Farzad Tasbihgoo

Some data-driven modeling approaches for detecting changes in nonlinear dampers

Various identification methods are compared for full-scale nonlinear viscous dampers, including a parametric approach using a simplified design model (SDM), the non-parametric Restoring Force Method (RFM), and the non-parametric Artificial Neural Network (ANN) approach. Advantages and disadvantages of each method are discussed for monitoring purposes. In the comparison, it is shown that the RFM is superior to other methods in regard to the following aspects: (1) no assumption is needed on the nature of the monitored systems; (2) the method is applicable to a wide range of nonlinear system types; (3) the same identification model can be used for the unknown system changes, including the change of system type as well as the change of system parameter values; and (4) physical interpretation of system changes are possible, using the identified values of the series expansion coefficients. A set of experiments was also conducted using magneto-rheological (MR) dampers to validate the feasibility of system change detection. For small changes in the magnetic field strength, the corresponding changes in the dynamic characteristics of the MR damper were detected, using the identified RFM coefficients.

Nonlinear Finite Element Analysis of a Threaded Pipe Connection

For the past several years, USC has been involved in a major research project to study the seismic mitigation measures of nonstructural components in hospitals funded by the Federal Emergency Management Agency (FEMA). It was determined that piping was the one of the most critical components affecting the functionality of a hospital following an earthquake. Consequently, a substantial effort was spent on quantifying the behavior of typical piping components. During the loading of the threaded joint, it was common to hear a loud popping sound, followed by a small water leak. It was assumed that the sound and leakage were due to the sliding of the mating pipe threads. To confirm this theory, and to provide a tool to help understand the failure mode(s) for a wide class of threaded fittings, a detailed nonlinear finite element model was constructed using MSC/NASTRAN, and correlated to the measured failures. In this paper, a simplified model is presented first to demonstrate the modeling procedure and to help understand the sliding phenomenon. Next, a symmetric half 3D model was generated for modeling the physical experiments. It is shown that the finite element analysis (FEA) of the threaded connections captures the dominant mechanism that was observed in the experimental tests.Copyright