Here at Stevens, technology is an integral part of the school: from the curriculum, the founding mission, and the name. As a part of Stevens’ multi-disciplinary innovation, research at Stevens leads the world in government, industry, and academic development. Recently, one example from the Department of Civil, Environmental, and Ocean Engineering could change the world of transportation to be safer for generations to come with the help of innovative technology: artificial intelligence (AI) analysis of the safety of bridges.
In a partnership with the University of Illinois at Urbana-Champaign, Stevens civil engineering professor Kaijian Liu and his colleagues have come together to research a revolutionary idea: a cutting-edge AI program that analyzes the health and safety of a bridge to determine its viability through an integration of different AI and machine learning processes along with multiple sources of data related to a bridge’s health.
Before professor Liu’s research, the National Bridge Inventory (NBI), an annual Department of Transportation-mandated evaluation that classifies bridges into broad condition categories, was the sole source of data for safety analyses of bridges. Professor Liu explained, “In our view, NBI data are certainly important but themselves alone are not sufficient to accurately predict dangerous conditions or failures […] because NBI data do [does] not capture information about bridge deficiencies and maintenance actions, which are detrimental to bridge conditions and deterioration.”
The difference, according to professor Liu, is using the entirety of the bridge inspection reports, which he said, “contain a wealth of detailed information about the types, quantities, severities, and causes of bridge deficiencies, as well as the types of performed maintenance and the applied maintenance materials.” Compared to NBI data, the reports allow for AI and computer programs to have much more variable data, which allows for a significant increase in the validity of long-term safety projection models.
The program the project is creating uses a recurrent neural network (RNN), where the large data sets from inspection reports and public data like temperature and weather The program virtually evolves bridge conditions far into the future and indexes the rate of safety from one, which means imminent failure predicted, to nine, which means excellent conditions. When the model was tested with over 2,600 bridges in Washington State, the program was accurate 90% of the time, nearly 20% more successful than other existing models.
The impacts of this new program are immense. Professor Liu explains that the program has significantly increased correct prediction rates and fewer outliers than standard models, allowing unsafe bridges to be identified before they fail. Additionally, by increasing the amount of data the system can use to analyze the safety factors, like traffic rates and weather, the system can only become more accurate in predicting bridge conditions. Through this data, the system can recognize patterns and predict the conditions of bridges. Professor Liu explained, “Bridges with similar geometric, structural and construction characteristics could have distinctive deterioration patterns, depending on the specific deficiency conditions of a bridge and the kind of maintenance it received — in addition to its ambient traffic and weather […] The NBI ratings alone do not sufficiently capture and differentiate the deterioration patterns of seemingly similar bridges.”
In civil engineering, bridges are perhaps the most involved, yet essential structures. Bridges like the George Washington Bridge allow nearly 300,000 people per day to cross the Hudson river. Yet, with so many cars, high wind conditions, and eroding water, bridges can be dangerous in unstable conditions. With the work of Professor Liu, unsafe bridges could be identified long before they become hazardous. Professor Liu offered, “This integrated data, from multiple sources, especially the inspection reports, make a real difference.”
