AbstractAn effective infrastructure asset management (AM) system is crucial for utilities, city officers, government agencies, and other asset-owning organizations to facilitate navigating the numerous challenges associated with operating and managing infrastructure assets. In this paper, the AM system itself is represented as a complex network (comprised of nodes and links) that describes the major components necessary for its operation within an organization and the information connections between such components. ISO 55001, a widely accepted international standard, specifies the requirements for an effective AM system and outlines the criticality levels of different system components—reflected in the corresponding network by the link weights. The main challenges facing managing an AM system (i.e., metamanagement) pertain to (1) information asymmetry (i.e., not relying on consistent information for decision making) between AM system components; and (2) information overload (i.e., excessive information undermining decision making) within the AM system components. These challenges cause systemic risks (possibility of dependence-induced disruptions) within the system network due to the connectedness of system components. Systemic risks can be mitigated through built-in network resilience by restructuring the system component connections. Such network reconfiguration presents a complex nonconvex optimization problem with multiple potential solutions depending on the number of new connections to be added to the AM system, the length of those connections, and the target risk mitigation level. Through this metamanagement (managing the management system) lens, a genetic algorithm approach was employed to explore the optimal AM network configurations considering different objective functions. These objective functions were based on different complex network measures including the betweenness-, closeness-, and eigenvector-centrality, as well as the vulnerability index. The devised objective functions were employed to the cases of adding 1 to 15 links only to limit network overconnectedness (i.e., information overload). Considering all objective functions evaluated, adding a small number of links (fewer than five) provided a significant reduction in systemic risk (18% to 49%). Finally, managerial insights are presented to explain how to employ the developed approach to mitigate the systemic risks within an organization’s AM system based on different metrics valuations and stakeholder inputs.Practical ApplicationsThe effective implementation of infrastructure asset management systems within organizations that own, operate, and manage infrastructure assets is critical to address the pressing challenges facing the infrastructure decision makers (e.g., infrastructure ageing and deterioration, maintenance backlogs, strict regulatory operating conditions, limited financial resources, and losing valuable experience through retirements). Infrastructure asset management systems contain connectivity between major operational components that can lead to systemic risks (i.e., dependence-induced possible disruptions). Such systemic risks are typically caused by information asymmetry, when (some) asset management stakeholders operate in silos, and/or information overload, when asset management stakeholders are overwhelmed by too much/irrelevant information thus, hindering their decision-making effectiveness. The approach described in this paper empowers infrastructure asset managers with the necessary means to reduce their organizations’ systemic risk exposure, and subsequently enhance their organization resilience to information asymmetry and/or overload. This study is meant to be applied by organizations that align their asset management system with an accepted system that includes 39 separate asset management subject areas.IntroductionOrganizations that own and manage infrastructure assets face many challenges in operating their businesses, including infrastructure ageing, evolving (usually more strict) regulatory operating conditions, limited renewal financial resources, and losing valuable experience through retirements (CIRC 2016; ASCE 2021). A well-defined asset management (AM) program can nonetheless minimize the impacts of these organizational challenges (Bertling Tjernberg 2018; INFC 2018; CIRC 2019; ASCE 2021). AM is the structured decision making and execution of plans adopted to achieve a balance between asset performance and disruption risk through the optimal allocation of available resources and the procurement of additional resources (Uddin et al. 2013; Ross 2019). Organizations that own, manage, and operate infrastructure assets typically implement AM systems to achieve their organizational strategic plan and objectives (Hodkiewicz 2015). An AM system is a set of interrelated and interacting elements of an organization; the system’s function is to establish the AM policy and AM objectives and the processes needed to achieve those objectives (ISO 2014b).As outlined by the Institute of Asset Management’s (IAM) Asset Management Anatomy document, a typical AM system is composed of six divisions and 39 subject areas (IAM 2015), as shown in Fig. 1. Each of the 39 AM subject areas was designed to illustrate the breadth of a certain set of activities within an AM system, the relationships between these activities and the need to integrate them, and the critical role for AM to align with and deliver the strategic plan of an asset-intensive organization (IAM 2015). As such, each AM subject area is a functional component necessary for the implementation and operation of an AM system within an organization that owns and manages infrastructure assets. Detailed descriptions of each AM subject area are given in the Asset Management Anatomy guide developed by IAM (IAM 2015). Goforth et al. (2022) introduced the concept of how a typical AM system can be viewed as a network of connected components, with nodes being the AM subject areas according to IAM (2015) and links being the information flow between connected nodes as defined by Global Forum on Maintenance and Asset Management (GFMAM 2014).The ISO 55001 standard Asset Management—Management Systems—Requirements specifies requirements for