ABSTRACT
When conducting finite element numerical analysis on complex structures, huge calculations are always required, and a complex implicit relationship between structural output and structural input is inevitable, which brings challenges to structural performance evaluation and safety management. At present, the Kriging model, as a kind of data-driven proxy model, has shown an excellent ability for structural response prediction. However, when the Kriging model is used to predict the structural response, the standard Kriging model is generally adopted by the existing methods, that is, the basic function of the regression term is set as the standard form in advance without distinction. This will lead to some unnecessary computations, and in many cases, it will have an adverse effect. Given this, this study proposes a novel structural response prediction method based on the Blind Kriging model. It is used to replace the traditional Kriging model, by which the important factor variables are automatically selected and the non-important factor variables are discarded, and thereby the efficiency of predicting structural response can be greatly improved. Finally, a case study has been conducted to verify the effectiveness of the proposed method is higher and the prediction error is about 1/3 of the standard method. The effectiveness and superiority of the proposed method are verified by comparison with the other two traditional Kriging models. The proposed method can further promote the development and application of the Kriging model in proxy prediction of structural response.
