Numerical optimization based control design for a ferromagnetic shape memory alloy actuator

Abstract

Ferromagnetic shape memory alloys (FSMA) based actuators are able to get very good precision, allowing, indeed, under nanometer positioning applications. To get those precisions, a good controller is needed, and different strategies have been developed. Experimental tests with a FSMA actuator designed by the research group have proven that the use of a controller operating in a “set-and-forget”mode and following an event-based control scheme allows for a position with a given precision and transient behavior to be achieved, with a reduced number of control actions and therefore a reduced energy consumption. However, the tuning of all control parameter can be hard, especially when taking into account the nonlinear characteristics of the actuator. It can be remarked that the NiMnGa alloy used in the actuator is highly hysteric, and that the control action is pulsed. In this work, a numerical optimization based design methodology is proposed, making easier the control design procedure. The tuning procedure use a model, that for the particular actuator considered in this work, is obtained using machine learning tools, in particular the Tensorflow/Keras framework. An application example shows the good results obtained, including experimental result.