Design of Neural Network for Nonlinear Seismic Control

摘要

Contemporary building structure is ordinarily made up of reinforced concrete or steel, and has the nonlinearity or hysteretic property, and it is obvious with strong shock or strong wind. it is necessary to identify the system for anti-seismic control of nonlinear system. The LM algorithm is the synthetically part of the body of Newton algorithm and gradient descent method. The means also may be supposed to modification of Gauss-Newton algorithm, it not only possesses local convergence as Gauss Newton algorithm but also possesses whole characteristic as gradient descent method. For number of practical times and the accurateness, the algorithm obvious is superior to general gradient method or Newton algorithm .It is applicable to the nonlinear system identification. The imitation result makes known that number of learning times of the neural network decrease to a large extent, and much better in accuracy and the stability. And the identifying effect is also ideal.