State-Of-Research on Performance Indicators for Bridge Quality Control and Management

Zanini, Mariano Angelo; Faleschini, Flora; Casas, Joan Ramon

doi:10.3389/fbuil.2019.00022

REVIEW article

Front. Built Environ., 13 March 2019

Sec. Bridge Engineering

Volume 5 - 2019 | https://doi.org/10.3389/fbuil.2019.00022

State-Of-Research on Performance Indicators for Bridge Quality Control and Management

$\r\nMariano Angelo Zanini*$ Mariano Angelo Zanini¹^*

Flora Faleschini^1,2

Joan Ramon Casas³

¹Department of Civil, Environmental and Architectural Engineering, University of Padova, Padova, Italy
²Department of Industrial Engineering, University of Padova, Padova, Italy
³Department of Civil and Environmental Engineering, Technical University of Catalonia, BarcelonaTech, Barcelona, Spain

The present study provides a review of the most diffused technical and non-technical performance indicators adopted worldwide by infrastructure owners. This work, developed within the European COST Action TU 1406—“Quality specifications for roadway bridges, standardization at a European level,” aims to summarize the state-of-art maintenance scheduling practices adopted by bridge owners, mainly focusing on the identification and classification of the most diffused performance indicators (PIs). PIs are subdivided in technical and non-technical ones: for this latter subclass, PIs are classified in environmental, social and economic-targeted. The study aims to be a reference for researchers dealing with performance-based assessments and bridge maintenance and management practices.

Introduction

Roadway infrastructure asset management aims at define the optimal maintenance strategies required in order to ensure the fulfillment of a desired performance level, thus achieving a pre-defined performance goal. Performance levels are usually assessed with the so-called performance indicators (PIs), representing an objective technical-based metric wherewith a rational ranking of maintenance intervention needs can be derived. PIs can be currently derived from visual examination, or based on results coming from the execution of non-destructive tests or the installation of temporary/permanent monitoring systems. In addition, PIs can be defined at different levels (i.e., component, system and network level) and can be classified in qualitative- or quantitative-based.

Once a PI has been assessed, it has to be subsequently compared with its related performance goal (PG), to evaluate if, for the analyzed bridge, the quality control (QC) plan is accomplished. Different approaches for the quantification and use of PIs are currently adopted in European countries regarding the quality control plans for roadway bridges. For this reason, the main aim of COST TU 1406 Action is to develop an overall homogenization among Countries, bringing together research and practicing communities, to establish a unified European guideline in this field.

Hence, the present study provides an extensive review about operational PIs currently in use for bridges, and research PIs under investigation by several research groups worldwide, that may have a feasible application for developing QC plans of roadway administrators in the next future. In particular, a special focus is given to not-purely technical PIs (e.g., environmental, social, economic) that might be considered in the development of a sustainable QC plan. Figure 1 shows how this paper is organized, where red text and numbers indicate the section and subsection labels. A total of 263 were surveyed to extract key information used to depict the present state-of-research: Table 1 lists the subdivision of documents by type, showing how more than two thirds of references come from international journal papers.

FIGURE 1

Figure 1. Workflow of the presented review on PIs for bridge QC and management.

TABLE 1

Table 1. Type of sources analyzed in the present review.

The first part of the work is devoted to describe tools currently used to quantify the technical PIs for existing bridges. The second part deals with a comprehensive review of the technical and non-technical PIs, used both in practical and research activities. Lastly, an overview of decision-making approaches currently in use for handling different types of PIs is detailed, and a discussion on present literature gaps and future developments in this field of research is addressed.

Performance Indicators' Assessment Tools

As we are mainly interested at reviewing those PIs that can be objectively quantified, it is first necessary to give some insights on the tools currently adopted for the assessment of technical PIs for bridge management. Four different macro groups can be identified: visual surveys, probing, non-destructive techniques (NDTs) and structural health monitoring (SHM). Figure 2 shows a sketch of the four categories of NDTs that may be used for assessing technical PIs of existing bridges.

FIGURE 2

Figure 2. Sketch of the tools currently in use for the assessment of technical PIs.

Visual Surveys

The use of visual surveys is the most diffused method for assessing structural damage among Bridge Management Systems (BMSs). Indeed, traditional BMSs are based on condition rating systems combining information from visual inspections on bridge components into overall bridge condition rates. Usually, trained technicians carry out inspections, compiling handwritten records during field surveys.

As reported in Moore et al. (2001), Caner et al. (2008), Avsar et al. (2012), and Tenzera et al. (2012), visual inspections suffer for important variability in the results due to subjectivity. Subjectivity in visual inspection outcomes may be mainly attributed to different experience levels among trained inspectors that lead to varying perception of the severity of an observed damage. In addition, environmental conditions that can be present at the time of the survey execution may also affect its outcomes, e.g., due to differences in light intensity and visual resolution of different colors.

Some efforts were carried out in this context to improve the effectiveness and accuracy of visual inspections. Sunkpho and Garrett (2003) illustrated the benefits coming from the use of speech recognition in the execution of visual inspections in snowy cold regions, instead of classic handwriting on paper, showing how it may reduce the risk that inspectors will meet with accidents. Kim et al. (2008) proposed other robotic solutions for damage detection, also with the combined use of wall-climbing and flying robots: such techniques allow carrying out safer damage detection operations, and are powerful tools when dealing with large bridges and hardly accessible viaducts. Oh et al. (2009) presented an interesting application of a robotic system able to detect automatically cracks, demonstrating the effectiveness of this solution able to overpass the problem of subjectivity of bridge inspectors. Koch et al. (2015) and Omar and Nehdi (2017) detailed interesting and up-to-date review of computer vision-based defect detection techniques for bridges.

Probing

Probing is required when a quantitative characterization of main material properties is required, and it commonly consists in the extraction of samples, to be subsequently characterized via laboratory tests (e.g., when dealing with the characterization of steel or concrete stress-strain curves) or with in-situ tests (e.g., single-double flat jack systems, see Pellegrino et al., 2014). Depending on the material of each structural element, different types of probing should be carried out. Reinforced concrete members may be investigated with pull-out tests, with concrete cover removal to check the presence of corrosion on steel reinforcement rebars, with concrete coring for assessing both mechanical and durability-related properties, and extraction of steel rebars. Masonry bridges can be investigated via the execution of mortar penetration tests, in-situ single/double flat jacks, and endoscopic tests. Probing for existing steel bridges consists in the extraction of dog-bone samples from steel profiles, and the removal of bolts to be further tested with laboratory tests.

