RePlanIT Ontology for Digital Product Passports of ICT: Laptops and Data Servers

1. Introduction

Information and communications technology (ICT) devices such as laptops and data servers are being used on average for 3–5 years (van Driel, 2020), whereas research shows that they can be used up to 7 years before being replaced (Bakker et al., 2014). This has accelerated the demand for ICT manufacturing, which has caused significant environmental pressure with greenhouse gas emissions and increased use of critical materials. Transitioning from a linear to a circular economy (CE) (Kirchherr et al., 2017), which encourages the repair, reuse, recycling, re-manufacturing, and re-purposing of materials and products, is a possible solution. To facilitate this paradigm shift, and boost the data economy and digital innovation in the field, the European Union has introduced the concept of digital product passports (DPPs), which should provide information about a product’s lifetime to bring more transparency into supply chains (European Commission, 2022). Although the research on DPPs is on the rise, the adoption is limited and there is a lack of guidelines for its easier facilitation. Motivated by this, we conducted interviews (McMahon et al., 2023) with ICT procurers, managers and sustainability experts from the telecommunications industry, the Dutch government and academia to further investigate the current barriers to CE’s adoption in the ICT domain. Data’s availability, accessibility and interoperability have been highlighted as underlying barriers to diverse decision making scenarios for ICT procurement. The lack of findable, accessible, interoperable and reusable (FAIR) (Wilkinson et al., 2016) data is also a technical challenge when building DPPs of ICT, which aim to support sustainable decision making by providing details about a product (e.g. an ICT device, its hardware components, materials and their lifetime) (Walden et al., 2021).

Ontologies have emerged as a possible solution due to their ability to interlink knowledge from different domains, represent it in a machine-understandable format and be used as unifying vocabularies and knowledge management infrastructures across organisations (Fensel, 2002; Maedche et al., 2003). Motivated by the need for FAIR ICT data sharing for the CE, we carried out an extensive survey (Kurteva et al., 2024) of ontologies for ICT, materials and the CE and investigated the role and value of semantic interoperability across these domains. Our analysis showed that currently, there is a lack of connectivity between these three domains, which are closely connected when looking through a CE lens. While ICT devices (e.g. laptops) and their hardware components have been represented by some ontologies, the materials they are made of and possible CE processes that can be applied are rarely present. The ontologies for materials also do not provide direct links to hardware components thus most ontologies remain domain-specific. In addition, most of the existing ontologies in these domains are not openly available and have not been documented according to best practices to aid future work on their reuse, extension and alignment with other ontologies.

This article is based on a real-world ICT DPP use case supported by the Circular Resource Planning for IT (RePlanIT) ¹ project: ‘The municipality of Amsterdam needs to procure laptops for their staff. The goal of the municipality is to become more sustainable in every area of its operations including laptop procurement. The experts, responsible for the decision-making, come from different backgrounds (technology, sustainability, management) and thus have different knowledge of the product’s (or device’s) sustainability and the factors affecting it. Currently, the knowledge on this is not easily accessible and interoperable by machines and humans. It is siloed between different people, departments, organisations and even the device’s manufacturers. Further challenge is the lack of tools that can support the experts’ decision-making process’. The main goals are to establish a unified vocabulary between all different expert groups of what ICT DPPs are, what data they should store and how they can be used to support CE’s implementation in the ICT domain. This can be realised by utilising the DPPs for ICT procurement and machine-based decision making such as predictive maintenance. Based on this use case and findings by McMahon et al. (2023) and Kurteva et al. (2024) regarding the current challenges to cultivating a more circular ICT ecosystem, the role of semantic interoperability and importance of FAIR data sharing, we present the RePlanIT ^{2 , 3} ontology for machine-readable DPPs of laptops and data servers. Our work goes beyond state-of-the-art by representing laptops and data servers at both hardware component and material levels to aid their CE lifetime monitoring. In addition, the ontology represents several categories of indicators that can be used to measure the sustainability of a device. As presented later on in Section 4, the RePlanIT ontology is currently utilised for building knowledge graph-based DPPs of laptops and data servers, which are the core data source for the RePlanIT tool (see Figure 1) aimed at motivating sustainable ICT procurement. To summarise, with this article, we make the following contributions:

–

A novel, open access, publicly documented and validated with a real-world use case ontology for ICT (laptop and data servers) DPPs, which interconnects the ICT, materials and CE domains.

–

A set of (semantically represented) functional, economic and sustainability indicators, derived from interviews with expert end-users and literature survey part of the ontology, which assist the sustainability evaluation of ICT devices during procurement.

–

A knowledge graph of ICT DPPs of over 120 diverse laptops and data servers that is openly accessible through APIs.

