基于神经网络的智能结构控制与试验研究

智能结构是指系统的几何特性和运动特性在运行中能根据工作的要求而进行调整的结构。由于振动激励的复杂性和不确定性使得智能结构控制具有高度的非线性,因此很难建立准确的动力模型,神经网络强大的非线性映射能力和自适应学习、记忆的特点使得它非常适合于解决智能结构控制问题。本文主要对基于神经网络的结构控制理论进行了研究,通过对压电智能结构基于神经网络的反馈控制进一步确定了控制算法和控制效果。最后进行了三层框架结构的振动台试验,通过对压电材料基于神经网络的反馈控制来主动调节压电层的轴向变形,以此对结构施加控制力,试验结果表明各层加速度都得到了有效地控制。

Intelligent Structural Control and Experimental Research Based on Neural Networks

Intelligent structure refers to that the geometric and kinetic characteristics of the system can be self-adjusted based on the needs of work when the structure is running.Because of the complexity and uncertainty of vibration,intelligent structure has a high degree of nonlinearity.So it would be difficult to establish an accurate dynamic model.But the characteristics of artificial neural networks such as powerful non-linear mapping capability,self-adjusting learning and memory make it very suitable to solve the civil engineering problems.The research in this paper focuses mainly on the theory of structural control on the basis of artificial neural networks.Control algorithm and effect can be confirmed further through feedback control of piezoelectric intelligent structure.Finally,the shaking table test of a three-story framework is performed.The feedback control of piezoelectric material which based on artificial neural networks can adjust the axial deformation of the piezoelectric layer for applying control force to the structure.Experimental results show that the acceleration at each floor has been effectively controlled.