Decision Making with Big Data

---

Bayesian Inference Methods for Model Identification

This project makes use of measurements from a structural system to learn the equations and parameters characterizing the system, further to damage detection and health monitoring. Bayesian inference methods are an attractive approach as they take into account uncertainties in the capacity and demand, providing predictions in a probabilistic format, thus enabling a more accurate performance assessment of the systems of interest. One of the major challenges in any damage detection algorithm that employs a base mathematical model is non-uniqueness of the obtained solution. In addition, noisy and incomplete measurements make the damage detection difficult in a structure. The Bayesian framework is well‐suited to address such non-unique, ill‐conditioned problems, where not all the parameters can be learnt from the available noisy data. This project capitalizes on using the strain data obtained from a full-scale experiment to inform the priors employed for damage detection of reinforced concrete girders using a hierarchical Bayesian framework.

---

Predictive Models for Life-Cycle Analysis and Management of Transportation Assets

The main objective of this project is to utilize life-cycle cost analysis (LCCA) as a decision-support tool for deteriorating bridges. Bridge management systems aim to assess investment decisions and identify the most cost-effective improvement alternatives to ensure the safety and serviceability of bridge networks. To achieve this goal, a suite of recently developed deterioration models paired with the cost data for highway bridges in the State of Iowa are utilized to conduct LCCA using both deterministic and probabilistic approaches. This predicts the deterioration of bridges over time and helps with planning the necessary maintenance and repair activities.

Back to Top