基于粗糙集——粒子群神经网络的建设项目安全预测研究

回顾施工项目安全管理和安全管理研究现状,建立建设项目安全管理指标体系。利用人工神经网络非线性函数逼近能力,对项目风险因素程度预测。针对该网络当数据量大时,其结构复杂、收敛慢,易陷入局部最优的缺点,引入粗糙集对影响建设项目安全目标的不确定性因素进行约简,找出最小不确定性风险因素集,大大简化网络输入信息的表达空间维数。并结合粒子群算法收敛速度快、全局最优的寻优能力强的优点,建立基于粗糙集——粒子群神经网络的建设项目安全预测系统。通过实例验证该系统的科学性和有效性。

Research on Safety Forecast for Construction Project Based on Rough Sets and Artificial Nerve Network with Particle Swarm Optimization

The present research status of safety management on construction project is reviewed, and the index system of construction project safety management is established. Risks in project construction are forecasted by using ANN (artificial neural network). Aiming at the shortages of ANN with complex structure, slow convergence and the liability to local optimal value when there is much data to deal with by ANN, Rough Sets (RS) are introduced to simplify the unascertained factors affecting the safety objective of construction project. Thus, the dimension of ANN is greatly simplified. At the same time, by introducing the Particle Swarm Optimization (PSO) with the advantages of fast convergence and global optimum, safety forecast system for construction project is set up based on RS and ANN with PSO. This system is illustrated very scientific and effective.