Non-destructive Techniques

Non-destructive techniques (NDTs) are another tool that can be used for the detection of cracks and deterioration from steel corrosion, providing information that cannot be derived with a simple visual examination of the structural elements. Nevertheless, they are usually less adopted, since they are more cost- and time-consuming with respect to visual inspections, especially due to the need of lane closures (Hiasa et al., 2017). In addition, sometimes difficulties may arise in data interpretation as evidenced in Kee and Gukunski (2016). More in general, it is of crucial importance to know the implicit uncertainties in the implementation of NDTs, since such issue can severely impact on a reliability of a risk-based inspection plan (see Hesse et al., 2015). In few countries, existing bridge condition assessment procedures include NDT surveys, but their periodical execution is still rare. NDTs results are generally used to assess in a more reliable way the most suitable condition state of a deteriorated component (Pushpakumara et al., 2017). In this regard, Gucunski et al. (2012), evidenced how NDTs can be used for a proper segmentation of a bridge structure in subparts characterized by similar condition states, with the final aim of helping inspectors to make a more objective rating. Omar et al. (2017a) presented an interesting study providing a procedure for an integrated defect-based condition rating procedure for RC bridges able to integrate information derived from the implementation of NDTs in terms of extent and severity of damage into a rating system via the use of a fuzzy approach.

However, NDTs are often badly correlated with material properties (like in the case of probing), and therefore their outcomes cannot be easily used in a further quantitative reliability assessment (Breysse et al., 2011; Breysse, 2012; Sbartai et al., 2012). A review of NDTs and strategies for the optimization of their implementation in BMS can be found in the results of the EC funded project “Sustainable Bridges” with particular emphasis to automatized solutions (Sustainable Bridges, 2007). In Jensen et al. (2006), a general discussion with possible implementation of NDTs in the framework of principal, general and special inspections was presented. In recent literature, Rehman et al. (2016) presented an interesting review of existing NDTs applied for defect detection in reinforced concrete bridges. Several comparative studies were conducted to emphasize merits and drawbacks of alternative NDTs. Among others, Yehia et al. (2007) conducted comparative tests on specimens using the three most commonly diffused NDTs in United States [i.e., Ground Penetrating Radar (GPR), Impact Echo (IE) and Infrared Thermography (IT)] evidencing how GPR seemed to be the most promising technique in terms of accuracy and time consumption. Oh et al. (2013) illustrated the results of an overpass condition assessment via the use of three NDTs (i.e., IE, IT, and chain drag) with the aim to compare their practicality and ability in detecting shallow lamination. Omar et al. (2017b) proposed a fuzzy methodology for assisting bridge stakeholders in a rational appraising of the most suitable NDT to be used given time- and cost-constraints. Some efforts were recently oriented in the development of automatized solutions. The US Federal Highway Administration in the framework of its Long Term Bridge Performance (LTBP) Program, developed a robotic assisted bridge inspection tool called RABIT (Gucunski et al., 2015), able to provide a comprehensive condition assessment of concrete bridge decks via the adoption of different NDTs integrated into a robotic platform.

Structural Health Monitoring

Structural Health Monitoring (SHM) is the most advanced type of NDT and it is usually implemented with the aim to characterize the main modal parameters, that will be following linked to a structural model in order to assess the safety level of an existing bridge against static and dynamic loading (Tecchio et al., 2013; Morbin et al., 2014; Prendergast et al., 2018). Extensive literature on SHM techniques and related algorithms can be found with reference to damage detection for bridges. Salawu (1997) presented a first review of existing SHM methods, followed by other state-of-art works (Carden and Fanning, 2004): among others, Hsieh et al. (2006) and Fan and Qiao (2011) reviewed different types of vibration sources and damage identification methods discussing in detail merits and drawbacks. Farrar and Jauregui (1998) compared the use of five alternative damage identification algorithms for a bridge case study. Over the last decade, a significant number of innovative sensors were introduced in civil engineering applications in order to implement continuous monitoring and real-time assessment of bridge performances. Among others, Nair and Cai (2010), Elfergani et al. (2013), and Behnia et al. (2014) illustrated the state-of-art on acoustic emission monitoring of bridges. Lopez-Higuera et al. (2011) and Ye et al. (2014) presented comprehensive reviews in the use of fiber optic-based sensors in SHM systems, Casas and Cruz (2003) described the application of intensity modulated and spectrometric fiber optic sensors for monitoring temperature, load measurements, as well as strain, corrosion and cracking of a pre-stressed bridge case study. Barrias et al. (2016) reviewed the current applications of distributed fiber optics for bridges and civil engineering structures. Lynch and Kenneth (2006) presented a review of different wireless sensors adopted in civil engineering applications. Digital image correlation (DIC) can also be used as alternative to traditional sensors in SHM applications, and consists in an optical measurement technique able to record tridimensional deformation via digital photography. One of the main advantages of DIC is that it does not require any equipment for the installation of components as well as wirings like in traditional SHM systems. Some applications of DIC in SHM applications can be found in Stephen et al. (1993), Bell et al. (2012), Nonis et al. (2013), Dworakowski et al. (2015), and Pan et al. (2016).

However, SHM systems are often costly, and therefore their application has to be carefully designed and justified in case of strategic bridges: in this regard, Guo et al. (2004) illustrated a possible use of genetic algorithms for the definition of the optimal placement of sensors in SHM applications. In addition, modern long-term SHM systems collect huge amounts of data that have to be adequately post-processed (Soyoz and Feng, 2009): the development of fast algorithms and new metrics represents therefore a relevant topic of research for advancing in this field. Among others, genetic algorithms and machine learning techniques (Nick et al., 2015; Liang et al., 2016) can be used to handle huge amount of data deriving from long-term SHMs, also taking into account effects of incomplete measurements (Marano et al., 2011) and operational/environmental variability over time (Figueiredo et al., 2011). Results of SHM are usually too linked to quantitative condition assessment: in this regard, interesting case study applications were presented by Catbas et al. (2008), Frangopol et al. (2008), Orcesi and Frangopol (2010), Liu et al. (2010), Orcesi et al. (2010) considering the reliability index as quantitative measure to be linked with monitoring data.

Technical Performance Indicators

Once classified tools available for assessing bridge condition, a detailed overview on existing technical PIs is presented. In the framework of a modern BMS, one of the key steps is damage assessment: usually it is expressed adopting suitable technical PIs, a metric for defining a qualitative/quantitative judgment via the use of the abovementioned tools on the bridge component/system condition state. In general terms, technical PIs can be subdivided in two main categories: operational and research indicators. While operational technical PIs are commonly used in practice by engineers of road agencies dealing with large stocks of bridges, research technical PIs are more refined metrics developed by academics, and whose practical application is still ongoing. Patidar et al. (1991) first gave a clear definition of technical PI in BMS applications, highlighting how it should have the following properties:

• Appropriateness: PI “should be an adequate reflection of at least one agency goal or objective” (Patidar et al., 1991);

• Comprehensible and defensible: PI “should be clear, simple, and concise in its definition” (Patidar et al., 1991) as well as in its method of computation;

• Comprehensive: PI levels “should cover the full range of possible consequences” (Patidar et al., 1991);

• Dimensionality: PI “should be able to capture the required level of each dimension associated with the decision-making problem, and it should be comparable across different time periods or geographic regions” (Patidar et al., 1991);

• Measurability: PI should be objectively measured;

• Predictable: it “should be possible to reliably determine future PI levels” (Patidar et al., 1991) like with the use of forecasting models;

• Realistic and operational: PI should be reliably measured without excessive effort or time;

• Unambiguous: PIs should be clearly defined and their metric should be directly related to the consequences.

