IABSE survey of implemented decision-making models used by public and private owners/operators of road- and railway infrastructures

IABSE Survey of Implemented Decision-making Models used by Public and Private Owners/Operators of Roadand Railway Infrastructures Alfred Strauss , Prof., University of Natural Resources and Life Sciences, Vienna, Austria; André Orcesi , Dr, Cerema, Research team ENDSUM, DTecITM/DTOA/GITEX, Champs-sur-Marne, France; Department MAST-EMGCU, Université Gustave Eiffel, IFSTTAR, Marne-la-Vallée, France; Andreas Lampropoulos, Prof., University of Brighton, Brighton, UK; Bruno Briseghella, Prof., Fuzhou University, Fujian, People’s Republic of China; Dan M. Frangopol, Prof., Lehigh University, Bethlehem, USA; Hélder S. Sousa, Dr, University of Minho, Minho, Portugal; Joan Casas ,Prof., Universitat Politécnica de Catalunya, Barcelona, Spain; José C. Matos ,Prof., University of Minho, Minho, Portugal; Kristian Schellenberg, Dr, Equi Bridges Ltd., Chur, Switzerland; Matias Valenzuela, Dr, Pontificia Universidad Católica de Valparaíso, Valparaíso, Chile; Mitsuyoshi Akiyama ,Dr, Waseda University, Tokyo, Japan; Poul Linneberg, M.Sc., COWI A/S, Kongens Lyngby, Denmark;Rade Hajdin , Dr, Infrastructure Management Consultants GmbH, Zürich, Switzerland;Thomas Moser, Dr, ASFiNAG GmbH, Vienna, Austria. Contact [email protected] DOI: 10.1080/10168664.2022.2154731 Abstract Infrastructure systems, such as bridges, are a driver for the economic growth and sustainable development of countries. Similarly, the development of operation and maintenance strategies for infrastructure systems may aim at optimal management using Key Performance Indicators (KPIs) such as reliability, redundancy, availability, safety, economy, environmental performance and resilience. Recent research and development projects, such as COST TU1406, highlight that infrastructure managers make decisions based on a mix of qualitative and quantitative data from various sources paired with models of various levels of complexity as well as expert judgement. Similarly, recent state-of-the-art academia reports on a variety of different decision-making models applicable to the optimal management of infrastructure systems may be used. Within IABSE Commission 5 on Existing Structures, Task Group 5.4 has performed a survey on implemented decision-making models among 23 infrastructure managers from 20 countries. It highlights some similarities in relation to KPIs, condition rating and limit state checks. This has stimulated the standardisation of decision making. The application of risk-based methods, performance prediction and intervention modelling are somewhat more scattered and may call for further research and development as well as training. The need to bridge the gap between implemented decision-making models and research is of paramount importance. Keywords: decision-making models; key performance indicators; risk; infrastructure systems; bridges Introduction Infrastructure systems comprise bridges, tunnels, waterways, roads, railways, dams, power plants and transmission lines among many others. These infrastructure systems are a driver for the economic growth and sustainable development of countries. Similarly, the development of operation and maintenance strategies for infrastructure systems may aim at optimal management using Key Performance Indicators (KPIs) such as reliability, redundancy, availability, safety, economy, and environmental performance. The integration of these KPIs calls for a risk-based approach addressing both the probability and consequences of various lifetime scenarios. In recent years, resilience as a KPI has been addressed by some infrastructure managers. Top level or executive asset management policies often address some or all the above KPIs. How these policies are translated within the owner/operator organisation into daily operation is not always transparent, and for a good reason—because this translation is not straightforward. It may be argued that infrastructure managers currently base their decisions on an implicit risk-based decision process. Often a combination of data from various sources is handled by models having various levels of complexity and paired with expert judgement. This process implicitly or explicitly addresses the probability of structural failure, operational failure or any other failure mode over time, alongside an assessment of the consequences if these failures should occur. This process is more often qualitative rather than quantitative (in this respect, management through condition rating is considered as a qualitative method). The qualitative nature of current infrastructure management and the use of expert judgement for these uncertain systems introduce a level of subjectivity, but most importantly they do not allow for fully transparent decision making. Budget allocation needs to be performed objectively based on the goals and criteria set by the operators. All types of infrastructure objects should be considered. Transparent decision making should be performed with different attributes (condition, costs, downtime, etc.) that are either part of the objective function or constraints. The quality of these attributes is of utmost importance. Increasing the level of transparency will be a benefit for many stakeholders and not only for the operator. Recent research and development projects such as COST TU1406 1 reveal that decision making varies between countries and that the performance assessment of roadway bridges (and other infrastructure systems) should be standardised. Standardisation will Structural Engineering International 2023 Scientific Paper 1 allow for an easier exchange of knowledge and will increase competition, to the benefit of multiple stakeholders and society at large. In parallel, recent state-of-the-art academia reports on decision-making models applicable to the optimal management of infrastructure systems, e.g. Refs. [2–9] summarise more than 400 important references related to infrastructure system decision making. Various decision-making models for infrastructure systems have been demonstrated in case studies in order to show their benefit. However, quite often, these models are not picked up by the industry. This can be due to various obstacles such as lack of data, reluctance to deviate (or difficulties related to deviating) from their current modus operandi, lack of trust that these models really add to the specific engineering understanding, lack of skills within the infrastructure organisationand/orlackofstandardprovisions for their application. To advance the current level of infrastructure decision making, it is important to bridge the gap between implemented decision making and the current state of the art. To this end, IABSE Commission 5 on Existing Structures