Abstract:
|
In the context of an important physical system, this paper illustrates how the estimation of model parameters can be improved by formally combining data from both the subsystems and the full system. The objective of this paper is to estimate the parameters of a stochastic shear wall system. Shear walls are components in a building to resist lateral force such as seismic or wind loads. Due to the complicated experimental procedures, current estimates of shear wall resistance capacity are based on very limited test data. In the current work, the authors introduce a method for estimating the parameters of a stochastic system composed of multiple subsystems by integrating the experimental data from the full system and the subsystems. In this work, the cold-formed steel shear wall model is considered as such a system, where the shear wall is the full system and the fasteners are the subsystems. It is assumed that the subsystem data are log-normally distributed and the full system data are normally (Gaussian) distributed. Maximum likelihood estimation (MLE) is used to estimate the capacity mean and variance of the shear wall as well as the strength mean of the fasteners.
|