Parameter Sensitivity and Inversion Analysis for a Concrete Face Rockfill Dam Based on CS-BPNN

• Published in: Advances in civil engineering 2019, Vol.2019 (2019), p.1-17
• Main Authors: Qin, Xiangnan, Shao, Chenfei, Gu, Chongshi, Chen, Yue
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Civil engineering; Concrete; Concrete dams; Deformation; Demand analysis; Engineering; Genetic algorithms; Mathematical models; Mathematical problems; Mechanical properties; Neural networks; Parameter modification; Parameter sensitivity; Rheological properties; Rheology; Rockfill dams; Search algorithms; Sensitivity analysis
• ISSN: 1687-8086; 1687-8094
• DOI: 10.1155/2019/9742961

The deformation behavior of rockfill is significant to the normal operation of concrete face rockfill dam. Considering both the nonlinear mechanical behavior and long-term rheological deformation, the E-ν model and modified Burgers model are coupled to describe the deformation behavior of the rockfill materials. The coupled E-ν and Burgers model contains numerous parameters with complex relationship, and an efficient and accurate inversion analysis is in demand. The sensitivity of the parameters in the coupled E-ν and modified Burgers is analyzed using the modified Morris method initially. Then, a new approach of parameter back analysis is proposed by combining back-propagation neutral network (BPNN) and Cuckoo Search (CS) algorithm. The numerical example shows that parameters K, Rf, and φ0 as well as G are more sensitive to the deformation of the rockfill body. The inversion analysis for these four parameters and η2, E2, and A as well as B in modified Burgers model is performed by the CS-BPNN algorithm. The numerical results demonstrate that the parameters obtained with the proposed method are reasonable and its feasibility is validated.