Force regulation and vibration control of flexible structures with twisted and coiled actuators based on neural network

Ultra-low frequency vibration is crucial for high-end precision instruments and manufacturing equipment. Twisted and coiled polymer actuators (TCPAs) have the advantages of large driving force, high driving accuracy and excellent flexibility, making them highly promising for ultra-low frequency vibration control. This paper establishes a history-dependent forward and inverse neural-network modeling framework for TCPA force regulation and flexible-structure vibration control. The available experimental data are used to train and validate the model, while additional numerical simulations are introduced to evaluate cross-condition and frequency-domain performance. The plant model and the inverse model of the neural network are then established, after which the open-loop vibration control of a flexible cantilever beam is investigated. The combined numerical and existing experimental results show that the proposed model-based open-loop control can reduce the low-frequency vibration of the flexible beam within the evaluated range, providing a basis for further TCPA-based vibration-control studies.