1. Introduction
Sustainability has become a pivotal factor in contemporary product development, driving innovation, reducing environmental impact, and ensuring long-term economic feasibility (Reference RelichRelich, 2023). Besides, product development processes are becoming increasingly complex due to rapid technological advancements and evolving market demands. Organizations are therefore under growing pressure to deliver high-quality products while simultaneously minimizing environmental impacts and optimizing resource use. Achieving sustainability extends beyond creating environmentally friendly products to also encompass optimizing the processes involved in their production. A frequently overlooked aspect of sustainability is the management and reuse of data and knowledge. Effective reuse of existing knowledge significantly contributes to conserving resources and improving efficiency by reducing the need for repetitive data collection and redundant analyses (Reference Xin and OjanenXin & Ojanen, 2022). The concept of knowledge reuse, widely recognized in the software industry for its role in saving time, reducing errors, and enhancing productivity (Reference Ko, Abraham, Beckwith, Blackwell, Burnett, Erwig, Scaffidi, Lawrance, Lieberman, Myers, Rosson, Rothermel, Shaw and WiedenbeckKo et al., 2011), is similarly applicable to product development. In particular, the reuse of decision knowledge offers significant potential to optimize decision-making processes by leveraging insights gained from prior decisions (Reference Chang, Abdel-Basset and RamachandranChang et al., 2019). Improving the documentation and accessibility of decisions holds significant potential for enhancing decision-reuse. A structured framework is necessary to capture, organize, and retrieve past decisions, enabling organizations to improve efficiency and make more informed decisions, ultimately leading to better outcomes (Reference Manteuffel, Avgeriou and HambergManteuffel et al., 2018).
To address these challenges, knowledge management solutions are essential to support the product developers in their daily tasks. Among these, ontologies represent a particularly suitable approach for structuring and managing knowledge to ensure decision-making processes are informed by well-organized and reusable information. Ontologies are defined as an explicit specification of a shared conceptualization (Reference Studer, Benjamins and FenselStuder et al., 1998). They enable the connection of knowledge elements, facilitating the inference of new knowledge and the establishment of relationships (Reference Maletzki, Rietzke, Grumbach, Bergmann, Kuhn, Di Francescomarino, Dijkman and ZdunMaletzki et al., 2019). Furthermore, ontologies enhance transparency in decision-making and can indirectly capture tacit knowledge, which is often lost in traditional workflows (Reference Wang, Wang and WangWang et al., 2018). Their application in product development includes areas such as general product design (Reference Catalano, Camossi, Ferrandes, Cheutet and SevilmisCatalano et al., 2009), requirements engineering (Reference Lin, Fox and BilgicLin et al., 1996) or decision-making (Reference Pickel, Gerschütz, Horber, Goetz and WartzackPickel et al., 2023).
This paper examines how ontologies support the reuse of decision knowledge in product development. It introduces the state of the art to clarify the research context, outlines the need for action, and formulates the specific research questions. A methodological approach is presented, including a step-by-step framework, a classification scheme for decisions, and a decision-reuse ontology to store and retrieve past decisions. Finally, the findings are discussed in relation to the research questions, concluding with insights and directions for future research.
2. State of the art
Decision-making research encompasses selection and evaluation techniques (Reference Wartzack, Bender and GerickeWartzack, 2021), which systematically assess alternatives against predefined criteria to ensure objective and informed outcomes. In this context, securing organizational knowledge is essential for long-term success, requiring models to prevent its loss and enable efficient reuse (Reference Levallet and ChanLevallet & Chan, 2019). These models should structure and identify reusable decision patterns, enhancing decision-making processes. Siikarla and Systä Reference Siikarla and Systä(2008) offer fundamental research on the automatic validation of decisions, which supports consistent transformation processes, reduces workload and enhances the decision quality. By structuring knowledge elements, knowledge reuse can further support decision-making and emphasizes the importance of standardized decision structures. For instance, decision-centred templates can be utilized to modularize, document and reuse recurring design tasks (Reference Panchal, Fernández, Paredis and MistreePanchal et al., 2004). Addressing challenges related to knowledge reuse, Liu et al. Reference Liu, Duffy, Boyle and Whitfield(2008) present the relationships among key elements of knowledge reuse for decision support in a Unified Modeling Language (UML) diagram. Similarly, Chang et al. Reference Chang, Abdel-Basset and Ramachandran(2019) propose a strategic framework that employs standardized decision patterns and predictive analytics to make decisions reusable in recurring processes. In their systematic literature review, Krishnan and Ulrich Reference Krishnan and Ulrich(2001) categorize product development decisions based on the phases of the development process, encompassing both project decisions as well as decisions in setting up a development project. Initial illustrations are utilized to exemplify the relationships among decisions, but these lack graph-based representations. Specifically, the absence of descriptions for the edges limits the identifications of dependencies between the nodes. While this categorization provides a suitable basis, its application in knowledge management for reuse remains unexplored. To further explore the complexities of decision-making and the categorization of decision outcomes, Luft et al. Reference Luft, Schneider and Wartzack(2015) focuses on the evaluation of conflicting complex decisions, aiming to identify and structure their potential consequences.
