PUB - Publikationen an der Universität Bielefeld

The quality and stability of stretching processes during the polyacrylonitrile-based precursor preparation could have strong influence on the properties of carbon fibers and subsequent processes efficiency. The distributed units existing in the stretching processes pose a serious challenge to the cooperative regulation of the whole process to achieve a proper ratio distribution among the different stretching units, each of which having a specific technological characteristics. In this paper, a novel cytokine-network-based stretching network (CNSN) is proposed to achieve the desired ratio distribution and control performance in large production lines composed of interconnected yet distributed units organized as nodes in both spatial and temporal layers. Based on the mechanism of the multi-layered cytokine network for regulating the whole immune system in human bodies, a dynamical model of stretching network is suggested in order to describe interaction between stretching units. The model consists of a control allocation layer, a distributed controller layer and an actuator layer. An allocation algorithm based on quadratic optimization has also been successfully proposed to enhance the system’s ability to redistribute stretching ratios to ensure the quality requirement of the final fiber. Numerical results show that the CNSN outperforms traditional control strategies for the multi-stage stretching process in fiber production lines. As the main components of the proposed CNSN are very generic, it can easily be extended to other large-scale production lines having multiple interconnected processes.