关键词: 气流引纬结构, 扩张状态观测器, 积分滑模面, 分数阶, RBF神经网络

Abstract:  With the continuous development of modern technology, emerging forces have been injected into the textile industry. The combination of advanced control theory and computer technology has created countless intelligent workshops. The intelligentization of workshop industrial lines has greatly increased the production of fabrics and saved the loss of human and material resources. However, the core technology mastered by China's textile industry still lags behind that of early developed countries, which is not conducive to the development of the domestic textile industry. As a highly automated equipment, air-jet looms require advanced control algorithms to further improve the quality of fabric and weaving efficiency. The airflow weft insertion system, as its core component, plays a crucial role in increasing the speed of the loom spindle. Improving the structure or algorithm of the airflow weft insertion system can improve the overall performance of the loom and respond to national policies.
To improve the weft insertion rate of jet looms, modern control theory was combined to improve the control algorithm of airflow weft insertion, so as to enhance the system's control accuracy and resistance to interference. This study aimed to develop a neural integral sliding mode fractional backstepping controller (RBFISMFOBC-ESO) based on an extended state observer to improve the efficiency of weft insertion. Firstly, a time-varying mathematical model of the weft insertion system was obtained through force analysis of the yarn, and it was transformed into a state equation. In order to improve chattering and enhance the control accuracy of the system, a new integral sliding mode fractional order backstepping control method was derived and proved through Lyapunov theory. At the same time, RBF neural network was used to compensate and correct the sliding surface coefficients, further improving the control accuracy of the system. In order to reduce the impact of time-varying factors such as pressure fluctuations and unmodeled data, finite time extended state observations were used to estimate the total disturbance of the system. In a simulation experiment environment, it was compared with PID control and traditional state observer based ISMFOBC control (ISMFOBC-ESO). The results showed that RBFISMFOBC-ESO control improved the robustness of the system, as well as the response speed and control accuracy.
The designed weft insertion control algorithm has been successfully verified through simulation and experiments, which effectively improves the robustness, stability, and control accuracy of the system. However, the process of yarn passing through the shed is completed by multiple units in collaboration, and the weft insertion control system is relatively complex, requiring more detailed research in the future from two main points. On the one hand, further exploration is needed for the duration of weft insertion and the magnitude of yarn tension under varying process parameters, so as to  screen out the optimal conditions for the RBFISMFOBC-ESO controller. On the other hand, further optimization is required for the controller's hardware. A controller with faster processing speed would further minimize yarn flight time, thereby enhancing the loom's weaving efficiency.

Key words: airflow weft insertion structure, expansion state observer, integral sliding mode surface, fractional order, RBF neural network