Variable tension dynamic control for filament winding of cylinder using neural network

As the key influencing factor in filament winding process, fluctuation of winding tension directly affects winding precision and productions' performance. In view of the dynamic change of winding tension and ensuring uniform circumferential residual stress of product, the method to dynamically control winding variable tension using a neural network is proposed. And considering the deformations of mandrel, the radial and circumferential stresses in winding layer under external pressure are obtained through analyzing the basis of anisotropic composite elastic theory and isotropic thick-walled cylinder elastic theory. Within the scope of the elastic limit, the analytic algorithm between residual tension distribution and winding tension is established based on the stress superposition principle. Based on the superposed characteristic of uniform circumferential residual stress, the variable tension during the winding process can be updated dynamically using a neural network with a given weight of output layer and error back propagation and amplification. Simulation and experimental results show that the proposed control method can dynamically optimize the variable tension of filament winding, and it can satisfy the desired requirements and is in line with the actual process of filament winding.