Model-based control of nonlinear wire tension in dynamic needle winding processes

A defined wire tension force at the nozzle exit is crucial for reproducible and precise wire deposition in a needle winding process. The estimation of wire tension with an Extended-Kalman-Filter (EKF), presented in a prior publication, was evolutionarily refined in this work. To advance emulation of nonlinear bending forces for tension simulation, neural networks were introduced. A comparison between estimation by an EKF with excluded updates and by a random walk EKF proved valid model quality. Loose wires were detected with an additional functional module. To prevent loose windings, wire tensions at the nozzle exit were controlled with a model predictive control approach. This controller was validated by dynamically adjusting the reference wire tensions during winding.