the AM system within the context of an organization (ISO 2014a). This standard was designed to be applied to all asset types and by all types and sizes of organizations (ISO 2014a). ISO 55001 groups the main requirements of an AM system according to the context of the organization, leadership, planning, support, operation, performance evaluation, and improvement (ISO 2014a). There are currently merely 27 ISO 55001 certified organizations in North and South America, indicating that they meet or exceed the specifications for an AM system as outlined in the standard (ISO 2021). In general, organizations aim at employing the principles and guidelines described within the ISO 55001 standard to improve their AM system operations (Woodhouse 2014; Hodkiewicz 2015; Konstantakos et al. 2019).An efficient AM system must include consistent information flow between the connected subject areas because miscommunication can lead to system dysfunction. In addition, an effective AM system minimizes the exposure of AM stakeholders to information overload caused by having access to too much out-of-scope information and data (Herrera et al. 2011; Prajogo et al. 2018). An AM system may be influenced by systemic risks (i.e., dependence-induced disruptions) related to information asymmetry caused by the malfunction of one or multiple specific subject areas or the interruption of information flow between subject areas (Goforth et al. 2022). Examples of information asymmetry in an AM system include stakeholders that use different and/or inconsistent information to support their AM subject area-specific decision process, not responding to other stakeholders’ decisions promptly, and the isolation of AM subject areas due to inadequate information-sharing procedures or protocols (Bergh et al. 2019).Systemic risks might also be caused by information overload within an AM subject area where there is too much (out-of-scope) information available and an AM stakeholder is overwhelmed and unable to make decisions (Goforth et al. 2022). These examples of information asymmetry and information overload are critical challenges for the effective operation of an AM system within an organization, and can potentially instigate cascading disruptions (systemic risks) throughout an AM system (Brunetto et al. 2014; Xerri et al. 2015; Pell et al. 2015; de la Pena et al. 2016; Golightly et al. 2018). However, the systemic risks caused by information asymmetry and information overload can be mitigated by increasing the system resilience to dependence-induced disruptions through adding new connections between related, yet unlinked, AM subject areas (Barabási 2016; Goforth et al. 2022).Improving the resilience of a network to dependence-induced possible disruptions (i.e., systemic risks) has been discussed in different fields such as financial systems, transportation engineering, and social science, with adding new connections between the different network components being suggested as a potential solution (Bloomberg Professional Services 2020; Crescenzi et al. 2016; Ohara et al. 2017; Pacreau et al. 2021; Papagelis 2015; Parotsidis et al. 2015; Wu 2015). Although introducing additional connections within a network may provide faster information transfer (Medvet and Bartoli 2021; Parotsidis et al. 2016), determining the optimal number and configuration of added connections is challenging because there may be multiple solutions depending on the number of links to be added, the length of those links, and the targeted reduction in systemic risk levels (Barbosa et al. 2018; Bhavathrathan and Patil 2018; Morshedlou et al. 2021; Nozhati et al. 2019; Vishnu et al. 2021). The connection-addition process has thus been formulated as an optimization problem, where previous studies have deployed different heuristic-based optimization techniques, including greedy algorithms (Crescenzi et al. 2016; Ohara et al. 2017; Parotsidis et al. 2015, 2016), path screening techniques (Papagelis 2015), and genetic algorithm (Medvet and Bartoli 2021; Paterson and Ombuki-Berman 2020; Pizzuti and Socievole 2018; Zhao et al. 2018).Therefore, the objective of the present study is to identify optimal link configurations to be added to the AM system network such that the expected systemic risks caused by information asymmetry and information overload would be significantly mitigated. This study first explains the structure of the AM network and its relation to ISO 55001. Next, a description of the network centrality measures used to determine the most critical AM subject areas in the AM system network is presented. Subsequently, the application procedure of genetic algorithm (GA) is described for different centrality measure–based objective functions. Results are then presented in terms of optimal link configurations to satisfy the different objective functions and the practical implications of the link additions to a real AM system. Finally, managerial insights are presented to allow for the comparison between the different link configuration scenarios according to the different centrality measures investigated.Asset Management Network StructureGoforth et al. (2022) conceptualized the AM system as a network of connected nodes to facilitate identifying the key elements within an organizational structure. Building on the same methodology, the typical AM system, developed by the IAM, is represented in this study by a weighted, directed network as shown in Fig. 2. This AM system network consists of 39 nodes, where each represents a specific AM subject area. Node labels in Fig. 2 correspond to the AM subject areas shown in Fig. 1.The valuation of the criticality of each subject area within an AM system depends on multiple metrics such as money, technology, data, and personnel. Fusing such metrics is challenging, and therefore specific quantification of the criticality of every AM subject area to the operation of an AM system is unattainable without detailed stakeholder input. Alternatively, the ISO 55001 standard (ISO 2014a) outlines the main requirements for a system to manage an organization’s infrastructure assets. The Institute for Asset Management defines functional relationships between the clauses of the ISO 55001 standard and each of the typical 39 