has formed a Task Group (TG) 5.4 dealing with decisionmaking models. TG 5.4 members represent various stakeholders, such as owners/operators, consultants and academics, covering the geographical areas of America, Asia, Australia and Europe. TG 5.4 has performed a survey among infrastructure managers on implemented decision-making models. The survey is analysed in the following sections. Figure 1 relates the structure of the article, which to a large extend follows the questionnaire, to the decision-making process. Finally, conclusions and outlook are presented with the aim of highlighting the current level of implemented decision-making models. Paired with the current state of the art from academia, it should be possible to narrow the gap for the benefit of industry, academia and society at large. Respondents To explore currently implemented decision-making models, TG 5.4 has circulated a questionnaire to public and private owner/operators of roadand railway infrastructures—more than 60 questionnaires were sent to more than 50 countries. During the period of 2019–2020, completed questionnaires from 23 respondents from 20 countries within America, Asia and Europe have been received. The respondents represent both private and public owners/operators dealing with transport infrastructure systems such as bridges, tunnels, retaining walls, cuttings, embankments and waterways. Some owners/operators manage only roadway infrastructure systems, while others manage both roadand railway infrastructures. Most of the questionnaires have been completed online directly by infrastructure owners/operators. Other questionnaires have been completed through interviews between a TG 5.4 member and an owner/ operator. The participating countries and organisations have been kept anonymous in this article in order not to favour specific countries or organisations that have the maturity and/or economic power to be quite advanced when it comes to implemented decision-making. The purpose of the article is to present best practices and emerging improvements as well as their diversity. In this way, the less advanced owners/operators may learn from the more advanced so that all are able to improve their approaches. Similarly, academia will gain a firm view of the level of maturity within the owner/operator organisations. This may be considered when proposing state-of-the-art decision-making models or their further development. The analysis of the responses related to decision-making models for infrastructure systems comprising bridges, tunnels, retaining walls, cuttings and embankments, among others, may be used as a blueprint for other sectors. When speaking of decision-making models, it is evident that different sectors can learn from each other. Basis for Decision Making Decision making is part of infrastructure managers’daily tasks. Decision making is made on all levels within an owner/operator organisation and often many years before the interventions. Decision making may comprise, for example, the selection of an optimal intervention strategy that meets technical and financial considerations as well as current regulations. As a basis for decision making, infrastructure managers rely on data as well as systems/tools developed and operated internally within their organisation or by external companies. The quality of information and the knowledge of the structures with their local peculiarities, damage processes as well as associated intervention costs and methods are crucial for good decision making. In the questionnaire, the respondents were asked about their structure’s management system and its basis. Twenty-one out of the 23 respondents carry out the analysis internally to Fig. 1: Relation between the structure of the article and the decision-making process 2 Scientific Paper Structural Engineering International 2023 determine the interventions to be performed. Approximately half (11) of the respondents also use support (e.g. detailed inspection reports and measurements), which is provided by one or more consultants. More than 50% of the respondents work with data provided by public or private companies in their decisionmaking process. In detailed reports, external consultants provide structural assessments, cost estimates and different maintenance strategies to support the decision-making process defined and guided by the owners. The decision-making process is also supported by a Bridge Management System (BMS) for 14 respondents and the remaining 9 respondents use databases for their decision support. The large majority use proprietary software owned by the owner/operator and adaptedtotheirownneedseitherin house or with the help of a service provider. Only three respondents use a commercial management system, one of them customised to their own needs. A summary from 2012 of BMS worldwide is provided by the IABMAS Bridge Management Committee. 10 In addition to the management systems described above, approximately half (12) of the respondents also use spreadsheets as a support for decision making. These spreadsheets may deal with deterioration modelling, cost calculations, risk assessment, traffic simulations, etc. In most cases, MS © Excel is used. In summary, the manager controls the decision-making process for the management of their structures. However, to a large extent, external support is also used to form the basis for final decisions. Key Performance Indicators (KPIs) The performance-based evaluation of engineering structures is becoming increasingly important and allows maintenance management and longterm planning to be carried out according to the needs of the user and owner. Often, performance evaluations make use of the term Performance Indicator (PI), which stems from economics and measures the success of an organisation or of a particular activity (such as projects) with which it engages. The application of this term to physical objects is coupled to their fitness for purpose. The PI measures the fitness for purpose of a physical object such as a bridge or one of its elements. Since the fitness for purpose (i.e. quality) can change over time, so does the value of a PI. Maintenance interventions can also change the value of PI. It is obvious that bridge performance relates to safety and serviceability, but other performance criteria can be useful as well. Generally, there is no clear distinction between PIs and KPIs. In this article, like the COST TU1406 project WG3 report, 11 KPIs relate to a whole bridge and are defined as follows (RASEE). .Reliability is the probability that a bridge will be fit for purpose during its service life. It is the complement of the probability of structural failure (structural safety), operational failure (serviceability) or any other failure mode chosen by the infrastructure manager or required