OPEN IN VIEWER

Figure 1.

RePlanIT UI: DPP-based laptop sustainability report (Noordhoek et al., 2023). UI: user interface; DPP: digital product passport.

The rest of the article is structured as follows. Section 2 presents an overview of related work on ontologies for DPPs and a summary of the findings. Section 3 presents the methodology for building the RePlanIT ontology, while Section 4 outlines the main ontology concepts and their relations. The evaluation and validation of the ontology, its utilisation as a schema for a knowledge graph of DPPs and its application to support human decision-making during ICT procurement are presented in Section 5. Conclusions are presented in Section 6.

2. Related Work

According to Kurteva et al. (2024), we conducted an extensive literature survey of existing semantic models in the ICT (focus on laptops), materials, and CE domains. Overall, as shown by Kurteva et al. (2024), 27 semantic models (taxonomies and ontologies) have been identified. However, the majority of them lack public documentation and are not openly available, which limits their reuse. This section presents a brief overview of the openly available ontologies in the above-mentioned domains.

In the ICT domain, the oneM2M Base Ontology (The oneM2M Consortium, 2019), DogOnt (Bonino & Corno, 2008), and the ontology by Corcho et al. (2021a, 2021b) represent a variety of ICT and its hardware at different levels of granularity and through different lenses. The oneM2M ontology, built in collaboration with several Internet of Things (IoT) standardisation organisations, is a general representation of operations, services and input/output data in the scope of the oneM2M device ecosystem. A device has been defined as a thing that can communicate electronically via a network. Specific types of devices have not been represented thus the definition remains abstract. Metadata has been defined as a class itself, however, there is no specification of the metadata types themselves. Due to its abstract nature, oneM2M can be utilised as an upper-level ontology to align more detailed ontologies or can itself be extended with more concrete examples of ICT hardware. The DogOnt ontology, on the other hand, represents more concrete and diverse types of devices (referred to as appliances) in the scope of smart domotic environments. Appliances have been categorised as white (washing machine and boiler) or brown goods (e.g. printer and computer) based on their size and functionality. The ontology is one of the few that not only represents specific devices but also represents different sensors (humidity, flood, light etc.) that can be embedded in them and measurements such as their active, reactive and energy usage. However, the high granularity of it can also be seen as excessive and when modelling cases focused on specific types of ICT devices such as ours. To mitigate this, while utilising the relevance of DogOnt, as shown later on in this article, RePlanIT reuses the concept dogont:Computer as a type of ICT device. In contrast, Corcho et al. (2021a) present a high-level ontology network for ICT infrastructures, which comprises nine interconnected ontologies that model different entities (organisations and data centres), hardware and software components and network security. Despite the modular nature of the work by Corcho et al. (2021a), which allows for the ontologies to be reused in a standalone manner, many interdependencies between them are still present. Reusing even one of the ontologies without following these dependencies can rupture the represented domain knowledge. As an extension of Corcho et al. (2021a), the authors present by Corcho et al. (2021b) an ontology for hardware items related to software development (dev) and operations (ops) (DevOps) infrastructures. The ontology represents specific hardware such as disks, and types of servers. To utilise the work by Corcho et al. (2021a, 2021b), RePlanIT currently reuses the concepts (devops-infra:VirtualServer, devops-infra:PhysicalServer) to represent types of servers.

An important aspect for successfully monitoring an ICT device’s lifetime in the CE are the materials used for its manufacturing and their CE lifetime. The materials domain has seen significant contributions in terms of ontologies as showcased by Kurteva et al. (2024). Cheung et al. (2008) present the MatOnto ontology, which aims to assist with data-driven material discovery. Through a chemistry perspective, the ontology represents types of materials (polymers, metals and glass) their chemical, biological and magnetic properties and the results of processes such as chemical material modelling and evaluation. Although MatOnto was built for a particular material-related activity, which limits its application, it still provides a generalisable categorisation of materials’ types. In our work, the class matonto:Material and its subclasses have been reused and extended with concrete examples of each material category. Hastings et al. (2015), on the other hand, focus on semantically representing a specific type of material – nanomaterials and their characteristics on a particle level. The presented by Hastings et al. (2015), eNanoMapper ontology, further represents different physicochemical and biological characteristics of engineered nanomaterials to support processes such as drug discovery, delivery and safety. eNanoMapper’s particular nature and scope restricts its reuse in more abstract use cases such as our, where materials have a pivotal role as well. In contrast to eNanoMapper, the EMMO (The EMMO consortium, 2021) ontology network represents 4D objects at different levels of detail. The top-level ontology represents types of quantum, physical and void items and collections of items, while the middle layer ontologies – specific domains that EMMO can be utilised for. While EMMO is not suitable for our use case due to the domain it models, it can be utilised as an upper-level ontology. The work by The EMMO consortium (2021) has shown promising results with regards to EMMO’s successful utilisation and alignment with the battery value chain (BVC) ⁴ and mappings ⁵ ontologies.

