ملخص الدراسة:

Estimating the construction labor productivity considering the effect of multiple factors is important for construction planning, scheduling and estimating. In planning and scheduling, it is important to maximize labor productivity and forecast activity durations to achieve lower labor cost and shorter project duration. In estimating, it is important to predict labor costs. A company may lose money in the execution of the project if the labor cost were wrongly estimated. On the other hand, if the estimate is high, the company may lose the contract due to overpricing. The aim of this research is to develop an artificial neural networks model for giving an expert opinion to predict the production rate for slabs works. ANN is new approach that is used in prediction labor productivity, which is able to learn from experience and examples and deal with non-linear problems. In this research, the effective factors that affect on labor productivity were collected from literature review. A questionnaire survey was done to determine the most effective factors by calculating the Relative Important Index (RII). The target group was determined as the contracting companies which have first, second, and third categories. 110 questioners were distributed, and 77 useful questionnaires were collected. Factors which have RII more than 0.75 are used as independent input variables affected on one dependent output variable "labor productivity" in neural network model. The most important factors that affect labor productivity are, number of labors, material shortages, floor height, tool and equipment shortages, labor experiences, weather, complexity due to steel bars, drawings and specifications alteration during execution, easy to arrive to the project location, lack of labor surveillance and payment delay. Areal data for the most important factors which used in model development was collected by the second questionnaire from 107 building projects in Gaza Strip. Neurosolution software version 5.07 was used to build up the models. The best model was obtained through the traditional trial and error process. However, over 1000 network structures were experimented and the satisfactory model was obtained. This model consists of input layer with 11 neurons, 2 hidden layer with 6, 4 neurons for first and second layer respectively, and 1 output neuron in the output layer. The results of the trained models indicated that neural network reasonably succeeded in estimating the labor productivity. The average error of test dataset for the adapted model was largely acceptable (6.7%). The performed sensitivity analysis showed that the number of labor and weather have the most influential parameters in productivity while payment delay and alteration in drawing and specification during execution have lowest impact on productivity.

توثيق المرجعي (APA)