by other stakeholders. .Availability is the proportion of time a structure is open for service. It does not include failure-related service outages, but those due to expected maintenance interventions. Alternatively, Availability can for instance include additional travel time due to an imposed traffic regime on a bridge. .Safety is the situation of life and limb being protected from harm during the service life of a bridge. Loss of life and limb due to structural failure is not included by this definition (since it would overlap with the Reliability). .Economy is related to minimising the long-term cost of maintenance activities over the service life of a structure. Herein the user costs for instance for bridges incurred due to detours and delays are not included. .Environment is related to minimising the harm to environment during the service life of a structure. In recent years, Resilience as a KPI has been addressed by some infrastructure managers. Resilience is associated with the ability of a structural system to deliver a certain service level even after the occurrence of an extreme event or series of events, and to recover the desired functionality as fast as possible. One can see that Resilience maybeseenasacombinationofthe above listed KPIs as a function of time. Following design provisions, the obvious choice for the KPIs for bridges would be safety and serviceability. Indeed, some suggestions for PIs, e.g. in Ref. [12,13], include safety and serviceability and combine them with other performance indicators. As presented in Ref. [12], the serviceability is combined with durability in the performance category “Structural Condition”, whereas safety is combined with stability to form the performance category “Structural Integrity”. The performance category “Costs”includes both agency and user costs. The user costs include delay, detour and accidents costs. Finally, the performance category “Functionality” includes clearance, ride quality, load ratings and restriction on use. Some relevant indicators of bridge performance are included in Ref. [12] but the classification merits some further consideration. For instance, their “structural integrity”is related only to sudden events, mostly natural hazards such as earthquakes and hurricanes. The observable deterioration processes, although they may compromise structural integrity, affect only “Durability”and “Serviceability”. “Durability”seems to be understood as a span of time in which neither “Safety”nor “Serviceability”is compromised, and this understanding is adopted in the present article as well. Within COST Action TU1406, 1 Working Group 2 (WG2) elaborated the proposal for KPIs based on the Dutch RAMSSH€EP approach Rijkswaterstaat. 14 The following KPIs were defined. .Safety, Reliability and Security (S, R, S)—a combined KPI. .Availability and Maintainability (A, M)—a combined KPI. .Economy €(i.e. Costs). .Environment (E). .Health and Politics (H, P)—a combined KPI. One notes that, whereas safety is considered directly, serviceability is not. The overall performance is represented in Ref. [15]asa“spider net / radar”diagram. Each axis represents a KPI and the most favourable value of a KPI is one (1). The larger the area in the diagram enclosed by the KPI values, the better the bridge performance. Clearly, a similar diagram can be applied also for several bridges or even a whole bridge inventory. Structural Engineering International 2023 Scientific Paper 3 The evaluation of KPIs from PIs and mere observations is still a matter for earnest discussion. In addition, it seems that the KPIs from Ref. [15] are not completely orthogonal (certain properties participate in more than one KPI). In particular, Reliability and Availability seem to overlap. Furthermore, some KPIs are difficult to assess, at least at bridge level, notably the KPIs Health and Politics, but also Maintainability. Maintainability is the ease with which a product can be maintained in order to repair damage or its cause, to repair or replace faulty components without having to replace still working parts and to prevent unforeseen maintenance interventions. This can be understood as a design aspect, and it is covered within Economy. Security is the degree of protection against vandalism and may be covered within Maintainability. Protection against terrorism, which especially in the US belongs to security, is not considered in this approach. The KPI Health is the absence of non-failure causes of illness (e.g. the use of asbestos), which is in most cases regulated. The KPI Politics includes the elimination of causes of public outcry, image protection, etc., and is a downstream performance indicator, i.e. fulfilled if RASEE goals are met. Given the latter, the definitions of the RASEE KPIs are adopted to comply with the aforementioned TU1406 WG3 framework. The primary goal of TU1406 WG3 was to develop a quality control plan that enables the systematic link between the best practice inspection systems and innovative asset management procedures. Within a quality control framework, the KPIs will be evaluated for different maintenance scenarios, looking for the most feasible one. Here, it must be underlined that the KPIs Reliability and Safety can be evaluated based on inspection and/or investigation at a point in time (i.e. static quality assessment) but can also be predicted over time (i.e. dynamic quality assessment). The KPIs Availability, Economy and Environment can only be reasonably applied as a function of time. Several case studies on KPIs and decision making are referenced in Refs. [1,16–18]. In the implemented survey, the importance of different KPIs in the scheduling of maintenance and repair of structures was queried. The interrogated stakeholders were given the chance to indicate a level of importance between 1 (not important) to 5 (very important) regarding the following criteria: reliability, availability, safety, economy, environment, special/extreme cases of loading or any other criteria to be described by the respondent. The criterion given higher importance was safety, closely followed by reliability. In both cases, 15 out of 23 of the answers gave a ranking of 5 to these criteria. The third criterion considered most important was availability, followed by economy and special/extreme cases of loading. From the given choices, environment was considered the least important, although in two cases a ranking of 5 was given to this criterion. Within, for example, the formulation of the United Nations Sustainability Development Goals in 2015, 19 the importance of the environmental KPI may increase in the future. Comparing the criteria considered most important (safety and reliability) with the least important (special/extreme cases of loading and environment), it is seen that the first ones are