Parameter estimation of an SMA actuator model using an extended Kalman filter

Abstract

Brinson model is one of the most-widely used models for shape memory alloy (SMA) wires. The parameters of this model must be determined experimentally. This paper has focused on identifying these parameters for a one-dimensional model of SMA wire. This model is identified in order to predict the outputs of a one-DOF SMA actuator. The parameters of Brinson model are estimated experimentally using an extended Kalman filter (EKF). The main contribution of this paper is designing a procedure to identify the model parameters based on experimental tests performed on the SMA actuator, without needing separate laboratory tests. In the designed procedure, the effects of unmodeled factors and sensor biases are taken into account. The estimation is performed in a pre-designed sequence to avoid the divergence of EKF and validated by comparing the outputs of the identified model with the actual output of the actuator. The results of the identified model match precisely with the experimental data in both phase transformation and linear regions. However, some mismatches are observed in transition regions that may be related to the Brinson model structure, where the governing equation suddenly changes in the transition region. The results show the capability of the EKF to estimate Brinson model parameters. This estimation can be performed directly on SMA actuators.