AM subject areas (IAM 2015). Therefore, this study approximates the criticality of each AM network subject area to the operation of an AM system as the number of ISO 55001 clauses related to each subject area as defined by the IAM (2015). These relationships are shown in Fig. 3, with the total number of connected ISO 55001 clauses (i.e., the AM subject area weight) being shown in the bottom row.Links between node pairs were defined according to GFMAM (2014), which described each AM subject area and identified its connections to others. Such connections are essential to provide information necessary to make decisions within each AM subject area and thus the overall AM system. Links in the AM network are directed as information transfer occurs from source to target nodes and are also weighted according to the criticality of source nodes. In Fig. 2, link direction is represented by an arrowhead at the target node, and link thickness reflects the corresponding weight, where thicker links indicate a greater ISO 55001 influence and therefore greater criticality to the overall AM system operation. The AM network shown in Fig. 2 is represented mathematically through an adjacency matrix A, which can be found in Appendix S1, with entries ai,j reflecting the link weight and direction between nodes i and j.Centrality MeasuresThe betweenness centrality is a measure that identifies the criticality of a specific node as the fraction of shortest paths passing through it (Freeman 1977). The betweenness centrality of a node i (BCi) is thus calculated as follows(1) BCi=∑j≠i≠kρjk(i)ρjkwhere ρjk = weighted length (i.e., sum of the link weights) of all shortest paths connecting nodes j and k; and ρjk(i) = weighted length of these shortest paths that traverse node i. Regarding the AM network, the BCi values reflect the importance of the corresponding AM subject area to the operation of the whole AM system and can thus be used to reflect its influence on the information flow throughout the AM system.The closeness centrality is a measure that identifies nodes that are key to rapidly process and relay information to other nodes in the network (Estrada and Knight 2015). The closeness centrality of a node i (CCi) is evaluated as (2) where d(i,j) = shortest path length between nodes i and j; and N = total number of nodes in the network. When applied to the AM network, the closeness centrality can be used to identify the AM subject areas that are critical for the rapid processing and transferring of information to other AM subject areas.The eigenvector centrality is a measure that identifies nodes that are highly connected to influential nodes within the network (Thai and Pardalos 2012). The eigenvector centrality of a node i (ECi) is (3) where λ = maximum eigenvalue of the adjacency matrix A. For the AM network, a greater value of ECi indicates that the corresponding AM subject area is connected to other highly influential AM subject areas and is thus critical to the information transfer within the AM system.The vulnerability index is related to the size of the network’s giant component (i.e., the largest connected set of nodes) and is used to identify the critical nodes that are highly sensitive to disruptions by quantifying the fraction of nonoperational nodes when a specific node is triggered to fail (Ezzeldin and El-Dakhakhni 2021). The vulnerability index of node i (VIi) is given by (4) where N′ = total number of operational nodes in the network’s giant component after node i was triggered to fail. The value of VIi is estimated considering the cascading disruption effect within the network as follows: (1) once a node i is triggered to fail, information moving through the network is redistributed based on the shortest paths available; (2) following the redistribution, nodes are considered operational when they can sustain their original information share in addition to those transferred from other nodes; and (3) the redistribution process continues until all nodes in the network are functional and the corresponding N′ value is then obtained. Therefore, calculating the values of VIi necessitates adopting the concept of overflow modeling (Motter and Lai 2002).Several studies have adopted the concept of overflow modeling in different fields, where the flow is simulated by the exchange of a single unit between node pairs along the shortest path connecting these nodes (Motter and Lai 2002; Ezzeldin and El-Dakhakhni 2021; Goforth et al. 2020; Alzoor et al. 2021). Within the AM network, this flow is the information or data shared between two AM subject areas. The betweenness centrality has been extensively employed to quantify the flow transmitted through a specific node (Kinney et al. 2005; Kourtellis et al. 2013; Mahyar et al. 2018). As such, the same methodology is adopted herein, where the BCi is assumed to be equivalent to the amount of information shared by node i (Li). In addition, the maximum amount of information (i.e., the capacity) that can be managed by a node i (CAPi) is assumed to be linearly proportional to its initial load Li(0), as follows: (5) where α = design capacity tolerance that represents the ability of a node to sustain additional information due to any disturbance in the network. The information overload of a node can be represented through Eq. (5). Therefore, within an AM system, the VIi represents the susceptibility of an AM subject area to information overload.Critical Subject Areas in the Asset Management NetworkThe centrality measures described previously were evaluated for the AM system network shown in Fig. 2. Fig. 4 presents the BCi, CCi, ECi, and VIi for the top 15 nodes (i.e., AM subject areas) because these were the most critical nodes to the AM system operation. Such nodes represent the most critical nodes within the typical AM system proposed by the IAM that can facilitate the systemic risk propagation within the AM system. The AM subject areas that are critical based on all four measures considered include Strategic Planning (S4), Resource Management (L3), and Life Cycle Value Realization (A4). This observation is well aligned with the definition of AM as the structured decision making and execution of plans (i.e., strategic plan) developed to optimize a balance