From Reactive to Proactive Infrastructure Maintenance: Remote Sensing Data and Practical Resilience in the Management of Leaky Pipes

Abstract

The introduction of remote sensing technologies, AI and big data analytics in the utility sector is warranted by the need to provide critical services with the least disruption to customers, but also to enable preventive maintenance, extend the life cycle of infrastructure components and reduce grid loss—or overall, to exhibit ‘durability’ and ‘resilience’ when faced with the certainty of breakage and decay. In this paper, we first explore the concept of ‘resilience’ and the nature of practice from a performativist perspective in order to set the scene for discussing the impact of ‘datafication’ on maintenance practices and infrastructure durability. We then describe an instance of introducing remote sensing technologies in district heating network surveillance and leak detection: drone-operated thermographic cameras and underground wire sensors. Based on insights from this case study, we discuss the specificity of data-driven infrastructure maintenance practices, and what it means to exhibit practical resilience in relation to how such practices unfold, interrelate and evolve over time. We reflect on how the use of remote sensing technologies and data analytics (1) potentially changes district heating workers’ epistemic worlds (i.e., how knowledge emerges, is negotiated and ordered in practice) and (2) provides opportunities for ‘messy’ pipe repair work to tacitly adopt proactive and preventive logics to meet continuously evolving organizational and societal needs.

1. Introduction

Maintenance practices are typically categorized as being either reactive or proactive [1,2]. Reactive maintenance involves repairing damages only after they occur, while proactive maintenance aims to predict and prevent damages beforehand. Clearly, the distinction between these approaches is significant, with proactive maintenance being preferable whenever possible due to its clear-cut advantages [3]. For instance, proactive maintenance allows for optimal resource allocation, including task prioritization, grouping of tasks, planning, and balancing task distribution among personnel, and the use of other resources [4]. Moreover, proactive maintenance enables the anticipation of damages, thereby averting both immediate and consequential losses.

The present paper seeks to nuance the distinction between reactive and proactive practices by exploring resilience in infrastructure maintenance as a phenomenon that can also be traced to the incremental evolution of a maintenance practice. Our account goes beyond the standard view on resilience in infrastructure systems, namely as a phenomenon that is predominantly designed. Specifically, we thematize resilience in the context of a Danish utility company that is undergoing a transition from a reactive to a proactive approach to district heating infrastructure maintenance—a change that occurred largely due to the use of sensor data and thermal imaging. What motivates our focus is a tendency in the literature to overemphasize the technical aspects of socio-technical systems. Surely, the technical constitution is crucial. For instance, Sirvio points out that proactive maintenance systems are necessarily ‘intelligent’ in the sense that they “require data collection, data transfer, data storage, data processing and Decision Support Systems to be in place” (p. 221, [1]). In fact, as we shall argue, technology forms a prerequisite for resilience. Nevertheless, with such an overemphasis, one runs the risk of not only rendering other constitutive elements and relations irrelevant to the practice, but also of ignoring how technologies are actually implemented and evolve through their usage. In fact, a focus on the technical dimension entails that one focuses on a particular dimension of resilience. Despite often being part of a strategy or a design process, resilience in human socio-technical systems is not necessarily always the result of a top-down process. Instead, it can also emerge from actual practice (bottom-up) and, hence, without being driven by an overall design scheme articulated by a manager or engineer. With respect to instances where such schemes are in play, we find not just frameworks that specify assessment criteria for exploring resilience in practice (e.g., [5,6]), but also frameworks that can be used for the design of resilient socio-technical systems. For instance, Li and colleagues [7] investigate the resilience of Urban Metro Systems (UMSs) and introduce a resilience framework in the form of their Multi-Dimensional Integrated Management Model. Their model comprises five dimensions of resilience, namely the natural, the physical, the social, the managerial and, finally, the economic. All of these factors, they argue, should be carefully integrated for the sake of ensuring resilience in mitigating flooding risks in UMSs. In similar terms, we highlight that resilience does not only relate to the actual ways of exhibiting resilience towards external factors (e.g., through the detection and repairing of leaks in heating pipes), but also to how the utility company has come to organize itself around being resilient. This is also part of Li and colleagues’ approach, for as they state, flood resilience in UMSs should develop “over time, exhibiting the response characteristics of absorption, resistance, repair, and adaptation” (p. 20, [7]). What we aim to show, however, is that such resilience can emerge in the absence of an integral management model. Rather, it can be understood as an ‘emergent’ property of the system in question once certain technical conditions are met. Thus, rather than being reducible to design or human intent, we show how an infrastructure maintenance practice can gradually develop and become more resilient, simply through practice and continuous adjustments of the available technologies and their practical applications.

