Assessing the potential value of bridge monitoring systems

Structural health monitoring (SHM) is becoming a standard feature of many major bridge projects (e.g., Catbas et al., 2013; Ko and Ni, 2005). High-profile bridges such as Stonecutters Bridge in Hong Kong have had extensive monitoring systems installed (Wong, 2004). Many bridges in the UK have or have had SHM systems installed; for example, Humber Bridge (Hoult et al., 2008); Nine Wells Bridge (Hoult et al., 2009; Schwamb, 2010; Webb, 2014); Ferriby Road Bridge (Hoult et al., 2010); Hammersmith flyover (Webb et al., 2014).

There have been numerous publications discussing how to deploy SHM systems and obtain data (e.g., Andersen and Vesterinen, 2006; BTS, 2011; Catbas et al., 2013; Feltrin, 2007; Gastineau et al., 2009; Koo et al., 2013; Kurata et al., 2013). However, the collection of data (in the bridge engineering context) is of little value in-and-of itself unless it can be used to inform and influence decision making. When discussing benefits to stakeholders, an early paper describing the role of sensor technologies for civil infrastructure management by Maser (1988) is worth revisiting. Maser (1988) wrote:

Maser's observation reveals that simply being able to measure a parameter does not mean that a specific need or piece of critical information is being obtained.

Five categories developed by Webb et al. (2015) are shown in Table 1 and describe different ways in which monitoring data can be used to provide different types of information. Table 1 also gives a description of what each type of deployment seeks to achieve. As was noted by Webb et al. (2015), many SHM deployments have aspects that can fit into more than one of the categories, namely: sensor deployment studies, anomaly detection, model validation, threshold check and damage detection.

In many monitoring situations there are a number of different stakeholder groups who may be able to benefit. In the work of Andersen and Vesterinen (2006) on the design and planning of SHM systems seven stakeholders were identified, namely: (a) authorities, (b) owners, (c) users, (d) researchers, (e) designers, (f) contractors and (g) operators. It is acknowledged that all seven stakeholders can benefit from SHM on bridges, but in different ways. Figure 1 has been proposed by the present authors to visually indicate the series of considerations that should be examined when deciding whether or not there is a prima facie case for SHM. It is important to note that the choice of technology itself, for example, fibre optic strain gauges or micro-electro-mechanical systems (MEMS) sensors, is a much later consideration compared with the more high-level considerations of ‘why’, ‘where’, ‘what’ and ‘for whom’ for example. If a prima facie case does exist, the next stage is to assess whether or not the proposed monitoring deployment could generate value to at least one stakeholder.

Having decided whether or not there is a prima facie requirement for an SHM system, the next stage is to determine if an installed system will in fact yield value. In principle, some value from any of the measurements will be delivered to some stakeholders; however, the proposed value assessment system is targeted pragmatically at owners and managers – who arguably are interested in a more immediate return on investment than, for example, researchers.

Yanev (2007) and Cole (2008) present more general reviews of bridge management and bridge assessment philosophy, respectively. A value assessment methodology for bridge engineers to decide if a proposed SHM deployment will deliver value to bridge owners is now presented. It is partly inspired by a recently published methodology for assessing the value of remote sensing technologies proposed by Vaghefi et al. (2012) and the procedure for the assignment of geotechnical reduction factors outlined in Poulos (2004).

To provide value for the bridge owner input is needed from three key players: SHM engineers, structural engineers and the bridge's owners.

• SHM engineers aim to provide instrumentation systems to produce measurement data from a structure, and quantify the expected level of accuracy and error rates.

• Structural engineers aim to provide appropriate structural models and data interpretation techniques along with meaningful ‘trigger values’ that have a sound technical basis. SHM engineers should then be confident of what they need to measure, why they need to measure it and the accuracy (or resolution) required to provide the information needed by the structural engineer. It should also be noted that structural engineers may be employed by contractors to assist with the assignment of critical values that are relevant to quality and speed of construction.

• Owners/asset managers aim to fix issues identified by the SHM system if empowered to act once informed that a trigger value or limit has been reached. They want maximum performance at minimum cost.

All three key players need to contribute to delivering value and the next section proposes a methodology that they can use in partnership. As previously stated, the focus here is on delivery of value to an asset owner and the methodology would need revision if the focus was on delivery of value to another stakeholder.

The user of the methodology assigns a value rating (VR) from 1 to 5 to the various criteria listed in Table 2. The most important phase of filling out Table 2 is arguably the writing of the comments justifying the score allocated. The VR-index is computed by simply averaging the scores given to the nine questions listed in Table 2. The computed VR-index is a simple metric that can be used to quantitatively compare different monitoring strategies and decide whether a particular SHM configuration may be worthwhile for a particular project (see Table 3).

To demonstrate the use of the value rating system that has been proposed, five bridge projects with which the authors are familiar have been used to provide hypothetical value rating scores of actual or proposed monitoring systems. All five case studies are also described in more detail in Webb (2014).

