Knowledge-based product and process design synthesis of additively manufactured non-assembly mechanisms

Abstract

Additive Manufacturing increasingly becomes a viable alternative to conventional manufacturing technologies. However, the knowledge and awareness to achieve a ready-for-manufacturing and quality-ensured design of additively manufactured products are mostly lacking. Knowledge from the manufacturing process on how certain process variables affect the product requirements must already be considered in product design to assure and potentially enhance the quality a priori, thus saving time and cost-intensive iterations. Therefore, this knowledge must be front-loaded as a basis to suitably adapt product and process design simultaneously. This contribution therefore proposes a framework for the simultaneous product and process design synthesis of existing designs by providing insight knowledge about product and process design specific to Fused Layer Modeling. This is achieved via a consolidated ontology-based knowledge representation and its integration in a design environment. Thus, the aim, in contrast to most existing approaches, is to provide this knowledge on a product- and application-specific basis to parallelise product and process design for non-assembly mechanisms and exploit currently unused potential concerning the interaction between the different domains. This can shorten development times and also compensate for a lack of background knowledge in certain areas. A practical use case highlights the general procedure, benefits, and applicability.

Keywords:

• Additive manufacturing
• fused layer modeling
• knowledge-based engineering
• product design
• process design
• ontology

1. Introduction and motivation

Notwithstanding the benefits that integrating Additive Manufacturing (AM) into industrial workflows has brought in recent years, like shape complexity and the independence of batch sizes, and its impressively proven maturity and continuing development, there are still some challenges that need to be resolved to efficiently and profitably deploy it (Wohlers et al. 2020). In particular, the Design for Additive Manufacturing (DfAM) compliant product design and the consideration of manufacturing-induced implications remain major hurdles. Creating complex 3D shapes suitable for 3D-printing requires a strong background in DfAM and manufacturing process-specific knowledge to achieve a ready-for-manufacturing design. Thereby, potential issues, like the need for support structures and inaccurate dimensions due to shrinkage and distortion, need to be considered in advance. For AM, the borders between design and manufacturing are blurring since AM is becoming more and more manageable for technologies like Fused Layer Modeling (FLM). Thus, the designer is increasingly taking on the role of the process planner and operator in the manufacturing process. In this context, the designer needs to know the quantitative relationships between product design, process variables, and resulting tolerances. This enables the prediction of effects on the final product quality and facilitates to perform adaptations before manufacturing (Roth et al. 2022). Therefore, design and process knowledge must be adequately linked and the interactions between these two domains must be considered in advance in upstream design activities. However, there is a lack of suitable design and process tools incorporating DfAM and process-specific knowledge simultaneously to support engineers in developing functional, additively manufactured products with the least possible effort (Booth et al. 2017). This joint consideration of product and process design aspects ensures that the final product is not only functional and manufacturable, but also optimised for the available manufacturing technology. This exploits the potential of the product while respecting its limitations. This results in a more efficient, cost-effective and time-saving development process.

In order to overcome the above-mentioned bottlenecks and to link the two domains of product and process design in a more streamlined manner, this paper proposes a holistic knowledge-based framework using the concept of ontologies for product and process design synthesis specific to additively manufactured non-assembly mechanisms (see Figure 1).

Figure 1. Illustration of a non-assembly revolute joint after manufacturing with support structures (left) and in operation (right). Adapted from Schaechtl, Schleich, and Wartzack (2021).

AM, and FLM in particular, enable the manufacturing of these fully functional non-assembly components within a single production step, making previously required assembly steps dispensable (Chen and Zhezheng 2011; Mavroidis et al. 2001; Schaechtl, Schleich, and Wartzack 2021). AM therefore offers opportunities as no trade-offs need to be made to simplify components for assembly. Instead, they can be optimised solely for their intended function as long as process-specific design and process restrictions are met a priori. The integration of data and knowledge from the FLM process enables control and adjustment of geometrical accuracy and achievable tolerances by considering process characteristics and opportunities in advance. This is achieved through an ontology-based knowledge representation. By integrating this knowledge into product and process design and providing it on an application- and product-specific basis, an efficient way to the final, quality-assured product can be enabled. Additionally, this integration and visualisation of specific knowledge aims to raise awareness of the significance and opportunity of these steps in the context of additively manufactured products alongside their holistic product development process.

After discussing the current state of the art and related works and presenting the resulting research questions in Sections 2, Section 3 discusses the systematic development of the above-mentioned framework for the product and process design of AM non-assembly mechanisms. Its exemplary application to an illustrative case study in Section 4 demonstrates the applicability and the benefits. It provides the basis for the subsequent discussion (Section 5). Lastly, a conclusion is given in Section 6.

