A central aspect of performance-based earthquake engineering
(PBEE) involves simulating the performance of facilities (broadly defined
to include the foundations, soil, and nonstructural components) during and
after an earthquake to assess impacts in terms of direct loss, functional
loss, and casualty loss. The current software for simulating the nonlinear
behavior of soils and structures is inadequate, with incomplete or outdated
models, problematic solution methods, and an inflexible software architecture
that inhibits innovative use of modern information technology and high-performance
computing. To address this shortcoming, PEER has embarked on a multiyear effort
to develop the Open System for Earthquake Engineering Simulation, or OpenSees.
OpenSees is an advanced performance simulation software framework for structural
and geotechnical facilities. The software is designed to integrate the implementations
of models for structural behavior, soil and foundation behavior, and damage
measures. Unlike traditional “codes,” OpenSees is designed to be
implemented in a modular, object-oriented manner with a clearly defined application
program interface (API). The modules for modeling, solution, equation solving,
databases, and visualization are independent, which allows great flexibility
in combining modules to solve classes of simulation problems.
One of the best features of OpenSees is its suitability to support a multidisciplinary
approach to PBEE simulation problems. The modular design allows researchers
from different disciplines to combine their software implementations. As one
example, OpenSees is linking new models in structural and soil behavior for
effecting a complete soil-foundation-structure interaction analysis. As another,
parallel and distributed equation solvers developed by computer scientists
and mathematicians are integrated into the framework for the simulation of
very large models. The framework has been combined with advanced ground motion
simulation software to investigate the earthquake behavior of structures distributed
through an urban region. Researchers are developing reliability modules using
the framework. In another example, other researchers are using OpenSees to
develop computational reliability modules for performance-based earthquake
engineering.
Fig. 1. Calculated results compared with experimental
results for a reinforced concrete column
The OpenSees framework will allow combining computational simulation
with physical testing. A component being tested on a shaking table or reaction
wall facility is essentially a physical element in a simulation model subjected
to simulated boundary conditions. Because these types of interfaces are supported
by OpenSees, the framework can serve as a basis for hybrid control of physical
experiments. This is an area of particular interest for PEER as it envisions
its role in the NSF NEES program (http://www.nsf.eng.gov/nees/).
In addition to its furtherance of PBEE, OpenSees is an important
educational tool. Students are excited about working with modern software.
When learning OpenSees, its combination of sophistication and accessibility
motivates students to learn more about computer science and advanced applications.
The software is “open source,” meaning that all parts of the code
are available for users to see, check, track changes, and make contributions
to. The website at http://opensees.berkeley.edu
not only provides a download center, but also supports a revision control
system, a method for submitting contributions, and a bulletin board for communication.
This is the first instance of an open-source, community code in earthquake
engineering.
Structural Modeling
The OpenSees framework is a valuable complement to the significant progress made in developing a family of models for beam-column elements for reinforced concrete members, including a library of constitutive models for concrete and steel. A recent extension is the inclusion of large-displacement geometry, based on the co-rotational formulation, for the accurate representation of P-D effects (Filip Filippou, UC Berkeley). Another approach based on generalized plastic hinge models is nearing completion and shows promising results (Gregory Deierlein, Stanford University). The models have been verified by experimental data and other solutions, when available. Figure 1 shows computed results by Filippou using a fiber model, compared with experimental data from Laura Lowes (University of Washington) and Jack Moehle (UC Berkeley).
Structural models have been developed with a general representation
of shear behavior in reinforced concrete. An accurate assessment of reinforced
concrete building damage must include shear and shear-flexure modes of behavior
in order to capture nonductile failure modes that can occur and to track the
degradation of shear strength and stiffness during an earthquake. Shear-flexure
interaction is represented in the models, and calibration of the models is
under way in a joint effort by Filippou and Gregory Fenves (UC Berkeley) on
the modeling side and by Moehle on the experimental and behavior side. A new
model of beam-column joints (Lowes) represents the effects of joint shear,
bond slip, and yield penetration.