given an importance value approximately 1.4 times higher compared to the least important ones. Figure 2 presents a comparison of the importance attributed to the above-mentioned criteria. It should also be noted that 9 out of 23 respondents indicated other criteria used for scheduling maintenance activities, namely feasibility, local community, redundancy, durability, consequences of failure (monetary consequences of failure compared to construction costs), reliability, network level availability related to structures (e.g. road category, annual average daily traffic), life-cycle costs, maintainability, security, health and politics (national governmental interest in safe and predictable transportation). For 65% of the group who also consider these other criteria in their management framework, the very important class (5) was indicated for the other criteria. This information also shows that the current RASEE KPI definitions given above are not yet fully recognised by operators in their detailed and comprehensive meaning, since the other criteria mentioned to a large extent are indirectly contained in the defined ones. There is therefore still a need for training regarding the RASEE KPIs. Similarly, there is also a need for research regarding the RASEE KPIs. Assessments Condition Rating Condition rating seems to be the most available method for the structural assessment of existing structures. This is due to its relatively easy deployment, mainly based on visual inspections, sometimes supplemented by some simple non-destructive testing in the case of standard bridges. In the case of complex, long-span and/or landmark bridges, permanent monitoring systems are also becoming more used in order to perform continuous monitoring of bridge performance (“Structural Health Monitoring”). The questionnaire evidences once again that Condition Rating (CR) is the most used assessment method regarding bridge performance. Therefore, it can be rated as a KPI. This is in accordance with the previously defined KPIs. In fact, due to difficulties Fig. 2: Importance of KPIs for scheduling the maintenance and repair of structures 4 Scientific Paper Structural Engineering International 2023 in evaluating the KPI Reliability, CR has been used as a substitute in standard assessment. Also, a clear relation exists between condition and safety, another KPI, and also, usually, CR has been used to quantify safety related issues (the condition of shoulders, parapets, handrails, expansion joints, etc.). Also, in all cases, respondents state that the calculation of the condition of whole bridges is based on a splitting into several parts (foundations, piers, abutments, decks) and that these parts are also split into different elements (for instance, a deck is divided into edge or interior beams, slab, etc.). The maximum number of condition states for an element varies from country to country with a maximum of 22, with 5 condition states being the most common. The number of states of a structure is also variable, but with less variability, with a minimum of 4 and a maximum of 7. Again 5 is the most common value (43% of cases). It is worth noting that 5 seems to be the optimum number of condition states, not only in the case of bridges, but also for other structural elements and structures (such as buildings) and in other scientific and technical fields. 20 Only 16 countries answered the question about the existence of a condition rating procedure for bridge assessment (e.g. a formula or a standardised procedure). Starting from the condition of the different elements, 9 of them use a formula or standardised procedure while6relyontheinspector’s assessment and engineering judgement. In summary, although other KPIs are used in a small number of countries, condition rating is the most and widely used KPI by stakeholders when looking at the assessment of the capacity of existing structures. In fact, all the respondents to the questionnaire answered that they use condition rating in their inspection and assessment procedure. However, 6 countries declare that a limit state check is also performed for the assessment, whereas only 3 use a risk assessment procedure (the probability and consequences of failure). This is further analysed in the following. Limit State Checks The design and performance assessment of structural elements, components and assets may be expressed in terms of limit state checks, allowing a more substantiated decision-making process. From that perspective, performance evaluation may be defined by the probability of a limit state function being violated. The failure, or non-acceptable performance, of an element is considered when the supply value (e.g. resistance, service limit) is exceeded by the demand value resulting from a loading scenario on that specific element. Therefore, limit state checks are effective tools for assessing the performance of an asset, being commonly divided into ultimate and serviceability limit states, depending on the aim of the assessment. For instance, to design and assess for both ultimate and serviceability limit states, diverse target values for limit state violation are established for various structural situations by considering different consequence classes, reference periods of time and relative costs of safety measures. In the survey, the stakeholders were asked what type of assessment they are employing based on inspection and inventory data. For that question, approximately 26% of all respondents stated that a limit state check was considered, whereas 100% considered condition rating and 17% considered risks concerning probability and consequences of failure. A combination of types of assessment was found in 39% of cases. In all cases where limit state checks were selected, condition rating was also part of the decisionmaking process. The sample of interrogated stakeholders who confirmed using limit state checks had a clear tendency to indicate reliability as the most important KPI for scheduling maintenance activities. Safety and availability were follow-ups on the importance scale, coming before economy and environment. In this analysis, it was found that these stakeholders determined the influence of extreme case loading to be the least important factor for scheduling maintenance activities. It is important to note that all stakeholders indicated that, in the decision-making process, they relied on assessment or analysis performed within their own organisations. In that respect, 83% of the stakeholders who confirmed using limit state checks referred to their use of simplified models and loading by design code or user specified limit states. Also 62% of them confirmed using detailed calculations for each structure or element. However, it is important to note that 50% confirmed using both procedures