A Neural Network Inverse Model for a Shape Memory Alloy Wire Actuator

Abstract

Tracking control of shape memory alloy (SMA) actuators is essential in many applications such as vibration controls. Due to the hysteresis, an inherent nonlinear phenomenon associated with SMAs, open-loop control design has proven inadequate for tracking control of these actuators. Aimed at eliminating the position sensor to reduce cost of an SMA actuator system, in this paper, a neural network open loop controller is proposed for tracking control of an SMA actuator. A test stand, including a titanium-nickel (TiNi, or Nitinol) SMA wire actuator, a position sensor, bias springs, and a programmable current amplifier, is used to generate training data and to verify the neural networks open loop controller. A digital data acquisition and real-time control system was used to record experimental data and to implement the control strategy. Based on the training data obtained from the test stand, two neural networks are used to respectively model the forward and inverse hysteresis relations between the applied voltage and the displacement of the SMA wire actuator. To control the SMA actuator without using a position sensor, the neural network inverse model is used as a feedforward controller. The experimental results demonstrate the effectiveness of the neural network open loop controller for tracking control of the SMA wire actuator.

Time-delayed dynamic neural network-based model for hysteresis behavior of shape-memory alloys

“…To compare DNN, TDDNN and SNN, we introduced the BP (back propagation) neural network (BP-NN) [37] as the example of SNN. The BP-NN contains a four-layered architecture with two hidden layers, one hidden layer which consists of four neurons and the other one which consists of five neurons.…”

“…Asua et al [40] just used the current state and two previous SMA states as three inputs to the NN to extract both the state-varying tendency and the frequency information. Most NNs mentioned above are static neural network (SNN) [11,[36][37][38][39]. It is well known that the basic architecture of the so called static neural networks (SNNs) always sums their weighted inputs x and obtains an output y though a nonlinear activation function g, i.e., y ¼ g wx þ b ð Þ, where w is a weight coefficient and b is a bias [41].…”

“…Researchers have attached great importance to the past state information and modeled it into neural networks [36]. Targeting at the Difficulty 1, Song et al [37] used a tag as a input to NN to distinguish the current state in an ascending branch of the loop from that in a descending branch of the loop. Ma et al [11] proposed an elementary inverse hysteresis operator (EIHO) to extract the change tendency of inverse hysteresis which includes increasing, decreasing and turning.…”

Position control of a shape memory alloy actuator using a four-term bilinear PID controller

“…Kumagi et al [14] have proposed a controller with a feedforward part that uses a neuro-fuzzy inference system and a PD controller. Song et al [15] have designed a neural network feedforward controller for open loop tracking control of a SMA wire actuator without a position sensor. Their neural network controller is an inverse model of the hysteresis that maps the relationship between the applied voltage and the actuator displacement.…”

“…The most typical configuration is a combination of feedforward and feedback controllers [13][14][15][16][17]. The main idea of the feedforward controllers consists of using an inverse model of the system to generate the appropriate control input to obtain the desired response.…”