A novel regularization framework for impact load identification on a high-speed train scale carbody via improved AIC and weighted sparsity

Abstract

The accurate identification of abnormal impact loads generated by under-vehicle equipment is recognized as critical for assessing structural integrity in high-speed trains. This inverse problem is addressed through a novel load identification approach, which leverages sparse regularization principles. The precision of impact load identification is enhanced through a weighted l₁-norm regularization framework, aimed at mitigating solution ill-posedness and minimizing peak force errors. This framework is efficiently solved by an optimized TwIST algorithm, into which an iterative reweighting technique is incorporated. Furthermore, an improved AIC is employed for optimal regularization parameter selection, and an analysis of sample size selection is complemented. The method proved accurate and robust, as verified by comprehensive simulations and scaled carbody experiments. The proposed method is demonstrated to significantly outperform classical l₁-norm regularization in identifying impact loads, with superior accuracy and increased robustness exhibited with respect to measurement point selection and noise contamination.