between asset performance and risk (i.e., life cycle value realization) using available resources or the procurement of additional resources (i.e., resource management).The AM subject areas identified within the top 15 based on three centrality measures include Asset Management Planning (S5), Operations and Maintenance Decision Making (A5), Stakeholder Engagement (R1), Maintenance Delivery (L1), Shutdown and Outage Strategy (A2), and Shutdown and Outage Management (L4), whereas those identified based on two centrality measures include Asset Management Strategy and Objectives (S2), Asset Performance and Health Monitoring (R2), Asset Information Strategy (I2), Sustainable Development (R5), Asset Costing and Evaluation (R9), Procurement and Supply Chain Management (O1), Technical Standards and Legislation (L5), Capital Investment Decision Making (A3), Risk Assessment and Management (R4), and Asset Operations (L2). The AM subject areas that have greater centralities for multiple measures are very critical for the functionality of the AM system because they can instigate cascade disruptions when they become dysfunctional. Therefore, adding new connections (i.e., links) between the AM subject areas (i.e., nodes) can enhance the resilience of the network to systemic risks (Goforth et al. 2022).Link Addition MethodologyThe link addition process is formulated as an optimization problem with the objective of identifying new connections that can reduce the criticality of highly important AM subject areas and therefore minimize the systemic risks within the AM system. There are 1,359 links that do not already exist in the AM system network and subsequently Cr1359 possible added link configurations, where r is the specified number of links added to the AM system network. Such an extensive search space, combined with the complex, nonlinear, and nonconvex nature of the problem in hand, increases the likelihood that mathematical linear or nonlinear optimization techniques will be trapped in local optima. Conversely, heuristic optimization techniques represent efficient alternatives that can rapidly provide solutions near the global optima for large optimization problems (Rodríguez et al. 2018) and when the objective function is nonconvex (Artoni 2019; Dolatnezhadsomarin and Khorram 2019; Mosa and Ali 2021).Of such techniques, GA has been successfully applied for complex system identification and optimization and was therefore employed in the present study to identify the optimum link configuration added to the AM system network, with the objective of minimizing the average centrality of the most critical AM subject areas. Accordingly, the following optimization problem has been formulated: (6) minx(aNt∑i=1NtBCi(G,x)BCi(G)+bNt∑i=1NtCCi(G,x)CCi(G)+cNt∑i=1NtECi(G,x)ECi(G)+dNt∑i=1NtVIi(G,x)VIi(G))subject to (7) (8) where Nt = number of critical nodes to be considered; G = directed graph representing the typical AM system network (Fig. 2) before adding the new set of links Lad with indices x; and a, b, c, and d = weighting factors of the betweenness centrality, closeness centrality, eigenvector centrality, and vulnerability index, respectively.In this optimization problem, the objective function is defined by Eq. (6), the constraints are defined by Eqs. (7) and (8), and the decision variable is the link index vector x that represents the set of links, Lad, added to the AM system network to minimize the objective function. Eqs. (6)–(8) were developed in such a way to ensure that (1) the original AM system nodes are always present in the optimal network configuration and directed links Lad are only added to the original AM system nodes; and (2) the AM system still provides the intended functionality of managing the organization’s objectives, processes, and assets. The optimization problem represented by Eqs. (6)–(8) can be solved for either an individual measures (i.e., betweenness centrality, closeness centrality, eigenvector centrality, or vulnerability index) or a weighted measure. When the former is of interest, the weighting factor corresponding to the considered centrality is 1.0, whereas those corresponding to other measures are set to zero. When the latter is of interest and a minimized weighted centrality is desired, a, b, c, and d should be chosen according to the target weighting scheme.The application of the GA starts with defining the desired number of links to be added (i.e., the size of the set Lad). A population of individuals is subsequently generated randomly, where each individual contains the possible link indices x to be added. Each individual is subsequently assigned a fitness value based on its ability to achieve the objective presented in Eq. (6). Individuals are evolved continually through a set of reproduction mechanisms, including (1) elitism, where individuals with high fitness values are replicated in the following generations, (2) crossover, where two individuals are selected based on their fitness and subsequently mixed to produce two offspring, and (3) mutation, through which the entries of a single individual are changed randomly. The reproduction process continues until a termination criterion is achieved. Such criterion may be a maximum number of generations, a certain fitness value, a specific computational time, or a combination of two or more criteria.In this study, the GA was applied for different link addition configurations with up to 15 links to assess the optimal combination of link additions that yielded the most improved centrality measure. It should be emphasized that adding links increases the information connectivity within the AM system network but can also lead to information overload for AM stakeholders and possibly a breakdown in system functionality (Herrera et al. 2011; Prajogo et al. 2018). Up to 15 added links were chosen arbitrarily as a representative number to illustrate the impact on centrality reduction with respect to links added to the AM system network without causing information overload. Information overload can be included within the link addition problem through (1) a direct incorporation into the objective function, or (2) applying a postoptimization sensitivity analysis of the effect of the number of added links on