In times of potential resource shortages and urgent need of a reduction in CO2 emission, reclaiming resources from end-of-life products becomes more and more relevant. This importance has already resulted in a growing number of legal regulations against the waste of resources and a strong push towards a circular economy. The design of a product has a strong influence on its environmental impact and its recyclability. In the project Methods and Technologies for an intelligent Circularity of Materials (MaTiC-M), we aim to develop a Design-for-Circularity methodology which accounts for product recyclability right from the design phase. However, lack of information for both product developers and recyclers about the entire life cycle of a product impedes recycling-friendly designs and eficient recycling processes. Overcoming this obstacle requires the integration of a wide range of data from diferent sources. Our goal is to develop a knowledge graph to facilitate the exchange of information between product designers and recyclers. In a first step, we analysed the requirements of various stakeholders on such a knowledge graph. In this paper, we present our procedure for the requirements analysis, as well as the resulting catalogue of competency questions.
In the manufacturing industry, we can currently observe two major trends: On the one hand, there is a
strong push towards net-zero emissions as witnessed, e.g., by the European Climate Law, which aims at a
climate-neutral European economy and society by 2050 [
The product design and development is of particular importance in this context as it influences the environmental impact along the entire life cycle of a product: It afects emissions, resource consumption, and waste generation during manufacturing, the extend of its use (including, e.g., robustness and durability but also the ability to repair it or replace broken components), and the degree to which its components can be recycled at its end-of-life. At the moment, product designers often lack proper information to make informed decisions on many of these sustainability aspects. In the same way, recyclers currently have little information available about products often causing suboptimal recycling results with respects to both ecological as well as economic regards as, e.g., rare and valuable materials cannot be specifically preserved.
In the project Methods and Technologies for an intelligent Circularity of Materials (MaTiC-M) [
A crucial aspect of this methodology is the information exchange between the various involved stakeholders, including the aforementioned product designers and recyclers but also, e.g., life cycle analysts, to rate diferent design options based on their environmental impact. In this paper, we report on the development of the knowledge graph that is at the heart of this information exchange. Here, we collect and interconnect all relevant information ranging from material characteristics, over life cycle assessments, to recycling strategies. The goal is to provide a decision support system for both product designers and recyclers that helps to make more eficient use of available resources, reduce the reliance on primary raw materials, and thus contribute to meeting the net-zero goals set out by politics.
This paper is structured as follows: In Section 2, we provide background information on circular economy and the Linked Open Terms (LOT) methodology. In Section 3, we introduce the method we applied to determine the requirements for the MaTiC-M knowledge graph. In Section 4, we present the results of the requirements analysis. In Section 5, we reflect on lessons learned with the requirements analysis in this projects. In Section 6, we provide a conclusion and give an outlook for the further goals of the project.
In this section we provide background information on key concepts for this paper.
A circular economy [
The Linked Open Terms (LOT) methodology [
An important obstacle to implement a circular economy is the lack of information of all stakeholders
along the product life cycle. In particular, product developers often can not assess the implications of
certain design decisions for the sustainability and recyclability of the product. Vice versa, recyclers
lack information on the individual parts and materials used in a product to select the optimal recycling
strategy. In MaTiC-M we aim to address this obstacle with an RDF knowledge graph that supports
the exchange of the required information. RDF supports us in the integration processes of the tool
landscapes of the involved stakeholders, e.g., via a rather seamless transition in communication protocols
using JSON-LD [
For the development of the MaTiC-M knowledge graph, we apply the Linked Open Terms (LOT)
methodology [
Based on an internal preliminary study, the use cases have been outlined in the MaTiC-M project plan: 1. provide information on recycling capabilities to enable an assessment of the recyclability of a product during its design as part of a Design-for-Circularity methodology, 2. provide information about designs and materials to enable the selection of an optimal decomposition and recycling at the end of life of a product.
However, as the Design-for-Circularity methodology itself is developed during the project, we expect further refinement of the use case specifications over the course of the project.