The reuse of knowledge and decisions has become increasingly significant, particularly within knowledge management solutions such as ontologies. Ontologies provide a robust framework for organizing knowledge into classes and defining relationships between them, creating a solid foundation for modularizing knowledge to support decision-making. Their construction often follows established guidelines such as Ontology Development 101 (Reference Noy and McGuinnessNoy & McGuinness, 2001). However, the relationships between defined classes within an ontology are sometimes insufficient. To address this limitation, Resource Description Framework (RDF)-star enables the modeling, management, an versioning of semantic data, including the storage of meta-information related to decisions (Reference Arenas-Guerrero, Iglesias-Molina, Chaves-Fraga, Garijo, Corcho and DimouArenas-Guerrero et al., 2024). Ontologies enable knowledge reuse, as demonstrated by Dworschak et al. Reference Dworschak, Kügler, Schleich and Wartzack(2021) in the area of design process knowledge. The advantage of reusing existing ontologies lies in their adaptability through various matching approaches, which typically involve reuse analysis of existing resources (Reference Euzenat and ShvaikoEuzenat & Shvaiko, 2013). Several ontologies offer specific schemes for structuring decision-relevant knowledge. For example, the Decision Ontology (DO) (Reference NowaraNowara, 2005) and the Core Ontology of Organization Know-How and Knowing-That (COOK) (Reference Ghrab, Saad, Kassel and GargouriGhrab et al., 2017) propose concepts and relations within decision-making contexts. Similarly, Konaté et al. Reference Konaté, Zaraté, Gueye, Camilleri, Morais, Fang and Horita(2020) developed an ontology for collaborative decision-making. An initial approach by Pickel et al. Reference Pickel, Gerschütz, Horber, Goetz and Wartzack(2023) combines decision-making with product development to support developers in their daily tasks. The potential of ontologies to structure and reuse knowledge for decision-making has been demonstrated. However, it does not comprehensively address decision-reuse. Kontopoulos et al. Reference Kontopoulos, Martinopoulos, Lazarou and Bassiliades(2016) highlighted their capacity to streamline future decision-making processes. He et al. Reference He, Hao, Wang, Li, Wang, Huang and Tian(2020) further emphasized how processes can be optimized by utilizing decision-related knowledge stored in ontologies. Additionally, Guo et al. Reference Guo, Zhou, Yu, Zhou, Wu and Hao(2024) introduced an ontology for reusing design knowledge specifically in maintenance tasks, underscoring the importance of leveraging existing knowledge to improve decision-making. Ming et al. Reference Ming, Wang, Yan, Panchal, Goh, Allen and Mistree(2018) addressed the explicit reuse of decision-making scenarios by developing a hierarchical ontology that accounts for the relationships between decisions, offering a consistent framework for reuse.
3. Need for action
Taking the related work into account, a research gap regarding the reuse of knowledge, particularly decision knowledge, in product development, is revealed. While the examined studies highlight the potential of reusing knowledge and decisions, their primary focus is on categorizing and modularizing this information. However, these efforts are not sufficiently integrated within advanced knowledge management solutions. Specifically, the existing research does not fully leverage the capabilities of ontologies to support product developers in their decision-making processes, nor does it provide mechanisms for integrating information and insights from past decision scenarios. Based on these considerations, the following research questions (RQ) are posed:
-
RQ1: What factors determine the similarity and reusability of decisions in different phases of product development?