A detailed overview of the operational and research technical PIs actually available is reported in the following, describing in detail all PIs reported in Figure 3.

FIGURE 3

Figure 3. Sketch of the investigated technical PIs.

Operational Technical Performance Indicators

Regarding operational technical PIs, the most commonly adopted PI is a qualitative condition rating (also called condition state or condition value, in a numerical scale, e.g., 0–5 or 0–9), and usually assigned during a visual survey (Gattulli and Chiaramonte, 2005). Several BMSs, like BRIME in Europe or PONTIS and BRIDGIT in the United States, were thus developed in past decades based on data mainly collected during visual surveys and adopting mathematical and statistical forecasting models for maintenance planning (Austroads, 2002). Several researchers (e.g., Catbas and Aktan, 2002; Mishalani and Madanat, 2002; Suksuwan and Hadikusumo, 2010; Adey and Hajdin, 2011; Aflatooni et al., 2013; Fernando et al., 2013; Nasrollahi and Washer, 2015; Denysiuk et al., 2016; Zanini et al., 2016a, 2017a; Quirk et al., 2017) proposed frameworks for the optimal maintenance planning on the basis of visual-inspection data, mainly considering environmental deterioration, and in some cases also consequences of natural hazards (e.g., Valenzuela et al., 2010; Fernando et al., 2015). Critical issues of condition rating-based BMSs are represented by the subjectivity in the judgments provided by the inspectors, non-linear relationships in damage progression over time, unbalanced availability of records, missing data. Visual inspections are highly subjective as noted by Tenzera et al. (2012), who pointed out that inspectors with different expertise levels can label the same bridge with a different rating. Hence, this implies that the same judgment from different bridge inspectors may result in a significant dispersion of condition rating results. Moore et al. (2001) analyzed this aspect, i.e., the reliability of visual inspection methods in USA, highlighting how external environmental aspects (e.g., presence of traffic, accessibility of the bridge members and connections, wind speed) can significantly affect dispersion in condition ratings. These observations highlight the need for improving inspector training programs and condition rating procedures. In addition, as explained in Wang et al. (2007), even with perfectly trained inspectors, different assessments can be obtained depending on weather conditions (inspection performed during sunny, cloud or rainy days). In this regard, Vanderzee (2004) evidenced how future BMS will tend to substitute a subjective process with an objective one, by switching the assessment tools from the classical visual survey to more sophisticated mixed NDT/SHM solutions. The use of probabilistic approaches (Zambon et al., 2017), also able to link the probability of damage detection with a certain condition rating (Frangopol et al., 1997; Kim et al., 2013), might represent viable solutions for properly handling such type of uncertainties. In this context, the implementation of fuzzy-logic tools (Zhao and Chen, 2002; Kawamura and Miyamoto, 2003; Sasmal et al., 2006; Pan, 2007; Sasmal and Ramanjaneyulu, 2008; Tarighat and Miyamoto, 2009; Liu et al., 2017) or evidential reasoning approaches (Wang and Elhag, 2008; Ayati et al., 2012; Bolar et al., 2013) were also explored in recent scientific literature, as alternative approaches for accounting uncertainties in a simpler way than via fully probabilistic analyses. Li and Burgueno (2010) presented and interesting study where different soft computing methods were applied to visual surveys, in order to reduce uncertainties and thus calibrating reliable deterioration forecast models. In addition, condition-rating systems cannot provide a clear structural safety judgment, since no quantitative evaluation is done both from resistance and loading sides. For this reason, it should be necessary to couple them with safety indicators in a process of multi-variate optimization, as proposed in Neves and Frangopol (2005).

Research Technical Performance Indicators

Several research technical PIs were proposed in the last decades, with the aim to optimize maintenance planning of aging bridges, mainly based on the quantitative evaluation of the structural safety, usually expressed in terms of probability of failure for a given limit state function, considering both load and resistance characteristics. Saydam and Frangopol (2011) and Zhu and Frangopol (2012) proposed reviews of the most valuable research technical PIs, also taking into account their time-dependency. Among the various proposals, main research technical PIs are:

• Structural reliability: it is one of the most common research technical PIs, quantifying the probability of failure for an investigated component/system (Tabsh and Nowak, 1991; Estes and Frangopol, 1998; Frangopol et al., 2001). Reliability can quantitatively take into account load models and resistance of structural elements, also considering deterioration phenomena that may affect safety over time (Val and Melchers, 1997; Stewart and Rosowsky, 1998; Vu and Stewart, 2000; Kong and Frangopol, 2003). Reliability can also be adopted when dealing with natural hazards like earthquakes (Choi et al., 2004; Duenas-Osorio and Padgett, 2011), flooding (Johnson and Ayyub, 1992; Muzzammil et al., 2008), hurricanes (Padgett et al., 2012; Ataei and Padgett, 2013). A comprehensive review of the reliability-based PIs is provided by Ghosn et al. (2016a);

• Cumulative probability of failure: a PI that quantifies the probability that the time to failure of a component is less than a generic time interval value t, and is calculated starting from the probability density function of the time to failure (Hoyland and Rausand, 1994; Okasha and Frangopol, 2010a);

• Survivor function: this PI estimates the complement of the cumulative probability of failure, and provides an estimation of the availability, i.e., the value of the probability that a component will not fail before a generic time instant t (Leemis, 1995);

• Hazard rate function: a PI providing a measure of the instantaneous failure rate of a structural component (Ramakumar, 1993), defined as the conditional probability that the component will fail in a future time interval, given the fact that it has survived until the present time instant (Mauch and Madanat, 2001);

• Structural redundancy: this PI represents an estimate of warning prior to system collapse (Okasha and Frangopol, 2010a), or in other terms, the ability of a structural system to continue carrying load after the failure of one structural component, e.g., like a column (Frangopol et al., 1992). Several literature works presented formulations for assessing redundancy, but no agreement has been reached in the most suitable metric to be used for such PI. Frangopol and Curley (1987) first proposed to adopt reliability metric for quantifying redundancy. Hendawi and Frangopol (1994) defined redundancy as the ratio between the reliability index of the intact system in a generic time instant and the difference of reliability indexes calculated for the intact and damaged system at the same time instant (see also Okasha and Frangopol, 2009, 2010b). Such difference can be seen as a measure of the availability of system warning before the entire structural system failure (Anitori et al., 2013). Ghosn et al. (2016b) extended the field of application of structural redundancy also at the network-level;