Walton Bridge is a newly constructed road bridge across the river Thames between Walton-on-Thames and Shepperton in London (further details are provided in Middleton et al., 2014; Webb, 2014). Completed in 2013, it is the sixth bridge to have occupied this site, constructed to replace two existing bridges. The new bridge consists of a steel arch spanning approximately 90 m and approach viaducts. There were originally no plans to install a monitoring system on the new Walton Bridge. This provided an opportunity to consider carefully whether SHM could add value to the project, as there were no preconceived ideas about what should be installed. Questions as to what could be measured to provide value for the client were asked. Deterioration of a number of critical components was considered. In this paper, the bridge's arch is considered in more detail.

The bridge's arch is constructed from steel and so, as with all steel structures, corrosion is a concern. Unfortunately, corrosion cannot yet be easily and reliably detected at all locations on a structure. One option is to measure relative humidity levels and if these are kept below 60% then corrosion of exposed steel is considered unlikely to occur (cf. Bloomstine, 2011). It was therefore suggested that relative humidity sensors could be installed within the steel arch and a threshold check could be used to alert the bridge's owners should the relative humidity increase to unacceptable levels. However, the arch had been designed with a protective coating expected to last 60 years and also with additional sacrificial thickness to allow corrosion to occur for an additional 60 years before causing problems. Since the risks due to corrosion have effectively been ‘designed out’ a relative humidity monitoring system would not affect the management of the bridge. Running such a system was not deemed to offer any clear benefits.

Another option considered was the use of fibre optic cables that could be installed within the arch to measure strains throughout its length. These could be used to gain a better understanding of how loads are transmitted through the bridge's arch and validate the design assumptions/predictions. However this would be of very little value to the bridge's owners, especially without a full knowledge of the input loading. Such data may prove interesting to researchers or code writers, but this would not justify the required expense to the bridge's owners.

Eventually no monitoring system was installed; however, perhaps with the benefit of hindsight, a model validation study (see Table 1) could have been useful to engineers and researchers to inform/influence future steel arch bridge designs. Table 4 shows the pre-monitoring assessment matrix filled out (albeit hypothetically) with a resulting VR-index of 1·9. Based on the rating from Table 3 the project ‘is unlikely to yield value to the asset owner/manager’. Given the results of the discussions this score seems reasonable.

The Nine Wells Bridge (Figure 2) was constructed to the south west of Addenbrooke's Hospital in Cambridgeshire in 2008 (further details are provided in Hoult et al., 2009, Schwamb, 2010; Webb, 2014). The bridge comprises three spans, each of approximately 30 m in length, and is constructed from precast, pre-stressed concrete ‘Y beams’ with an in situ concrete deck. Both the intermediate piers and the abutments are supported on piled raft foundations. The instrumentation installed on this bridge consisted of a number of fibre optic cables, which can be interrogated using a technique called Brillouin optical time domain reflectometry (BOTDR) to measure strains (e.g., Kurashima et al., 1993). This would be classified as a ‘category 1: sensor deployment study’ according to Table 1. Six of the beams in the most western span of the Nine Wells Bridge were constructed with optical fibres cast into the concrete, allowing distributed strain measurements to be taken along the lengths of each beam. Some of the details of the installation are discussed by Hoult et al. (2009) and Schwamb (2010). Unfortunately, damage during installation to some of the fibre optic cables used in this experimental deployment limited the data that could be acquired, while subsequent vandalism to the cables had rendered the system inoperable after about two years. However, the main benefit derived was from learning how to interpret the data from the new sensor systems (cf. Webb, 2014), although this took some years to realise and the real beneficiaries of this are design engineers who may be involved in future endeavours rather than to the owners/operators of this bridge. Table 5 shows the pre-monitoring assessment matrix filled out (albeit hypothetically, but after an actual project was completed) with a VR-index of 2·1 resulting. Based on the rating from Table 3, the project ‘may yield value to the asset owner/manager’. This slightly better result than for Walton Bridge is because some useful research data could become available for model validation, and hence the scores for the three ‘structural engineer questions’ were higher.

The Humber Bridge near Hull was completed in 1981 and is the UK's longest suspension bridge (Hoult et al., 2008). A dehumidification system had been installed in each anchorage chamber (where the main cables of the suspension bridge are not coated with a protective sheath, see Figure 3) to ensure that corrosion is inhibited by keeping the relative humidity below a threshold level of 60% (cf. Bloomstine, 2011). Relative humidity sensors were installed in the chambers to control the dehumidification system. Humidity levels reported by these sensors could also be viewed on liquid-crystal display (LCD) panels from within the anchorage chambers. In July 2007, a wireless monitoring system consisting of a number of temperature and relative humidity sensors was installed in the northern anchorage chamber to supplement the existing wired system (Hoult et al., 2008). Sensors were placed adjacent to the existing sensors, as well as in other locations such as suspended directly adjacent to the steel cables. This system provided the bridge manager with hourly data by way of a website, with the potential for alerting the user to any failures of the dehumidification system in near real time. Additionally it was discovered that the dehumidifiers in the two sides of the northern anchorage chamber had been configured very differently in terms of their switching threshold levels and that significant energy and cost savings could be made by optimising their operation. Although primarily conceived as a threshold check, this example also demonstrates an anomaly detection study providing a serendipitous benefit. This monitoring system has been in operation since July 2007 with minor upgrades. Although there have been outages, these were mainly attributable to battery exhaustion before a regular maintenance regime was put in place. With annual battery changes and maintenance, the system could remain operational for many further years. Table 6 shows the pre-monitoring assessment matrix filled out (albeit long after the project was initially installed) with a VR-index of 3·8 resulting. Based on the rating from Table 3, the project is ‘likely to yield value to the asset owner/manager’. This good result is due to the fact that a well-defined relative humidity threshold was (is) able to be checked reasonably reliably.

