Machine learning-based self-sensing of the stiffness of shape memory coil actuator

Self-sensing actuation (SSA) assists in sensing the vital property of the shape memory coil which can be used to monitor and control the actuation in smart equipment as well as robotics. The stiffness characteristic of the shape memory coil (SMC) is sensed during actuation which plays a significant role in development of intelligent robotics in defense systems. The electrical property of SMC such as electrical resistance changes due to martensitic phase transformation which is further used to sense the mechanical properties such as strain, stress, temperature, length, and force of the equipment. Nowadays electrical properties are used to sense the stiffness of the shape memory coil. As of now, there is no well-established analytical model to predict the stiffness of SMC during actuation accurately. Therefore, a soft model based on machine learning is proposed in this paper for autosensing of the stiffness in SMC. The experimental facility has been developed for the collection of data with respect to diverse Joule heating currents. To determine the experimental data values of stiffness and electrical resistance of SMC is a cumbersome task. Hence, we have proposed an automated method to predict the stiffness of the SMC using soft computing-based methods. The Classical Polynomial and Bayesian optimization-based Feedforward Neural Network (FFNN) models are developed for analyzing the stiffness of the SMC. It is found that the hybrid FFNN model outperforms the other ML-based model by attaining 95.2650% accuracy. The FFNN model is also able to explain almost all the predicted stiffness values which are experimentally recorded.