-
RQ2: How can ontologies support the management and reuse of decision knowledge in product development to enhance both current and future decision-making processes?
Building on the research questions, the study addresses a research gap in methods for classifying decisions within the product development process and the operationalization of ontology-based decision management. This includes the retrieval of similar past decisions, conducting SPARQL queries, and analysing the resulting data along with the potential impacts of those decisions. To enable meaningful reuse, decisions should be semantically structured and reversed where necessary.
The objective of this paper is to investigate how past decisions can be reused effectively in similar scenarios to foster learning, reduce iteration cycles, and save time. An ontology-based framework is proposed to assist product developers by utilizing insights from previously made decisions. To facilitate this, decisions must be systematically categorized, using modularized phases of the product development process as a basis. Furthermore, product development knowledge and decision-making information will be interconnected and made accessible through a linked ontology. This approach aims to create a comprehensive system for managing and reusing decision knowledge to enhance decision-making efficiency and sustainability.
4. Methodical approach
Based on the proposed research questions, a methodical approach is outlined and illustrated in Figure 1. The process begins with the presentation of a given decision problem (step one). For example, determining which bearing should be selected to meet the specific requirements in the scenario of transmission system design. In the second step, the decision problem is categorized using a predefined classification scheme, which consists of three hierarchical levels offering increasing granularity. If the initial classification results in an overly broad set of potentially relevant decisions, the scheme can be refined to a more detailed level. Conversely, if the scope is too narrow, a broader level can be employed. The details of the classification scheme are elaborated in Chapter 4.1. In the third step, the categorized decision is integrated into a decision-reuse ontology, further explained in Chapter 4.2. The ontology introduces the structure, including its classes, individuals, and slots, capturing knowledge from past decisions. The primary purpose of the ontology is to provide relevant historical decision knowledge to support the current decision-making process. In step four, specific SPARQL queries are used to extract and present relevant past decisions based on the classification scheme. Each retrieved decision includes a detailed description of its classification, the SPARQL query results, associated conditions (e.g., time parameters, decision criteria, and stakeholders), and the outcomes of the decision (e.g., the decision made and its impacts). This information enables the decision-maker to evaluate the relevance of past decisions to the current problem and determine whether they can be reused. The applicability of retrieved decisions depends on the similarity of their context and boundary conditions to the current problem. Decisions are considered irrelevant if they include context-specific or temporary criteria that cannot be generalized or applied to the current scenario. If none of the provided decisions are deemed relevant, the decision-maker can refine the query by navigating to a broader classification level. This process is visually represented by the inverted arrow connecting step three to step four, indicating the iterative nature of querying.
Figure 1. Overview of the methodical approach for reusing decisions in product development
Building on the previous steps, the methodical approach advances to step five, where additional SPARQL queries can be directed to a decision-support ontology (Reference Pickel, Gerschütz, Horber, Goetz and WartzackPickel et al., 2023). This query leverages the information and knowledge of similar decision problems retrieved earlier to provide targeted support for the current decision. Thus, the context of the decision, e.g., including stakeholders and influencing factors, can be considered, improving reuse effectiveness. Although the decision-support ontology and the decision-reuse ontology are interconnected, their distinct purposes require them to be applied in different steps for clarity. Once the product developer reaches a decision, the information and outcomes of the current decision are integrated into the decision-reuse ontology in step six. This integration uses the same classification scheme, ensuring that future decision problems can benefit from this newly acquired knowledge. Furthermore, any prior decisions that influenced or supported the present decision are explicitly linked within the ontology, creating a traceable network of interconnected decision knowledge.
It is important to note that the methodical framework does not imply an infinite loop between the decision-support ontology (step five) and the decision-reuse ontology (step three). Instead, the process highlights the importance of reintegrating the insights and outcomes of the current decision into the decision-reuse ontology, including its detailed descriptions. This structured approach ensures a systematic and adaptable reuse of decision knowledge, empowering decision-makers to effectively utilize prior insights while simultaneously enriching the knowledge base for future scenarios.