information overload. Although these approaches are not demonstrated herein, they could be nonetheless operationalized when all proprietary information is made available by stakeholders. In other words, the link addition methodology presented in this study outlines the meta-management approach that can be applied to real AM systems when all relevant data is available.As the number of added links increased, a larger population and a higher number of maximum generations were employed to enhance the likelihood of achieving a globally optimal solution. The GA’s convergence to a globally optimal solution is primarily governed by the population size and the maximum number of generations employed (Yosri et al. 2021). Therefore, the global optimality of a GA solution can be evaluated through (1) employing initial populations with different sizes for the same maximum number of generations and subsequently evaluating the variability across the obtained solutions, (2) utilizing different values for the maximum number of generations for the same initial population and compare the resulting solutions, or (3) using fixed-sized sets of randomly generated initial populations and assessing the variability in resulting solutions for the same maximum number of generations. The third approach has been employed in this study to evaluate the global optimality of the GA solutions, where the largest population size employed was 5,500 when 15 links were added, and the most generations used was 32. Convergence was defined when the evaluated function changed by less than 1×10−3 from the previous GA iteration.Analysis ResultsFig. 5 shows the average centrality value of the most critical 15 AM subject areas (i.e., Nt=15) for the different numbers of added links when the betweenness centrality, closeness centrality, eigenvector centrality, and vulnerability index are considered individually in Eq. (6). To identify the true optimum, a balance between the link addition cost and the improvement in the objective function would need to be obtained as described previously. The cost of a link addition would be in terms of resources (e.g., money, people, data, and technology) needed to establish an information connection, which would be provided by an organization, but such cost details were proprietary to specific organizations at the time of the development of this study. Therefore, Fig. 5 presents a Pareto front with respect to the objective function value and the number of added links, and the following discussion identifies commonalities across different link addition scenarios for each centrality measure used to define the objective function. A full Pareto analysis would be completed internally by a specific AM organization using link addition costs unique to their organization.Betweenness CentralityFig. 6 presents the connections that were identified to be added to the typical AM system for up to six added links when minimizing the systemic risk due to critical betweenness centrality AM subject areas is of interest, and Appendix S2 provides the labeled connections in a table for 1 to 15 link additions.By including a mere four additional links (i.e., increasing the number of links in the typical AM system from 123, as suggested by the IAM, to 127 would decrease the mean betweenness centrality of the most critical 15 AM subject areas by 38%. The information connection formed between Asset Management Strategy and Objectives (S2) and Resource Management (L3) is an optimal link addition for one to four link additions, and the information connection formed between Demand Analysis (S3) and Management Review, Audit, and Assurance (R8) is including in the optimal set when adding two, three, and four links. This indicates that the S2→L3 and S3→R8 links are critical in reducing the average betweenness centrality in the highly critical AM subject areas when only up to four link additions are available to implement.Beyond four added links, the S2→L3 and S3→R8 links are replaced by a combination of other links that stem from similar source nodes. The S2→L3 and S3→R8 connections would be feasible for introducing an information- or data-sharing process in a real AM system. For example, the AM strategy and objectives would specifically provide information on how the resources of the organization should be managed and the analysis of asset demand would provide information related to the review, audit, and assurance processes.Closeness CentralityFig. 7 presents the connections that were identified to be added to the typical AM system for up to six added links when reducing the systemic risk due to nodes with high closeness centrality values is desired, and Appendix S2 provides the labeled connections in a table for one to fifteen link additions.The information connection formed between the Strategic Planning (S4) and Asset Decommissioning and Disposal (L11) areas is optimal among all link addition scenarios from Scenario 1 to Scenario 6. All optimal link additions identified originated from the Strategic Planning (S4) subject area, highlighting the importance of links originating from this node in reducing the average closeness centrality within the AM system (i.e., reducing the distance of the 15 critical AM subject areas to the other AM subject areas in the AM system network). The S4→L11 link would be feasible within a working AM system if the strategic plan provided additional information to describe the process and the feedback required for asset decommissioning and disposal.Adding the three links identified as optimal within the typical AM system decreased the average closeness centrality of the most critical 15 AM subject areas by 30%, whereas adding the six links identified as optimal provided a 52% reduction. Both scenarios provide a large reduction in the closeness centrality–based objective function without greatly increasing the number of links, and therefore the possibility of information overload, within the AM system network.Eigenvector CentralityFig. 