CAREER: Model Updating Cognitive Systems (MUCogS)

Funding agency: 
National Science Foundation
08/10/2009 - 00:00 to 07/31/2014 - 00:00

Model Updating Cognitive Systems (MUCogS) provide a new paradigm on model updating of structural systems. Commonly used numerical modeling techniques, such as finite or boundary elements, provide accurate representations of simple structures but often fail to represent complex structural systems. Model updating techniques are used to enhance these numerical models based on experimental data. Current model updating techniques obtain a single model that best matches the existing structure by minimizing the error between experimental measurements and data produced by the numerical model. The analyst is limited to the selection of an initial model, adjusting few algorithm parameters or “knobs” and has little control on the process performed by the algorithm itself. This research changes this paradigm by designing advances algorithms to detect several plausible solutions to the model updating problem. A trained engineer can use his/her engineering judgment to select one or several appropriated models for subsequent analysis. Through this framework MUCogS formally include the analyst on the model updating technique creating a cooperative human-computer system. MUCogS can be further enhanced by including software agents to aid the analyst in the decision making process.
Conceptual Design

Four fundamental parts are essential to fully synthesize the human-computer cooperative capabilities to create effective MUCogS as shown on the figure above:

  1. External interface
  2. Computational core
  3. Human-computer interface
  4. Human sub-system

Unlike traditional model updating strategies that follow a linear strategy (i.e., collect data, clean data, update model), MUCogS present a two-way interaction scheme between the four components.

The computational core is one of the main components of MUCogS and is the focus of fundamental research for MUCogS. The computational core will use the data obtained from sensors and create a series of plausible updated models and statistical data for the analyst. Similarly, the computational core will take instructions from the analyst to perform further calculations, request more information from the external interface and make changes to the updated model. This connects with other research currently under development at the lab.

An analyst or group of analysts driving the search will be part of the MUCogS human sub-system, which is the focus of educational activities such as the Environments For Fostering Effective Critical Thinking. The analyst will identify promising models based on the statistical information, sensor data, a pool of plausible solutions and engineering judgment. The analyst can also make changes to the numerical model of the structure, request further information from previously performed analysis, request further analysis or request information or data from the sensors. Finally, the analyst will select a model or a set of models as the final updated model.

The human-computer and external interfaces are the components that allow for the synthesis of the computational core and the human sub-system. The human-computer interface will streamline the communication between the analyst and the computer algorithms through the use of visualization tools and graphical user interfaces. Different layers of information will be used in the human-computer interface to aid the analyst’s decision making process. This information ranges from real-time raw sensor data to statistical information and plausible structural systems. Additionally, the human-computer interface will allow the user to modify the structural model, request information from the structure and perform further calculations. The external interface is designed to communicate the system with the external world. This includes data acquisition as well as communication with external databases and other MUCogS.

Broader Impacts

This research has direct application in any field of structural engineering that incorporates numerical modeling of existing structures (e.g., earthquake engineering, wind engineering and structural health monitoring). The successful completion of this research will significantly enhance an engineer’s ability to validate innovative designs, evaluate the performance and detect damage of existing structures. This advancement directly impacts state, federal and private budgets dedicated to infrastructure maintenance and replacement. The educational component of this research includes the development of engineering judgment among undergraduate students by nurturing critical thinking through the implementation of inquiry-based teaching in structural engineering classes. In addition, undergraduate research assistants will be working on the research activities during the duration of the project.