The Study Of Shaking Table's Control Based On Multi-variable Control Kalman Neutral Network

Application of seismic experimental method to detect structural model and parameters for analysis of complex building structures seismic performance, the assessment of new materials and construction measures for cutting-edge one of the topics. Large-scale shaking table test structure is an important development direction of seismic testing. How to analyze the vibration model of hydraulic servo system to overcome the nonlinear characteristics of the system and, based on state changes and the specimen table vibration effects adaptive control of great significance. This article analyzes the work done by shaking table of the operating characteristics, the establishment of adaptive control methods in order to improve the work of shaking table performance, promoting the development of electro-hydraulic control technology.This paper carried out the following several aspects of the research work:First introduced the significance of seismic experiments and experimental methods, experimental study of dynamic, new technologies, new materials applied in the experiment, analysis of electro-hydraulic shaking table of the physical properties and electrification control theory, introduced the test room electro-hydraulic servo shaking table performance parameters and equipment constitute the proposed vibration table design requirements.Studied the discrete-time systems of data processing and model identification methods. Starting from the finite element model, the introduction of state-space vector, to be the structure of continuous-time state space model is derived discrete-time state space model and parameter identification for the physical structure of the mathematical model, using parameters based on state space identification methods for recognition. To accurately set the model parameters and to build self-adaptive control algorithm provides a theoretical basis;For the control system itself and the non-linear dynamic components, the system frequency response curve is not flat case, this article in the original system, within the closed-loop displacement control based on the design of a multi-parameter control based on the external closed-loop control circuit, will soon force, displacement, speed, and acceleration have joined the control loop, setting the appropriate initial weights, the formation of the waveform based on the acceleration of the outer tracking closed-loop, integrated control, in order to provide flat frequency response characteristics;For the experiment, the specimen structure, the changing characteristics of physical parameters will be added to the Kalman neural network multi-parameter control over the value of the adjustment process to achieve adaptive real-time multi-parameter control to adjust the weight. And the various neural network connection weights between neurons as the learning of the value of the state vector Kalman filter for optimal estimation problem, using the extended Kalman,filter learning algorithm for neural network approximation of connection weights is given the minimum square estimation to speed up the process of iterative learning neural networks, and ultimately the acceleration waveforms and high-bandwidth high-accuracy tracking control.