4.1. Classification of decisions in product development process
To facilitate the reuse of past decisions, a classification scheme has been developed (refer to step 2 in Figure 1). The hierarchical structure of this classification is depicted in Figure 2 and consists of three layers: level, problem typology, and work results. The first layer, level, aligns with the overarching phases of the product development process as defined by VDI 2221 (Verein Deutscher Ingenieure, 2019) and Pahl et al. Reference Pahl, Beitz, Bender and Gericke(2021), namely task clarification, conceptual design, embodiment design, and detail design. These phases are further elaborated through the second layer, problem typology, which categorizes decisions based on their specific focus. Strategical documents, such as market analysis, technology roadmaps, and resource plans, guide target decisions, while requirements define decisions related to specific requirements, resources, or framework conditions. Technological problems address principle decisions, technology selection, or material choices, while assessment and selection focus on evaluating and choosing the most suitable concepts. Design and draft concern structural and interface decisions. Quality and test relate to quality assurance, testing, and validation. Finally, production decisions determine the manufacturing process.
Figure 2. Categories of decision-classification inspired by Verein Deutscher Ingenieure (2019)
The third layer, work results, details the classification by linking each problem typology to specific outputs based on guidelines from VDI 2221 (Verein Deutscher Ingenieure, 2019) and Pickel et al. Reference Pickel, Bickel, Goetz and Wartzack(2024), whereby they can be assigned to more than one. Notably, strategical documents are not directly included in the product development process according to VDI 2221 (Verein Deutscher Ingenieure, 2019) but are recognized as critical influences throughout decision-making.
This classification scheme can be further refined, if necessary, by incorporating a more granular level of classification when a substantial number of past decisions are available. In this case, the categorization of product development decisions within a project and in setting up a development project can be applied (Reference Krishnan and UlrichKrishnan & Ulrich, 2001). Conversely, when fewer past decisions are accessible, it may be beneficial to employ a higher level of abstraction, ensuring that a sufficient breadth of historical data is incorporated for informed decision-making.
4.2. Ontology-based framework for reusing decisions
According to Pahl et al. Reference Pahl, Beitz, Bender and Gericke(2021), essential skills for managing decision processes include the ability to recognize dependencies, assess importance and urgency, as well as maintain consistency and flexibility. To address these aspects, an ontology is utilized that supports these capabilities. A central component of the methodological approach (see step 3) is the decision-reuse ontology. In general, an ontology is constructed using a class schema and the semantic triple format: <subject> <predicate> <object> (Reference Schreiber and RaimondSchreiber & Raimond, 2014), as defined by the RDF format (Reference Staab and StuderStaab & Studer, 2009). This allows for the representation of interrelationships, such as <stakeholder> <defines> <decision criteria>. The purpose of the ontology is to provide access to past decisions that are relevant to the current decision problem. Reuse is facilitated through querying the reuse ontology using the proposed classification scheme. Additional decision properties are assigned to the scheme, including SPARQL queries, conditions, and outcomes (see Figure 3), which are represented as ontological classes. These can be further subdivided into relevant subclasses. The classification class documents the categories outlined in chapter 4.1 - namely level, problem typology, and work results. Past SPARQL queries of similar decisions provide examples of how to query the ontology in subsequent steps. The specific queries and its corresponding results are stored as annotation properties of the individuals SPARQLquery and ResultSet, which offers information about the output knowledge derived from the query. A decision is dependent on its underlying conditions, so it is essential to provide details about the stakeholders (e.g., the decision maker) to facilitate consultations. The decision criteria significantly influence the decision process, meaning that, when multiple past decision situations are available, the one with the most appropriate criteria can be selected. This involves considering how well past decisions align with the categories of classification and its descriptions. This can be achieved through careful querying, comparison of decision properties, and continuous feedback loops to improve the decision-making process. The time parameter, particularly the date of the decision, is captured as meta-information, as described in detail below. To facilitate a learning effect from past decisions, it is also crucial to link the outcomes. While the decision-support ontology (Reference Pickel, Gerschütz, Horber, Goetz and WartzackPickel et al., 2023) primarily provides decision support, the actual decision and its impact on subsequent processes and products must also be documented. Both positive and negative influences, along with their interdependencies, are considered in this context.