when needed. In one case, limit state assessment based on detailed calculations is considered when the condition rating has reached its worst condition level. Also, a similar assumption is made in one case where the limit state assessment is performed when the observed condition state calls for a more refined analysis. Therefore, it was found that, in those cases, the limit state assessment is only considered when deemed necessary after analysis of the condition state (condition rating) of the asset. It was also found for one respondent that, although using existing codes for limit state checks, adaptations are made to the design factors for existing structures considering the condition state (or observed damage). Only one entry mentioned the use of advanced models based on proof loading and the application of probabilistic methods for assessing their assets. Nevertheless, also in this case, simplified models are applied if the condition rating does not call for more complex analysis. From the analysis of the results of the survey, it is concluded that limit state checks are generally considered to substantiate the decision-making process when condition rating indicates a bad condition state or inadequate performance of the asset. In that case, the level of complexity of the analysis may vary depending on its objective, and adaptations may be made considering the assessment of existing structures. Risk Since resource scarcity amplifies the requirement that funds should be used as efficiently as possible, there is a strong need to establish inspection and maintenance strategies depending on the magnitude of risk, to prevent performance deterioration and/or enhance the durability of existing structures. Risk is a concept that can be defined as ajoint measure of the occurrence of a hazard and the consequences (direct as well as indirect, related to safety, Structural Engineering International 2023 Scientific Paper 5 economy, environment, etc.) induced by its realisation. 21,22 Risk assessment, then, requires identifying and characterising these hazards and their consequences. In practice, a risk-based methodology is useful for making transparent and consistent decisions concerning the priorities for upgrades and/or repairs considering multiple constraints. It should qualify and quantify risks in order to state proposals for further investigations, for detailed structural assessment and for maintenance/repair strategies. As an example, guidelines quantifying the value of structural health information for decision support have been developed for operators, practising engineers and scientists in Ref. [16–18]. It is worth noting that four owners consider the following consequences in their management of structures. .Casualties and injuries (three respondents). .Types of stakeholder affected (one respondent). .User availability, e.g. additional travel time for traffic infrastructure (three respondents). .Additional operational costs (two respondents). .Repair costs (two respondents). .Environmental consequences (three respondents). Since the failure of an element can cause severe consequences, it is important in structure management to establish scenarios including loss of functionality, loss of life or injury and other economic and social impacts. Only 4 out of 23 owners/operators from 20 countries filled out answers to the questions on risk assessment performed in the management system. These four countries declare in the response to “Assessments performed in the management system” that risk assessment is performed based on inspection and inventory data. Three of them consider both interceptable and non-interceptable processes. Their interceptable processes include “damage processes (e.g. corrosion, alkali aggregate, carbonation)”and “demand (e.g. traffic volume, traffic loading, overweight)”. Meanwhile, all four owners consider sudden events of the non-interceptable processes as man-made hazards (e.g. vehicle/ship impact, explosions and fire). In addition, sudden events as natural hazards (e.g. flood, earthquake and avalanche) and non-observable action (e.g. fatigue) are also considered by three owners. It is noted that no one selected demands associated with climate change, although climate change is becoming a critical issue in the field of structural engineering. Risk assessment incorporating climate change effects seems to be still in its infancy in practical management systems. Performance Prediction Modelling of Deterioration and Effects of Interventions The wide range of processes that affect any infrastructure system during its lifetime leads to a requirement for periodic assessments based on inspection and sometimes monitoring. When necessary, the performance of these infrastructure systems and their components is updated through carefully optimised maintenance interventions. Notwithstanding such control mechanisms and adjustment procedures, it is of utmost importance for the structure owner to be able to predict both shortterm and long-term behaviour reliably in order (a) to assign both short-term and long-term interventions efficiently, (b) to plan the relevant expenses, and (c) to provide an acceptable degree of reliability throughout the structural lifetime as defined by the relevant standards and within the limits accepted by society. Hence, forecasting models for the prediction of an infrastructure’s deterioration process plays a significant role in the estimation of optimal maintenance, rehabilitation and replacement strategies. For the modelling and representation of deterioration processes of structural components and systems, stochastic processes have gained increasing interest and they offer a suitable approach for modelling the long-term performance of structures. Owners generally rely on statistical analysis of condition scores (obtained during inspections) and expert opinions to build some deterioration models. 23 The idea is to characterise a continuous deterioration process by the lowering of a discrete condition state. In practice, rather simple mathematical models are used to predict the future condition state based on an inspection score database. Several management systems use Markov chain models for deterioration prediction, assuming an exponential distribution of the sojourn time in condition states. Other models have been investigated to reflect better that the transition probability is likely to increase with the sojourn time spent in a given condition state. 24,25 One can cite in particular the semi-Markov (assuming a Weibull distribution for the sojourn time) and hidden Markov models together with Artificial Neural Networks (ANNs), which have been reported in the literature as reliable deterioration prediction models. 