Analysis Software

State-of-the-art software for structural and geotechnical analysis will be assembled from the work of the Project Team and also work within other NEES facilities.  However, the platform for the project development may be different.  It will be integrated with the testing and control software to deliver capabilities of modeling complex structures and materials alongside foundation materials, as well as deformation and failure of the foundation. This will drive the test in an online computer-controlled manner using recent development in algorithms that take into account the characteristics of the loading system (actuators, servo-valves, etc.).  Multi-function data visualization and knowledge interpretation tools will be developed for the NEES site with the Automated Learning Group of the National Center for Supercomputing Applications (NCSA).  This will be accomplished in four stages:  (i) visualization, (ii) integration and interpretation of multiple-source test data, (iii) integration of test and analysis information, and (iv) model adjustment and optimization. Teaming with NCSA in developing visualization tools ensures the efficacy of the project.  Finally, the control and telepresence system does a three-level algorithm comprising network (supervisory), link, and servo levels, affording full teleobservation and teleoperation capabilities in an open, easy-to-modify architecture.  The three-level control system provides high levels of safety in terms of teleoperation of such large facility.  The PIs and the Project Team have an established record in all aspects outlined above (online dynamic testing, instrumentation, analysis, visualization, and control).

Candidates for the structural analysis software are the INDYAS and ZEUS-NL software packages which the Principal Investigator and his researchers have been developing since 1987. These are a fully adaptive 3D static and dynamic analysis packages capable of modeling steel, concrete and composite structures. The main characteristics of these analysis tools are described in the literature. The accuracy and stability of these packages have been extensively verified by comparison with closed form solutions for static cyclic loading of members and for full scale 3D structures tested at the Joint Research Centre of the European Community. The packages have the advantage of being able to undertake fully adaptive pushover analysis where the following features are included:

• Automatic switch from load- to displacement-control and back
• Continuous updating of the load or displacement profile.
• Updating of the action profile according to the current eigen vectors of the structure including all sources of inelasticity and second-order deformations.
• Updating may use any number of modes, as specified by the user.

These advanced pushover features of the packages makes them ideally suited to applications in an online computer-controlled testing set-up. When undertaking a sub-structured test, where for example the soil and foundations are being tested, the software will be used to model the structural response and couple the stiffness matrix of the structure with that measured for the soil-foundation system.  The suite of geotechnical software that could be employed for this NEES facility is the non-linear site response analysis code DEEPSOIL which will be used to develop the input motion used in the simulations to account for site effects.  In the initial phase of the project, commercial software packages will be considered for modeling the soil and foundations as necessary for the sub-structured test.  Software packages under consideration include ABAQUS, FLAC3D, SASSI and QUAD4M.  MERLIN, a computation methodology/software that is currently under development by one of the PIs in collaboration with one of the Project Team members, will also be extended as part of this project.  The methodology combines conventional boundary value problem numerical analysis (e.g. finite element analysis) with autoprogressive training and neural network constitutive models. This methodology is able to simulate soil-structure interaction as well as directly "learn" and update the numerical model from experimental observations during the performance of a single test and from one test to another.  This approach may be valuable for repetitive testing of different structures on the same type of soil, or vice versa. Consideration is also being given to the use of OpenSees (Open System for Earthquake Engineering Simulators); an analysis environment developed at the PEER Center.

Visualization

Development of Data Visualization and Knowledge Integration Software

This section provides a description of tools that will be developed for the display of test data and the integration of knowledge sources for the purpose of load control and model verification and validation. Structural engineering researchers at UIUC will join forces with NCSA’s Automated Learning Group to produce these multi-function data visualization and knowledge integration tools.  In their simplest form, the tools will have the capabilities of a post-processor for a finite element analysis package; while in their more mature form, the tools will provide means for real-time numerical model updating and validation.  A description of the tools in many stages or levels in their development of this program is now provided.

In addition to other deficiencies in traditional data collection practices, a major hindrance to structural engineering research is associated with the tremendous effort that must be devoted to the analysis of raw experimental data. This step in a test is usually conducted well after the completion of the test and the results of data analysis have no impact on the current experiment. To overcome this problem, a data visualization and data analysis tool will be written that not only presents the data in a clear and remotely accessible format, but also that enables this data to be readily compared with the predictions of simulation models and further utilized to set loading protocols. To accomplish the feedback of data analysis results to the ongoing experiments (e.g., set loading protocols during the experiment), near real-time performance defined by tens of seconds per feedback update (e.g., load-step separation) will be targeted. This rate also satisfies the requirements of the planned experiments, for instance, 1,000 load steps in one to two days.