For computation, OpenSees incorporates variable time-stepping pushover analysis
procedures that increase the robustness of the solutions. Armen Der Kiureghian
(UC Berkeley) and Joel Conte (UCLA) have introduced reliability-based methods,
using first-order and second-order reliability analyses on structures. An
initial implementation of parameterizations for models and sensitivity calculations
gives not only the response for a ground motion but also the first-order sensitivity
to design parameters.
For performance evaluation, OpenSees can be used with a new database for response
results to evaluate damage parameters and visualization (Kincho Law, Stanford
University). Additionally, a higher-level scripting language and graphical-user
interface have been completed for users to create building models quickly
(George Turkiyyah, University of Washington).
Geotechnical Modeling
A current emphasis in PEER is the addition of geotechnical
capabilities to OpenSees. Ahmed Elgamal (UC San Diego) has implemented two-dimensional
soil plasticity models that have been calibrated with experimental data developed
by Juan Pestana, Raymond Seed, and Ann Marie Kammerer at UC Berkeley. Boris
Jeremic (UC Davis) has implemented a three-dimensional solid element into
OpenSees that enables fully three-dimensional modeling of soil-structure systems.
Pestana has contributed to the modeling of deep foundations through refinement
of material models and development of simplified one-dimensional elements
describing the complex soil-pile response interface. The various models have
been implemented in a project by Frieder Seible and Elgamal (UC San Diego)
and Joel Conte (UCLA) that models an existing bridge site. Figure 2 shows
the computed permanent soil deformations (amplified) for the bridge under
uniform base excitation. Further modeling by Jeremic includes a three-dimensional
model of the soil-foundation system.
Fig. 2. Computed permanent soil deformations under existing
bridge
Many dynamic response problems involve waves propagating in
semi-infinite domains, which is especially complicated if semi-discretization
techniques are used to analyze the soil domain. Pedro Arduino (University
of Washington) has implemented a generalized viscous boundary condition and
specification of input motion.
PEER/CMU/MSU
Collaboration
The most ambitious use of the OpenSees framework is perhaps
PEER’s joint work with Carnegie Mellon University and the Engineering
Research Center at Mississippi State. The goal of this NSF-sponsored project
is to advance the science of simulating the seismic performance of entire
urban regions. Jacobo Bielak at CMU is developing the ground motion simulation,
including fault modeling and sedimentary basins; PEER is developing structural
and soil-structure-foundation interaction (SSFI) models and simulation methods,
and MSU-ERC is developing the middleware and visualization tools to link the
models and computations in a web-based environment that can be used by stakeholders
and decisionmakers. In the pilot project, OpenSees has been integrated with
MSU-ERC’s visualization tools to show the spatial variation of damage
in a region due to a point seismic source and a moving source. Figure 3 shows
one set of results, specifically the spatial distribution of ductility demand
for buildings in a 20 km x 20 km region subjected to a point-source. In this
simple example buildings are modeled as an elasto-plastic single-degree-of-freedom
system with elastic vibration period of 1 second and the strength is based
on the mean spectrum for the region reduced by a factor of R=4. The results
show the large spatial variation of the ductility demand for buildings in
the region because of the variability of the ground motion. Results such as
these are being used to provide detailed information on the spatial variability
of building performance using different zonation and design procedures. The
long-range goal is to provide realistic simulations of performance of the
built infrastructure in urban regions with realistic models for the fault
mechanisms, regional geological structures, site effects, and building and
lifeline inventories.
Fig. 3. Distributions of maximum ductilities and orientations,
Period =1 sec, R=4
Future Plans
With the models for soils, concrete structural components, and their interactions well under way, the next phase of OpenSees development will shift in two directions. The first is application of the models to several test bed structures; these are real facilities that will be modeled and studied in a collaboration between PEER researchers and its Business and Industry Partners. The testbed studies will examine the effectiveness of the new models and solution strategies, and will provide information on the impact of earthquakes on constructed facilities. The second new direction is to link computed performance parameters with losses. A primary objective of performance-based earthquake engineering is to provide a probabilistic statement of the likely performance of a facility in terms of losses, functionality, and casualties. The modular structure of OpenSees will facilitate the development of this new capability in the next development phase.