The Ferriby Road Bridge (see Figure 4) is a three-span reinforced-concrete slab bridge on the northern approach to the Humber Bridge (some of the monitoring efforts are described in Hoult et al., 2010). During a visual inspection in 2002, several defects were noted. Firstly, cracks were spotted on the soffit of the slab which, although small in size, had the potential to grow and allow moisture ingress and possibly corrosion. Secondly, many of the bridge's elastomeric bearings were slightly inclined, although it was not known whether these had been inclined since they were first installed or whether they had become inclined over time. These defects did not present an immediate cause for concern, but there was a potential benefit to the bridge manager in a monitoring system which would allow any changes or deterioration over time to be observed. A monitoring system was therefore installed to measure changes in the size of three of the soffit cracks and inclination of a number of the bearings. The crack width measurements have shown the crack openings remained constant over the three years during which the monitoring system was operational. Unfortunately in this installation the bearing inclinometers were less successful since the bearings were found to expand and contract due to temperature changes, regardless of whether the bridge deck itself was moving. It would therefore have been more useful to directly measure the displacement of the bridge deck with respect to the column, rather than attempting to infer this movement from the inclination of the bearings. This highlights the need for careful consideration during the design of any monitoring system. Table 7 shows the pre-monitoring assessment matrix filled out (albeit hypothetically but after an actual project was completed) with a VR-index of 2·3 resulting. Based on the rating from Table 3, the project ‘may yield value to the asset owner/manager’. This is a slightly better result than Walton Bridge, but clearly a less convincing result than the Humber anchorage deployment.

The Hammersmith flyover is a 622 m long, 16-span, pre-stressed concrete bridge constructed in the early 1960s (further details are provided in Webb et al., 2014) (see Figure 5). Details of the original design are discussed in Rawlinson and Stott (1962) and Wroth (1962). Longitudinal movements owing to temperature variations are accommodated by roller bearings situated at the base of each pier. The bridge is fixed at each abutment with a single expansion joint located in a span towards the centre. Various monitoring techniques have been used on this bridge since late 2009, including an extensive wired system installed by a contractor and a wireless sensor network (WSN) installed by researchers working at the University of Cambridge. Further details of the analysis of some of the monitoring data have been presented in Webb et al. (2014).

The structure's owners were concerned that deterioration of the roller bearings at the base of each pier could result in restraint against longitudinal movement. If true this meant that expansion of the bridge deck due to temperature changes would induce extra bending moments in the piers and the bridge deck. These additional stresses could potentially exacerbate other problems caused by corrosion of the deck pre-stressing tendons.

Displacement gauges were installed to monitor the longitudinal movements of each bearing in addition to a number of temperature gauges distributed throughout the structure. Analysis of a structural model determined that restraint of any single bearing would have very little effect on the movements of the remaining bearings. This allowed damage detection to be implemented by analysing the correlation between temperature and movement for each pier. It became apparent that two of the bearings were moving to a far lesser extent than would have been expected. Subsequent visual inspections discovered that movement at these bearings was indeed being restricted, confirming the monitoring system's ability to detect a problem. It should be noted that a detailed analysis of the expected behaviour of the structure was a prerequisite to the success of this monitoring system; simply installing sensors and plotting the data obtained would have produced very little information of use. Table 8 shows the pre-monitoring assessment matrix filled out (albeit after the actual project was completed) with a VR-index of 3·6 resulting. Based on the rating from Table 3, the project is ‘likely to yield value to the asset owner/manager’. This positive result is due to the fact that a good diagnosis of bearing performance was sought and achieved.

This paper proposes a set of questions that allow for an assessment of the potential value of bridge monitoring systems. The developed value assessment methodology has been trialled hypothetically on five bridge monitoring studies with which the authors are familiar. The need for the SHM engineer, structural engineer and owner all to contribute to the pursuit of potential value is central to the methodology.

For the five case studies reviewed in this paper the computed average value ratings are seen to range from 1·9 to 3·8. The relatively low scores across these deployments are a reminder that it remains difficult to get value from bridge monitoring. In spite of this, useful information can sometimes be obtained and it is suggested that this is more likely to occur if due consideration is given to the needs of the asset owner, as discussed in this paper.

This work was supported by Engineering & Physical Sciences Research Council (EPSRC) grant no: EP/I019308/1 and EP/K000314/1: Innovation Knowledge Centre for Smart Infrastructure and Construction. The authors note that this paper relies on no underlying datasets.