2. Related work and state of the art

The main aspects of knowledge-based DfAM and the principles of manufacturing process design, including tolerancing aspects for Fused Layer Modeling (FLM), are presented in Sections 2.1 and 2.2 and discussed in Section 2.3.

2.1. Knowledge-based design for additive manufacturing

The DfAM terminology can be differentiated between restrictive, opportunistic, and joint approaches, unifying the preceding two (Kumke 2018; Laverne et al. 2015). Opportunistic approaches aim to exploit the geometrical and material-related opportunities, e.g. free form surfaces, lattice structures, and non-assembly parts provided by AM. In contrast, restrictive approaches usually only consider AM-specific constraints such as the maximum overhang angle and minimum achievable dimensions (Kumke 2018; Laverne et al. 2015).

Formalising DfAM knowledge for these different approaches has proven its general potential to enable computer-aided assistance during the design process. However, a sufficient formalisation of DfAM rules and strategies is missing. Existing design rules are primarily based on experiential data gained through cost-intensive trial-and-error processes (Ko et al. 2021). This formalisation builds the basis for a computer-aided DfAM and describes the process of acquiring and collecting AM-specific design and process knowledge. To achieve a computer- and human-interpretable representation of this knowledge, ontologies can explicitly document the complex relationships in design (Chandrasegaran et al. 2013; Štorga, Andreasen, and Marjanović 2010). An ontology, defined as an ‘explicit specification of a conceptualisation’ (Gruber 1993), can therefore be utilised to capture data, information and knowledge in a generic and formal manner. It facilitates this generic and formal representation to be reused and shared among different applications and stakeholders (Štorga, Andreasen, and Marjanović 2010), e.g. between design and manufacturing divisions. Therefore, it serves as a common cross-domain vocabulary for linking different domains which is not easily possible with multiple-domain specific data bases (Amrouch and Mostefai 2012). To share and reuse knowledge, the concept of ontologies provides schema-level knowledge that can be used to structure and query a knowledge base.

Ontologies have already proven their maturity in providing manufacturing knowledge for specific design tasks (Zirngibl et al. 2022) and DfAM-specific knowledge (Dinar and Rosen 2017; Formentini et al. 2022; Hagedorn, Krishnamurty, and Grosse 2018; Haruna et al. 2024; Kim, Park, and Park 2023; Kim et al. 2019; Mayerhofer et al. 2021; Wang et al. 2024). The current state of the art concerning AM-specific ontologies, described in detail in Kim et al. (2019), provides requirements for their systematic development (Dinar and Rosen 2017) and proposes an ontology for innovative design, but without taking into account particular manufacturing aspects (Hagedorn, Krishnamurty, and Grosse 2018). Additionally, it poses a framework for finding, listing, and collecting DfAM design guidelines within an ontology (Formentini et al. 2022). DfAM ontologies, providing a holistic information structure, can therefore be used to perform manufacturability analysis of AM parts (Kim et al. 2019; Mayerhofer et al. 2021), for example, through applying design rules expressed as Semantic Query-Enhanced Web Rule Language (SQWRL) (Kim et al. 2019). However, these manufacturability analyses are mostly focused on restrictive guidelines by checking the design using simple restrictions like the maximum overhang angle and providing general design solutions.

2.2. Process design and tolerances in additive manufacturing

Even if a theoretically suitable AM design was virtually created and tested for manufacturability, the implications of the subsequent manufacturing process and their impact on the final product must be considered to ensure the desired quality and functionality. Geometrical deviations due to the AM technology's high variability in the specification of process parameters and non-linearity between process parameters and resulting geometrical accuracy are still major challenges in AM (Gibson, Rosen, and Stucker 2015). The investigation of the impact of process parameter settings and perturbance factors (e.g. temperature and material variations) has therefore been studied extensively in the past few years to research and quantify their relationship (Dey and Yodo 2019; Singh et al. 2020). A key finding is that product design in terms of geometry and orientation, and aspects of process design in terms of machine speed and layer height are the most significant technology-independent process parameters affecting build time and overall quality (Alafaghani, Qattawi, and Ablat 2017; Calignano et al. 2017; Saqib and Urbanic 2012). Further studies show that this can be resolved to a large extent by specifying suitable process parameter settings through optimisation (Dey, Hoffman, and Yodo 2020) and applying sophisticated part orientation algorithms (Matos, Rocha, and Pereira 2020) as well as slicing and path planning techniques as integral parts of process design within the pre-processing (Dolenc and Mäkelä 1994; Schaechtl, Schleich, and Wartzack 2022; Wasserfall, Hendrich, and Zhang 2017).

