Study on Influencing Factors of Dynamic Load Identification Based on Least Squares Support Vector Machine

Data-driven dynamic load identification methods have seen rapid development in recent years, with one of the methods being based on Least Squares Support Vector Machine (LSSVM). Currently, this method is still in its early stages of development, with many issues yet to be addressed. This paper focuses on studying the influencing factors of dynamic load identification concerning the kernel parameters and regularization parameters of LSSVM. Firstly, it derives the construction process of dynamic response datasets for Bernoulli-Euler beam structures and develops a regression model based on Least Squares Support Vector Machine. Subsequently, utilizing beam structures as a basis, the paper analyzes the effects of kernel parameters, regularization parameters, and noise amplitude on dynamic load identification. Finally, possible optimization methods are proposed based on the analysis results. The factor analysis presented in this paper provides guidance for parameter selection when applying Support Vector Machine methods.