Parameter Identification and Linear Model of Giant Magnetostrictive Vibrator

Abstract

A linear magnetization model is built to replace the Jiles–Atherton model in order to describe the relationship between the magnetic field intensity and the magnetization intensity of the giant magnetostrictive vibrator (GMV). The systematic modeling of the GMV is composed of three aspects, i.e., the structural mechanic model, the magnetostrictive model, and the Jiles–Atherton model. The Jiles–Atherton model has five parameters to be defined; hence, its solution is so complex that it is not convenient in application. Therefore, the immune genetic algorithm (IGA) is applied in the identification of the five parameters of the Jiles–Atherton model and it showed a higher stability compared with the identification result of the differential evolution algorithm (DEA). The identification parameters of the two algorithms were employed, respectively, to calculate the excitation force and it was found that the relative error of IGA was evidently smaller than that of DEA, indicating that the former was more reliable than the latter. According to the identification results of IGA and based on the least square method (LSM), curve-fittings to the magnetic field intensity and magnetization intensity were conducted by using the linear function. And the linear magnetization model was built to replace the Jiles–Atherton model. Research results show that the linear model of the GMV can be established by combining the linear magnetization model with the structural mechanic model as well as the giant magnetostrictive model. The linear magnetization model, which has great engineering application value, can be applied in the open-loop control of the vibrator.

Performance Analysis of Giant Magnetostrictive Actuator and Its Application on Active Suspension of Train

“…Figure 4 shows the influence of GMA's excitation frequency on output force, and the excitation current is 0.5A. In our previous theoretical research work, it could be known that the GMA's output force was not related to the magnitude of the frequency (Wang et al 2021), but in the process of our experiments, we found that this was not the case. When the excitation frequency was less than 120Hz, the GMA's output force would decrease with the increase of the excitation frequency.…”

“…The response time of the actuator will affect the time delay of the controller, which will affect the overall performance of the train's active suspension. According to our previous theoretical study (Wang et al 2021), the theoretical response time was only 1.25ms, but it was almost impossible to achieve such a short response time in the actual experimental verification process, since the response time was also related to some external factors. Figure 5 shows the preload's influence on response time of the GMA.…”

An explainable neural network integrating Jiles-Atherton and nonlinear auto-regressive exogenous models for modeling universal hysteresis

“…Regarding the energy dissipation, the differential expression of the irreversible magnetization irr M is given as [46]:…”

A Particle-Swarm-Optimization-Algorithm-Improved Jiles–Atherton Model for Magnetorheological Dampers Considering Magnetic Hysteresis Characteristics

“…Trapanese [17] introduced chaos theory and a simulated annealing algorithm to the classical genetic algorithm, which solved the problem of the classical genetic algorithm [18] and improved the accuracy of identification, but the calculation speed is slow and the convergence time is long; Chen [19] proposed an improved J-A hysteresis model, so that the number of parameters to be identified increased from five to seven, and the key parameters were identified using a differential evolutionary algorithm, which was able to identify the parameter values more quickly, but it had a large error in accuracy and the algorithm had a complicated calculation process. Wang [20] used a neural network to identify the key parameters of the J-A hysteresis model, and the identification results were highly accurate, and the fitted hysteresis curves were in good agreement with the measured curves of the real test, but the neural network relied too much on the training dataset, and it could not work when the data were insufficient, which could easily lead to the loss of information [21]. In addition to this, there are many emerging intelligent algorithms being studied.…”

“…Initialising relevant parameters,Input iteration d, the population size N, and the population dimension DRandomly initialize the position and velocity of the particles Differential evolution on particle swarms according to equation(20) …”