The effect of deviations on the AM parts' geometrical accuracy and quality can generally be specified by tolerances. Thus, tolerancing is part of process design since the correlation between process parameters and settings and the resulting quality in terms of accuracy can be described using tolerances. Thereby, mostly simple relationships are described in the literature. For example, the flatness tolerance is primarily influenced by layer height and build orientation and can be minimised if the build vector resides in the X-Y-plane or if it is parallel to the Z-axis (Budinoff, McMains, and Rinaldi 2018). However, the exact correlation is mostly challenging to quantify, and accurate predictions about achievable manufacturing tolerances are hard to establish for AM (Budinoff and McMains 2018). Both design and tolerancing activities can be better streamlined if the knowledge about the interrelation between process parameters and resulting geometrical accuracy is exploited suitably (Budinoff, McMains, and Rinaldi 2018). This can be achieved, for example, by combining general guidelines as above with quantitative data from experiments and suggestions of favourable parameters and settings to obtain optimum tolerances.

For achieving a process-oriented design, a strong incorporation of manufacturing knowledge from process design into product design is thus inevitable. Since the AM technology is becoming more and more manageable concerning the hardware, the affordability, and the usage, the clear boundary between design and manufacturing increasingly blurs. This allows even inexperienced designers to realise their virtual product directly. Thus, the designer is therefore gradually taking over the role of the process planner and machine operator as well. However, specifying appropriate process parameter settings and applying other pre-processing techniques in process design to ensure overall quality in terms of manufacturability and functionality is a tedious task for designers and requires knowledge and awareness of the technology's potential and implications. In general, approaches and techniques for process synthesis have mostly not been considered in the literature in knowledge-based DfAM approaches for the analysis and assurance of manufacturability and functionality. To harness the full potential and raise awareness for the benefits and applicability of these approaches and techniques in pre-processing, insights about FLM-specific process design must therefore be front-loaded into earlier phases. Previous studies (Schaechtl, Schleich, and Wartzack 2022) have already demonstrated the potential to improve part quality by incorporating insights from process design (design constraints, process variables and hardware limitations) into product and tolerance design by harmonising aspects of pre-processing with part geometry. Finally, the potential of these activities can be fully exploited in the context of product, tolerance, and process design.

2.3. Discussion and research questions

A considerable variety of approaches and methods to ensure and enhance quality in AM in product and process design has been established in the literature and in commercial CAD software tools like Siemens NX and Autodesk Fusion 360 with their respective extensions for AM (Autodesk 2023; Siemens 2023). However, in the current state of the art concerning knowledge-based and -driven DfAM, there are only a few methods and approaches coping with the challenge of linking product and process design knowledge for their simultaneous use in Computer-Aided-Design (CAD) and pre-processing tools. Commercial tools mostly focus on integrating aspects like generative design, topology optimisation and process simulation into the general design workflow for AM (Autodesk 2023; Siemens 2023). But, currently there is a shift in DfAM from process-focused guidelines to sophisticated guidelines incorporating the link between the process and the product design domain, covering their interactions to ensure manufacturability for the certain technologies (Qi et al. 2018). However, their formalisation and utilisation in knowledge-based approaches is currently yet missing. Most commonly, restrictive guidelines are made available for a manufacturability analysis within virtual design tools. Approaches and techniques for further enhancing the quality of the final AM part in process design are thereby neglected. In the recent literature there is a lack of making formalised product and process knowledge for AM simultaneously available in common virtual design environments, further blurring the boundaries between product design and final realisation. Therefore, a streamlining and merging of the individual product and process design activities including tolerancing is required. By formalising product- and process-specific knowledge, an automatic transfer of design information and requirements (in terms of design tolerances) into a precise production plan can be enabled through a tolerance transfer (Hong and Chang 2002; Weill 1988). Thus, a suitable framework, incorporating the above-mentioned shortcomings needs to be developed to enable the consideration of interactions between these activities and their front-loading into the product development process. The development of such a framework, and its integration and visualisation in common design environments, raises designers' awareness of the potential and limitations of the technologies.

