Neural Network Self-Tuning Control for a Piezoelectric Actuator

Area of Science:

• Control Systems Engineering
• Materials Science
• Nanotechnology

Background:

• Piezoelectric actuators (PEA) are crucial for ultra-precision manufacturing.
• Hysteresis nonlinearity in PEAs significantly degrades positioning accuracy due to rate dependency and multivalued mapping.

Purpose of the Study:

• To develop a model-free control methodology for PEA nanopositioning systems.
• To address and mitigate the negative impact of hysteresis nonlinearity on control accuracy.

Main Methods:

• A neural network self-tuning control approach is proposed.
• PEAs are modeled as nonlinear equations with unknown variables.
• Neural network identifiers approximate unknown variables with adaptive parameter adjustment, avoiding offline identification.

Main Results:

• Experimental validation on a commercial PEA product.
• The proposed control method effectively suppresses hysteresis nonlinearity.
• Significant enhancement in trajectory tracking performance was observed.

Conclusions:

• The neural network self-tuning control is an effective strategy for PEA nanopositioning.
• This method improves accuracy and overcomes limitations posed by hysteresis nonlinearity.