A fast and accurate piezoelectric actuator modeling method based on truncated least squares support vector regression

In order to improve the applicability of piezoelectric actuators (PEAs) in precision positioning, least squares support vector regression (LS-SVR) is applied to model hysteresis in PEAs due to its high modeling accuracy and fast convergence speed. However, low robustness of LS-SVR makes modeling accuracy susceptible to noises, which makes LS-SVR hysteresis models difficult to be applied in engineering environment. In this article, a robust truncated least squares support vector regression (T-LSSVR) is proposed. With the truncation strategy, redundancy in the training set is reduced and robustness is improved. Parameters required for T-LSSVR are optimized by particle swarm optimization and cross optimization algorithms. To test the proposed approach, it is applied to predict the hysteresis of PEAs. Results show that the proposed method is more accurate and robust than other versions of LS-SVR when the training set is polluted by noises, and meanwhile reduces the sample size and increases computational efficiency.In order to improve the applicability of piezoelectric actuators (PEAs) in precision positioning, least squares support vector regression (LS-SVR) is applied to model hysteresis in PEAs due to its high modeling accuracy and fast convergence speed. However, low robustness of LS-SVR makes modeling accuracy susceptible to noises, which makes LS-SVR hysteresis models difficult to be applied in engineering environment. In this article, a robust truncated least squares support vector regression (T-LSSVR) is proposed. With the truncation strategy, redundancy in the training set is reduced and robustness is improved. Parameters required for T-LSSVR are optimized by particle swarm optimization and cross optimization algorithms. To test the proposed approach, it is applied to predict the hysteresis of PEAs. Results show that the proposed method is more accurate and robust than other versions of LS-SVR when the training set is polluted by noises, and meanwhile reduces the sample size and increases computational effi...