In the present paper, we draw on a performativist approach to Science and Technology Studies (STS) as our main theoretical backdrop. We do so not only to avoid attributing an overly privileged role to the technical systems in use, but also to explore how the technology changes as it is being used. In this connection, effective implementation requires not only advanced technology, but also the continuous adaptation and evolution of practice-defining relations across heterogenous elements such as organizational structures, human skill sets and collaborative processes. In this connection, we argue, resilient change need not be something that is consciously designed or implemented, but rather an emergent system property that is linked to actual performances of maintenance practices. Performativists support the dictum ‘essence is existence’, which was originally formulated by the French sociologist Bruno Latour [8]. This underscores a pragmatic approach to studying technology wherein technology is understood in terms of what it ‘actually does’ (cf. [9]), rather than what its thought potential is (i.e., what a piece of technology was actually designed to do). Although this is not always explicitly recognized in engineering science, there are several exceptions. For instance, Lu and colleagues [10] present a case study on the application of remote sensing technology for investigating landslides in southwestern China. Their work, which also involves drone technology, describes a surveillance practice that functions in conjunction with a broader geological practice and urban decision-making. Similar to the current study, the implemented technologies (i.e., drones) are not regarded as stand-alone tools. Rather, their effectiveness is shown to depend on human expertise and how they are used together with other elements of the practice. Indeed, as Lu and colleagues put it, “geological experts must conduct field investigations to confirm the accuracy of InSAR [interferometric synthetic aperture radar] identification, assess current deformation levels, and predict future trends” (p. 10, [10]). In other words, the authors clearly show that technological systems and their features are but one aspect of the otherwise complex practice of investigating and predicting landslides. In fact, whether InSAR technology is useful for ensuring resilience in urban decision-making depends on how the technology is used and the practice it is actually implemented in. What we take from the above examples is that the technology comprises a necessary but not sufficient element for establishing resilient practices and enabling proactive infrastructure operations.

Against this backdrop, the purpose of the present paper is to conceptualize the relation between reactive and proactive practices in the context of a performativist approach to technology-mediated utility infrastructure maintenance. More specifically, this paper aims to explore the notion of ‘resilience’ in relation to how such sociomaterial practices actually function, interrelate and evolve over time. To do so, this study follows district heating workers in Denmark on their transition from a sole reliance on vigilant citizens and workers’ first-hand experience to data-driven maintenance and repair planning, as maintenance work is getting increasingly supported by remote sensing technologies and big data that render leaks from underground pipes ‘visible’.

The structure of this paper is as follows: In Section 2 and Section 3, we introduce our conceptual framework which, as mentioned, is based on insights from performativist STS. We show what a performativist ontology entails and, further, how it contrasts with a realist ontology—on the basis of which resilience is normally thematized. Therein, we introduce general STS concepts and relate these with recent organizational perspectives on technology in use. The purpose of doing so is to stress compatibility between these accounts, and to build upon these with regards to conceptualizing resilience in a performativist context. Section 4 presents our case study, which takes the form of an ethnographic study on leak detection in a Danish utility company. We show how the proactive and reactive elements of maintenance practices exist side by side and, further, how resilience emerges as a property of proactive leak detection without having been designed or conceptualized in advance. Section 5 discusses the practical and theoretical implications of our study, while Section 6 makes several concluding remarks.

2. Overcoming the Realist Hurdle

It is far from straightforward to explore the ‘nature of practices’ (i.e., as being either proactive or reactive) and a phenomenon such as resilience in a performativist context. This is because both notions are underpinned by realist (or even potentially essentialist) assumptions (see also [11]). The issue concerns the fact that, from a performativist view, practices cannot be essentially characterized as ‘things-in-themselves’. In a realist ontology, practices are taken to be something that can be clearly defined and delineated in their own right, and that continue to exist in a self-identical manner despite facing perturbations or other kinds of changes. However, the performativist alternative is somewhat straightforward in that practices can be understood as recurrent socio-technical constellations expressive of either reactivity or proactivity. What this entails is that one avoids considering proactivity and reactivity as foundational traits of practice. However, the hurdle appears trickier to overcome in the case of resilience, since the concept is typically defined in realist terms, whereby entities are conceived of as privileged substances with inert traits and qualities, and are thus studied in isolation from their environment. In a realist ontology then, resilience is seen as an objective trait pertaining to an entity, be it human or non-human. Indeed, the Aristotelian dictum that “[…] individuals are substances—and substances are deeper than their accidents and relations to other things, and capable of enduring despite changes in these inessential features” (p. 14, [12]) underlies at least four ways in which resilience is typically thematized across various disciplines such as ecology, biology, organization theory, psychology and engineering science ([13,14,15,16,17,18,19]). Namely, resilience can be understood as a system’s capacity to (1) withstand pressure without undergoing internal changes and modifications, (2) overcome a perturbation or threat by returning to an initial, so-called ‘naïve’ state, (3) overcome a threat but with this resulting in a weakened self and, lastly, (4) overcome the threat thus evolving into a modified or improved self, such as in the case of cell hormesis:

“In many cases, cells do not return to the naïve state after a toxic insult. The phenomena of ‘pre-conditioning’, ‘tolerance’ and ‘hormesis’ describe this for low-dose exposures to toxicants that render the cell more resistant to subsequent hits.” (p. 247, [16]).