• Structural robustness: PI quantifying the ability of a structural system to suffer damage induced by the occurrence of an extreme action (Ghosn and Moses, 1998; Liu et al., 2000; Saydam and Frangopol, 2011; Sorensen et al., 2012). Starossek and Haberland (2011), Anitori et al. (2013) and Cavaco et al. (2013) summarized different measures (deterministic, probabilistic and risk-based) for this indicator. Recently the concept of robustness has been extended to the case of systems under deterioration occurring progressively due to aging and environmental effects (Baker et al., 2008). For instance, Biondini (2009) proposed a time-dependent measure of robustness intended to quantify the susceptibility to damage increases at during the service-life of a structural system. On the other hand, Cavaco et al. (2013, 2017, 2018) proposed a time-independent measure of robustness with the aim to quantify the susceptibility to damage in the whole service life of the structure. According to Maes et al. (2006) present three robustness measures: one related to the residual system strength, one to the residual structural reliability and the third one taking into account the risks of all the system consequences of failure. A practical application of the first approach considering the system capacity as the performance objective to the case of a railway bridge is presented in Wisniewski et al. (2006);

• Structural vulnerability: this PI is a key-measure used to define the susceptibility of a structural component or a system to some external natural or man-made action (Agarwal et al., 2003; Haimes, 2006). Different approaches for quantifying this PI were proposed in literature: Lind (1995) proposed a structural vulnerability index based on the ratio between the failure probability of the damaged system and the failure probability of the intact system; The concept of vulnerability is often treated in literature studies focusing on the structural response of bridges subject to natural hazardous actions like floods, earthquakes, hurricanes (e.g., see Morbin et al., 2015; Zampieri et al., 2016). Some research in this field was oriented on the assessment of the interaction between natural aging and structural vulnerability, with the aim to quantify the increase of vulnerability related to the development of natural deterioration phenomena (Choe et al., 2009; Ghosh and Padgett, 2010; Simon et al., 2010; Choine et al., 2012; Zanini et al., 2013; Kumar and Gardoni, 2014);

• Structural risk: a PI quantifying the combined effect of actions, probability of failures and related consequences or disaster in a given context (Adey et al., 2003; Ellis et al., 2016). Saydam et al. (2013) presented a risk-based methodology able to take into account direct and indirect losses for highway bridges considering a Markov model for the prediction of bridge performances over time. Many efforts were done in recent years in assessing risk for bridges, in particular with reference to the occurrence of natural hazardous actions. As examples, considering seismic events, Ghosh and Padgett (2011), Alipour and Shafei (2016) and Zanini et al. (2016b, 2017b) investigated the time-dependent variation of structural risk due to seismic actions;

• Structural resilience: this PI estimates the ability of a system to recover its original functionality after the occurrence of a hazardous event. With specific reference to seismic events, Bruneau et al. (2003) defined seismic resilience as “the ability of social units (e.g., organizations, communities) to mitigate hazards, contain the effects of disasters when they occur, and carry out recovery activities in ways that minimize social disruption and mitigate the effects of future earthquakes,” illustrating also a framework for quantitatively assessing it. Other approaches to quantify resilience can be found in Cimellaro et al. (2010), Zobel (2011), and Bocchini et al. (2013). Recently, Zhang et al. (2017) have proposed two resilience metrics, namely the total recovery time (TRT) representing the rapidity of the restoration process and the skew of the recovery trajectory (SRT), directly linked to the efficiency of the restoration process. The latter aims to capture the characteristics of the recovery trajectory that relates to the efficiency of those restoration strategies considered. Based on them, the optimal restoration schedules of a bridge network following extreme events are decided. It is worth mentioning that not only the total time to recover is important, but also the sequence of recovering activities that should prioritize the repair and, hence, put into service the critical nodes of the network at the early stages after the extreme event.

Also for research technical PIs, probabilistic approaches are required for accounting uncertainties in the definition of the main input parameters. Several researchers stressed the importance of using a probabilistic approach when dealing with the assessment of bridge performances: among others, Ellingwood (2005) highlighted how probabilistic risk analysis methods allow estimating and thus managing uncertainties when dealing with structural safety. Knowledge of the probability of satisfactory performance over a certain time interval is a key PI, and it needs to be compared with some performance objectives. However, the definition of acceptance criteria is still not adequately investigated, since also performance goals have to be expressed in terms of probabilities, losses, or some combinations of these metrics. In this regard, an interesting fully probabilistic reliability analysis taking into account also deterioration forecasts and synthetizing results in terms of different reliability indexes through the use of radar charts was carried out by Strauss et al. (2014).

Correlation Between Different Technical Performance Indicators

Given the substantial differences between operational and research technical indicators, many researchers tried to find correlations between them. Anitori et al. (2014), as example, analyzed potential relationships between robustness and condition ratings of existing bridges, with the aim to correct rating with data derived by robustness assessment to make it dependent on the system behavior. Deco and Frangopol (2010) proposed a condition-based approach describing lifetime deterioration of reinforced concrete (RC) bridge decks. The study evidenced how the combined use of condition and reliability indices is a powerful tool, especially when it is applied to RC bridge decks under corrosion. Furthermore, in the case of RC decks under corrosion, the correlation between condition and reliability was demonstrated. Load rating factor is also a commonly used operational performance indicator for bridge capacity and not only condition. Estes and Frangopol (2005) performed both a load rating analysis and a reliability analysis on the same highway bridge, concluding how a direct correlation between the two methods is lacking, since reliabilities are strongly dependent by assumed failure modes and load models, whereas load ratings do not account for redundancy in a structure or correlation between failure modes. In the research presented in Estes and Frangopol (2003), it was shown how routine visual information related to condition rating of composite highway bridges and used in the PONTIS bridge management system can be used to update the reliability of these bridges subject to corrosion, demonstrating a clear interaction between operational and research performance indicators.

Other Technical Performance Indicators

Other types of technical PIs could be mentioned as not fully belonging to operational/scientific subclasses, rather mainly based on statistical theories applied on pure operational/scientific technical PIs. Among others, condition indexing based on the concept of apparent age can be associated with operational technical PIs, whereas the credibility index can be proposed on the research technical PIs. The condition index based on the concept of the apparent age was proposed by Zonta et al. (2008), based on the underlying theory that “the apparent age of a standard element is the most likely age of the element given its condition state, assuming theoretical age distributions that are consistent with the normal deterioration model adopted for the element.” Biondini et al. (2010) proposed the adaptation of a statistical estimator called “credibility indicator,” and originally proposed by Grandori et al. (1998), to the case of the deterioration modeling of bridge structures. In particular, the effect of the epistemic uncertainty associated with deterioration modeling on the bridge service life prediction was evaluated through the calculation of the credibility indicator, comparing two models (quantitative and qualitative) and identifying the most reliable one. Avsar et al. (2012) proposed a prioritization index based on the combination of visual inspection results and vulnerability estimates. Lastly, McCarten (2016) presented a review of bridge failures, showing how risk events whether by natural causes or man-made have the greatest impact on bridge performance and highlighting the need to have risk-based indicators as well as condition indicators: the study proposed a Bridge Critically Indicator (BCI) reflecting four basic indicators: risk, robustness, redundancy, and resilience.

