McC BarsMcCormick School of Engineering HomeMcCormick School of Engineering HomeNorthwestern University Homepage

Infrastructure Technology Institute

ITI Researchers discuss Statistical Modeling of the Hurley Bridge


Professors David Corr and Pablo Durango-Cohen

Why are the two of you a good team? What specialized skills do you each bring to this project?

Pablo Durango-Cohen has been working in management of infrastructure for some time, particularly on developing statistical techniques to support decision-making for pavements. A natural progression of this work was to move to more complex infrastructure elements such as bridges and tunnels, but the physical behavior of these elements is quite a bit more complex and less accessible than for pavements. Dave Corr is an expert on structural analysis and diagnostics, and he has been working on structural health monitoring with ITI’s Research Engineering Group, both of which help bring physical meaning to the statistical work that Pablo has initiated. Dave also has studied statistical and probabilistic methods in the past, and he has an appreciation for the power they have to facilitate informed decisions. It also helps that Dave and Pablo’s offices are adjacent, so new research ideas never need to travel far.

If you weren’t doing this type of analysis, what kind would you be doing? Why is doing things in this way better? What are the advantages?

Typical structural health monitoring research focuses on the installation of hardware, the process of data collection, and uses of the resulting data, such as calibration of finite element models to give deterministic predictions of structural behavior. This approach provides a direct link between the data and physical source of the data, but it is hard to interpret variability in the data and predictions. By adding statistical methods, we can filter out sources of variability that might otherwise manifest themselves as errors in the finite element approach. However, on their own, statistical methods lack a connection to the physical behavior of the system that produces the data stream. The approach we take incorporates elements of both approaches to monitoring and management of infrastructure.

How do deployment partners benefit from this type of data analysis? How does ITI benefit?

Deployment partners who work with us on health monitoring projects quickly gain the advantage of robust data streams that provide a large amount of data. Before this project, however, we left interpretation of that data to the deployment partner, and there was always the potential to overwhelm the end-user with an impenetrable mountain of data. The techniques we are developing takes the data streams one step further towards the deployment partner’s goal: decision-making and management of infrastructure. Now, instead of simply providing a stream of data, we provide a stream of data and a rational way of distilling and interpreting what it has to say. Throughout its history, ITI has supported research both in health monitoring, and in infrastructure management. This project is an integrator of two areas for which ITI has been a leading research institution for over a decade.

Will these methods advance the way we monitor structures now?

These methods are at the forefront of how we monitor structures. Sensor and data acquisition technologies are so advanced at this point that they can measure a vast array of physical quantities, and at extremely high frequency, and the resulting data meets most end user needs. It is the initial selection of the desired data streams through physical interpretation, and the post-processing of these data through statistical methods, that push the envelope of how structures are monitored by infrastructure professionals today.