Resilience in all of these instances presupposes that there is an identity of the agent which is either maintained or slightly modified. We find the same in, for instance, attempts at maximizing the so-called “expected value of resilience” (EVR) of integrated energy infrastructures such as electric power systems (see [20]). Here, the system in question is defined as a single, homogeneous entity which can be managed and controlled. In other words, from such a view, the resilient system exhibits durability. Furthermore, such a view places privileged emphasis on the agency of those managing the system, thus considering the system as something that can be acted upon and failing to acknowledge complex dynamics and one’s own entanglement with the system. Thus, this view neglects how systemic innovations may evolve as emergent features in the absence of deliberate planning and design. However, both points contrast with the kind of flat ontology of Latour which effectively “grants all actants an equal right to existence, regardless of size and complexity” (p. 17, [12]) and which also criticizes the assumed concept of durability by rejecting it as a mere ‘figure of speech’ that has no actual correspondence with the actualities of actants (ibid.). In this view, the technical infrastructure system itself is just as much an agent that acts upon its human counterparts and, in so doing, is functionally inseparable from other (e.g., human, economic and technical) agents.

In light of this, it would thus seem that resilience cannot gain traction in a performativist context for the simple reason that agents always exist in a relationship where they exert their influence on each other. Yet, as Harman further notes, there is an ambivalence in how agents and their traits are to be understood in a Latourian (and hence, performative) sense [12]. For Latour, agents (or what he calls ‘actants’) are neither fleeting power plays or social constructions, nor are they mere concrete entities or systems, nor inert substances. In a sense then, Latour partly holds to Aristotle’s notion of substance. However, in another sense, Latour’s principle of irreduction entails that there is no a priori, privileged substance from which all else derives (i.e., nothing is irreducible to something else). Thus, the view challenges the privileged position that human agents ascribe to themselves and others in their capacity of being engineers, managers or even scholars. It does not deprive them of their powers and rationality; rather, it emphasizes the messy network that their powers and rationality play out in and, moreover, that the environment and technical systems also exert influence on their human counterparts. Rather, an actant’s attributes and traits are traceable to (and emerge from) what the agents actually do. Thus, they are traceable to their performances.

Consequently, in performativist terms, resilience cannot be restricted to a particular actant, nor can it be seen as a foundational trait of any sociomaterial system. Instead, the relation between the actant and the system (or actor–network) is one of performative co-evolution, where resilience emerges from a multitude of intersecting, heterogeneous relations that constitute practices. Yet, this does not mean that planners and designers cannot approximate a more performance-based view on resilience. However, such a view would necessarily entail a consideration of the system as not just heterogeneous, dynamic and less predictable, but also as fundamentally entangled. Thus, the separation of factors as in the case of Liu and colleagues [6] is one step in the right direction for designing resilient systems, but one needs to also consider these factors in terms of how they interrelate in practice and, further, how they afford agents the opportunity to develop and refine resilient strategies. For instance, as shown by Lambrou (p. 10, [21]), the development of resilient climate strategies entails a practice-based approach and, further, that one thinks of resilience as something that relates to different elements and, hence, of ecological resilience as something that interplays with economic and social resilience.

However, as we want to show below, resilience can also be understood and explored in terms of how practical elements of proactivity, both human and non-human, come to either replace or add on to those of reactive relations. In these cases, resilience might not be a planned feature, the result of a managed and structured process nor, for that matter, something which pertains strictly to the technical makeup of an infrastructure system. Before addressing this, however, we turn to a closely related topic, yet one deserving of full consideration: that of “technology in action”. We do so in order to set the context for an exploration of how the use of remote sensing technologies and other types of practical activities can be said to interrelate with the way in which resilience is organized in infrastructure maintenance practices.

3. Attuning to ‘Technology in Action’

The recognition that human and non-human entities are not mere substances possessing static traits that can be studied in isolation from one another also marks an important intellectual turning point in the study of technology in organizations. In the last few decades, the idea that technologies have unforeseen or even unpredictable outcomes for organizational structures, human skills and work practices has been firmly cemented in organizational studies (e.g., [22,23,24,25,26,27,28,29]). The active and consequential role of technology in organizations has been discussed using various conceptual underpinnings, such as Bourdieu’s [30] practice theory, Latour’s [31] actor–network theory, Pickering’s [32] mangle of practice and Orlikowski and Scott’s [26] sociomateriality. Empirically, technologies have been shown to redefine activities and roles [33,34], redesign knowledge boundaries [33,35,36], challenge worker control and expertise [27,37] and even to significantly reshape workers’ epistemic worlds—i.e., how knowledge is constructed and used, and what counts as valuable knowledge [38,39].