The semantics of materials are also quite divergent across domains. For example, the Building Information Modelling-based holistic tools for Energy-driven existing Residences (BIMMER) (The BIMERR Project, 2021) ontology represents materials in the building domain. The ontology focuses on property building and renovation processes and represents a building board as a type of material. However, it does not capture information on the type of material the board is made of. In contrast, Voigt and Kalidindi (2021) present an ontology that focuses on materials themselves, their chemical structure, properties and performance. In addition, the developed ontology is able to capture provenance information on a material level, which is useful for deriving process steps such as heat treatments and soaking and their sequence of execution. The information can provide valuable insights into the current lifetime (e.g. highlight errors, and incompatibilities on chemical level) of a material and support future material engineering. Lambrix et al. (nd) further look into the materials domain and present the Materials Design Ontology (MDO). The ontology captures knowledge about materials from the lens of solid-state physics and condensed matter theory and can support various computational methods for material design. Representing provenance information about each material’s history is achieved by reusing the PROV (Lebo et al., 2013) ontology. Although MDO is still under development, it has proven to successfully facilitate the harmonisation and federation of several materials databases (Lambrix et al., nd). A more recent work, which is currently under development, is Materials and Molecules Basic Ontology (MAMBO) by Piane et al. (2021). One of the main goals of MAMBO is to support processes related to material engineering at the nanoscale. To achieve this, the authors propose capturing knowledge of materials at a molecular level. Each material is represented in terms of the material units (e.g. particles and atoms) in its structure and is associated with various measurements and calculations as part of the material’s engineering process.

Last but not least, there has been a rise in the ontology work in the CE domain. According to Kurteva et al. (2024), we identified six semantic models, amongst which three ontologies (i.e. Blomqvist et al., 2023; Echefaj et al., 2023; Sauter et al., 2019). Sauter et al. (2019) present the circular exchange ontology (CEO) and the circular materials and activities ontology (CAMO), which focus on the circularity of textiles in the retail sector. In comparison to the previously overviewed ontologies for ICT and materials, CEO and CAMO represent cross-domain knowledge. The authors have recognised the numerous factors affecting a product’s lifetime in the CE and have represented concepts of different types of materials and CE processes and activities. Despite its limited scope, the work by Sauter et al. (2019) can be seen as a baseline and guideline for future CE ontologies. The importance of cross-domain data interoperability to drive the implementation of the CE has been also recently acknowledged by Blomqvist et al. (2023). By following a modular ontology engineering approach to aid reuse, the CEON (Blomqvist et al., 2023) ontology network has been proposed. Although it is more of a high-level representation of products, materials and CE strategies for lifetime extension (e.g. repair, reuse and refurbishment), this is one of the most up-to-date ontologies that covers a wide range of cross-domain concepts. Last, but not least, Echefaj et al. (2023) investigate the CE domain from a resource supply chain perspective. To assist organisations in selecting the most optimal suppliers in terms of, for example, sustainability, cost, and CE awareness, the authors have semantically represented a set of criteria (e.g. sustainability and economic) to aid the process. While a valuable resource and contribution to the CE domain, the developed ontology is not openly available and its limitations need further investigation.

To summarise, most of the existing ontologies are domain-specific, have limited scope and interpret the ICT, materials and CE domain through different lenses. Cross-domain relationships are rarely represented and when defined they are often quite abstract. The limited availability of the ontologies further restricts their reusability. Representing DPPs of ICT devices for the CE in a machine-interoperable format necessitates overcoming these challenges to facilitate cross-domain knowledge exchange. Motivated by the need for better data interoperability in the CE and decision support (for both humans and machines) and the lack of an ontology that can be used as a guideline (or baseline) for representing DPPs of ICT devices in detail when it comes to their functional characteristics, hardware components, material composition and CE lifetime, we propose the RePlanIT ontology. To showcase the novelty with regards to the cross-domain knowledge that RePlanIT captures for building ICT DPPs, we compared it to existing ontologies as shown in Table 1. The data types in Table 1 were derived from collaboration with industry on the RePlanIT project, interviews with ICT managers (McMahon et al., 2023) and literature review (Kurteva et al., 2024) in the scope of our use case. Definitions and examples for each term from Table 1 are presented in Table A1 in the Appendix. A ‘✓’ signifies that an ontology represents or can represent the information via inference of classes and their subclasses. Table 1 shows that apart from RePlanIT, the CEON (Blomqvist et al., 2023) ontology is the only existing ontology that represents diverse concepts across domains relevant to a product’s lifetime in the CE. However, CEON represents these concepts at an abstract level. It represents only high-level concepts such as a product, component and material and not specific types for each (e.g. laptop as a type of a product, specific hardware components such as central processing unit, hard disk, materials such as plastics, metals and ceramics), which makes it suited to be used as an upper-level ontology to align domain-specific ones. In comparison, RePlanIT can represent knowledge (i.e. DPPs) about an ICT device at different granularity levels to support its sustainability and circularity evaluation – two complex processes that require the interpretation of various heterogeneous data about the device’s hardware components, materials used for their manufacturing, and both of their lifetimes in the CE.

