Design and research of benchmark fixed speed computer control system for Carbon Fiber Putting Machine based on deep learning algorithm

In the field of machine design and control systems for carbon fiber production, achieving precise control of filament conveying speed is crucial as it directly affects product quality and production efficiency. The existing approximate methods for this problem include traditional PID control systems and manual control mechanisms. These methods have some drawbacks, such as poor handling of nonlinear dynamics, long response times, and inability to optimize production in real time. To partially overcome these drawbacks, this manuscript proposes an implementation of a method based on deep learning algorithms, particularly convolutional neural networks and long-term short-term memory networks. This method improves the performance of carbon fiber shaping machines by predicting and adjusting the filament conveying speed with higher accuracy and responsiveness. Compared with traditional PID control systems, the positive results of this method include a 38.2% improvement in control accuracy, a 22.3% reduction in response time, and a 10.5% increase in production efficiency. However, there are still some limitations, such as the challenge of integrating the system into existing manufacturing infrastructure and the need for further improvements in noise reduction technology. Future work will focus on addressing long-term system performance and maintenance, including periodic retraining of models to adapt to evolving production conditions. Regular maintenance of the system infrastructure, such as sensor calibration and data quality checks, will be necessary to ensure stable and efficient operation over extended periods.