Session: SYMP 3-5: Actuator Systems

Paper Number: 139592

139592 - Neural Network-Based Sensorless Control of a Shape Memory Alloy Actuator 

Shape memory alloy (SMA) wires can be used as actuators when heated, as their length reduces because of a phase transformation from martensite to austenite. During phase transformation, changes in electrical properties such as the wire resistance can be correlated to changes in wire length, thus enabling the self-sensing operation mode. This effect can be used to implement sensorless closed-loop control, where the mechanical feedback is provided by a position estimation obtained through the wire resistance. Self-sensing based control has a significant advantage over sensor-based control, as it removes the need for an additional electro-mechanical transducer such as laser sensors, into the control system. Due to strong nonlinearities intrinsic to the SMA behavior, the relationship between resistance and wire elongation is often complex and hysteretic, making the practical implementation of self-sensing architectures in SMA rather challenging. 

In this paper, we present a self-sensing based control of an SMA wire actuator where a neural network is used to estimate the displacement by using electrical properties. An experimental setup is developed in which the SMA wire is acting against a linear spring. This spring-loaded SMA wire mechanism is used for analyzing displacement achieved by the SMA wire actuator based on various input waveforms. Characterization data is collected using an embedded system with a microcontroller STM32H743ZI2, and then used for training and validating a neural network. Such a network estimates the displacement of an SMA actuator using electrical properties, i.e., wire resistance and input power. The trained neural network is then implemented on the STM32H743ZI2 and used to estimate the displacement in real-time. The developed embedded system is used to implement and evaluate the sensorless closed loop position control architecture and compare the obtained self-sensing performance with a laser sensor-based closed loop control system. For a closed loop position control architecture, it is necessary to provide the current position of the actuator as feedback to a controller. For implementation of sensorless closed loop control on the experimental setup, estimated displacement is used as feedback to a PI control system. The performance of the self-sensing based PI control is compared against laser-sensor based PI control in which feedback signal is provided using laser-sensor.  

The presented results motivate the potential usage self-sensing based control in development of size-, weight-, and cost-efficient systems for accurate positioning. In future works, self-sensing will be used in combination with advanced control strategies for complex SMA actuator systems like antagonistic architectures. 

Presenting Author: Domenico Bevilacqua University of Saarland

Presenting Author Biography: Domenico Bevilacqua was born in September 1995 in Trani, Italy. He graduated from a scientific high school in Italy in 2014 and earned a Bachelor's Degree in Computer Science and Automation Engineering from Polytechnic University of Bari in 2017. He then completed a Master's Degree in Automation Engineering with honors at Polytechnic University of Bari in 2020. During this time, he conducted his final thesis research as Erasmus+ student at the University of Saarland in Saarbrücken, Germany.
In 2020, Domenico transitioned to a full-time doctoral candidate in System Engineering at University of Saarland, working as scientific researcher at the Center for Mechatronics and Automation Technology (ZeMA) in Saarbrücken, Germany. Here, he focuses on advancing the fields of bio-inspired mechatronics and smart-materials actuation.

Authors:

Krunal Jagdishbhai Koshiya
Gianluca Rizzello
Paul Motzki
Domenico Bevilacqua