基于集成神经网络的板下地基脱空判定

通过有限元法建立了地基脱空的混凝土路面计算模型.提出采用基于信息融合理论的集成神经网络技术对混凝土路面脱空状况进行识别,通过路面脱空输入特征向量的组合,用各子神经网络对混凝土路面脱空进行初步缺陷识别,然后对识别结果进行决策融合.给出了系统的实现策略和子网络的组建原则.数值模拟结果表明,采用这种识别方法合理地选取了各种输入特征向量,具有更好的识别效果.

Study on void identifying of rigid pavement based on integrated neural networks

This paper developed a numerical calculating model of rigid pavement which can take the void of foundation. The integrated neural networks void under rigid pavement plate identification technology was put forward based on the information fusion theory. Taking the sub-neural networks as primary separation identification from different sides, the conclusions were gained through decision-making fusion. The realizable policy of the identification system and established principle of the sub-neural networks were given in the paper. It can be educed from the numerical emulation examples that it takes full advantage of diversified characteristic information, and improves the diagnosis rate.