Active Control of Structures Using Neural Networks

该项目侧重于开发和测试使用神经网络的结构控制方法。研究活动包括理论和实验两部分。实验研究将在一个九层钢结构的八分之一比例模型上进行。所有测试将在位于伊利诺斯州尚佩恩的美国陆军建筑工程研究实验室（CERL）的振动台上进行。CERL技术人员将参与该项目。选择神经网络作为控制器是本研究的一个重要特点，因为神经网络具有适合控制应用的能力。初步的实验和理论研究已经证实了神经网络在控制问题中的适用性和能力。重点研究了主动支撑系统和变刚度两种结构控制方法，并对新的控制布置进行了试验。不同的结构控制设计将在数值模拟中开发和研究，然后将在CERL的振动台上实施和测试。将主动控制结构的响应与无控制、有斜撑和无斜撑结构的响应进行比较。

The project focuses on the development and testing of structural control methods using neural networks. Research activities include both theoretical and experimental components. Experimental research will be done on a one-eighth scale model of a nine story steel structure. All tests will be performed on the shaking table at the U.S. Army Construction Engineering Research Lab (CERL) at Champaign, Illinois. CERL technical staff will participate in this project. The choice of neural networks as the controller is an important feature of the research as neural networks have capabilities suitable for control applications. Preliminary experimental and theoretical studies have confirmed the applicability and the power of neural networks in control problems. The study emphasizes two structural control methods, active bracing systems and variable stiffness methods including experimentation with new control arrangements. Different structural control designs will be developed and studied in numerical simulations, and then will be implemented and tested on the shaking table at CERL. The response of the actively controlled structure will be compared with the response of the structure without control, with and without diagonal bracing.