26–28 Gamma process representations also offer an alternative to discrete Markov models for the description of degradation processes. Gamma processes are continuous-time stochastic processes with independent, non-negative increments that follow gamma distributions with typically identical scale parameters and a time dependent shape parameter. As such, they are well suited for modelling damage that gradually accumulates over time in a sequence of small increments. 29 In addition to the above mentioned models based on condition scores, there are also physically based models for predicting degradation. For instance, there are fatigue, chloride-, carbonationand corrosion prediction models used in specific situations for the remaining lifetime predictions. In parallel, new technologies have been developed in recent years for the identification of physical processes such as corrosion, fatigue, corrosion protection, creep and shrinkage, among others. In addition, the availability of data for calibration and validation, which can be gained through monitoring, is increasing rapidly. This also creates a solid foundation for reviewing the effectiveness of long-established inspection procedures for infrastructure systems, which are paramount for achieving the expected performance and safety of structural systems. The systematic collection of a wide range of sitespecific data, if made available to the engineering community, is an essential element for the verification and improvement of current models and design concepts. In practice though, the use of physically based models requires the collection of more extensive model input data and cannot be operated solely on the basis of information from visual inspections. Furthermore, questions on topics such as spatial variability are still largely not 6 Scientific Paper Structural Engineering International 2023 standardised for such models, whose level of complexity prevents them being widely used in management systems up to now. Based on this previously outlined knowledge about model predictions, one focus of the survey was also on the modelling of deterioration and the effects of interventions, see also Interventions. The methods presented above and the data material available from the survey allow the following analyses of the application of the models. .A tool for modelling the deterioration of an asset on deterministic assumptions is used by eight of the ten respondents who answered that question. .A deterioration model within an asset management system is included by four respondents, an independent expert software or tool is used by five other respondents, and in one case the model is mixed as it is partially within the system and partially using other expert software. .Changes in the physical condition of the structure/elements are modelled by nine out of ten respondents, and only in one case are changes in the KPIs used. .Modelling the effects of interventions was performed in eight cases. The modelling is processed in a deterministic manner. .From those countries modelling interventions, in 38% of the systems the model for intervention effects is implemented in it, and another 50% use external expert software/tools. Again, one country (the same as for the case of deterioration modelling) reports a mixed system. In all cases, the changes in the physical condition of the structure/elements are modelled. In addition, the owners reported budgetary and technical limitations associated with a comprehensive development of deterioration and intervention models for effective use in practice. In particular, issues like the updating of old databases for compatibility with new software products are the limiting elements. In summary, the agencies are willing to include performance prediction modules in their systems, although this would require an expertise that is in some cases not yet available. In consequence, it also requires the service of external consultants and additional costs that must be accommodated in budgetary specifications and constraints. Interventions Intervention Triggers Infrastructure managers launch interventions (e.g. inspections, maintenance, repair and replacements) in order to keep their asset at a performance level that matches a given set of KPIs. These interventions may have different time horizons depending on their urgency and budgetary and/or network level considerations. The first question was about intervention triggering. Almost all respondents mentioned that the main trigger for an intervention is the current condition state/rating of an asset. Only in second place (eight respondents) is the theoretical condition, based on the damage process, considered as an intervention trigger. This is probably because the damage process is explicitly or implicitly evaluated as part of the condition rating process and can give information on the urgency of the invention. It is highlighted that the respondents were not given a direct opportunity to choose financial resources as a trigger. For eight of the respondents, the time and thus the age of an asset was also stated as an intervention trigger. It can be concluded that interventions are launched due to a mix of triggers. Only one respondent reported that interventions would be executed based on the age of structure only. In this case, one or more interventions are defined independently of the condition using life-cycle models. The survey identified the following interventions. .Element level—only some bridge elements will be maintained. .Structure level—for better availability in the future, in most cases the whole asset will be maintained. .Multiple structure level (corridor interventions)—also for better availability in the future, whole network sections (more than one structure) will be maintained. The idea behind this distinction is to show the different strategies at the structure and network levels. A multiple selection was allowed for this question. About 80% of the respondents stated that interventions would be related to the element level. Another approximately 65% said that interventions would not only be related to the element level but also to the level of the asset itself. However, 4 out of the 23 respondents stated that interventions are basically only implemented at the asset level. The idea is to maximise availability by implementing one large intervention rather than many smaller ones. Network availability is considered by 8 out of the 23 respondents. The reduction of work sites is also a priority. This approach requires a very delicate comparison between the costs related to asset intervention and the costs due to network unavailability. Depending on the weighting of availability, different results can be expected. For example, interventions on elements that can be delayed involve additional long-term costs. However, if these costs are lower than the economic damage caused by disturbance to traffic, this intervention can still be beneficial. Intervention Strategies An important step in the infrastructure management process is related to the identification of optimal intervention strategies. The interventions should be adopted by minimising some KPIs (e.g. cost and environmental impact) and maximising others (e.g. availability and safety). In recent years, works 1,30,31 to predict intervention strategies, including inspections, monitoring, maintenance and/or repair actions, for optimal management have been implemented and should be evaluated specifically for each asset. The developed approaches sometimes integrate reliability or risk as well as life-cycle cost assessment— some even with multi-objective optimisation techniques to determine optimum-infrastructure, or infrastructure-network, management plans to assist the decision maker. Optimisation can be considered as an essential tool for providing optimal decision support in the management of infrastructure processes. 