Figure 3. Class scheme of the decision-reuse ontology
Figure 4 presents an overview of the ontological structure for decision-reuse, as well as the domain decision problem, which is described through the previously discussed (sub-) classes and slots, represented by the relationships between the classes. The classes and slots are illustrated in Figure 4, accompanied by a legend that explains the slots. In addition to describing the slots using object properties, meta-information regarding the semantic triples must also be stored. This particularly pertains to the temporal aspect of the <stakeholder> <made> <past decision> relationship. Based on the RDF schema, RDF-star (Reference Arenas-Guerrero, Iglesias-Molina, Chaves-Fraga, Garijo, Corcho and DimouArenas-Guerrero et al., 2024) can be employed by adding the versioning element <hasTimestep> <date> to the triple (as shown in Figure 4 on the left). This enables the inclusion of the decision’s validity period. RDF-star also has the potential to incorporate provenance, i.e., the source of a statement, although this is currently captured through relationships with the stakeholder. By versioning decisions, the development of similar decision situations and the accumulation of knowledge over time can be documented. Upon completion of a decision by the decision-maker, the associated information must be reintegrated into the decision-reuse ontology. It is essential to document which past decisions influenced or were referenced in the current decision-making process. To enable this linkage, past decisions are defined as recursive (rec), which is highlighted in yellow.
Figure 4. Conceptualized decision-reuse ontology with classes (yellow diamonds), slots (blue arrows), description of meta-data (on the left) and associated legend of slots (on the right)
A notable advantage of ontologies is the ability to reuse their structures. Beyond to the decision-reuse ontology, which concerns the decision problem, two additional ontologies are linked for broader coverage. The Friend of a Friend (FOAF) Ontology is used to describe stakeholders (Reference Kalemi and MartiriKalemi & Martiri, 2011), while the decision-support ontology (Reference Pickel, Gerschütz, Horber, Goetz and WartzackPickel et al., 2023) characterizes past decisions to assist future decision-making. Thus, the decision-reuse ontology facilitates the current decision-making process by applying the classification scheme and integrating the results of the current decision for future decisions.
With regard to the specific decision problem presented in Figure 1, “Which bearing do I choose for requirements XY?”, it is classified as an assessment (work result) within the assessment and results (problem typology) at the embodiment design level. Figure 5 demonstrates how a SPARQL query could be executed on the decision-reuse ontology (only the main excerpt from the query is represented). This query retrieves a set of past decisions with the same classification, providing not only the relevant SPARQL queries applied but also their corresponding result sets, conditions, and the actual outcomes. In this example, only a single past decision is presented for illustrative purposes.
Figure 5. Application of the decision-reuse ontology
By manually selecting the most pertinent decision problems, their SPARQL queries can be adjusted and submitted to the decisions-support ontology (Reference Pickel, Gerschütz, Horber, Goetz and WartzackPickel et al., 2023). This process provides decision-relevant knowledge, such as bearing catalogs, thereby assisting in the decision-making process. The selected bearing and the relevant information are then returned to the decision-reuse ontology for future reuse. Through these ontological connections, it becomes possible not only to identify which similar decisions have been made in the past but also to understand the reasoning behind these decisions, allowing for targeted retrieval of relevant knowledge.
The present decision-making problem is supported by facilitating interaction with the ontology through the provision of past SPARQL queries. These queries do not need to be developed entirely from scratch but only adapted to the specific application scenario. Furthermore, the potential consequences of decisions can be better assessed by reviewing previous decisions along with their respective outcomes. Additionally, transparency is ensured by clearly indicating which stakeholder made each decision, thereby enabling targeted follow-up in critical decision-making situations.