Stage/Level 1: Visualization of Experimental Test Data
The contact and non-contact instrumentation methods used in the MUST-SIM Facility will produce “continuous” and discrete strain and displacement data at thousands of locations on the surface and within the body of the test specimen. Similar to how a post-processor for a finite element program is used to display the predicted states of strain, deformation, cracking, etc., from an analysis, the MUST-SIM visualization will enable the display (with graphical manipulations) of the measured response. The visualization will make this information available to on- and off-site researchers in near real-time.

Structural and Geotechnical Software will be assembled to deal with structure-foundation-soil interaction problems.

The objective of this stage is to develop an interactive visualization tool that makes use of data mining, application portals, grid-based tools, 3D animation and other software visualization tools that become available or are developed during the lifetime of the project. The MUST-SIM project scope includes (i) assessment of existing visualization environments used in structural engineering (ii) the most effective medium, or a number of media, will then be selected for the development, (iii) development of the software modules for the above and (iv) verification of the above on the measured strain and displacement data.

Stage/Level 2: Integration and Interpretation of Multiple Experimental Test Data Sets
The measured response of each test specimen will include redundant and thus contradictory test data.  In addition, the measurement grids and individual measurements will not equally cover all regions of the test specimen.  As a result, methods will have to be developed that are able to integrate, interpolate, and extrapolate the results from all forms of measurement (knowledge sources) to create the clearest picture of the measured state of the test specimen at each stage in the loading history.  The procedure used for doing this must account for the anticipated accuracy of each form of measurements as well as selected levels of researcher certainty. The visualization will also make this information available to on- and off-site researchers in near real-time.  The interpreted information may also be used to identify critical mechanisms and set loading-control protocols.  The NCSA system of data management called Data-to-Knowledge (D2K) is proposed for this task as well as for other visualization and analysis tasks. D2K is a next-generation desktop workspace used to create data analysis and visualization applications and carry out data analysis functions. The necessary optimization and smoothing techniques will be implemented in D2K as a post-processing step prior to data display.

Stage/Level 3: Integration of Experimental Test Data with Predictions of Analysis Tools
In conjunction with the experiment, numerical simulation models will be used to predict the response of the test specimen to the imposed loadings. These predictions will provide essentially complete information for the response (strains, displacements, etc.) of all regions and points of the test specimen. Using a common reference for the geometry of the specimen in the analysis and experiment, the predictions of the analysis tool will be compared with the experimental test data at selected points, between selected points, or over selected regions. Effective measures will be developed for describing the fit or differences between the predicted and measured response. These measures will be used for subsequent numerical model verification and validation activities.

Long-Term Goal: Model Updating and Model Assessment/Verification
In the MUST-SIM sub-structuring testing facility, there will be a complete numerical model of both the test specimen and of the surrounding structure. In defining this numerical model, numerous material properties and other parametric values must be selected to define the test specimen and surrounding structure. For example, in a reinforced concrete test specimen, this may have included a specific compressive stress-strain relationship, Poisson’s ratio, shear retention characteristics, tension stiffening characteristics, assumed crack spacings, etc. After a low level of loading is applied to the test structure, it will immediately be seen that the measured response of the test specimen will be somewhat different from that which was predicted. In a truly integrated experimentation/analysis facility, a system is envisioned with a real-time analysis of this data from which the initially assumed parametric values are updated for both the model of the test specimen and the model of the surrounding structure. This approach will enable the determination of more accurate boundary conditions in the analysis routines as the surrounding structure is more accurately modeled. It will also enable more accurate next-loading-step updates to – and evaluation of – the numerical model of the structure. A challenge in developing this part of the visualization will be to create the near real-time post-processing capabilities that enable the extraction of these parametric values from the results of complex test specimens subjected to time-varying loading states. Ultimately, it is this type of integration of analysis with experimentation that will drive the development of improved and verifiable numerical models for structural behavior.

Not all of the capabilities described above for the three stages/levels are not anticipated to be developed by October 1, 2004, but rather that the visualization tools that are developed by that time will form the basis for future efforts, as well as provide direction to the research community at large for integration of experimental and analytical research activities.