Interviews were conducted with domain experts, to identify the file formats currently used for storing
or exchanging data between stakeholders along the product life cycle. These interviews revealed a
range of relevant data formats, including tabular data stored in CSV and Excel files. Furthermore, 3D
models of the product designs must be analysed by several stakeholders. However, these analyses
require diferent computer-aided design ( CAD) data formats. The standard for the exchange of product
model data (STEP) format [
The purpose and scope of the knowledge graph is already outlined in the project proposal. However, the specific methods the knowledge graph needs to support are still under active development. As a consequence, the purpose and scope definition could not be detailed so far. This is a unavoidable deviation from the LOT methodology. Therefore, the following knowledge graph development is based on the outlined assumptions, which might change to a certain degree over the course of the project, though. Nevertheless, some decisions have already been made. For example, the primary language of the knowledge graph was decided to be English, supplemented by German as secondary language.
We specified the functional ontology requirements in the form of competency questions ( CQs). In a ifrst step, all involved domain experts were asked to provide CQs that are relevant for the use cases in their domain and area of expertise. Overall, 14 domain experts from seven institutes of diferent domains within the German Aerospace Center (DLR) contributed to the collection of CQs. To prepare the domain experts for the task, example questions were provided along with an introduction of the purpose of the CQs. This resulted in several independent list of CQs (one per stakeholder or group of stakeholders involved in the project; in total seven independent lists). However, the level of granularity and compliance with the question format difered. In particular, we experienced a clear diference between domain experts with and without prior experience in developing or using knowledge graphs. Initially, the CQs were collected and edited in German.
In a second step, we consolidated the collected CQs into a single list. During this consolidation and in close collaboration with the domain experts, we also clarified and aligned the CQs. This included the following activities:
Some initial CQs were only single keywords instead of full questions or the questions were consecutive questions that are understandable only in the context of the previous ones. We rephrased these CQs to comply with the intended question format and to be unambiguous and independent of one another.
Some initial CQs were rather long or complex and involved a multitude of independent aspects.
We split these CQs into multiple CQs to obtain a more even level of granularity.
Each CQ contains important terms. We highlighted these important terms to identify relevant concepts. Further, this helps to keep an overview during the question consolidation.
We noticed that there is couple of central topics among the CQs. Therefore, we grouped the CQs based on their main topic of concern. This activity is merely for organisational purposes and does not imply any priority or rating among the CQs. It helps to keep an overview during the question consolidation and we expect it to help during the ontology implementation, too.
Some equivalent or very similar questions were part of or derived from multiple initial CQ lists.
We merged these CQs to avoid redundancy.
We identified terminology gaps between diferent domains and contributors and clarified the
meaning of used terms. For example, we decided that we consider (a) the term “material” as the
chemical composition and structure of things like in “materials science” and therefore rejected
the interpretation as things that are used to produce another thing like in a “bill of material” [
During these activities, we maintained the links between initial CQs and their according consolidated CQs to track the provenance of the latter. Due to splitting and merging during the consolidation, there is a many-to-many relation between initial and consolidated CQs. The provenance of each CQ helped to ensure that all initial CQs have been considered entirely in the consolidated CQs and that the consolidated CQs do not extend the scope of the initial CQs. Further, it eased the validation of consolidated CQs in the next step. These activities resulted in a CQ catalogue draft.
To complete the CQ catalogue, we conducted a verification of the catalogue draft. All involved domain experts were asked to check whether the consolidated CQs cover all their information needs and that
Topic material scrap collection disassembly recycling 202 individual part / assembly / product 301 connection / joining material source manufacturing / production What properties (e.g., tensile strength, thermal conductivity, heat capacity, density) does a material have? Which parts and assemblies does an assembly / product consist of? Which connections of which connection type / which joining method exist between two components? In what amount is a material available from a primary / secondary material source? How much electrical energy does a production process / manufacturing process require per workpiece or amount of material? What material does an amount of scrap consist of? What personnel is required for the collection of the scrap of a product or assembly? Which disassembly steps are to be carried out for a product? How much electrical energy does a recycling step require per product or component or amount of material? What is the lifespan of a device? How many person hours with which qualifications does a personnel requirement comprise? none of their intents was misinterpreted. This revealed a few gaps in the consolidated CQs that have subsequently been addressed. In one case, one of several alternative concepts in a question was missing. In other cases, the links between the initial CQs and the consolidated CQs were incorrect.