Non-Technical Performance Indicators

Technical PIs are also considered for the prediction of deterioration over time and thus plan future restoration interventions and their effectiveness over time (Zhu et al., 2017). Given a certain observed/predicted condition, the owner can define the optimal restoration strategy to be implemented. Technical PIs allow quantifying the effectiveness of a restoration strategy: however, its implementation involves a series of social, environmental and economic consequences that may significantly affect the decision-making process. In addition, more solutions can be developed (e.g., with different techniques, materials, costs, Gantt charts), and thus there is the need of identifying the best one, not only focusing on technical PIs, but analyzing also a set of non-technical PIs to rationally find the best restoration strategy. In the following, an extensive review of non-technical PIs retrievable in literature is provided. Figure 4 illustrates a sketch of the investigated environmental, social and economic PIs currently retrievable in scientific literature.

FIGURE 4

Figure 4. Sketch of the investigated non-technical PIs.

Environmental Performance Indicators

Regarding environmental PIs, many parameters, including energy consumption, use of non-renewable resources, traffic disruption, durability and the reuse or recycling of components and materials have to be considered (Wallbank et al., 1999). In order to make sense of all these different factors, Life Cycle Assessment (LCA) was defined as a methodology for quantifying environmental impacts and burdens associated with an item over its entire life (ISO, 2006a,b). Depending on the systems boundaries, analyses may be carried with a cradle-to-gate or cradle-to-cradle approach, thus integrating all the life phases of the structure, i.e., material manufacture, construction, maintenance and operation, and lastly the end of life, with its alternative solutions being dismantling and landfilling or materials recovery and recycling. LCA concentrates on environmental aspects—resource use, ecological consequences and human health—and does not normally address economic or most social aspects (Klöpffer and Grahl, 2009). Rodrigues et al. (2017) presented an application to evaluate the environmental friendliness of TCC bridges with respect to traditional solutions. Pang et al. (2015), Penadés-Plà et al. (2016), and Dabous et al. (2017) presented comparative LCA analyses between alternative bridge maintenance scenarios. Some researchers proposed probabilistic LCA approaches able to account for uncertainties in main input parameters (Shen and Lepech, 2017). Regarding reinforced concrete systems, some studies were also conducted with the aim to highlight the sustainability potential paying attention to the composition of the mix-design and its environmental implications (Hammervold et al., 2013; Hooton and Bickley, 2014; Muller et al., 2014; Ali et al., 2015) performed LCA on Norway bridge types highlighting how main environmental impacts can be related to the production of materials for bridge decks and abutments, as these components require large quantities of materials. The following environmental indicators, that practically cover some of the main mid-point indicators of the CML2002 method (Guinée et al., 2002) and energy demand, were proposed:

• Primary Energy Demand: a PI quantifying the amount of energy directly withdrawn from the hydrosphere, atmosphere or geosphere (Keoleian et al., 2005a), that derive from the CED (cumulative energy method) method (Boustead and Hancock, 1979). It is distinguished from non-renewable and renewable resources: for the former (i.e., fossil fuels and uranium), the amount is expressed in MJ- equivalents, whereas in case of the latter it is quantified in biomass kg- equivalents. For hydropower, Primary Energy is instead assessed as the amount of energy needed for an equivalent change in the potential energy of the water (Kreißig and Kümmel, 1999). According to Du et al. (2014), it has been shown that CED is largely dominated by the material manufacture phase, regardless of the structural type analyzed, and it is responsible for 68–80% of the overall energy demand during bridge life cycle. Maintenance phase instead is highly dependent on the bridge type, given the same deteriorating scenario;

• Abiotic depletion: this PI represents one of the most discussed impact categories since currently a reliable method for its assessment is still lacking (van Oers and Guinee, 2016). The issue of depletion of abiotic resources can be viewed as a decrease of the resource itself, or, an incremental change in the environmental impact of extraction processes at some point in the future (Heijungs et al., 1997; Guinée et al., 2002). The impact category of “abiotic depletion” is measured in kg Sb-equivalents/kg extraction, and is estimated as the product between the amount of extractions/fossil fuels and some characterization factors (ADPs). It is worth recalling that this indicator is mostly used to evaluate mineral resources, whereas it still do not find large applications when dealing with the extraction of virgin bulk resources, e.g., aggregates used for concrete manufacture (Habert et al., 2010). For the above reasons, it is not typically included in LCA analysis when dealing with RC structures analysis. Recently, this indicator has been used as a basis to capture also the influences on land occupation and transformation, associated to degradation problems related to soil depletion, topography alteration and visual damage to landscape (Milài Canals et al., 2007);

• Global Warming Potential: this PI is probably the most well-known, on which the communities and hence decision makers put more attention, and it describes the mechanism of the greenhouse effect, measured in kg CO₂- equivalents (Padgett and Tapia, 2013; Mara et al., 2014). Greenhouse gases that are considered to be caused or increased anthropogenically are for example carbon dioxide, methane, and chlorofluorocarbons (CFCs), which are converted into equivalent of CO₂ by well-known conversion factors, e.g., 1 kg CH₄ = 25 kg CO₂. For a bridge 320 m long and 22.5 m wide, carrying two traffic lanes in each direction located in Västra Götaland, Sweden, Du et al. (2014) have analyzed five alternatives realized with steel box girder composite bridge, a steel I-girder composite bridge, two solutions with post-tensioned concrete box girder bridges, and a balanced cantilever concrete box girder bridge. In all the cases, the order of magnitude of GWP indicator was about 6 × 10⁶ kg CO_2−eq, along the whole service life of the structure. Particularly, the maintenance operations are responsible for < 10% of such emissions, whereas material production still remains by far the main contribution, in a range between 72 and 94%, depending on the analyzed solution;

• Ozone Depletion Potential: this PI quantifies anthropogenic emissions, such as fluorine-chlorine-hydrocarbons (CFCs) and nitrogen oxides (NOX), and it is measured in kg R11-equivalents. These emissions are considered responsible of potential ozone depletion, and main factors of the increase of the hole in the stratosphere layer (Habert et al., 2012). For the above case-studies, Du et al. (2014) have estimated that a range between 2 and 3 × 10⁻¹ kg CFC-11_eq may be produced during the entire life cycle of a structure;

