Non-stationary turbulent wind field simulation of bridge deck using non-negative matrix factorization

Numerical simulation of turbulent wind field is essential for the time-domain buffeting analysis of long-span bridges. However, recent field measurements captured strong non-stationary features during extreme wind events, which may pose questions on the current bridge aerodynamic analysis based on the stationary assumption. Therefore, it is both necessary and realistic to perform non-stationary simulation of turbulent wind fields of long-span bridges. In this study, a spectral-representation-based non-stationary process simulation algorithm, which takes advantage of the FFT technique, is proposed using non-negative matrix factorization. In addition, an appropriate scheme is recommended to search the initial matrices for the factorization. After factorizing the decomposed spectrum, the FFT technique can be utilized to enhance the non-stationary simulation efficiency. Numerical examples demonstrate the feasibility and effectiveness of the factorization of the decomposed spectrum. Then, the non-stationary turbulent wind field on the long-span bridge deck is simulated. The computational efficiency, accuracy and applicability of the proposed method are compared with alternative approaches. Results show that the factorization and simulation agree well with the target ones. Therefore, the proposed approach can be used to simulate the non-stationary turbulent wind field in practice.