钢管砼轴心受压柱的可靠性研究

Reliability Study on Concrete Columns Filled Steel Tube Subjected to Axial Pressure Based on Artifical Neural Networks and Monte-Carlo Method

规程所给钢管砼轴心受压柱的计算公式比较复杂 ,给可靠度的计算带来一定的困难 ,尝试利用神经网络 -蒙特卡罗法计算钢管砼柱可靠度 .利用一实例柱的轴心受压试验结果 ,训练了四层 BP网络 ,进行了柱轴压承载力的预报 ,预报值和试验值吻合良好 .进而利用实例柱网络模型结合蒙特卡罗法 ,算出柱的可靠度 。

The complex formula by the code for carrying capacity of the concrete column filled steel tube makes it very difficult to calculate its reliability index in the traditional reliability theory. Artifical neural networks Monte-Carlo method is used to overcome these difficulties.With experimental data of the axial pressure-resistant behavior of columns,a four-layer back-propagation network is trained to predictthe carrying capacity of columns and the predicted results match well the test results.In further analysis,Monte-Carlo method with the trained network model is used to calculate the reliability index of the column as a representative example.Therefore a new approach is proposed to analyze the structure reliability.