December 1996 € Volume 23 € Number 6

An OR Bridge to the Future

Pontis reorganizes structures' upkeep, improvement and replacement

By David Greenfield

Most real world applications of operations research serve to save either a company or particular industry a large sum of money. However, when OR is applied to federal and state-funded projects, we all stand to profit. A prime example of the money-saving benefits that can be derived from well-applied OR is evident in Pontis -- a system designed to optimize bridge maintenance, repair and rehabilitation (MR&R) throughout the United States.

Pontis (a moniker derived from the Latin word pons, meaning bridge), was developed as a network-wide optimization system for bridge management through a joint venture between Optima Inc., San Francisco, and Cambridge Systematics Inc. in Massachusetts. Optima is an operations research firm that develops large-scale planning systems, particularly for infrastructure management. Kamal Golabi, president and chief scientist at Optima, served as principal investigator and modeler for the Pontis project.

Optima was responsible for problem definition and the OR portion of the project, while Cambridge Systematics was responsible for developing the software package for Pontis. Other contributors included the Urban Institute, which qualified users' costs, and the Technical Advisory Committee (TAC) -- comprised of representatives from the Federal Highway Administration (FHWA), the Transportation Research Board (TRB) and six states which undertook the task of defining bridge elements and feasible management actions.

Optima and Cambridge Systematics got the nod from the government to proceed with the Pontis project after years of government spending on alternative methods of bridge management had produced unsatisfactory results.

"Prior to Pontis, states handled bridge MR&R in a subjective way based on a method the FHWA had established years earlier," Golabi says. According to this method, if a sufficiency number (defects rating) of a bridge was below a certain level, the bridge became eligible for federal funds for rehabilitation. If the rating fell even lower, that bridge could became eligible for replacement. "This process was unproductive and unfair because it encouraged a lack of maintenance on the part of states and a non-optimal spending pattern of government funds," he says.

The Approach
Golabi approached the Pontis project in an entirely different way from that which had been previously used to determine bridge MR&R. "We did not try to use the data on bridges collected over the years according to the FHWA rating method or any other rating method, nor did we simply introduce an optimization technique to address the network problem as defined by individual bridges," says Golabi.

Instead Golabi's approach involved the implementation of several key ideas:

Abandoning the FHWA rating method in favor of requiring more detailed information on the conditions of all elements.
Separating maintenance optimization, which is a dynamic process, from improvement, which is a static process. (Actions such as widening or increasing the height of a bridge are improvements, while actions that fix an element that has deteriorated over time is considered maintenance.)
Defining a set of bridge elements from which all bridges in the United States can be "built."
Developing a self-learning prediction model.
Carrying out maintenance optimization by first considering the network of bridge elements and then combining the results to produce recommendations for individual bridges. "We didn't look at individual bridges," Golabi says. "We looked at the network elements of bridges. For example, instead of looking at California (the first state used for implementation of Pontis) as a combination of 24,000 bridges, we looked at it as if it were a network of girders, decks, railings, etc. This way the system arrived at optimal policies for each of the elements within the state. So we were then looking at the 160 elements that describe fairly accurately all the bridges in the U.S."
Coordinating maintenance and improvement activities prescribed by separate models.

Modeling
"As with every large scale infrastructure model I've designed," Golabi says, "you have to have a set of optimization and prediction models. You have to be able to predict how the element deteriorates and be able to quantify that. And given that you can predict the rate of deterioration and that defects can be corrected -- and knowing that every corrective action costs money -- you are able to arrive at how to best distribute your funds through the optimization model."

However, in dealing with bridges, inherent problems inevitably arise that lead to difficulties in coming up with an effective MR&R system. First, no two bridges are the same, and there are some 600,000 bridges in the U.S. In addition, every bridge consists of many different elements which affect the types of actions that can be taken and the interaction between those actions and the elements themselves. Designing Pontis so that MR&R is independent from the number of bridges in the inventory leaves the model free from the computational restrictions associated with earlier efforts.

By considering one unit of an element, the number of possible conditions that the unit can possibly be in at any given time can be defined. This makes it possible to define the possible actions that can be taken for the element in each condition, thereby effectively overcoming the dimensionality problem.

Some of the data used to initiate the Pontis project were available nationwide, such as the National Bridge Inventory (NBI) databases that each state maintains. The detailed data required to establish the Pontis models, however, were non-existent.

Because no data existed for each of the elements and how they would deteriorate, a statistical analysis was impossible. "But Pontis has a way of eliciting expert engineering judgment, translating that into transition probabilities, and improving this information via its learning mechanism," Golabi says. "As data is collected, the model gets updated in a Bayesian context and produces new deterioration probabilities, leading to more accurate results over time."

Golabi considers this one of the major accomplishments of Pontis -- that it has produced a new set of data requirements about how bridges have to be inspected and how the data is collected to increase engineers' understanding about bridge deterioration.

Affecting the Future
The success of Pontis became an impetus for the Intermodal Surface Transportation Efficiency Act (ISTEA), which requires every state to have six management systems, including a bridge management system and a pavement management system. Following ISTEA implementation, the FHWA told states that if they did not have Pontis, they must get something that is as good or better -- in effect, standardizing the evaluation of bridge efficacy. Forty-four states currently use Pontis, and others are expected, following FHWA directives, to adopt it soon.

Pontis has been adopted by the association that comprises the different state transportation agencies -- AASHTO (American Association of State Highway Transportation Officials). Adoption by AASHTO proves that Pontis is widely applicable and a success, according to Golabi. "Pontis now plays a part in the distribution of billions of dollars of federal funds for bridge MR&R and related activities," he says.

Pontis also has the ability to adapt, as it is designed so that each state can change the elements that fit their bridges and budgets the best. And although Pontis has been designed for the U.S., it can be adapted to other countries' requirements.

"Many other countries don't have the amount of data we have on our bridges," Golabi says. "However, because Pontis is based not so much on predata, but on inputting current data and keeping up with future evaluative data collection, it can be used in countries without a lot of preexisting data."

Golabi added as evidence of Pontis' adaptability that several European countries are looking at Pontis for possible implementation.

Words to the Wise
"Going into such a project, one should look at every problem individually and not try to fit a problem into a preexisting model," Golabi advises. "For example, I have developed another system for pavement management that works in somewhat the same way as Pontis, and people have tried to use this for bridge management, but they have failed. Bridge management is a different problem from pavement management, and you have to approach them differently. Basically, you have to develop a new model to address the specifics of each individual problem."

The major obstacles Golabi faced in completing this project were political ones. "There were a number of states involved that were anxious to get their hands on the product. Many were worried that since California was the pilot state, that California would get a tailor-made system and they would be left waiting. I had to assure them that once Pontis was completed, everybody would be able to use it. Originally, my plan was to design it specifically for California and then adapt it to the other states. But through interaction with TAC, it gradually became clear that the system would be designed in a way that would work for everybody."

Bottom-line: The success of Pontis in refocusing the process of bridge MR&R in the United States hinged on the age-old political necessity of meeting the expectations of everyone involved.

David Greenfield is the managing editor of OR/MS Today.

E-mail to the Editorial Department of OR/MS Today: orms@lionhrtpub.com

Lionheart Publishing, Inc.
2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com