8 presents the connections that were identified to be added to the typical AM system for up to six added links when reducing the systemic risk due to the eigenvector centrality is the focus, and Appendix S2 provides the labeled connections in a table for 1 to 15 link additions.An 18% reduction in the systemic risk due to the disruption of AM subject areas with high eigenvector centralities can be obtained by adding just four links to the AM system network. The information connection formed between the Risk Assessment and Management (R4) and Sustainable Development (R5) areas is optimal for each link addition scenario in Fig. 8, and the information connection from Risk Assessment and Management (R4) and Stakeholder Engagement (R1) areas is optimal among all link addition scenarios from Scenario 2 to Scenario 6. Such occurrences highlight the criticality of each link in the reduction in eigenvector centrality–based systemic risk (i.e., reducing the reliance of the AM system on information transfer through important AM subject areas connected to other important AM subject areas).Both the R4→R5 and the R4→R1 added links are feasible to be implementing in an AM system if appropriate information or data transfers were available. For example, the risk assessment and management process could provide details on a risk-based plan for sustainable development and for engaging with AM stakeholders. The Risk Assessment and Management (R4) subject area is a common source node for many links in Fig. 8, indicating its importance to the reduction in eigenvector centrality–related systemic risk of the most critical AM subject areas.Vulnerability IndexFig. 9 presents the connections that were identified to be added to the typical AM system for up to six added links when reducing the systemic risk due to the average vulnerability index of the top 15 subject areas is the only goal, and Appendix S2 provides the labeled connections in a table for 1 to 15 link additions. The α value from Eq. (5) was specified as 0.05, indicating that if a node’s information load exceeded its capacity by 5%, the node was considered to be in a state of information overload; 5% was chosen as a representative value to illustrate the application of the previously described methodology and it has been used in other vulnerability index applications (Ezzeldin and El-Dakhakhni 2021; Goforth et al. 2020).The optimal added links presented in Fig. 9 do not have significant commonalities among each of the six link addition scenarios. There are only three links (i.e., L1→R1, L5→R8, and L8→S4) that were optimal among two consecutive link addition scenarios. This lack of link addition commonality highlights the importance of applying this optimization methodology when looking to reduce the systemic risk related to vulnerability index (i.e., reducing the potential for information overload scenarios within the AM system network) because a small number of specific links do not provide a consistent reduction, as was the case with the previous centrality measures. There is still great value in implementing the proposed link addition configurations because adding the identified four optimal links can lead to a 49% decrease in the average vulnerability index of the most critical 15 AM subject areas.Weighted Combination of MeasuresWhereas the previously described applications considered minimizing each of betweenness centrality, closeness centrality, eigenvector centrality, and vulnerability index separately, minimizing a weighted combination of such measure is also crucial because each centrality measure evaluates a different aspect of the network’s exposure to systemic risks. As such, the values of a, b, c, and d in Eq. (7) were each assumed as 0.25, and the resulting objective function evaluations are shown in Fig. 10 for link additions from 1 to 15.Fig. 11 presents the connections that were identified to be added to the typical AM system for up to six added links and Appendix S2 provides the labeled connections in a table for 1 to 15 link additions.The information connection between the Strategic Planning (S4) and Resource Management (L3) areas is common among most link addition scenarios, highlighting its importance in reducing the combined systemic risk according to the weighted objective function (19% reduction). The importance of the S4→L3 link is also consistent with the criticality of each AM subject area because both S4 and L3 are critical for all centrality measures as determined from Fig. 4. Adding just four links can provide a 40% reduction in the weighted objective function value from the original AM system network. This large reduction in the objective function evaluation further highlights the impact of a small number of added links in reducing systemic risk within the AM system network.Managerial InsightsIn each of the aforementioned cases, it was found that adding even only a small number of links (fewer than five) provided large reductions in each of the evaluated objective functions [18% (EC) to 49% (VI)]. This result is valuable for AM organizations because they could reduce their systemic risk exposure with minimal added links and therefore minimize the exposure of their AM stakeholders to information overload. However, there is little commonality in added links among each of the individual centrality-based objective functions. This highlights the importance for organizations and managers within such organizations to evaluate the relative importance of each centrality-based systemic risk to their organization and run the optimization methodology accordingly.The optimal added links associated with the weighted combination, outlined in Fig. 11, were the same as those for the closeness and eigenvector measures. Specifically, the Strategic Planning (S4) to Resource Management (L3) link was important for the reduction in closeness centrality and it was also important to the weighted combination reduction. Additionally, the links that originated from the Risk Assessment and Management (R4) subject area were critical for the reduction in the weighted combination, which was similar in the eigenvector centrality reduction. This demonstrates how employing a weighted combination of the described measures allows individual organizations to optimize their own AM system network for their desired systemic risk reduction focus. This facilitates implementation of this optimization methodology in AM organizations across all infrastructure classes.ConclusionImplementing an effective