balance problem, Hegazy (1999) used the genetic algorithm method in solving the problem of resource allocation and leveling. In their modeling, they followed single-objective optimization and considered the decision maker's preferences in choosing options only by weighting the time and cost parameters.

In this thesis, using the genetic algorithm and multi-objective model, the balance of time, cost and resource allocation was done, while the uncertainties in the duration and cost of each project activity were included.

Chapter

1-2- Introduction:

All human actions in nature are the result of his decisions and his knowledge of all possible options and possible results based on these types of activities. Considering the type of problem and its possible complications, it seems practically impossible to accurately predict the most favorable answer. In practice, the decision-making process is usually related to several different and non-equal weight and often non-co-directed objective functions. This means that the valuation process of different functions cannot be essentially the same and in many cases, the value of one objective function cannot increase without decreasing the value of the other objective function. Therefore, in order to make a final decision, a kind of balance between different goals is needed, and the way of this compromise is very important in decision making. In this regard, risk, profit, implementation cost and social interest can have a significant impact on this compromise. Such problems are known as multi-criteria decision making problems (MCDM [1]).

2-2- Principles of multi-objective decision-making

Due to the impossibility of reaching the optimal values ??in all objective functions simultaneously, the multi-criteria decision-making problem will usually lead to the selection of an option among a number of candidate solutions. In the end, the final choice will be a compromise and a balance between the objective functions, and finally, the preference of the decision maker will determine the final answer among the set of candidate answers. Many decision-making problems include a large number of decision variables, which is practically impossible to compare all of them and all possible choices. Therefore, according to this point, such optimization problems will become a search problem with the approach of choosing the optimal solution based on the process of eliminating undesirable solutions. Solving such problems is known as multi-objective decision making or multi-objective optimization.

A common classification for solving such problems is the method presented in 1969 based on how to apply the decision maker's priorities. Based on this classification, there are four ways to introduce the final solution in multi-objective optimization problems: 1- Not considering any kind of superiority (only the search is performed) 2- Taking into account the superiority (preference) of the decision maker before the search process (determination of preference before the search)

3- Providing information related to the superiority (preference) of the decision maker in a dynamic form (determination) Preference at the same time as the search) 4-Meaning the superior information (preference) of the decision maker after completing the search process The first method: It includes methods during which the search is performed without considering any priority from the decision maker. This means that the decision maker's wishes and preferences are attributed to the objective functions in the form of weights. The third solution includes methods that allow the decision maker to actively influence his priorities in the search process. In the last method, it is possible for the user or decision maker to make a choice based on their preferences after completing the search process. In this method, after the search is completed, sets containing acceptable answers are produced and given to the decision maker to apply the opinion. 2-3- History of studies in the field of multi-objective evolutionary algorithms Goldberg suggested that multi-objective optimization can be investigated using the Pareto ranking process. This means to assign a relative value of fitness to each member of the population according to the dominance of the Pareto members. This process, which is known as non-post sorting [2], is the cornerstone of all multi-objective evolutionary optimization algorithms (Figure (1-2)).

Goldberg himself did not implement his theory, but shortly after, the NSGA algorithm [3] was introduced based on this thinking and was considerably accepted by the scientific community.

Figure (1-2): Different approaches to Pareto ranking

In the figure on the left, relative fitness values ??have been assigned to the members of the present population based on the Goldberg method. The fitness value of 1 is assigned to the non-post answers of the population, and the answers with this fitness are removed from the population. Then fitness 2 is assigned to the remaining non-post answers in the current set and they are also removed. This procedure continues until all members of the population have their own fitness. In the figure on the right, the method of assigning the value of relative fitness is in such a way that if there is no solution that overcomes our desired solution, the value of 1 is given to that solution, if one solution overcomes it, it is given the value of 2, and if there are two solutions that overcome it, the value of 3 is considered for its fitness. This process continues until all solutions have a fitness value. 2-3-1 Multi-Objective Evolutionary Algorithms (Multi-Objective Evolutionary Algorithms) After Goldberg's very efficient proposal, a number of Multi-Objective Evolutionary Algorithms (MOEAs) have been proposed in the last decade, some of them are listed below.