Machine Learning-Aided Nonlinear Dynamic Analysis of Engineering Structures

Publisher:
Springer Nature
Publication Type:
Conference Proceeding
Citation:
Lecture Notes in Civil Engineering, 2023, 356 LNCE, pp. 347-352
Issue Date:
2023-01-01
Full metadata record
A machine learning (ML) technique was used to assist in the dynamic analysis of mixed geometric and material nonlinearities of real-life engineering structures. Various types of inputs of system properties were considered in the 3D dynamic geometric elastoplastic analysis, giving a series of realistic nonlinear descriptions of complex, large deformation structural behaviors. To resolve the numerical challenges of solving the mixed nonlinear problems, a newly established ML technique using a new cluster-based extended support vector regression (X-SVR) algorithm was applied. With this technique, a surrogate model can be built at each time step in the Newmark time integration process, which can then be used to predict the deflection, force and stress of the relevant structural performance at different loading time stages. To demonstrate the accuracy and efficiency of the proposed framework, practical engineering applications with linear and nonlinear properties are fully demonstrated, and the nonlinear behavior of the structure under predicted working conditions in the future was predicted and verified in numerical studies.
Please use this identifier to cite or link to this item: