The overarching goal of this United States Department of Transportation (USDOT) project is to integrate data from three commercial remote sensing and spatial information (CRS&SI) technologies to create a novel data-driven decision making framework that empowers the railroad industry to monitor, assess, and manage the risks associated with their aging bridge inventories. First, automated wireless sensing technology provides a cost-effective means of remotely monitoring the capacity of operational rail bridges in a network. Second, wheel impact load detection (WILD) data provides a measure of load demand in the form of axle weights in the rail network. Third, global position system (GPS) data of train movement in combination with the WILD data provides a spatial mapping of the movement of live load within the rail network. A data-to-decision (D2D) framework is established to automate the processing of the CRS&SI data to convert data to actionable information that empowers risk-based decision-making. Within the D2D framework is the use of reliability methods to rationally combine structural capacity and demand information to calculate reliability indices (i.e., betas) for critical bridge components monitored. Lower limit states are established on the reliability indices to provide a quantitative basis for quantitatively assessing the condition ratings of a bridge span. When adding the consequences associated with bridge performance during natural hazard events (e.g., flooding, earthquakes), the data-driven decision making framework provides bridge engineers a powerful risk assessment strategy.

Figure 1: Overview of Project Thrusts

Monitoring of the Harahan Bridge

The Harahan Bridge spans the Mississippi River near Memphis, Tennessee and is part of the Memphis subjunction of the Union Pacific (UP) network. The Harahan Bridge finished construction in 1916 and is a five-span cantilever truss bridge. The carbon and alloy steel structure supports two rail lines carrying trains 4913 feet (1497.5 m) across the Mississippi River just west of Memphis. The main bridge (five spans) stretches 2550 feet (777.2 m) but to the west of the five main spans is a tower girder and viaduct that continues 2363 feet (720.2 m). The Harahan Bridge is exposed to multiple hazards: it is located in the New Madrid fault zone, may experience scour or flooding events from the Mississippi River, and is susceptible to barge collisions. Weathering and loading can also contribute to deterioration of individual structural members including corrosion and fatigue; these effects are deemed to be an aging “hazard”. Thus, both global and local sensing approaches are needed. A wireless monitoring system was installed in the main truss of the bridge and is designed to monitor global responses in addition to the response of local truss elements. Fourteen Narada wireless sensor nodes are installed to measure 36 channels of data: 12 channels of strain measured from weldable strain gages on vertical truss hanger elements including eyebar elements, 10 channels of acceleration to measure the truss response to ambient, wind, seismic and train loads, and 6 channels of acceleration measuring the transverse response of 6 parallel eyebar elements. Data is communicated by Zigbee to two base stations installed at the bridge where cellular connections exist. The base stations transmit recorded data to servers at the University of Michigan for data processing. Data is collected by triggering when geophones at the base stations measure low levels of vibrations on the bridge. Finally, the sensor nodes and base stations are solar powered using solar panels.

Global bridge accelerations are used to perform modal analysis of the bridge for updating finite element models of the bridge developed in CSi SAP2000. The strain measurements are used assess the local response of four truss elements: two vertical eyebar tensile truss elements and two vertical built-up box truss elements. The six acceleration signals on one of the two eyebar elements are used to assess the primary modal frequency of the individual eyebar links to identify tensile loads in the links and to assess the proportionality of tensile load across the six parallel elements. Tenile loads and local strain measurements are then used to assess the reliability of the eyebar elements to various limit states including fatigue accumulation.

Monitoring of the Parkin Bridge

The Parkin Bridge is a short-span railroad bridge located in Parkin, Arkansas. The bridge was constructed in 1935 and has undergone several renovations through its history. The bridge carries one railroad track between Bald Knob, Arkansas and Memphis, Tennessee and is owned and operated by UP. The entire bridge spans 56 feet (17m) with 3 spans. The two wing spans are built as thick reinforced concrete slabs while the mid-span consists of four steel girders. The mid-span is 24 ft (7.32 m) supported by reinforced concrete piers with Tyronza Street running underneath (with a clearance of 9.75 ft (2.97 m)). Only the mid-span is of the interest in this project as it experiences the largest structural responses and is more vulnerable to over-height truck collisions that can occur from the road beneath. The motivation for monitoring the Parkin Bridge is to provide bridge response data for estimation of train speeds, static axle loads, dynamic load factors, and to count the number of rail cars. These parameters will serve as inputs for performing a data-driven load rating for the bridge. The potential hazards that the Parkin Bridge is exposed to include: excessive train loads, aging (i.e., fatigue, corrosion), seismic loads, and vehicular collision (i.e., traffic colliding with the over-pass bridge girders). Among those hazards, excessive train loads and collision were the two primary hazards investigated in this project. A wireless monitoring system was installed on the center short span of the bridge to monitor global responses in addition to the response of the four steel girders. Sixteen Narada wireless sensor nodes are installed in the Parkin Bridge to collect data from 30 sensor channels. A total of 18 weldable strain gages are installed on the bridge girders to measure flexural girder responses to train loads and collissions. Eight channels of acceleration are installed to also measure the span response to train loads, collissions, and seismic events. Finally, two piezoelectric elements are installed at the girder bearings to identify individual train axles entering and exiting the bridge. Similar to the Harahan Bridge, the wireless sensors communicate data via Zigbee to a base station where a cellular connection is establihsed. All sensor nodes and the base station are powered by solar panels.

The strain gage measurements during train loads are used to use the instrumented bridge span as a weigh in motion system that can estimate the axle loads of the train similar to a WILD station. This is used by performing filtering and fitting an influence line extracted from a finite element model of the Parkin Bridge constructed using ABAQUS. The strain measurements are also used to extract dynamic load factors (i.e., impact factors). Using estimations of train axle weights, train speed, and dynamic load factors, a data-driven load rating is performed for each train event. The load rating analysis is used to estimate the residual capacity of the bridge for further risk analysis. The project also utilizes the data-driven load rating analysis to assess how the railroad could increase the speed and weightage crossing the bridge to generate more revenue thereby providing a quantitative cost-benefit analysis for the investment in instrumentation.

Project Accomplishments

1. Two permanent wireless monitoring systems were designed and deployed on the Harahan and Parkin Bridges. These bridges are critical elements of the Memphis sub-junction and were selected due to their exposure to natural (e.g., seismic, aging) and man-made (e.g., vehicular collisions) hazards. The two bridges were instrumented with 36 and 30 wireless sensing channels, respectively, ranging from strain to acceleration measurements. As part of these monitoring systems, a real-time alert service was implemented to provide email alerts to UPRR engineers and inspectors regarding structural responses exceeding pre-defined response thresholds.

2. A relational database was developed to provide a comprehensive data repository to store inspection data, CRS&SI sensor data and analytical models of railroad bridges. Unifying these three forms of data/information empowers more extensive data analysis to assess the health of railroad bridges and aid railroads with risk management of their networks.

3. A reliability framework was created to assess the health of bridge components monitored. The reliability index calculated using the first-order reliability method was found to be an ideal scalar metric for assessing component health. Lower limit states were defined to trigger inspection and maintenance of bridge elements using monitoring data. The reliability index, when coupled with the consequences of exceeding defined limit states, allows owners to assess and manage the risks associated with their bridges and rail networks. The reliability framework was successfully validated on the Harahan Bridge.

4. The loads imposed on short-span rail bridges were quantified using long-term monitoring data. Specifically, the maximum static response and dynamic load factor as a function of train velocity were assessed for the Parkin Bridge. A novel approach to bridge load rating was developed using structural monitoring data; results indicated AREMA-specified load rating procedures are likely conservative for well-maintained rail and bridges.

5. An extensive return-on-investment (ROI) analysis was performed to assess the return rate for railroads that elect to invest in structural monitoring systems. Using the data-driven load rating methodology developed, the revelation of greater bridge capacity allows railroads to carry larger loads at higher train velocities. This ensures the return on investment is positive and reaped rapidly after technology adoption.