圆筒纤维缠绕变张力神经网络动态控制

 缠绕张力作为纤维缠绕成型中的关键影响因素,其波动直接影响缠绕精度和制品的性能。针对缠绕张力的动态变化,且保证制品等环向残余应力,提出神经网络动态控制缠绕变张力方法。考虑芯模变形影响,基于各向异性缠绕层弹性变形及各向同性内衬厚壁筒理论,给出外压作用下缠绕层的径向应力及环向应力;在弹性范围内采用应力叠加原理建立剩余张力与缠绕张力之间的解析算法。基于制品环向残余应力叠加特点,采用给定输出层权值的神经网络算法,通过误差反向传播及放大方法,对等环向残余应力制品纤维缠绕过程中的缠绕变张力进行动态更新。仿真与实验结果表明:该控制方法对纤维缠绕变张力起到动态优化作用,可以达到预期要求,且更符合实际缠绕过程。

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.