Deflection Control of an Active Beam String Structure Using a Hybrid Genetic Algorithm and Back-Propagation Neural Network

Abstract

A beam string structure is an efficient hybrid system comprising beams, cables, and struts. This study proposes an active beam string structure that adapts to external loading for deflection control, achieved by replacing passive struts with telescopic ones. An optimization-based model is created to minimize deflection, with the maximum deflection of beams serving as the optimization objective. A deflection control framework is constructed by using a hybrid genetic algorithm and back-propagation neural network. The former combines the strengths of the genetic and gradient descent algorithms, and the latter trains a prediction network applying mechanical responses, resulting in quick output of control schemes. To assess the control framework’s performance, a scaled model is designed and fabricated, including a measuring system for deflection and stress, an actuating system with telescopic struts, and a PC-based decision-making system. Experimental and numerical studies are carried out for the model. The control schemes using the hybrid genetic algorithm and back-propagation neural network successfully reduced the deflection responses by at least 80% in simulations and experiments. The results validate the accuracy of the algorithm and reliability of the network, further demonstrating the effectiveness of the control framework. In addition, the deflection control process also optimizes the internal forces of the beam, with a maximum decline rate of stress response approaching 60%.