A Temperature-Dependent Back Propagation Neural Network Modelling for a Giant Magnetostrictive Actuator

Giant Magnetostrictive Actuator (GMA) exhibit significant precision degradation in dynamic thermal environments due to hysteresis nonlinearity and thermal drift effects. Conventional mathematical models struggle to characterize the temperature-hysteresis coupling effect. To address this challenge, this study proposes a Temperature-Dependent Back Propagation (TD-BP) neural network model by constructing a three-dimensional input space that integrates the excitation current, temperature and hysteresis operator outputs. This model effectively overcomes the limitations of traditional Back Propagation (BP) neural networks in handling multivalued dependencies while incorporating temperature dynamics. Experimental validation under air output 16–22°C demonstrates that the TD-BP model achieves a Root Mean Square Error (RMSE) 0.0784 µm in displacement output. This demonstrates that the TD-BP neural network model can not only accurately describe the hysteresis static characteristics but also precisely characterize the dynamic characteristics under the influence of temperature.