Research On Dynamic Management System Of Construction Site Safety Evaluation Based On MATLAB Neural Network

From the historical process of all countries’ social development, the construction industry does always be an important pillar of national economy which effects people’s livelihood. The development of the construction industry has opportunities in the era and many severe challenges. Due to the characteristics of construction industry, there are many insecurities of construction site, so the possibility of accident occurrence is high. A little carelessness will cause terrible losses to national economy and people’s life and property safety. The problems of safety management in construction site are a major issue that needs long-faced and constantly improve. In recent years, with the healthy development of construction industry and the steady improvement of safety management in China, the number of accidents decreased significantly. But compared to western countries, there is still a significant distance. Based on safety management, the theory and methods about artificial neural network are used in the management of quantitative safety evaluation. In this paper, a dynamic management system of construction site safety evaluation has been established in order to control the real time security status and evaluate safety management of the construction site during the construction process.This paper firstly reviews domestic and foreign research of construction safety management, introduces the theories of safety management, analyzes the common security evaluation methods, compares their advantages and disadvantages, and presents the concept of building a dynamic safety evaluation management system of construction site. Secondly, it introduces basic principles and the process of BP neural network, analyzes and improves the algorithm, then establishes the mathematical basis of dynamic management system of safety evaluation by using improved BP algorithm. Then the neural network model program of the dynamic management system of safety evaluation is written by MATLAB after choosing proper network structure and parameters. There are 10 evaluation samples from expert in this paper for training neural network. The trained neural network model will be used in real construction cases, and show the effect of artificial neural networks in the dynamic management system of safety evaluation.