Robust inversion for material parameters identification from correlated outlying observations

Abstract

In this paper, a novel robust inversion method for correlated observations (RIMCO) is proposed to determine the material parameters from correlated observations under the effect of outliers and leverage points. This method is based on a full equivalent weight matrix established from the original measurement weight matrix and an adapted full weight matrix with hard rejection to outliers. This equivalent weight matrix plays key role to refine the stochastic model, while keeping the original correlation of measurements unchanged on the one hand, and ensuring simultaneously high robustness and statistical efficiency of the proposed method, on the other hand. The performance of the proposed method is demonstrated by considering a rockfill dam as an example, where the material parameters are identified from geotechnical and geodetic measurements after achievement of the construction, and during the first filling up of reservoir. Results of comparison of RIMCO with least squares and M Huber methods concerning their robustness and efficiency are presented for various configuration options.