This contribution therefore aims to achieve a simultaneous product and process design synthesis of existing virtual product designs through an ontology-based knowledge representation and the integration into product development. The aim is to achieve a design compliant for 3D-printing, specifically for non-assembly mechanisms on a product-specific basis. The scientific challenge and novelty of this contribution lies in considering and incorporating interactions between the product and process design while considering aspects of tolerancing and the final consolidation of these different domains within a common knowledge representation. This knowledge is integrated into a commercial CAD tool via an interface to ensure applicability and enable a semi-automatic product and process design synthesis of virtually existing products. Thus, the aim of this contribution, in contrast to most existing approaches, is to provide this knowledge on a product- and application-specific basis to parallelise individual phases of product development and exploit currently unused potential concerning the interaction between the different domains. This can shorten development times and also compensate for a lack of background knowledge in certain areas. Along the way to achieve the overall goal, the following research questions (RQ) must first be clarified:

• RQ1: Which AM-specific information and knowledge must be front-loaded into the product development process to achieve the overall vision?

• RQ2: How can design- and process-relevant information and knowledge be used for DfAM-compliant product and process design adaptation simultaneously?

• RQ3: How can this information and knowledge be made accessible for an efficient application on a product-specific basis?

3. Product and process design synthesis of additively manufactured mechanisms

The general product development cycle starts with the initial concept, through product design, tolerancing, to process design for the realisation of the physical product (see Figure 2). Within the product design, the intended function of the product is first conceptualised and the nominal dimensions are specified subsequently. Since every manufacturing process cannot be perfect, variations are unavoidable, accompany the whole product development cycle and effect the overall product quality. Therefore, tolerancing, as connection between product and process design, aims to counteract the resulting problems of variations. Tolerancing can be defined as the sum of several activities to manage geometrical variations from the first concept to process design and planning up to inspection to finally evaluate the fulfilment of the initially defined product quality requirements (Dantan et al. 2008). The activities basically involve tolerance specification, allocation, analysis and the final tolerance synthesis (see Figure 2). The next step, the process design, involves all steps required for producing the physical product. For the AM technology, the pre-processing typically includes data transfer, part positioning and nesting, slicing and path planning, the definition of suitable process parameters, and finally the setup of the machine itself. A product identified as non-conforming after inspection and testing therefore means that either individual or all product and process design steps have to be repeated sequentially. Figure 2 illustrates this sequential process.

Figure 2. Sequential procedure of product development from concept to use including product design, tolerance design and process design. Inspired by Schleich (2017) and Roth (2024).

Figure 3. Holistic framework for quality assurance of additively manufactured non-assembly mechanisms.

To accelerate this tedious and sequential process and thus shorten development cycles, this contribution proposes a comprehensive framework for a simultaneous product and process design synthesis while incorporating aspects of tolerancing. The potentials and implications of the FLM technology for all the three mentioned domains are thereby considered simultaneously in advance to compensate for potentially missing background knowledge of designers. For this purpose, a stepwise framework is proposed (see Figure 3), that systematically leads to a ready-for-manufacturing and quality assured non-assembly product. Within this framework, an ontology-based knowledge representation is developed in which, on the one hand, design guidelines and restrictions (referred to as DfAM) are incorporated as a function of process parameter settings. On the other hand, process design guidelines are formalised to improve the overall quality of the final product in terms of manufacturability and functionality by considering tolerancing aspects.

Figure 4. Extraction of product information from existing CAD models for the subsequent analysis of constraints and requirements.

3.1. Analysis of existing product designs

In the first step, the main product information (assembly constraints, geometrical features, general product, and manufacturing information (PMI)) must be extracted from the initial CAD model to derive associated constraints and requirements. This information is subsequently processed and stored, providing the basis for the query within the knowledge representation for retrieving design and process guidelines specific to the present product. Figure 4 illustrates this process with the case study of a knuckle joint. Thereby, the product information concerning Design features, PMI and Constraints are automatically extracted directly from the CAD model using software scripts. It exemplarily shows the assembly constraint ‘Cylindrical’ and additionally several design-relevant features like a ‘Hole with diameter 25 [mm]’, ‘Threads’, and dimensional as well as geometrical tolerances. ‘GDT’ thereby stands for Geometric Dimensioning and Tolerancing (GD&T). According to the specific representation for the CAD tool Siemens NX, used in this case, the ‘&9’ represents a parallelism tolerance. The following sections will show how this basic information will be further processed and used to query and retrieve specific guidelines from the ontology-based knowledge representation.