The overall methodology for building the RePlanIT ontology comprises eight steps as presented in Figure 2. The work began with an exploration of the main use case’s scope, namely building DPPs for laptops and their hardware components, and setting a hypothesis for the types of data that need to be modelled by the ontology. In the next step, we conducted interviews (McMahon et al., 2023) with 11 experts (over five public and private organisations) from the domains of organisational decision making for the procurement, maintenance, repair and disposal of ICT equipment. The domain experts were asked about the existing procedures for each of these activities, the success or failure of new initiatives to introduce circularity into these activities, and the experienced or anticipated barriers to that introduction of circularity for each ICT-related activity. Further details and insights are presented by McMahon et al. (2023). Next, we conducted an extensive literature survey (Kurteva et al., 2024) of semantic models for ICT devices, their hardware components, materials and CE processes. The gathered information on the topic (from the interviews and the literature survey) led to deriving a set of competency questions (Table A2 by Kurteva et al., 2024). The set of questions was refined (see Table A2) and used as guidelines for the data that our ontology needs to represent semantically. The survey further helped identify existing ontology concepts (classes, data and object properties) that can already be reused within RePlanIT to answer the competency questions. In summary, some of the ontologies that we have reused are MatOnto (Miller, 2021) for representing materials due to its clear and easy to comprehend specification of the different types of materials (see Figure 3 by Miller, 2021), the well known and widely used PROV-O (Lebo et al., 2013) for defining types of agents involved in the lifecycle of an ICT device in the CE, the Data Catalog Vocabulary (DCAT) ⁶ (Consortium et al., 2014) for defining the agents’ roles and SOSA (Janowicz et al., 2019) for sensors, their observations and results.

OPEN IN VIEWER

Figure 2.

Overall methodology.

OPEN IN VIEWER

Figure 3.

RePlanIT ontology class overview (reused concepts, object and data properties are presented together with their existing namespaces, while newly defined ones have the name space ‘replanit’ (omitted for readibility purposes)).

OPEN IN VIEWER

Figure 4.

RePlanIT ontology overview: class ICTDevice (classes in blue and data properties in grey). ICT: information and communications technology.

OPEN IN VIEWER

Figure 5.

RePlanIT ontology overview: class indicator (classes in blue and data properties in grey).

OPEN IN VIEWER

Figure 6.

RePlanIT ontology overview: class HardwareComponent (classes in blue and data properties in grey).

OPEN IN VIEWER

Figure 7.

RePlanIT ontology overview: class Material (classes in blue and data properties in grey).

OPEN IN VIEWER

Figure 8.

RePlanIT ontology overview: class CEStrategy (classes in blue and data properties in grey). CE: circular economy.

OPEN IN VIEWER

Figure 9.

RePlanIT ontology overview: class unit (classes in blue and data properties in grey).

OPEN IN VIEWER

Figure 10.

RePlanIT ontology overview: class activity (classes in blue and data properties in grey).

OPEN IN VIEWER

Figure 11.

RePlanIT APIs: Swagger documentation.¹⁰

OPEN IN VIEWER

Figure 12.

RePlanIT UI: DPP-based laptop comparison (Noordhoek et al., 2023). UI: user interface; DPP: digital product passport.

