ScholarWorks@숙명여자대학교

초록

Shape memory alloy (SMA) actuators, known for their high force output despite being lightweight and compact, are increasingly attracting attention in fields such as robotics, medical devices, and wearable devices. Due to these advantages, many studies have been conducted on SMA actuators driven by Joule heating under static air conditions. However, ensuring precise and responsive control of these actuators in diverse and dynamic environmental conditions remains a significant challenge. In particular, the implementation of efficient cooling mechanisms—such as fan-based or liquid-based cooling—for rapid operation is critical for practical applications but still underexplored. Herein, we introduce a durable and adaptive SMA actuation system that operates under randomly generated current conditions. This system enables reliable prediction of SMA output force using machine learning models, even as environmental conditions change. A Long Short-Term Memory (LSTM) neural network is trained to predict the output force of SMA coils using time-series inputs of current, electrical resistance, and environmental temperature. This model is validated under both static air and flowing water conditions with varying temperatures and flow rates. Furthermore, we demonstrate a closed-loop feedback control strategy for the SMA actuator, highlighting its potential for precise and robust operation across diverse environments.

키워드