• Human/Terrestrial Toxicity: this PI estimates the Human Toxicity Potential (HTP), i.e., the potential harm of a unit of chemical released into the environment (Habert et al., 2012). HTP includes both inherent toxicity and generic source-to-dose relationships for pollutant emissions, and it is measured in kg 1.4-DB equivalents. The overall HTP score of an emissions' profile is calculated as the sum of the release of each chemical multiplied by the respective equivalency factor (McKone and Hertwich, 2001). For the above case-studies, Du et al. (2014) have estimated that a range between 1 and 2 × 10⁶ kg DB_eq may be produced during the whole life cycle of a bridge structure, and particularly the solutions with post-tensioned concrete box girders with a balanced cantilever concrete box girder promote a reduction by 30% of the impacts, due to the reduced impacts of selecting concrete instead of steel as main structural materials. Although this reduction, within the whole life cycle, the impact of maintenance operations seems still negligible if compared to the materials manufacture phase, but it worth recalling not negligible in terms of absolute values;

• Acidification Potential: this PI quantifies the acidification of soils and waters originated by the transformation of air pollutants, such as sulfur dioxide and nitrogen oxide into acids (H₂SO₄ und HNO₃), and it is measured in kg SO₂- equivalents (Kim et al., 2013). This leads to a decrease in the pH-value of rainwater (i.e., “acid rain”) and fog, harming ecosystems and causing forest dieback. As an indicative example, LCA analysis carried out on Mälkiä Canal Bridge by Rantala (2010), one of the biggest bridge under construction in Finland at year 2009 (a continuous composite girder bridge with steel girders as main bearing structure and reinforced concrete deck, with seven spans and total length of 318.8 meters), has estimated that operation, repair and maintenance operations will be responsible for 6.5% of whole SO₂ emissions, thus being 2,500 kg of SO_{2, eq}. Within maintenance operations, waterproofing renovation, zinc coating of steel girders, bearing and expansion joints reparation and substitution are examples of the operations that were considered;

• Eutrophication Potential: this PI describes the enrichment of nutrients in a defined place, either aquatic or terrestrial, and it is measured in kg ${PO}_{4^{-}}$ equivalents (Kim et al., 2013). It is caused by waste water, air pollutants and fertilization in agriculture. In the analyzed case of Mälkiä Canal Bridge by Rantala (2010), 300 kg ${PO}_{4 eq}^{-}$ are estimated to be released during operation, repair and maintenance operations, thus representing about 4% of the overall emission within this category;

• Photochemical Ozone Creation Potential: this PI quantifies the photochemical ozone production in the troposphere, commonly known as summer smog, and it is measured in kg C₂H₄⁻ equivalents (Du and Karoumi, 2013). Ground-level ozone is produced by combination of radiation from the sun and the presence of nitrogen oxides and hydrocarbons. Rantala (2010) estimated that operation, repair and maintenance may lead to 10.8% of the overall emissions in this category for the Mälkiä Canal Bridge, being about 160 kg ${C_{2} H_{4}^{-}}_{eq}$ , compared to steel parts manufacturing that are responsible for more than 30% of the emissions alone;

• Solid waste production: this PI estimates potential impacts due to the production of construction and demolition waste (Keoleian et al., 2005a). The value of this indicator is sensible dependent on the end-of-life (EOL) scenario that might be experienced by the structure. If recycling/reuse strategies are adopted, such emissions can be highly reduced, and are caused by the sole transportation ones. However, typical, concrete parts might be contaminated for instance by chlorides, due to the use of de-icing salts during winter or in cold climates; hence, it is necessary to evaluate carefully the amount that could not be re-used, or that should be used for low-value applications.

Specific software programs containing suitable dataset for the LCA analysis (e.g., SimaPro7, 2008) can be used for the assessment of the abovementioned environmental PIs.

Social Performance Indicators

Regarding existing social PIs, they can be subdivided in quantitative and qualitative ones. Among quantitative social PIs, the most significant social effects associable with bridge deterioration and consequent maintenance works and/or bridge failures are linked to indirect consequences like traffic delays with related increase of traveling times for roadway users and potential human losses due to accidents (Keoleian et al., 2005b). Delays during maintenance works to bridges on busy roads can be relevant, with economic impacts many times outweighing the pure rehabilitation costs: Koch et al. (2002) estimate the user costs due to traffic delays and lost productivity to be more than ten times the direct cost of maintenance, repair, and rehabilitation of bridges. For this reason, the owners have to minimize such effects, thus reflecting in a substantial improvement in the level of sustainability of the rehabilitation (Wallbank et al., 1999; Wallbank, 2002; Liu and Frangopol, 2006). An interesting discussion on user cost models was provided by Thoft-Christensen (2009), in which the author also explained how to estimate user costs associated to the repair activities of a deteriorated bridge structure. The quantification of user costs is often a hard task, since in most of the cases several aspects have to be accounted for: user costs can be estimated as the sum of different components (Wilde et al., 2001). Goh and Yang (2010) tried to give a categorization of cost components via the execution of a questionnaire survey in Australia. In general terms, three main categories can be identified: vehicle operating costs (VOC, including the costs of fuel, tires, engine oil, maintenance, and depreciation), traffic delay (TD, considering speed delay, detour time and consequent loss of opportunities) and accident costs (AC, fatal accidents, non-fatal injury accidents, property damage accidents) (Najafi and Soares, 2001). VOC estimation models are usually based on the definition of main road characteristics (e.g., roughness, geometry, traffic rates), vehicle attributes (e.g., weight, age, horse power, price of maintenance works) and environmental climate characteristics (Ben-Akiva and Gopinath, 1995). TD is often the most relevant component and must be differentiated when dealing with cars or trucks (Sobanjo and Thompson, 2004): its estimation is based on the product between the increase in travel time and the travel time cost per hour (Gao and Zhang, 2013). Travel time cost represents a measure of value of time to road users, and usually it is based on average hourly wage and income level in a city or region, with differentiated values for passenger cars and commercial trucks (Matthews and Allouche, 2010). AC cost models try instead to capture social impact related to the potential increase of vehicle accidents during the period in which rehabilitation is taking place, disrupting normal traffic flow and reducing the vehicle capacity of a bridge (Ehlen, 1997). Some studies dealt with the calibration of models taking into account main bridge parameters: among others, Thompson (2002) presented a user annual accident count model taking also into account geometrical and traffic flow characteristics of a generic road link; Lounis and Daigle (2010) considered average accident costs and the normal accident rate taken from statistics published by Wilson et al. (1994); Transport Canada (2006). Considering the lack of statistical data on accident in work zones, a rule of thumb of three times the normal accident rates is used as suggested by Walls and Smith (1998). Bai et al. (2010) presented a framework for comprehensive estimation of user cost for bridge management, synthesizing the existing state of practice of user cost estimation and techniques to address a number of considerations in such estimation. Chen et al. (2017) recently calibrated multivariate models able to quantify accident rates in relation to road-surface condition state.

In case of the occurrence of natural hazards like earthquakes, tornadoes, etc., damage can affect multiple bridge structures, thus influencing the functionality of the transportation network itself. In such cases, social PIs that can be monitored are related to the variation of gross domestic product due to the occurrence of the natural hazard (Carturan et al., 2014). At the network-level, it is also important to take into account the potential residual functionality of the transportation network due to bridge damage/failures and also buildings like in the case of historical centers (Zanini et al., 2017c). In this context, connectivity indexes are usually considered as social PIs for the evaluation of the functionality of the infrastructure both for analyzing issues related to the accessibility by rescuing operations and the safe evacuation of citizens (Hadas et al., 2015; Nahum et al., 2017).