The ontology development itself carried out with Protégé ⁷ , followed a top-down modular approach similar to the one by Shimizu et al. (2021). We begun by defining high-level concepts (e.g. ICT Device, Hardware Component, Materials, CE Strategy and Indicators) that form modules of knowledge that are interconnected in RePlanIT (details in Section 4). For consistency, we have used ‘isA’ relationship as a guideline when defining classes and subclasses before the actual ontology implementation. This helped as we collaborated with domain experts who did not have experience in ontology engineering. We then translated the ‘isA’ relationship to the more formal ‘rdfs:subclassOf’ in the ontology. Once all concepts for each module were defined, the relationships interconnecting the modules were specified. The ontology was then evaluated with standard ontology evaluation tools such as the OntOlogy Pitfall Scanner! (OOPS!) (Poveda-Villalón et al., 2014), the Pellet (Sirin et al., 2007), and HetmiT (Glimm et al., 2014) reasoners in Protégé. The evaluation was also performed with the set of predefined competency questions (presented in Table A2). An iteration of the ontology to fix inconsistencies was performed, followed by ontology documentation with WIDOCO (Garijo, 2017) and its public release. These steps comprise the ontology development process. To further validate the ontology, its utilisation as a schema for a knowledge graph that represents DPPs of laptops and data servers was investigated. Through the use of APIs, the built knowledge graph was further used as the main data source for the sustainability calculations provided by the RePlanIT user interface (UI) (Figure 1). Details on the RePlanIT ontology’s evaluation and validation based on its application(s) are presented in Section 5.

The RePlanIT^2,3 OWL ontology interlinks the ICT, materials and CE domains to represent machine-readable DPPs of ICT devices, namely laptops and data servers. Currently, the ontology (Figure 3) comprises of 295 classes, 72 object properties and 143 data properties, reuses concepts from 15 existing ontologies and defines new ones to represent highly granular DPPs at both device and device component levels. Figure 3 presents an overview ⁸ of the main ontology concepts represented as classes and the relationships between them. A detailed ontology documentation is available online.²

5. Evaluation

The RePlanIT ontology was evaluated following best practices for ontology evaluation – against a set of competency questions based on an extensive literature survey and interviews with experts from the sustainability and ICT domains in the context of the use case, with the OOPS! (Poveda-Villalón et al., 2014) ontology pitfall scanner throughout its development and with the HermiT (Glimm et al., 2014) and Pellet (Sirin et al., 2007) reasoners in Protégé. The ontology was validated through its successful utilisation for (i) building a knowledge graph of ICT DPPs and (ii) by designing and implementing a UI, which utilises the knowledge graph-based DPPs to support more sustainable ICT procurement in companies. The usability of the UI and the usefulness of the DPP data visualised on it, have been evaluated with end-users from the Municipality of Amsterdam in the scope of the RePlanIT project. Last but not least, RePlanIT not only reuses but also has already been reused for alignment with other existing ontologies (Section 5.4) to enable semantic interoperability between data spaces for CE’s monitoring. The following sections present more detail on the ontology evaluation in terms of use case modelling, ontology engineering, usability and reuse.

6. Conclusions

In this article, we presented the RePlanIT^2,3 ontology which by interlinking the ICT, materials and CE domains supports the representation of machine-interoperable DPPs of ICT devices such as laptops and data servers. RePlanIT’s evaluation and validation with a real-world scenario have shown that it represents useful data about devices, their components, material composition and possible circular strategies that can be used to extend lifetime and that it can be utilised successfully for various sustainability calculations that support decision makers in ICT procurement.

While we acknowledge that our work is limited with regards to the use case it covers (i.e. laptops and their sustainable procurement), it can already be reused for representing DPPs of ICT hardware components depending on the level of granularity of one’s use case. The hardware components that RePlanIT represents are often part of other ICT devices such as data servers thus we see a wide range of applications for our work. The reuse of RePlanIT for supporting cross-data space data interoperability for CE’s implementation in the batteries domain has already been investigated as well (see Section 5.4 and Chirvasuta et al., 2025). An improvement that we plan is the temporal aspect of our ontology. More specifically, validating that temporal changes for each data type are successfully captured and managed for processes such as predictive maintenance. Due to the intricate nature of laptops and the various factors affecting their recyclability, including material composition, device condition and economic considerations, we recognise the need for a more detailed representation of the recyclability characteristics of devices and materials in our ontology. While this was not in the scope of our work initially, it becomes an important aspect to consider for the future iterations of the ontology.

Following the RePlanIT ontology, we have also already built prototypes of data server DPPs and provided APIs for their querying with SPARQL. We are also actively monitoring the developments ²⁸ on DPP standardisation and believe that Open Access work such as ours can support their wider adoption and technical implementation. As for future work, a research direction that we are currently investigating is RePlanIT’s enrichment or alignment with existing lifecycle assessment ontologies to fully capture the sustainability and circularity of all ICT components and their materials through their lifetime from mining, through manufacturing, distribution, to use and specific CE strategy use. Last but not least, we are currently investigating the utilisation of knowledge graph-based DPPs for more explainable predictive maintenance and automated sustainability recommendations throughout a device’s lifetime.