The Study On Neural Network Control For Wind-Induced Vibration Of Long-Span Spacial Latticed Structure

Due to their outstanding advantages,large-span spacial structures are widely used in many important buildings, which are usually with low frequency and quite sensitive to pulsating wind loads. Wind loads may result in remarkable structural vibration of such structures and cause problems in its general conditions of use and safety. However, the studies on wind-induced vibration and its control of the spatial reticulated shell structures rarely appear in the literatures. Moreover, due to the huge size, complicated shape and time delay, the vibration control for such structures with traditional methods are quite difficult. Hence, it is important to set up effective control methods for the wind-induced dynamic responses of long-span spatial reticulated shell structures.In this study, the characteristic of some structural vibration control method was clarified, and the applicability of neural network control was pointed out. The characteristic, network structure, and learning algorithm are detailedly presented. The improved neural network control method for wind-induced vibration of long-span spatial reticulated structures and time delay compensation for the structural system were deeply studied. Also, other key problems in neural network control system were briefly investigated.In the improved neural network control for wind-induced vibration of long-span spatial reticulated structures, the wind-induced vibration and its control were reviewed. The wind load simulation technique was detailedly analyzed. The structure of neural network was optimized with both empirical formation and experimental comparison. The Back-Propagation Algorithm of traditional neural network was improved with Levenberg-Marquardt Algorithm, and the convergence efficiency was enhanced with one more order. Based on the combined application of two neural networks, an improved neural network predictive control strategy was introduced to predict the effective control forces for structural dynamic responses.In the study of time delay compensation with neural network in control systems, the principle of time delay and traditional time delay compensation methods were introduced. The single neural network was applied to replace the control system of two neural networks, indentifying structural dynamic responses and predicting control forces, so as to minish the time delay. Then the neural network was used to make up for the shortcomings of traditional time delay compensation methods such as numerical computation difficulty and failure for multiple-degree-of-freedom systems. Multi-step prediction was conducted with neural networks to compensate the time delay in the wind-induced vibration control systems of long-span spatial latticed structures.