Other quantitative research-based social indicators can be used: among others, Pandey et al. (2006) proposed the use of the so-called Life Quality Index (LQI) as a rigorous basis for program evaluation to assist decision-makers in directing expenditures where they may most effective. The LQI is equivalent to a multi-attribute utility function being consistent with the principles of rational decision analysis. It is further refined to consider the issues of discounting of life years, competing background risks, and population age and mortality distribution. Rackwitz et al. (2005) expanded the LQI framework and applied it to determine optimal safety levels in civil engineering infrastructures. Maes et al. (2003) applied LQI for optimizing the life-cycle cost of structures.

The above-mentioned Social PIs requires a higher level of analysis, moving from the so-called system-level (i.e., focused on the single bridge) to the network-level: such type of analyses, allow to capture the impact of a bridge damage/failure/restoration on the overall transportation system accounting for reliable traffic demand models and detailed network topology transportation graphs. In this way, it is possible to identity those bridges that can mainly impact at the network-level and thus give priority to them in the following implementation of restoration/replacement actions.

With reference to qualitative social PIs, aesthetic impact, prestige and historic value and political implications can be evaluated. One important point is in fact that bridges are often located directly in the urban space or are seen as prestigious landmarks. Therefore, their aesthetic impact and acceptance is another important performance aspect (Barelli et al., 2006). Furthermore, historic bridges can constitute a part of the cultural heritage and hence, preservation can become an important aspect. Dette and Sigrist (2011) proposed an aesthetic indicator for concrete bridges called time of unsatisfactory appearance (RTUA), which is defined as the fraction of the service lifetime in which the condition of the concrete surface is below a certain aesthetic threshold and thus the appearance is impaired. Other visual quality characteristics can be considered as aesthetic indicators (Smardon and Hunter, 1983; FHWA, 1988; Rahman, 1992), as:

• Visual Pattern Elements: this PI describes “how well the bridge fits into the overall landscape on the basis of primary visual attributes” (Patidar et al., 1991) of objects like color, form, line, texture;

• Scale Contrast: this PI describes “the extent to which a bridge blends into its environment on the basis of its relative size with respect to those of other features in its environment” (Patidar et al., 1991);

• Spatial Dominance: this PI is similar to the previous, “but pertains to a larger dimension; is the extent to which the bridge elements would be dominant in views of larger landscape and cityscape” (Patidar et al., 1991);

• Diversity: this PI describes “a function of the frequency, variety, and intermixing of the visual pattern elements of the bridge with its setting. Also termed as setting contrast (the extent to which project's visual pattern elements contrast with or blends in with its existing natural or man-made background)” Patidar et al. (1991);

• Continuity: this PI evaluates “the uninterrupted flow of pattern elements in a landscape and the maintenance of visual relationships between landscape components that are immediately connected or related” (Patidar et al., 1991);

• Variety: this PI estimates “the richness/diversity of physical objects and interrelationships within the landscape” (Patidar et al., 1991);

• Visual quality: this PI rates “the excellence of the viewing experience. Visual quality may be assessed using one of several approaches, e.g., via using opinion surveys of viewers, or judgments of visual quality metrics like vividness, intactness, unity” (Patidar et al., 1991).

From the abovementioned list of visual quality attributes, an owner can select a set of PIs to evaluate alternative bridge projects and rehabilitation solutions.

Regarding politics, no clear indicator was found in literature. However, some suggestions are herein reported on how to define it in a consistent way. In the authors' opinion, the aim of a political indicator is to try to measure benefits associated with the implementation of a rehabilitation intervention on a bridge in terms of improvement of the social consensus for a politician. Hence, when dealing with the comparison of different solutions, the one characterized by the highest consensus is the best one. For the assessment of this indicator, the most suitable tool is represented by opinion polls via interviews or other ways: the key issue in this context is to define a proper sample of citizens to be queried, reflecting the real distribution of population potentially interested/afflicted in case of adoption of a specific decision.

Economic Performance Indicators

Bridge owners can also use economic PIs to assess economic efficiency of alternative bridge projects and rehabilitation strategies (Virtala, 1997; Pitonak and Pepucha, 2016). Among others, Robert et al. (2004) evidenced the role of economic analysis in engineering and political decisions regarding transportation investments, focusing on the use of computer systems in the execution of systematic economic analyses for United States roadway projects. In the context of BMSs, it seems evident that the main economic PI is the direct cost of a rehabilitation intervention, i.e., the sum of all costs associated with the material execution of the intervention (Zanini et al., 2016c). Zanini et al. (2016d) presented an extensive cost analysis for an asset of reinforced-concrete and masonry bridges with the aim to derive some relationships between restoration and seismic retrofitting costs with condition ratings obtainable from visual surveys. However, when dealing with the identification of the optimal rehabilitation strategy among different alternatives, direct cost cannot has to be integrated also with an economic quantification of social and environmental impacts estimated with respective PIs, in order to obtain the total cost for a given decision-making.

After estimating the total cost of a certain rehabilitation strategy, other economic PIs that can be used to capture the entire financial dimension of an investment (Valuch and Pitonak, 2015), as:

• Payback period: this PI estimates the number of years needed for equalizing benefits and costs for restoration of bridges;

• Net present value: this PI represents the actualized difference between social benefits and costs. In case of more sophisticated build-operate-transfer contracts, usually developed when dealing with the identification of the best solution among a set of alternative projects, other additional terms related to royalties, business income taxes and earnings must be subtracted to the net value (Chang and Po-Han, 2001);

• Internal rate of return: this PI quantifies the value of the rate of return required for obtaining a net present value equal to zero;

• Benefit-cost ratio: this PI compares “the economic net present value of all the social benefits and costs of the project life cycle and its cost of acquisition” (Chang and Po-Han, 2001), quantifying in such a way the profitability (Hofer et al., 2018) of a rehabilitation investment;

• Viewpoint of equity: this PI accounts for “the equity invested in the construction period and the total net profit before dividends are given to stockholders in the operating period. The total net profit here comes from the statement of cash flows, which considers financing-related items, such as loans, interest, stocks, dividends, and so forth. In this way, the concessionaire will know how long it will take for their investment to be recovered with the total net profit in the operating period” Chang and Po-Han (2001);

• Viewpoint of dividends: this PI considers “the equity invested in the construction period and the dividends paid to stockholders in the operating period. This viewpoint also provides information to stockholders about the period of time during which the dividends given to stockholders in the operating period can recover the equity investment in the construction period” (Chang and Po-Han, 2001);

• Debt coverage ratio: this PI computes “the ratio between earnings before interest and taxes (EBIT), including depreciation. However, depreciation is not a real cash outflow. It represents the wearing-out of the equipment. Therefore, to present the concessionaire's available capital to pay debt, depreciation is added back to EBIT. Debt coverage ratio shows the concessionaire's ability to pay debt. The debt coverage ratio influences the willingness of banks to loan money to the concessionaire” (Chang and Po-Han, 2001). Generally speaking, a debt coverage ratio at least equal to or larger than 1.0 is acceptable (Brigham and Gapenski, 1997).