Overall, studying ‘technology in action’ [9] has produced important contributions to an understanding of the various ways in which technology is consequential to work and organizing, and has shown that individuals do not merely adapt to what the technology is designed to do, but actively engage with it and “bring to life novel relations and objects” (p. 83, [40]). However, the material turn in organizational studies has also produced an intellectual split between, on the one hand, organizational scholars adhering to a sociomaterial approach grounded in a being ontology (e.g., [34,41]), and, on the other hand, those adhering to a sociomaterial approach grounded in a relational (or becoming) ontology (e.g., [22,26,42]). Studies grounded in a being ontology, while ‘attuned’ to the social importance of materiality, “consider ‘structures’ as exterior constraints or affordances for the individual to act in relation to” (p. 97, [43]). Another way to put this is to say that critical realism in organizational studies perpetuates a commitment to dualities such as social–material, human–technology or structure–agency, and to abstract representations of practice. This contrasts Latour’s [44] dictum that society is defined by the heterogeneous elements of practical activity, and not by descriptions that are far-removed from unfolding activity. Instead, organizational studies grounded in a relational ontology adopt the notion of performativity in a Latourian sense and show how human and non-human agents do not merely interact in order to produce organizational phenomena, but rather are themselves continuously (re)produced in emergent and sometimes arbitrary relational enactments. Put in a different way, a relational or performative ontology allows us to “move beyond the dichotomous view of structure and agency and [recognize] how categories such as identities are radically unstable and open to modification and alteration” (p. 97, [43]). From this perspective, resilience cannot be attributable to either technology, humans or their ‘imbrication’ (cf. [34]), and should only be understood as the outcome of performative iterations of practice that condition possibilities for human and non-human agents [43]. It is from this perspective that we explore how remote sensing technologies interrelate with the way in which resilience manifests in infrastructure maintenance practices.

4. Case Study: Messy Networks, Leaky Pipes and Remote Sensing Technology

Empirically, this paper focuses on the evolving work practices for leak detection and repair at the maintenance department (MD) of a large Danish utility company. The MD is responsible for managing leaks in about 3000 km of heating pipes that run under the city of Copenhagen and its surroundings (Greater Copenhagen area). There are approximately 40 employees in the department, handling anywhere from approximately 60 to 100 active leak cases in a given week. District heating infrastructure maintenance represents a large and critical part of the operations of the utility company, since challenging conditions such as high temperature, pressure and humidity mean that the district heating network is under continuous mechanical and thermal stress, leading to problems like corrosion, leaks and rainwater infiltrations. Large sections of the network are also approaching the end of their technical life cycle (some pipes are over 100 years old), whereby thousands of undetectable tiny leaks reduce network efficiency and water quality (through infiltrations of mineralized water that further corrodes pipes). Additionally, the widespread distribution of pipes across different urban areas (including developing areas) increases the risk of accidents caused by third-party construction activities or natural disasters, which often cause massive grid losses and require immediate action. Owing to these increasing challenges, traditional reactive maintenance strategies (i.e., the “wait until it breaks” approach) pose significant risks to system efficiency and sustainability, but also to human safety and the environment. This is why, in recent years, the MD has implemented two remote sensing technologies: thermographic cameras mounted on drones and underground wire sensors. The wire sensors provide real-time leak information through an internet-based monitoring platform (PGWeb), while drone-collected thermographic data are analyzed using interactive orthomapping software (TeraPlan). Together, these technologies grant the MD access to an unprecedented volume of data on both potential and actual leaks across the network, potentially affording them a more stable operational environment and more proactive and resilience-enhancing maintenance strategies.

4.1. Research Design and Method

Instead of tracing resilience to the durability of a particular actant, we consider it to be a question of the relational constellations between a myriad of human and non-human actants that are practically linked in continuously changing and evolving socio-material networks of agencies. This exploration draws on an ongoing ethnography of ‘technology in action’ [9] conducted at the MD, where we engaged with the data relationally—rather than categorically [43]—in order to understand how the materiality of remote sensing technologies shape the social organizing at the MD. The fieldwork presented here is conducted as part of a larger research project concerned with understanding how organizations can mitigate and minimize the impact of significant changes in their internal or external environment (e.g., disruptive technologies) through mutual adaptation and coevolution after being subjected to some form of disturbance. The research centers on the concept of ‘resilience’, which refers to the capacity of an organization (or other forms of social organizing) to adjust to change by slowly adapting its practices through mutual interference with the objects of change, while interacting with individuals’ cognitive dispositions and attitudes—i.e., the extent to which peoples’ thinking and actions are intertwined with here-and-now materiality and emerging social dynamics.

The data for this study were gathered over a period of 2 years, between 2022 and 2024, and the data collection focused on examining how practitioners interact with material elements in their workplace, following a performativist approach to technology studies [44]. Unlike the traditional interpretation of work that focuses on workflows and task analysis [45], this approach to studying work practices allowed us to capture the practical significance of both human and non-human actors and to investigate how maintenance work is actually carried out on a daily basis. Due to the practice-oriented approach of the research design, the study was exploratory in nature, incorporating ethnographic methods such as participant observations and interviews that allowed for an ongoing engagement with maintenance practices at the department, with the aim of uncovering their emergent sociomaterial aspects (see

Table 1. Fieldwork activities and data sources.

