Study on the adaptability of dynamic load identification based on least squares support vector machine

Dynamic load identification is a crucial aspect in the field of structural dynamics. In recent years, the identification method based on the Least Squares Support Vector Machine (LSSVM) has garnered significant attention. The LSSVM method offers high accuracy, does not require extensive response information, and exhibits strong algorithmic stability, meaning the load identification results remain consistent across different training sessions with the same training and prediction sets, without cumulative error. However, research on identification using this method is still in its infancy, with many issues yet to be addressed. This paper investigates the adaptability of this method, exploring the effects of factors such as the position of sinusoidal excitation, amplitude, frequency, and load type on the identification results. The computational results indicate that the dynamic load identification method based on the least squares approach has strong adaptability and holds certain value for engineering applications.