AM system is critical for organizations that own, manage, and operate infrastructure assets to ensure that their assets can provide the greatest life cycle value and the best service to their customers or users. A well-managed asset management system can help organizations alleviate the challenges induced by infrastructure ageing, strict regulatory operating conditions, limited financial means, and losing valuable experience due to retirements. Through a metamanagement lens, this study synthesized the AM system, developed by the IAM, as a network of connected components (i.e., nodes and links). Nodes within the network represent the AM subject areas, whereas links between node simulate the information transfer between subject area pairs.The ISO 55001 standard specifies the best practice for organizations in their implementation and operation of AM systems. Based on this, the AM system network link weights were defined in this study as the number of ISO 55001 clauses that pertain to the source node. The main challenges of AM system operation are caused by information asymmetry between and information overload within system components. These main challenges create systemic risks (i.e., dependence-induced possible disruptions) within the AM system network, which can lead to cascading disruptions throughout the system. To mitigate such risks, this study manages the AM system through a network reconfiguration approach.Betweenness centrality, closeness centrality, eigenvector centrality, and vulnerability index were employed to identify the different AM subject areas most critical for information transfer. Identifying such subject areas is essential because their disruption can initiate a systemic risk situation within the AM system and may lead to system-level malfunction. Subsequently, a GA was deployed to reduce the systemic risk within the AM system through minimizing the measures considered. Different objective functions were considered to minimize the betweenness centrality, closeness centrality, and eigenvector centrality, as well as the vulnerability index (individually or weighted combinations thereof), and optimal link configurations were presented visually for one to six added links. Considering all evaluations, adding a small number of links (fewer than five) can significantly enhance the AM system’s ability to combat system risks (reduced the EC and VI by 18% and 49%, respectively).Managerial insights were also presented highlighting that there were only very few added link commonalities considering the different objective functions. This observation underlines the importance of clarify organization-specific systemic risk reduction goals and highlights the value of exploring the use of weighted combinations of different network measures. It should also be recalled that the valuation of an AM subject areas depends on multiple metrics such as money, technology, data, and personnel. Fusing such metrics is challenging, and therefore relevant detailed stakeholder input is key to operationalize the developed metamanagement approach.Overall, organizations can employ the approach developed in this study to reduce the exposure of their AM system to systemic risks due to the potential cascading disruption of highly connected and critical AM subject areas. The connection addition methodology presented herein enhances the information transfer throughout the AM system while also ensuring that AM stakeholders are not overloaded with too much information. Future work may apply the developed methodology at different stages of AM deployment within organizations and with specific link addition costs, allowing a full Pareto analysis to minimize the centrality values while also minimizing the cost of adding new links to the AM system network.Data Availability StatementSome data (network data), models (MATLAB and R models), and code (optimization code) that support the findings of this study are available from the corresponding author upon reasonable request.AcknowledgmentsThe financial support for the study was provided through the Canadian Nuclear Energy Infrastructure Resilience under Systemic Risk (CaNRisk)—Collaborative Research and Training Experience (CREATE) program of the Natural Science and Engineering Research Council (NSERC) of Canada. The INTERFACE Institute and the INViSiONLab support in the development of this study is also acknowledged.References ASCE. 2021. 2021 report card for America’s infrastructure. Reston, VA: ASCE. Barabási, A.-L. 2016. Network science. Cambridge, UK: Cambridge University Press. Barbosa, F., A. De Sousa, and A. Agra. 2018. “Topology design of transparent optical networks resilient to multiple node failures.” In Proc., 2018 10th Int. Workshop on Resilient Networks Design and Modeling, 1–8. New York: IEEE. Bergh, D. D., D. J. Ketchen Jr., I. Orlandi, P. P. M. A. R. Heugens, and B. K. Boyd. 2019. “Information asymmetry in management research: Past accomplishments and future opportunities.” J. Manage. 45 (1): 122–158. https://doi.org/10.1177/0149206318798026. Bertling Tjernberg, L. 2018. Infrastructure asset management with power system applications. Boca Raton, FL: Taylor & Francis Group. Bloomberg Professional Services. 2020. Supply chain momentum strategies with graph neural networks. New York: Bloomberg. CIRC (Canadian Infrastructure Report Card). 2016. Informing the future: The Canadian infrastructure report card. Montreal: CIRC. CIRC (Canadian Infrastructure Report Card). 2019. Monitoring the state of Canada’s core public infrastructure: The Canadian infrastructure report card 2019. Montreal: CIRC. Crescenzi, P., G. D’Angelo, L. Severini, and Y. Velaj. 2016. “Greedily improving our own closeness centrality in a network.” ACM Trans. Knowl. Discovery Data 11 (1): 1–32. https://doi.org/10.1145/2953882. Dolatnezhadsomarin, A., and E. Khorram. 2019. “Two efficient algorithms for constructing almost even approximations of the Pareto front in multi-objective optimization problems.” Eng. Optim. 51 (4): 567–589. https://doi.org/10.1080/0305215X.2018.1479405. Estrada, E., and P. Knight. 2015. A first course in network theory. Oxford, UK: Oxford University Press. GFMAM (Global Forum on Maintenance and Asset Management). 