The main objective of this research is to develop a constitutive model for clay that simulates realistically the laboratory and field observations of soil behavior, such as anisotropy, overconsolidation, and time-dependent effects. The proposed model is first developed within the framework of time-independent bounding surface plasticity and anisotropic critical state concept. An isotropic, anisotropic/rotational and distortional hardening rules were adopted under associated flow conditions. The proposed anisotropic time-independent theory was further expanded into coupled elastoplastic-viscoplastic bounding surface theory based on Perzyna
formulation. The unified constitutive model was verified against the experimental results for three different types of clay (Kaolin clay, San Francisco Bay Mud and Boston Blue Clay) with different stress paths and loading conditions.

The proposed time-independent model was incorporated into a two-dimensional finite element procedure for coupled stress-flow analysis. A containment dike constructed over soft clay foundation was selected as case study. The agreement between the model and actual measurements was satisfactory. The importance of anisotropy is depicted from the results of this analysis.

This presentation will summarize the progress of current FHWA seismic vulnerability research  program in existing and new highway constructions. Research studies of this program will be briefly introduced by group areas: Seismic Hazards and Ground Motion, Geotechnical Engineering, Intelligent and Protective Systems, Structures and Systems, Earthquake Reconnaissance, Demonstration Projects and Workshops and Conferences. This program will result in the development of an improved understanding of the seismic hazard in the eastern and central United States, the behavior and response of foundations and soils under seismic excitation, and the overall response of highway structures and systems during earthquakes. This project is also developing retrofitting technologies appropriate for bridges and other highway structures in low-to-moderate seismic zones and design criteria and philosophies intended to reduce the vulnerability of future highway construction nationally.

Artificial salt marsh ecosystems have been introduced over the past ten years as a measure to control areas of high soil contamination since some species found in this environment tend to trap heavy metals and digest toxic elements. This ecosystem is, however, very sensitive to saturation and humidity. Therefore, when introduced artificially, water-retaining structures are necessary in order to control the inflow and the outflow of water from the system. These structures often consist of an outer layer of large diameter aggregate and a core of concrete blocks. In this work, a method to model the inflow and outflow of water through concrete block interfaces is presented. This model uses the classical theory of contact mechanics to account for the deformation of the block interface, under the influence of the weight of the blocks and buoyancy effects. A FEM approximation to the solution of the interface mechanical problem is calculated in order to provide numerical limits of the model parameters and yield conditions. Given the outer side tidal level, Couette flow is assumed in the deformed cavities of the block interfaces to obtain an approximation of the progression of water level on the protected side of the breaker.

A method for detecting anomalous gear behavior in rocket engine turbomachinery during acceptance hot firing is presented. The diagnostic procedure is based on a cepstrum analysis of steady-state gearbox accelerations. In this application, the cepstrum is defined as the inverse discrete Fourier transform of the log of the two-sided autospectral density. The vibration measurements used in the analysis are acquired during static hot fire tests from accelerometers mounted on the external surface of the turbopump gearbox. Following the ground tests, the cepstrum technique is used to provide insight into the differences between turbopumps that have functioned normally and those with out-of-family signatures. The effectiveness of the method is demonstrated by comparing analysis results from an engine in good condition with a similar engine that suffered complete gear failure during its development test. This example is used to suggest that the cepstrum method can not only help detect out-of-family vibration characteristics, but can also provide insight into the nature of a defect. A source mechanism is proposed to explain the unique spectral characteristics that appear due to the presence of the gear fault.

Continuous domains can be modeled using either the finite or boundary element methods.  Both require test functions to be used as interpolants.   A shape function can be viewed as a specific solution to an elliptic equation on a given polygonal domain.  For convex n-sided polygons these functions can readily be found in closed form as rational polynomials. However, convexity is too restrictive a requirement for smooth modeling of many continuous domains. Examples include analysis of biological entities, soil-structure interaction, and high precision graphics rendering. The concavity restriction is alleviated by analytically enforcing continuity of the displacement, slope and curvature tensor fields. In particular, for the simplest concave shape, the four-noded element, the shape function associated with the concave node is calculated.  The three remaining shape functions are generated using the three patch test requirements, which require reproducibility of any arbitrary linear field.

A clearly formulated method is proposed to use multilayer feedforward neural network with sigmoidal activation functions to conduct effective general nonlinear function approximation. This approach is inspired by a geometrical interpretation of the capability of multilayer perceptrons as pattern classifiers. It is proved and demonstrated that neural networks can be constructed corresponding to certain mathematical formulations (such as polynomial regression analysis) with weights and biases related to mathematical quantities. This analytical feature distinguishes this study from most research on neural networks; a neural network can thus be closely related to more conventional computational tools rather than being simply a powerful, but mysterious “black box”. The understanding of the internal workings of this class of neural networks allows them to be used for damage detection purposes.

The proposed technique is successfully applied to a system identification problem to approximate the nonlinear restoring force of a dynamically excited SDOF system. Compared with other existing identification approaches, using neural network in this specified manner provides a simpler, more unified and flexible way of handling this type of identification problem. The potential/advantage of this proposed approach is also shown by its application to a published ‘Volterra-Wiener’ Neural Network which is used to predict structural nonlinear hysteretic response.

In addition to this topic, a brief discussion will be given regarding some system identification work conducted on the Vincent Thomas Bridge in Los Angeles, CA. This long-span bridge is modeled based on its recorded earthquake response.