An NRBO-BP-PID-based feedforward-feedback control strategy for high-precision auto-leveling in drawing frames

The auto-leveling control system of the drawing frame is essential for enhancing sliver uniformity during the spinning process. Existing open-loop auto-leveling control systems lack feedback regulation, which limits their ability to compensate for disturbances, whereas closed-loop control systems remain unable to regulate sliver variations between the detection and leveling points. To overcome these limitations, we propose a feedforward-feedback control strategy for auto-leveling. At the feedforward stage, a model based on the law of mass conservation is developed to describe the relationship among middle roller rotational speed, input sliver thickness, input sliver speed, and output sliver thickness. For the feedback control, a strategy that combines a Newton–Raphson-based optimization (NRBO) algorithm with a backpropagation neural network PID (BP-PID) controller is introduced. The NRBO algorithm optimizes the BP neural network weights, thereby enhancing its self-learning capability and control accuracy. The control performance of the NRBO algorithm is verified through numerical simulations, and an experimental platform is constructed for further validation. The results demonstrate that the proposed NRBO-BP-PID-based feedforward-feedback control strategy provides superior dynamic response, stability, and convergence, yielding the best overall control performance. Compared with conventional feedforward control method and PID-based feedforward-feedback control method, the proposed control approach reduces the mean absolute error of the produced sliver by 38.48 and 31.64%, respectively, while maintaining excellent leveling performance across different target weights and sliver types. This work provides a theoretical and technical reference for the development of intelligent control algorithms in drawing frame auto-leveling systems.