5. Discussion
Reusing decisions mitigates inefficiencies caused by time-consuming analysis processes, enabling access to empirical knowledge. This promotes a learning curve that reduces errors and conserves resources. The use of an ontology further enhances consistency, critical in the inherently complex product development process. The first research question (RQ1) aimed to identify the factors that influence the similarity and reusability of decisions across different phases of product development. A classification scheme was developed to structure decisions chronologically and technically within the product development process, following the VDI 2221 (Verein Deutscher Ingenieure, 2019), with its phases - task clarification, conceptual design, embodiment design, and detail design. Iterative search mechanisms allow users to refine their queries by exploring broader classifications or investigating more detailed subclasses. This prevents relevant decisions from being overlooked due to overly broad or specific initial searches. To improve clarity and universality, keywords and synonyms can be incorporated to ensure consistent representation of classes. This can be achieved by integrating open thesauri, such as WordNet (Reference Fellbaum and BrownFellbaum, 2006) or ConceptNet (Reference Speer, Chin and HavasiSpeer et al., 2017), which enable synonym recognition and allow users to perform SPARQL queries without requiring deep familiarity with the ontology. By leveraging these resources, decisions described with alternate phrasing can be retrieved, ensuring comprehensive access to relevant past decisions. Efficient navigation is supported through SPARQL queries for filtering by parameters such as context, outcomes, or timeframes, alongside metadata tagging (via RDF-star (Reference Arenas-Guerrero, Iglesias-Molina, Chaves-Fraga, Garijo, Corcho and DimouArenas-Guerrero et al., 2024)) to provide additional context for easier identification. OntoGraf (GitHub, 2016) visualizations further simplify interaction with the classification system, offering intuitive exploration of decision hierarchies and relationships. These tools help users navigate the ontology without extensive prior knowledge, ensuring accessibility. However, the classification of decisions relies on the developer’s expertise and manual execution, including the formulation of SPARQL queries. While these processes are supported by leveraging previous decisions and their queries, they still require active user involvement. Moreover, the approach is limited to the product development stages and does not cover the entire product lifecycle. In summary, the classification scheme establishes a robust foundation for decision-reuse by combining structured categorization, iterative search mechanisms, and tools like SPARQL filtering, metadata tagging, and OntoGraf (GitHub, 2016) visualization. These features ensure that users can efficiently locate and reuse decisions while addressing challenges related to the diversity of decision contexts and classification sensitivity.
The second research question (RQ2) explored how ontologies can support the management and reuse of decision knowledge in product development, thereby enhancing both current and future decision-making processes. A decision-reuse ontology was developed to manage and retrieve decision knowledge effectively. By linking the primary classes - classification, SPARQL queries, conditions, and outcomes - and connecting this ontology to an existing decision-support ontology, insights from past decisions support concrete decision-making. The ontology improves efficiency by reducing development iterations and minimizing misinformation through relevant, context-specific knowledge. A feedback mechanism allows users to refine and enrich reused decisions, enhancing their applicability and relevance for future scenarios. The ontology is particularly beneficial in repetitive decision-making scenarios, such as material selection, where experienced decision-makers can use it as a memory aid and newcomers as a learning resource. It is also valuable for high-stakes decisions, where reviewing past outcomes mitigates the risks of errors. The ontology serves as a support system, highlighting decision patterns and best practices, while leaving the final decision to users, ensuring human judgment remains central. While limitations exist in the need for manual processes and a focus on product development stages, this approach enhances decision-making efficiency, transparency, and reusability in diverse contexts of product development.
6. Conclusion and outlook
This paper introduces a novel ontology-based approach for decision reuse in the product development process. A classification scheme is developed for past decisions, which is integrated into a decision-reuse ontology. This enables the retrieval of relevant information about potentially similar past decisions. In addition, the decision-reuse ontology is linked to a decision-support ontology, allowing not only access to past decision knowledge but also supporting informed decision-making in subsequent stages of the development process. The final decision outcomes, including their classifications and associated metadata, are then fed back into the ontologies, creating a continuous cycle of knowledge enrichment and ensuring ongoing support for future decision-making.
Future research should focus on improving the automation of the classification and the decision-reuse ontology to enhance the efficiency of the proposed schema, while keeping it adaptable for further lifecycle phases and domains. Natural Language Processing (NLP) could be utilized to automate the classification of decision problems based on their descriptions, removing the reliance on manual classification and predefined schemas. Additionally, automating the integration of decision classifications, SPARQL queries, conditions, and outcomes into the decision-reuse ontology could streamline the reuse process. By automating the transfer of decision information between interconnected ontologies, the system’s interoperability, traceability, and consistency would be significantly improved, ultimately enhancing the overall functionality and ease of use.
To assess the benefits of the proposed framework, the next step requires an industry-related study that identifies the trade-off between the reuse benefits and the effort involved in storing and retrieving past decisions. This will include testing the practical application of the framework to uncover both its practical advantages and limitations. Additionally, the study will evaluate the classification scheme’s effectiveness and assess whether the framework is universally applicable across different product developers facing similar decision problems. It will also examine user-friendliness and identify areas for optimization to improve both functionality and ease of integration into existing workflows.
Acknowledgements
The research has been funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - 400342876.
Open access