4 All elements of this process (i.e. reliability, maintenance, condition, safety, structural health monitoring, inspection, Structural Engineering International 2023 Scientific Paper 7 cost, redundancy and robustness) interact and sometimes conflict. From all possible solutions, optimisation should extract the best solutions that maintain the optimum compromises among these elements. For example, the role of optimisation can be to identify the most effective retrofit strategy for aging bridges and the optimal times for retrofit actions. When performing optimisation, the process needs to be transparent so that stakeholders understand the basis for budget allocation. In this survey, 14 responses were available concerning the intervention strategies used by the respondents. In Table 1, a summary of interventioncriteria results is reported. The predefinition of a time horizon is used by almost all respondents. A relevant number of respondents also evaluate life-cycle cost and/or perform risk analysis. One could argue that, when an inspector formulates an intervention strategy as part of the inspection, the way of thinking is often somewhat similar to a risk analysis. Nine respondents use more than one criterion and three of these nine respondents use all three intervention criteria. Five respondents report that they use some kind of optimisation process. Three respondents use only one criterion, being the definition of the time horizon in two cases and life-cycle cost analysis in the other. In some countries, life-cycle cost analysis has been used for decades, where discounted maintenance costs over a certain time period are compared for different intervention strategies, e.g. the technically optimal strategy and a postponed strategy. Sometimes user costs are monetised and included in the analysis. The results of the questionnaire show the need to implement intervention strategies based on optimisation criteria in the management systems. Work Programmes Four different criteria were provided as the key drivers for the selection of work programmes. These criteria were linked to the scheduled time frame for the completion of the work (Predefined time period) and the allocation and types of work packages (Project packaging), while cost implications (Budget constraints) and the tools and techniques for optimisation of the process (Optimisation methods) were also considered. The main purpose of this part is to identify the most influential factors that should be taken into consideration for the development of work programmes as well as predictive models for the evaluation of the effectiveness of the examined strategies. Seventeen responses were obtained concerning the criteria used to define the work programmes. The results are presented in Table 2. Five respondents utilise at least three out of the four criteria, also four use two criteria at the same time and the rest (eight) only use the Budget constraints, Project packing or Optimisation methods to define their work programmes. The results in Table 2 show that the most crucial parameters for the development of work programmes are budget constraints, which are the criteria included in nearly all the responses. Predefined time period and Project packaging are the next most important parameters (both roughly 40%). In addition to these criteria, information considered as “short timeframe length”for the development of work programmes was requested for the duration. Also, an important parameter for the development of an improved decision-making process is effective financial forecasting, and for this reason information about the “long time-frame length”for future budgets and the development of management policies was also requested. Short time-frame lengths for the development of work programmes range from one to five years, being one year in 57% of cases. Long time-frame lengths for the prediction of future budgets and the development of management policies vary from one to fifty years, the most frequent values being five and ten years. It should be highlighted that financial forecasting for long-term budget predictions involves a high degree of uncertainty. Therefore, it is important to develop predictive models using appropriate tools, and to update forecasting models continuously where re-calibration of the models will be applied to consider recent data and the latest market conditions. Conclusions and Outlook In this article, IABSE TG 5.4 has performed an analysis of the results of a survey of implemented decisionmaking models used by owners/operators of infrastructure systems, bridges in particular. The main conclusions and outlook for the future are as follows. .The relative importance of different Key Performance Indicators (KPIs) is similar among the respondents, with safety and reliability being the most important KPIs. This may promote standardisation. Environment as a KPI is considered to be very important by only a few respondents. The orthogonal set of KPIs (Reliability, Availability, Safety, Economy and Environment) are not fully recognised as KPIs. Training as well as further research may stimulate this process. .Infrastructure managers control the decision-making process for their structures. However, to a large extent external support is also used as a basis. Some examples are structural assessments and cost estimates by public or private companies. .Condition rating, based on visual inspection and non-destructive testing, is the most available method for the structural assessment of existing structures. .Limit state checks are overall considered to substantiate the decision-making process when condition rating indicates a poor condition state or inadequate performance of the asset. In that case, the level of complexity of the Criterion Response Percentage Time