Many researchers deal with another PI called Life-Cycle Cost (LCC) (Chang and Shinozuka, 1996; Frangopol et al., 1997, 2001; Kong and Frangopol, 2003; Soliman and Frangopol, 2014; Wessels et al., 2014; Biondini and Frangopol, 2016; Rossi et al., 2017), i.e., the sum of all the costs that an owner has to sustain during the entire service-life of a bridge structure. LCC seems to be a promising economic PI in bridge management practices, even if some cautions have to be used when actualizing future cash flows if the time horizon is too long.

Decision-Making With Different Types of PIs

As illustrated above, different types of PIs can be used in ordinary bridge maintenance practice, each characterized by a specific qualitative or quantitative metric. Therefore, there is the need to rationally handle various metrics accounting for different aspects in the identification of the optimal rehabilitation strategy for a bridge structure. In this regard, some researchers proposed indicators to comprehensively accounting for such different aspects: Hendy and Petty (2012) presented the so-called “Sustainability index” based on radar charts ranging between 0 and 1, and used it for comparing different solutions for a new bridge project. Radar charts are often used when dealing with qualitative/quantitative indicators characterized by different metrics, as in the case of sustainability analyses (Rezayat, 2009): Umer et al. (2016) presented an index called “Sustainometer” based on radar charts and able to incorporate uncertainties with a fuzzy logic toolbox for the evaluation of sustainability of roadway projects. Yadollahi et al. (2016) presented a fuzzy factor analysis aimed at identifying the most significant factors involved in the definition of bridge rehabilitation projects with the aim to improve sustainability of bridge maintenance operations. Radar charts allow also considering the contribution of each individual impact to the overall rating via the use of weighting systems. In general terms, there is the need of refined decision-making approaches in order to use a rational approach in the selection of the optimal restoration strategy accounting for all the different economic, social and environmental aspects. Penadés-Plà et al. (2017) presented an interesting review of previous literature studies dealing with the implementation of methods based on the Multiple-Attribute-Decision-Making (MADM) for a sustainable bridge design: among others, scoring methods (Podvezko, 2011), distance-based methods (Tamiza et al., 1998; Ballestero, 2007), pairwise comparison approaches like analytic hierarchy/network processes (Gorener, 2012; Ali et al., 2015; Rashidi et al., 2016), outranking methods (Behzadian et al., 2010; Govindan and Jepsen, 2016; Jajac et al., 2017), multi-attribute utility/value theory (Sarabando and Dias, 2010; Sabatino et al., 2015) can be viable solutions for a rational decision-making process. Recently Yoon and Hastak (2017) presented a multitiered prioritizing method based on urgency scales able to take into account two hierarchical selection steps based on urgency and total prioritization scales. Another interesting work was proposed by Lounis and McAllister (2016) that illustrated the proposal of a risk-based decision-making framework for bridges subject to different hazards able to account both technical and sustainability requirements.

Discussion

The issue of bridge quality control and management has been significantly deepened in past and recent literature studies, and has been widely addressed in the present review. Looking to the current state-of-research on this field, some considerations can be carried out with the aim to highlight present gaps and potential future research developments.

In particular, significant efforts are still required in order to try to find reliable correlations between visual inspections and NDTs, as well as between the latter and probing outcomes. As regards SHM techniques, one open issue is related to develop smart algorithms able to handle big data coming from SHM systems, and more generally, trying to reduce as possible the amount of data collected and costs for the implementation of SHM. Given the substantial standardization of the bridge types, it would be desirable in the near-future to develop low-cost and less-computationally onerous permanent SHM systems to be directly installed at the time of construction of the bridge.

As regards technical PIs, one challenge is represented by the conversion of the current visual-inspection based BMSs, adopted by the majority of public/private infrastructure owners, to more refined quantitative technical PIs, like those reported in section Operational Technical Performance Indicators. This passage it is not an easy task, and for this reason, scientific community may contribute with more research activities aimed to investigate correlations between qualitative and quantitative technical PIs. Another relevant issue that may be addressed deals with the integration of the existing BMSs, mainly focused on the collection of reports on surveyed damage due to deterioration phenomena, with PIs like vulnerability and risk that are usually related to sudden natural or man/made events. In particular, more research is required on the combined effect of natural aging and sudden events in order to prevent potential relevant damage/failures.

Lastly, the field of non-technical PIs as well as that of decision-making approaches are still little known among infrastructure owners, and currently no relevant research presents a detailed and comprehensive application of all technical, socio-environmental and economic PIs to a bridge case study. It is therefore strongly recommended to scientist dealing with research in bridge quality control and management to put efforts in developing case-study applications with a multidisciplinary approach in order to allow engineers and infrastructure owners to familiarize with such issues.

Conclusions

This work focused on the identification and classification of the most diffused technical, environmental, social and economic performance indicators (PIs) adopted in existing Bridge Management Systems, and those still under progress and that could be eventually included in the next future. A wide literature review was, therefore, performed, with the aim to provide a comprehensive state of the research actually available in the identification of PIs to be used in the development of sustainable QC plans for roadway administrators in the near future. First, a focus on tools actually adopted for the quantification of PIs for existing bridges was presented with emphasis on visual surveys, probing, non-destructive techniques and structural health monitoring, reporting related advantages and drawbacks. In the following, a detailed overview of technical and non-technical PIs used both in practical and research activities was carried out, explaining their quantitative/qualitative metrics, and providing references to literature studies in which are presented realistic applications. In the last part, a section illustrating a wide selection of decision-making methods able to take into account different types of PIs and related metrics is illustrated. The present work can therefore be considered as a useful support for researchers and practitioners involved with bridge maintenance operations and development of QC plans, providing a global perspective to the issue of measuring the performance of existing bridges via the use of suitable PIs.

Author Contributions

MZ performed the review of technical indicators and social and economic non-technical indicators. FF reviewed the environmental non-technical indicators currently in use. JC analyzed method for taking into account multiple indicators and organized the structure of the manuscript.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

The authors would like to acknowledge the Management Committee of the COST Action TU1406—Quality specifications for roadway bridges, standardization at a European level for the financial support provided for the development of the STSM programme ECOST-STSM-TU1406-031016-079244.

References

Adey, B., Hajdin, R., and Bruhwiler, E. (2003). Risk-based approach to the determination of optimal interventions for bridges affected by multiple hazards. Eng. Struct. 25, 903–912. doi: 10.1016/S0141-0296(03)00024-5