Nonlinear identification of IPMC actuators based on ANFIS–NARX paradigm

• We present an accurate dynamic nonlinear model to predict the mechanical displacement features of IPMC actuators.
• Model includes two main parts, an ANFIS and a NARX structure.
• Provides a powerful system to create an accurate and transparent identification method for estimation of IPMC behaviors.
• Results prove the ability of the proposed model to capture the mechanical displacement features of the IPMC actuator.

Ionic Polymer Metal Composite (IPMC) is a new smart material that bends in response to a relatively low applied voltage. Due to its special properties such as low density and high toughness, very large stimulus strain, lightness and inherent vibration damping, IPMC is different compared to the other smart materials especially in biological applications. This paper presents an accurate dynamic nonlinear black-box model to predict the mechanical displacement features of IPMC actuators. This model includes two main parts which are an adaptive neuro fuzzy inference system (ANFIS) and a nonlinear auto-regressive with exogenous input (NARX) structure. The combination of universal approximation capability and transparency of fuzzy inference system and training ability of neural networks with adaptive and predictive potential of NARX structure provide a powerful system to create an accurate and transparent identification method for estimation, prediction and interpretation of IPMC behaviors. For the model verification, an IPMC actuator is set up to investigate the IPMC displacement features as well as to generate training data. Validation results proved the ability of the proposed model to capture the real mechanical displacement features of IPMC actuators.