Automated Systems for Infrastructure Condition Assessment

Infrastructure asset management systems have evolved for highways, bridges, water and sewer systems, and railroads in order to organize and provide rational decision support for budgeting and prioritizing maintenance and rehabilitation. Decisions involving maintenance, repair, and rehabilitation of infrastructure systems require accurate and current information describing the conditions of these systems and their components. This information is required not only to characterize current conditions but also to project future performance and remaining life. The past $15\text{years}$ has seen significant developments in the automation of the assessment of infrastructure condition. These developments have been motivated by the recognition that infrastructure systems are too vast to be adequately covered by manual inspection techniques, and by the need to supply this condition data to asset management systems. The advances in sensor and computer technology that have occurred over this time period have enabled these developments to take place. Advances in sensor technology, such as with visual and infrared cameras, lasers, ground-penetrating radar, and ultrasound, have provided a means to capture surface and subsurface condition data at high speeds. Advances in computer technology have paralleled those of sensor technology to enable rapid digital recording of the data and high-speed processing for data interpretation. With these combined developments, infrastructure system managers now have a suite of condition survey technologies capable of rapidly and accurately characterizing the condition of their system elements. These survey technologies are now being used to automate visual inspection, to expedite the transmission and recording of this data into asset management systems, and to automate the evaluation and interpretation of this data.

The papers in this special issue highlight recent advancements in the automation of infrastructure condition assessment. Wang and Gong describe a system for collection and real-time automated interpretation of pavement video images collected at highway speed. A key feature of this work is the implementation of automated image interpretation using advanced computer processing techniques. This level of automation can minimize data storage requirements, and can rapidly produce valuable pavement surface condition data for agencies that manage pavements. Automation can also reduce cost, and thus allow for more frequent condition updates than can be collected by traditional means.

Shehab and Moselhi describe a similar system for automating the interpretation of sewerline videologs. The focus of their paper is on a neural network for detection and classification of infiltration. While pavements and pipelines are continuous linear systems that lend themselves to the types of automation described above, other systems, such as bridges, are represented by discrete units, each of which contains a similar collection of elements. To automate the inspection of these types of systems, Jauregui et al. describe a system for incorporating visual data collected on these individual elements into the bridge inspection process. The process focuses on standardization of element characterization and on bringing visual data directly into the asset (bridge) management system.

Since these automated systems generate large quantities of data, there is a need to provide some type of quality assurance to address the possibility of malfunctioning sensors or other unanticipated data-quality issues. Buchheit et al. address this issue by proposing an automated data-quality assessment procedure. The procedure seeks to identify deviations from expected data patterns, and to distinguish the causes as either actual occurrences or faulty data.

The final paper, by Sunkpho et al., exploits the generic commonalities of infrastructure inspection tasks by proposing a common platform for developing specific inspection support systems. These systems are used to gather, organize, store, and operate on condition data collected on infrastructure systems and their components. The work by Sunkpho et al. shows how specific inspection support systems can be automatically generated in response to agency specification. The paper presents a computational structure through which user agencies can implement their own criteria and standards within an overall asset management system framework.

These papers illustrate some of the potential for automation in infrastructure condition assessment. Given the need for condition information, we can expect to see more of such systems developed and implemented in the near future.