Identification of Material Parameters by Particle Filter Using Observation Data Obtained during Construction of Rock-Fill Dam

Abstract

When an earth dam is constructed, the mechanical behavior must be monitored in order to realize and confirm the safe construction of the dam. In addition, a numerical analysis is commonly conducted before and during the construction to predict the future behavior of the structure. Often, however, not all the material parameters needed for the numerical analysis are available; parameter identification has therefore been an important problem in numerical simulations performed to predict future behavior. This case study targets a rock-fill dam in the Kyushu area of Japan, and its objectives are to enhance the prediction of dam deformation and to demonstrate the identification of the unknown constitutive parameters needed for the actual dam construction process. The particle filter, incorporated into the soil/water coupled FEM, is used for the parameter identification. The identification results show that the numerical results obtained from the identified parameters agree considerably well with the observation data and that 10,000 particles are sufficient for making an accurate prediction.