Data Source	Focus	Materials
Observations	Department-wide observations of maintenance practices, including contractor meetings, internal meetings and office work Shadowing case coordinators during office work and site visits	Ethnographic field notes (58 typed pages 1.5 line spacing, Times New Roman, 12)
Interviews	Formal semi-structured interviews Informal interviews, in connection to participant observations	Twelve formal interviews amounting to approx. 16 h of audio material (average length per interview: 80 min) Seven informal interviews (varied length, from 15 min to 3 h)
Secondary sources	Internal documents on leak cases covering the observation period	Weekly contractor meeting reports with overview of open cases (89 reports) Photo and video material from site visits and meetings Case-specific printouts from the Geographical Information System (GIS), PGWeb and TeraPlan

for an overview of fieldwork activities and data sources).

The fieldwork included both ethnographic observations and interviews. The observation process involved shadowing case coordinators (operations engineers and technicians tasked with overseeing entire leak cases) during both site visits and office routines, participating in ‘Operations’ meetings (da. ‘Driftsmøder’) between team leaders and case coordinators, and attending weekly ‘Damage’ meetings (da. ‘Havarimøder’) with contractor representatives (the MD uses external contractors for digging, welding and insulating pipes) to discuss ongoing cases. These ‘Damage’ meetings facilitate the creation of provisional weekly plans through collaborative discussions and the spontaneous assignment of responsibilities, allowing all parties to address issues and seek clarification. The primary focus during the observation process was on the interactions among district heating workers, their colleagues and other stakeholders, and on the significance of material assets such as types of pipes, tools, software and computers that they engaged with in their daily tasks.

The interview process involved a range of MD employees, including team leaders, operations engineers and technicians responsible for coordinating leak repairs with contractors (only 9 out of 40 employees hold case coordination roles). In total, 12 in-depth interviews were conducted, with some respondents participating in follow-up sessions. The interviews varied in length but averaged around 80 min. Whenever possible, the interviews were held ‘in situ’ to ensure that participants remained immersed in their work environment. For instance, interviews were conducted alongside physical interactions with leak case files, damaged pipes, tools and software. These took place in meeting rooms, boiler watch rooms and during site visits, allowing for a richer understanding of the everyday practices of district heating workers. The integration of material objects and work settings during interviews highlighted the importance of these elements in shaping how workers engage with and assign practical meaning to various heterogeneous factors in their work, including how the leak data from the two remote sensing technologies become ‘organizationally real’ [46] as these slowly become interwoven with the maintenance practices at the department. Alongside formal interviews, informal conversations took place during field observations, focusing on immediate tasks and emerging issues. Herein, the case coordinators were asked to explain their tasks as they were performing them and the meaning behind their engagement with different tools and digital representations of network leaks.

4.2. Empirical Observations and Analysis

Despite the increasing use of sensor technologies and remote sensing data, the maintenance practices of district heating workers at the MD remain far from being purely proactive. Through our engaged fieldwork, we observed that several structural, technological and organizational factors prevent the MD from experiencing a full transition away from the traditional reactive maintenance approach—factors that may well be overlooked when designing data-driven organizational resilience and managing digitalization programs through a top-down approach. Amongst others, factors such as the technical heterogeneity of the network, the unpredictability of failures, organizational habits and routines, and not least past managerial decisions and financial considerations ensure that reactive elements remain a critical part of the department’s maintenance strategy. Yet, the hybrid approach that is emerging allows the MD to manage its aging infrastructure adaptively, combining proactive planning with reactive interventions to ensure enhanced system efficiency and resilience.

The district heating network overseen by the MD comprises over 3000 km of heating pipes, varying significantly in age—some over a century old—creating complex conditions for maintenance. During field observations, practitioners frequently noted how this age disparity leads to significant differences in the structural integrity and technical compatibility of different parts of the network. Some newer sections of the infrastructure are equipped with pre-insulated pipes that can support underground wire sensors, which provide real-time leak data through the PGWeb monitoring platform. However, many older sections, still in use, lack this capability due to their outdated materials and structural limitations. One engineer from the MD described the challenges of working with such inconsistent infrastructure: “We have one of the oldest networks in the world… but our net is very different. Some areas are only a few years old, while others are nearly 100 years old”.

This infrastructure variability, coupled with the unpredictability of accidents and heightened likelihood of pipe corrosion, means that district heating workers remain much dependent on manual inspections and reactive interventions. In these cases, leaks are normally identified and reported by district heating workers on maintenance duty or by citizens who notice green water (the utility company adds a green tracer to the district heating water for easy identification). This reality is illustrated in Figure 1, which depicts the MD’s traditional reactive approach to leak detection. Here, the process begins with visible signs of damage (center picture), such as the appearance of green water (left-hand side picture) on streets or in basements, prompting the creation of a case referral by the Process and Documentation department of the utility company. Upon receiving the referral, maintenance teams consult the GIS (right-hand side picture illustrates a case coordinator using a GIS printout to localize a suspected leak) and other static data sources to localize the problem, but the process often involves on-site inspection, relying heavily on workers’ practical knowledge, haptic perception (temperature, humidity) and experiential skills. As one operations engineer commented during an emergency response, “Now! Water is coming in somebody’s basement!” This vividly illustrates the urgent, last-minute nature of the reactive approach, where action is driven by visible signs of failure rather than predictive data.