To produce the ontology requirements specification document (ORSD), we decided to create a table containing the consolidated CQs, stored as a CSV file. As the non-functional requirements can not be comprehensively specified yet due to the project’s setup, this is suficient to store the known requirements. Each CQ is stored both in German as well as in English. We used the translation service DeepL Pro1 to generate a draft translation from German into English, which was manually reviewed and corrected afterwards. Further, we assigned a numeric identifier to each CQ.
At this stage of development, we identified a total number of 97 CQs that are grouped regarding their
central concepts into 11 topic groups: Material contains questions around characteristics of individual
materials employed. Individual part / assembly / product centres around which components make a final
product whereas connection / joining includes question of how they are assembled. Material source and
manufacturing / production provide necessary information for the life cycle assessment [
Unfortunately, we can not publish all consolidated CQs for now to protect intellectual property rights
of our project partners. Therefore, we omitted or generalised some of the question in the consolidated
CQ table for publishing. The resulting table contains 65 CQs. It is publicly available [
As mentioned in Section 3, we had to adapt the LOT methodology according to the framework of the project. Even though very helpful, it did not fit perfectly in our situation, where the application environment of the knowledge graph is not yet largely defined. In such case, a less linear but more agile methodology might be a better option.
Maintaining the provenance of CQs turned out to be very helpful for later activities during the requirements specification. But as we used a wiki platform to collect and consolidate the CQs, maintaining the links between initial CQs and consolidated CQs was a rather cumbersome manual task.
During the collection of the initial CQs we experienced several misconceptions regarding their purpose and form. In particular, there was a misconception that the CQs are about information needed within the project as opposed to information needed to apply the Design-for-Circularity methodology. We also noticed a clear diference between domain experts with and without prior experience in developing or using knowledge graphs. As a take-away, we need to put more emphasis on the introduction of the requirements specification activities to the domain experts in future projects.
The implementation of a circular economy requires stakeholders along a product life cycle to exchange information about the product. For instance, product designers need more information about the processes at the end-of-life/end-of-use of a product, to ease the product recycling and recyclers have to know the structure and involved materials of a product to decide for an optimal recycling strategy. We investigated the requirements for a knowledge graph that supports product designers and recyclers to increase the recyclability and the actual recycling rate of products on the path to a circular economy. Based on the identified requirements, we will develop the MaTiC-M knowledge graph to support the Design-for-Circularity methodology currently under development. We intend to make extensive use of existing ontologies and knowledge graphs to (a) reduce the development efort and (b) ensure compatibility with other eforts in the domain.
Many thanks to Mohamad Abdallah (DLR Institute of Vehicle Concepts), Elmar Beeh (DLR Institute of Vehicle Concepts), Katrin Bugelnig (DLR Institute of Materials Research), Jan Haubrich (DLR Institute of Materials Research), Sree Sameer Kumtamukkula (DLR Institute of Vehicle Concepts), Tom Lorenz (DLR Institute of Low-Carbon Industrial Processes), Miriam Löbbecke (DLR Institute of Materials Research), Nicole Carina Neumann (DLR Institute of Future Fuels), Korbinian Nottensteiner (DLR Institute of Robotics and Mechatronics), Diana Peters (DLR Institute of Data Science), Robert Reiser (DLR Institute of System Dynamics and Control), Ismael Valentin Rodriguez Brena (DLR Institute of Robotics and Mechatronics), Ralf Sturm (DLR Institute of Vehicle Concepts), Bernhard Thiele (DLR Institute of System Dynamics and Control), and Sven Torstrick-von der Lieth (DLR Institute of Lightweight Systems) for their valuable contributions to the requirements analysis.