horizon predefined 10 71 Life-cycle costs 8 57 Risk analysis 8 57 Table 1: Summary of intervention criteria Criterion Response Percentage Predefined time period 635 Project packaging 741 Budget constraints 15 88 Optimisation methods 317 Table 2: Summary of criteria used to define the work programmes 8 Scientific Paper Structural Engineering International 2023 analysis may vary depending on its objective, and adaptations may be made considering the assessment of existing structures. .The respondents are willing to include a performance prediction module in the system, although this will require expertise that in some owner/operator organisations is not yet available. In consequence, it also requires the service of external consultants and additional costs that must be accommodated in budget specifications and budget constraints. .The results of the questionnaire show the need to implement intervention strategies based on optimisation criteria in the management systems. The respondents mention, for example, life-cycle costs as an optimisation criterion, but other KPIs could be included. .It should be highlighted that financial forecasting for long-term budgetary predictions involves a high degree of uncertainty. Therefore, it is important to develop predictive models using appropriate tools and rolling forecasting models where re-calibration of the models will be applied to take into account recent data and the latest market conditions. .Multi-attribute decision making related to uncertain systems (based on KPIs) calls for a risk-based approach considering both the probability of a failure mode, whether structural or operational, or other and associated consequences. Risk assessment incorporating climate change effects seems to be still in its infancy in practical management systems. .The decision-making process needs to be transparent, so that budget allocation is understandable to all stakeholders. The decision-making process needs to be flexible, so that unstable budgets can be met by changing/updating strategies and the assessment of possible impacts. The decision-making process needs to be robust, so that application within owner/operator organisations is efficient and provides clear answers (unambiguous communication to various internal and external stakeholders). Decision making varies between countries, between organisations and sometimes also within organisations. In particular, it may differ within an organisation dependent on the level at which decisions are made (from high level policy making to daily operation). As for decision-making models, it is evident that different sectors can learn from each other. .There is a need for bridging the gap between implemented decisionmaking models and research models and methods. .Data-driven asset management is high on the agenda within owner/ operator organisations. This comprises the application of Building Information Modeling (BIM) and Artificial Intelligence (AI), among others. A transition from qualitative condition indicator towards quantitative or semi-quantitative data informed risk-based decisionmaking will be a natural next step for many owner/operator organisations. Researchers should play an important role, when identification and prioritisation of data collection is discussed by the industry. The concept of “Value of Information” 16–18 could be implemented but needs substantial refinement to be practice ready. Acknowledgements The authors would like to acknowledge the huge support from owners and operators of infrastructure systems, especially those who have completed the proposed survey. Similarly, the authors would like to acknowledge IABSE Commission 5 on Existing structures for supporting this project. Disclosure Statement No potential conflict of interest was reported by the author(s). ORCID Alfred Strauss http://orcid.org/00000002-1674-7083 André Orcesi http://orcid.org/00000001-7011-0940 Joan Casas http://orcid.org/00000003-4473-4308 Jose Matos http://orcid.org/00000002-1536-2149 Mitsuyoshi Akiyama http://orcid. org/0000-0001-9560-2159 Rade Hajdin http://orcid.org/00000001-5453-1841 References [1] COST TU1406, Quality specifications for roadway bridges, standardization at a European level. [2] Biondini F, Frangopol DM. Life-cycle performance of deteriorating structural systems under uncertainty: review. J Struct Eng. 2016;142:F4016001. [3] Sánchez-Silva M, Frangopol DM, Padgett J, et al. Maintenance and operation of infrastructure systems: review. J Struct Eng. 2016;142: F4016004. [4] Frangopol DM. Life-cycle performance, management, and optimisation of structural system under uncertainty: accomplishments and challenges. Struct Infrastruct Eng.2011;7 (6):389–413. [5] Thöns S. On the value of monitoring information for the structural integrity and risk management. Comput-Aided Civ Infrastruct Eng. 2018;33(1): 79–94. [6] Straub D, Chatzi E, Bismut E, et al. Value of information: a roadmap to quantifying the benefit of structural health monitoring. ICOSSAR-12th International Conference on Structural Safety & Reliability, 2017. [7] Faber M, Stewart M. Risk assessment for civil engineering facilities: critical overview and discussion. Reliab Eng Syst Saf. 2003;80(2):173– 184. [8] Straub D, Faber M. Risk based inspection planning for structural systems. Struct Saf. 2005;27(4):335–355. [9] Rackwitz R, Lentz A, Faber M. Socio-economically sustainable civil engineering infrastructures by optimization. Struct Saf.2005;27(3): 187–229. [10] The IABMAS bridge management committee, overview of existing bridge management systems, 2012. [11] Hajdin R, Kusar M, Masovic S, et al. (2018). WG3 technical report, establishment of a quality control plan. COST TU1406. ISBN: 978-867518-200-9. boutik.pt, Braga, Portugal. [12] Brown MC, Gomez JP, Hammer ML, et al. Long-term bridge performance, high priority bridge performance issues. Report No. FHWAHRT-14-052. s.l.: FHWA, 2014. [13] Grischa D, Sigrist V. Maintenance of important infrastructures—objectives and strategies. IABSE Symposium Report, 2010. [14] Rijkswaterstaat. Leidraad RAMS—sturen op prestaties van systemen (in Dutch). Den Haag: Ministerie van Verkeer en Waterstaat, 2012. [15] Ademovic N, Matos J, Stipanovic I, et al. Performance goals for roadway bridges of COST TU1406. (WG2 Technical Report). s.l.: COST TU1406, 2017. [16] Sousa H, Wenzel H, Thöns S. Quantifying the value of structural health information for decision support –guide for operators. COST TU1402, 2019. [17] Diamantidis D, Sykora M, Sousa H. Quantifying the Value of Structural Health Information for Decision Support –Guide for Practicing Engineers, COST TU1402, 2019. [18] Thöns S. Quantifying the Value of Structural Health Information for Decision Support –Guide for Scientists, COST TU1402, 2019. Structural Engineering International 2023 Scientific Paper 9