While this approach to leak detection allows the MD to keep the network operational during peak demand periods, it does so at a great cost to the customers and the environment. This is because leaks are readily observable only after the water has reached the surface or has flooded building basements, meaning that the identified pipe fractures are often old and have reached a critical point with considerable water loss and/or extensive material damage.

Another factor constraining the transition to a purely proactive maintenance approach is the unpredictability of network failures, which still necessitate reactive maintenance. Despite the improvements made possible by thermographic drones and underground wire sensors, these systems cannot predict all potential failures. The pipes in the district heating network are subject to continuous thermal and mechanical stress, which can lead to sudden failures that proactive measures might not detect in time. During several field observations, case coordinators shared stories of third-party construction accidents and natural disasters (such as floods) that damaged the pipes unexpectedly. One coordinator explained, “We still have situations where a contractor accidentally hits a pipe, or there’s sudden water burst, and we have to react quickly”. Such situations do not allow for much planning and pre-emptive coordination, since fast-response interventions are often required to deal with the situation at hand. It is in these circumstances that the reactive aspect of maintenance is crucial to ensuring that the network remains operational.

The complexity of the district heating network also poses significant limitations on how extensively proactive technologies can be applied. While thermographic cameras mounted on drones have proven to be effective tools for detecting heat anomalies indicative of leaks in copper piping, these are less effective for steel piping structures that are buried deep underground and even under Copenhagen’s canals. Additionally, underground wire sensors—which detect leaks based on electrical resistance—can only be used in sections of the network with straight, pre-insulated pipes, limiting their applicability. Yet, sections with intricate piping layouts, concrete pipes or those passing under buildings remain challenging for either of the two remote sensing technologies to serve. These technological limitations mean that the heterogeneous nature of the network requires a mix of reactive and proactive approaches. As a team leader explained, “We can detect [a leak] fast if we have automatic surveillance, then we can actually catch it the same day. This is the best solution we have. But we don’t have this in all our piping because some people didn’t want to invest in this from the beginning. And I think this will be the future, we’re just not mature enough for it yet. We have one of the oldest nets in the whole world, but our net is very different. We have areas that are only a few years old, and we have ones that are close to 100 years old now, or more actually.” (Team Leader, MD). He continues: “And this is where the drones come in. And for most [drone-detected] leaks we end up fixing the pipes the best possible way without doing it the right way because the technology is old, and we are bound to making a one-to-one exchange. Because when we have to renew the piping, we actually need the Project Department from our headquarters to do the calculations, dig it all up, remove it… let’s say it’s a few hundred meters, and put new piping and technology in the same route. So, we [the MD] have to make something temporary, in between, in order to satisfy the citizens’ needs”.

This gradual and patchy transition to proactive leak detection using remote sensors is visualized in Figure 2. The proactive approach contrasts with the reactive approach in that it entails considerably more context-detached work, as well as different skill sets and knowledge bases. Figure 2 illustrates the proactive strategy, where drone-operated thermographic cameras scan the surface for heat signatures (right-hand side picture), while wire sensors embedded in the pipe insulation (left-hand side picture) provide continuous data on potential leaks based on principles of electrical resistance and conductivity that can signal the real-time presence of water in the insulation of the pipe. This data-driven approach allows the MD workers to detect leaks before they become visible, helping to plan repairs ahead of time and reducing the reactive burden.

However, as mentioned above, these technologies only cover specific sections of the network, and context-detached detection still leaves significant gaps, especially in older or inaccessible areas. However, their incorporation in the daily routines at the MD suggests a shift towards a proactive maintenance strategy at the MD, enabled by the previously unavailable data on thousands of potential and actual leaks throughout large and critical parts of the network. These tools are central to the new approach, providing district heating workers with the ability to monitor the network continuously and address issues before they escalate into major failures.

The PGWeb (https://pgweb.pipeguard.se/isopluspgweb/?lang=da_DK) software (left-hand side picture) is the department’s internet-based monitoring system, designed to collect and display real-time data from underground wire sensors embedded in the insulation of the heating pipes. These sensors work by measuring electrical resistance and conductivity—parameters that change when water infiltrates the insulation, indicating potential leaks. Workers are notified of anomalies through automatic alerts generated when the sensors detect water where there should be none, and they can access PGWeb from their desktops both for visualizing entire sections of sensor-fitted pipes in a geospatial interface and for monitoring a specific section that gives abnormal readings. The system integrates data streams from sensors distributed across newer sections of the network, providing live updates on the state of these pipes.