2014. The asset management landscape. Toronto, ON, Canada: GFMAM. Goforth, E., M. Ezzeldin, W. El-Dakhakhni, L. Wiebe, and M. Mohamed. 2020. “Network-of-networks framework for multimodal hazmat transportation risk mitigation: Application to used nuclear fuel in Canada.” J. Hazard. Toxic Radioact. Waste 24 (3): 04020016. https://doi.org/10.1061/(ASCE)HZ.2153-5515.0000493. Golightly, D., G. Kefalidou, and S. Sharples. 2018. “A cross-sector analysis of human and organisational factors in the deployment of data-driven predictive maintenance.” Inf. Syst. e-Bus. Manage. 16 (3): 627–648. https://doi.org/10.1007/s10257-017-0343-1. Herrera, F., G. Chan, M. Legault, R. M. Kassim, and V. Sharma. 2011. The digital workplace: Think, share, do. New York: Deloitte & Touche. Hodkiewicz, M. R. 2015. “The development of ISO 55000 series standards.” In Vol. 19 of Engineering asset management-systems, professional practices and certification, 427–438. Cham, Switzerland: Springer. IAM (Institute for Asset Management). 2015. Asset management—An anatomy. Bristol, UK: IAM. INFC (Infrastructure Canada). 2018. Investing in Canada: Canada’s long-term infrastructure plan. Ottawa: INFC. ISO. 2014a. Asset management—Management systems—Requirements. ISO 55001. Geneva: ISO. ISO. 2014b. Asset management—Overview, principles and terminology. ISO 55000. Geneva: ISO. Konstantakos, P. C., P. T. Chountalas, and A. I. Magoutas. 2019. “The contemporary landscape of asset management systems.” Qual.-Access Success 20 (169): 10–17. Kourtellis, N., T. Alahakoon, R. Simha, A. Iamnitchi, and R. Tripathi. 2013. “Identifying high betweenness centrality nodes in large social networks.” Social Network Anal. Min. 3 (4): 899–914. https://doi.org/10.1007/s13278-012-0076-6. Mahyar, H., R. Hasheminezhad, E. Ghalebi, A. Nazemian, R. Grosu, A. Movaghar, and H. R. Rabiee. 2018. “Identifying central nodes for information flow in social networks using compressive sensing.” Social Network Anal. Min. 8 (1): 1–24. https://doi.org/10.1007/s13278-018-0506-1. Medvet, E., and A. Bartoli. 2021. “Evolutionary optimization of graphs with GraphEA.” In AIxIA 2020—Advances in artificial intelligence, edited by M. Baldoni and S. Bandini, 83–98. Cham, Switzerland: Springer. Morshedlou, N., K. Barker, A. D. González, and A. Ermagun. 2021. “A heuristic approach to an interdependent restoration planning and crew routing problem.” Comput. Ind. Eng. 161 (Nov): 107626. https://doi.org/10.1016/j.cie.2021.107626. Mosa, M. A., and A. A. Ali. 2021. “Energy management system of low voltage dc microgrid using mixed-integer nonlinear programing and a global optimization technique.” Electr. Power Syst. Res. 192 (Mar): 106971. https://doi.org/10.1016/j.epsr.2020.106971. Nozhati, S., Y. Sarkale, B. Ellingwood, E. K. P. Chong, and H. Mahmoud. 2019. “Near-optimal planning using approximate dynamic programming to enhance post-hazard community resilience management.” Reliab. Eng. Syst. Saf. 181 (Jan): 116–126. https://doi.org/10.1016/j.ress.2018.09.011. Ohara, K., K. Saito, M. Kimura, and H. Motoda. 2017. “Maximizing network performance based on group centrality by creating most effective k-links.” In Proc., Int. Conf. on Data Science and Advanced Analytics, 561–570. New York: IEEE. Pacreau, G., E. Lezmi, and J. Xu. 2021. Graph neural networks for asset management. Amsterdam, Netherlands: Elsevier. Papagelis, M. 2015. “Refining social graph connectivity via shortcut edge addition.” ACM Trans. Knowl. Discovery Data 10 (2): 1–35. https://doi.org/10.1145/2757281. Parotsidis, N., E. Pitoura, and P. Tsaparas. 2015. “Selecting shortcuts for a smaller world.” In Proc., SIAM Int. Conf. on Data Mining 2015, 28–36. Philadelphia: Society for Industrial and Applied Mathematics. Parotsidis, N., E. Pitoura, and P. Tsaparas. 2016. “Centrality-aware link recommendations.” In Proc., 9th ACM Int. Conf. on Web Search and Data Mining, 503–512. New York: Association for Computing Machinery. Paterson, J., and B. Ombuki-Berman. 2020. “A hybrid approach to network robustness optimization using edge rewiring and edge addition.” In Proc., IEEE Int. Conf. on Systems, Man, and Cybernetics, 4051–4057. New York: IEEE. Pell, R., R. Svoboda, R. Eagar, P. Ondko, and F. Kirschnick. 2015. Effective infrastructure asset management—A holistic approach to transformation. Hong Kong: Arthur D. Little—Prism. Pizzuti, C., and A. Socievole. 2018. “A genetic algorithm for improving robustness of complex networks.” In Proc., IEEE 30th Int. Conf. on Tools with Artificial Intelligence, 514–521. New York: IEEE. Prajogo, D., J. Toy, A. Bhattacharya, A. Oke, and T. C. E. Cheng. 2018. “The relationships between information management, process management and operational performance: Internal and external contexts.” Int. J. Prod. Econ. 199 (May): 95–103. https://doi.org/10.1016/j.ijpe.2018.02.019. Rodríguez, N., A. Gupta, P. L. Zabala, and G. Cabrera-Guerrero. 2018. “Optimization algorithms combining (meta)heuristics and mathematical programming and its application in engineering.” Math. Probl. Eng. 2018 (Sep): 1–3. https://doi.org/10.1155/2018/3967457. Ross, R. 2019. Reliability analysis for asset management of electric power grids. Singapore: Wiley. Thai, M. T., and P. M. Pardalos. 2012. Handbook of optimization in complex networks: Communication and social networks. Edited by M. T. Thai, P. M. Pardalos, and Laboratory of Algorithms and Technologies for Networks Analysis. New York: Springer. Uddin, W., W. R. Hudson, and R. Haas. 2013. Public infrastructure asset management. New York: McGraw-Hill. Vishnu, C. R., S. P. Das, R. Sridharan, P. N. Ram Kumar, and N. S. Narahari. 2021. “Development of a reliable and flexible supply chain network design model: A genetic algorithm based approach.” Int. J. Prod. Res. 59 (20): 6185–6209. https://doi.org/10.1080/00207543.2020.1808256. Xerri, M. J., S. Nelson, and Y. Brunetto. 2015. “Importance of workplace relationships and attitudes toward organizational change in engineering asset-management organizations.” J. Manage. Eng. 31 (5): 04014074. https://doi.org/10.1061/(ASCE)ME.1943-5479.0000306. Zhao, L., B. Li, H. Chen, and Y. Yao. 2018. “An assembly sequence optimization oriented small world networks genetic algorithm and case study.” Assembly Autom. 38 (4): 387–397. https://doi.org/10.1108/AA-04-2017-049.