为了更合理、方便地控制土木工程结构地震动力反应,提出基于反向传播(back propagation,BP)神经网络的结构振动模态模糊控制算法。以结构地震动力反应数据训练神经网络建立结构分析模型,以时域模态坐标作为被控变量,实现系统降阶,使建...为了更合理、方便地控制土木工程结构地震动力反应,提出基于反向传播(back propagation,BP)神经网络的结构振动模态模糊控制算法。以结构地震动力反应数据训练神经网络建立结构分析模型,以时域模态坐标作为被控变量,实现系统降阶,使建立模态模糊控制规则所需要的模糊推理数量处于可接受范围内,并以体系能量最小作为控制目标制定控制规则。建立结构动力反应模糊控制数值模型,根据计算地震动力反应评价所提出算法的减震效果。结果表明:经过训练的BP神经网络可以准确地预测结构的地震动力反应,并可以据此建立模糊控制规则。仅对结构第一阶振型采用模态模糊控制就能达到满意的减震效果。采用主动质量驱动(active mass driver,AMD)最优控制力幅作为各楼层控制力的论域时,模态模糊控制减震效果与其存在差距;增大控制力的论域,可以得到更好的减震效果。展开更多
In order to reduce the errors of the reliability of the retaining wall structure in the establishment of function, in the estimation of parameter and algorithm, firstly, two new reliability and stability models of ant...In order to reduce the errors of the reliability of the retaining wall structure in the establishment of function, in the estimation of parameter and algorithm, firstly, two new reliability and stability models of anti-slipping and anti-overturning based on the upper-bound theory of limit analysis were established, and two kinds of failure modes were regarded as a series of systems with multiple correlated failure modes. Then, statistical characteristics of parameters of the retaining wall structure were inferred by maximal entropy principle. At last, the structural reliabilities of single failure mode and multiple failure modes were calculated by Monte Carlo method in MATLAB and the results were compared and analyzed on the sensitivity. It indicates that this method, with a high precision, is not only easy to program and quick in calculation, but also without the limit of nonlinear functions and non-normal random variables. And the results calculated by this method which applies both the limit analysis theory, maximal entropy principle and Monte Carlo method into analyzing the reliability of the retaining wall structures is more scientific, accurate and reliable, in comparison with those calculated by traditional method.展开更多
文摘为了更合理、方便地控制土木工程结构地震动力反应,提出基于反向传播(back propagation,BP)神经网络的结构振动模态模糊控制算法。以结构地震动力反应数据训练神经网络建立结构分析模型,以时域模态坐标作为被控变量,实现系统降阶,使建立模态模糊控制规则所需要的模糊推理数量处于可接受范围内,并以体系能量最小作为控制目标制定控制规则。建立结构动力反应模糊控制数值模型,根据计算地震动力反应评价所提出算法的减震效果。结果表明:经过训练的BP神经网络可以准确地预测结构的地震动力反应,并可以据此建立模糊控制规则。仅对结构第一阶振型采用模态模糊控制就能达到满意的减震效果。采用主动质量驱动(active mass driver,AMD)最优控制力幅作为各楼层控制力的论域时,模态模糊控制减震效果与其存在差距;增大控制力的论域,可以得到更好的减震效果。
基金Project(2013CB036004) supported by the National Basic Research Program of ChinaProjects(51178468,51174086) supported by the National Natural Science Foundation of ChinaProject(201102) supported by the Open Foundation of Hunan Key Laboratory of Safe Mining Techniques of Coal Mines,China
文摘In order to reduce the errors of the reliability of the retaining wall structure in the establishment of function, in the estimation of parameter and algorithm, firstly, two new reliability and stability models of anti-slipping and anti-overturning based on the upper-bound theory of limit analysis were established, and two kinds of failure modes were regarded as a series of systems with multiple correlated failure modes. Then, statistical characteristics of parameters of the retaining wall structure were inferred by maximal entropy principle. At last, the structural reliabilities of single failure mode and multiple failure modes were calculated by Monte Carlo method in MATLAB and the results were compared and analyzed on the sensitivity. It indicates that this method, with a high precision, is not only easy to program and quick in calculation, but also without the limit of nonlinear functions and non-normal random variables. And the results calculated by this method which applies both the limit analysis theory, maximal entropy principle and Monte Carlo method into analyzing the reliability of the retaining wall structures is more scientific, accurate and reliable, in comparison with those calculated by traditional method.