However, beyond reactive responses to detected leaks, PGWeb also enables proactive predictions. By analyzing long-term sensor data, workers can observe patterns of deterioration in certain sections of the network—indicating which pipes are likely to fail soon. This predictive capability allows the MD to plan targeted repairs and the clustering of leaks before these manifest on the surface, effectively minimizing downtime and water loss.

While PGWeb is vital for the monitoring of specific sections of the network equipped with sensors, TeraPlan—the interactive orthomaps software—expands proactive maintenance capabilities by analyzing thermographic data collected via drone-operated cameras. Drones equipped with infrared cameras fly over designated areas of the network, scanning the ground for temperature anomalies that indicate heat escaping from underground pipes—a telltale sign of leaking water. Once the thermographic data are collected, they are processed using TeraPlan (center and right-hand side pictures), which overlays the heat map onto Google Maps-based orthomaps of the district heating network. This creates a visual interface where district heating workers can explore color-coded temperature gradients that highlight areas of concern. For example, regions showing unusually high temperatures are flagged as potential leak sites, prompting further investigation or pre-emptive repairs.

In the MD’s workflow, TeraPlan serves as a crucial tool for planning interventions in parts of the network that are not covered by PGWeb sensors, particularly in older or more complex sections of the infrastructure. During field observations, an operations engineer demonstrated how the thermographic data are used alongside other sources of information, such as GIS layers and historical repair records, to prioritize repair work. “This drone data showed us heat escaping in this area here”, he explained, pointing to a cluster of red patches on the orthomaps. “We wouldn’t have found this leak until it caused real damage, but now we can plan to dig it up when the materials arrive”.

Moreover, TeraPlan’s capacity for data visualization allows district heating workers to gain a holistic view of the network. By combining thermographic data from multiple drone flights, workers can compare trends over time, identifying which sections of the network are experiencing the most stress. This helps in creating long-term maintenance schedules and enables the clustering of cases within bigger ‘network renovation’ projects, as explained by one of the operations engineers at the department: “So back to something that I said earlier that we do not plan ahead a whole lot. But [employee responsible for drone cases] is actually the only one planning ahead a whole lot, because he knows he has to share the diggers, and he has to share the welders with us, so he has to plan, and mark areas where he maybe takes two As and a B [leaks identified by drones are ranked based on severity of water loss, where A is the most severe] and say: ‘These three are in close proximity to each other, please go and dig all three up’ […].” (Operations Engineer, MD).

Together, these technologies provide district heating workers with the ability to detect leaks early and make informed decisions about where to allocate resources. For instance, based on both PGWeb alerts and TeraPlan orthomaps, the MD team now makes clustered repair plans—addressing multiple leaks in one geographic area at once, rather than treating them as isolated events. This clustering approach optimizes the use of excavators, welders and other resources, ultimately reducing costs and minimizing the disruption to customers caused by repeated dig-ups in the same area.

Ultimately, our ethnographic study reveals that the MD has developed a hybrid maintenance strategy that blends both reactive and proactive elements. The remote sensing technologies introduced in recent years have significantly improved the MD’s ability to detect and address leaks before they cause major damage. Importantly, this is not something that was envisioned by management or pre-designed into the technology assemblage. Rather, the proactive mindset emerged in the ongoing interrelation of skillful workers, practical expertise, technologies and other heterogeneous elements of practical activity (e.g., shortage of excavators, citizen complaints, traffic disruptions, liability disputes, the pressure to produce greener energy). This means that while leaks were previously dealt with only as isolated occurrences, they are now also handled in clusters depending on factors such as the geographical proximity of leaks, similarities in needed components, or contractor availability. Another emerging change in the maintenance practice can be observed in the way leaks from citizen notifications and scheduled maintenance are handled, as the MD employees devised new and creative ways of using remote sensing data to, e.g., corroborate data from traditional sources, enhance the precision of digging areas or negotiate project accountability and budget allocation with contractors and the municipality.

These gradual, subtle, yet significant changes in the maintenance practice attest to how the MD is making itself more resilient as it continuously develops and evolves proactive practical elements which make the overall practice potentially more predictable, stable and manageable, and, hence, less prone to getting reactively overwhelmed. Yet, it is not a process that is designed or envisaged through a masterplan. By contrast, it seems to emerge spontaneously, thus being traceable to the fact that the maintenance practice itself is characterized by “a high degree of self-organization, precisely because (a) it has the freedom to do so and, (b) it is organized around sets of complex tasks (or problems) which are complex and thus do not lend themselves to straightforward solutions based on the successful resolving of tasks in the past, fixed training regimes, or SOP.” ([47], p. 4).

5. Discussion

The main insight of our study—that resilience1 can emerge and develop in practices over time in the absence of pre-conceptualization and planning—makes several important contributions to the literature and to practice. Firstly, we demonstrate that proactiveness is not inherently tied to the predefined use of a particular technology as intended by its designers. Instead, it arises from how individuals and groups organize themselves around the technology, enabling its utilization in unforeseen ways (such as the clustering observed in Teraplan, or the triangulation of data across Teraplan and PGWeb). Rather than categorizing the practice as solely reactive or proactive, our study uncovers that (a) these exist on a continuum rather than as distinct categories (at least within the transformative practice we examined), and (b) there are inherent tendencies towards reactive behaviors that serve as fundamental conditions and cannot be entirely eliminated, especially during transitional phases. Furthermore, becoming proactive and resilient as an organization by boarding the ‘datafication bandwagon’ cannot be done by relying solely on predefined roadmaps or overarching schemes. Rather, the emergence of proactive elements, which occurs as a subtle undercurrent, is continuously reinforced by ongoing work practices and emergent problem-solving activities that afford the gradual intertwinement of reactive and proactive maintenance approaches (e.g., transitioning from leak case classification to clustering for efficiency, or from traditional static report sheets to integrating heterogeneous software such as GIS, Teraplan and PGWeb that enable a more dynamic and preventive system maintenance approach).

Moreover, we show that the resilience of the maintenance organization is inseparable from the resilience of the network of heating pipes (cf. [11]). Indeed, the relation between the two is one of coevolution, where resilience emerges through specific entanglements of the proactive and reactive elements of the leak detection practice. Beyond the particular case in our study, this testifies to the fact that resilience should not be considered a particular attribute of an organization, or that of a piece of technology. Instead, in line with a performative approach, resilience should be seen as a relationally constituted phenomenon that emerges through sociomaterial enactments, granting actants their practical–material significance.

Lastly, our case study testifies to the fact that resilience can be an emergent property of a socio-technical system, understood in this particular case as a heterogeneous infrastructure maintenance practice. Here, we show how resilience interrelates with a variety of factors which effectively relate to how the district heating workers use the available technology and gradually change the nature of their work. One might wonder to what extent our case study is generalizable to other contexts. In order to answer this question, one would need to take into account some of the specific conditions under which the proactive practice has evolved. This will make it clear that the maintenance practice functions under constraints which, ironically, also comprise the enabling conditions of proactivity. First, it is important to note that the utility company is a publicly owned company operating under some form of monopoly. This means that the company is the sole provider of district heating in a particular geographical area, and it has overall ownership over the physical assets of the district heating network in the area. Thus, due to a lack of competitors, the company lacks a direct incentive to innovate or to optimize its operations. Furthermore, since fixing leaks is a mandatory (and critical) part of managing the district heating network, it is simply expected that such leaks will inevitably occur and that the network is full of them. Both of these points boil down to the fact that there is no internal or external pressure on the MD to radically enhance its efficiency, thus creating ample room for the workers to innovate the practice ‘from within’ in the sense of trying novel solutions whenever they think it makes sense to do so.

Second, the MD is not organized as a tightly coupled system [47,48]. This we touched upon in the previous section. The organization is loosely coupled, meaning that workers have a certain degree of freedom with regards to selecting their tasks and performing them. And since most tasks cannot simply be solved by means of standard operating procedures given their special conditions, it follows that the maintenance operations are somehow in a process of constant evolution and adaptation.

Third, the remote sensing technologies function as the overall technical enabler of the proactive practice. Indeed, the technology does provide the crucial enabling conditions for proactive leak detection since it allows for the effective screening of the piping network for leaks and, therefore, provides a surplus of data on suspected leaks in the PGWeb and Teraplan software [49]. Nevertheless, we cannot be technologically deterministic. The workers make sense of the data and, subsequently, fix the identified leaks in a way that is afforded by their existing expertise and by the technical configuration of the department’s software and hardware infrastructure.

These three points are not, by any means, exhaustive, but they do highlight several conditions that apply in the case of the utility company and that have made it possible for the department to gradually supplement (and, possibly, in the long run, substitute) their predominant reactive approach to maintenance with a more proactive one.

6. Concluding Remarks

In many ways, what we are witnessing at the MD is an unfolding story of how organizational practices and structures are at once flexible and persistent (or resilient) in dynamically evolving contexts; changes in its environment require the MD to strike a balance between the stability of existing reactive practices, on the one hand, and new action possibilities introduced by technologies, on the other hand. While the emerging proactive leak detection practice required some adjustments to existing practices, the organization did not experience change in a disruptive manner. Rather, the new leakage detection practice has come to co-exist with other (reactive) domains of the maintenance work, in an open-ended and evolutionary interplay between old ways of doing things and emergent action possibilities introduced by remote sensing data. For us, this story invited questions about the relation between resilience and reactive and proactive practical elements in the context of continuously evolving societal needs, technology capabilities and increased infrastructure ageing and deterioration. In this respect, our paper demonstrates that resilience emerges in a bottom-up fashion when the necessary conditions for proactivity are met and nurtured. In our case, these conditions were the affordances introduced by remote sensing technologies that allowed district heating workers to gradually change (but not entirely replace) their ways of doing things and to establish a more stable operational environment, thus minimizing the risk of unexpected failures.