3.2. Ontology-based knowledge representation for product and process design synthesis

Developing a specific ontology-based knowledge representation requires high initial effort. However, both a consistency check through reasoning and inference of new knowledge within the ontology are facilitated, which usually amortises this one-time invest. This is especially useful when ontologies from different domains, such as product design, process design and tolerancing are merged as for this contribution. New knowledge can thus be inferred by suitably linking the different domains and additionally provides a common vocabulary. Another reason for choosing the concept of ontologies over relational databases, for example, is the possibility of easily adding new knowledge and information due to the open world assumption which is especially useful for a rapidly evolving technology like AM. The development of the ontology-based knowledge representation for both DfAM-compliant FLM products and the associated process design bases on the three main requirements according to Dinar and Rosen (2017). First, it should incorporate and represent domain, experiential, and experimental knowledge. Second, it should facilitate reasoning linked with descriptive logic (Dinar and Rosen 2017). Third, it should be a basis for the integration within a CAD tool – or at least it should be easily implemented into it (Dinar and Rosen 2017).

For the systematic development of the ontology-based knowledge representation for product and process design synthesis, the steps according to Noy and McGuiness (2001) were followed. Figure 5 illustrates the overall procedure schematically.

Figure 5. Systematic procedure for developing the ontology-based knowledge representation for product and process design specific to the FLM technology. Steps for ontology development process according to Noy and McGuiness (2001).

Following Step 1, the domain of the aspired combined ontology-based knowledge representation involves the product and process design of FLM-manufactured non-assembly products considering aspects of tolerancing. The specific benefit of the ontology is aimed to achieve a ready-for-manufacturing and quality-assured non-assembly mechanism. The targeted user is the designer. To represent the domains and the intended purpose of the developed ontology, following so-called competency questions (CQs) can be formulated:

• What guidelines for product and process design exist for the specific product?

• Which parameter settings influence the current design?

• Which parameter settings can lead to an improvement of the final product?

• What measures can be taken to optimise the overall quality of the product?

These CQs represent examples to illustrate the overall scope and purpose of the ontology-based knowledge representation. For developing the ontology and deriving answers to these CQs, specific knowledge and data in the context of DfAM, relevant to the FLM technology and non-assembly mechanisms, need to be systematically gathered. Thereby, existing ontologies in the mentioned domains were considered (Step 2). Table 1 includes the listing and categorisation of existing ontologies in the mentioned domains for possible reuse.

Table 1. Listing and categorisation of existing ontologies for possible reuse.

Domain	Ontologies	Considered
DfAM product design	Schaechtl et al. (2023)	Yes
	Kim et al. (2019)	Yes
	Mayerhofer et al. (2021)	Yes
	Ko et al. (2021)	Yes
	Dinar and Rosen (2017)	Yes
	Hagedorn, Krishnamurty, and Grosse (2018)	No
	Sanfilippo, Belkadi, and Bernard (2019)	No
	Kim, Park, and Park (2023)	No
	Wang et al. (2024)	No
	Haruna et al. (2024)	No
Tolerance design	Goetz et al. (2021)	Yes
	Zhong et al. (2014)	Yes
	Zhong et al. (2013)	No
	Qie et al. (2017)	No
	Qin et al. (2018)	No
Process design	Kim, Park, and Park (2023)	No

Different knowledge sources were examined for incorporating domain, experiential and experimental knowledge (Dinar and Rosen 2017) in the manufacturing-specific ontology. These sources included implicit knowledge of experts in this field (derived through personal conversations, analysing blog posts and videos from the Maker community), standards (e.g. ISO/ASTM 52910:2018), case studies from state-of-the-art literature (Cuellar et al. 2018; Klahn, Singer, and Meboldt 2016; Lussenburg, Sakes, and Breedveld 2021; Schulz, Schlattmann, and Rosenthal 2017) and own preliminary experiments (Hallmann et al. 2018; Schaechtl, Schleich, and Wartzack 2022). Based on the findings and after standardisation in a common terminology, important terms for the structure of the ontologies can be enumerated (Step 3). This subsequently allowed classes to be defined within a hierarchy (Step 4). The explicit class hierarchies of the ontologies are shown in the respective sections (Sections 3.2 and 3.3). By scanning the different knowledge sources, important verbs are identified as slots, and are subsequently assigned to matching classes (Step 5 & 6). The following example from literature is intended to illustrate the procedure outlined above (see Figure 5).

‘Printing orientation affects number of parameters of the printed item, these include dimensional accuracy, surface finish quality, strength, build time, support structure, and eventually cost’ (Sossou et al. 2018).

By analysing this quote, the terms in bold can be defined as classes and the term in italic can be defined as slot. Additionally, a class hierarchy can be built since some parameters can be grouped as quality criteria. This example is only intended to show the general procedure from Figure 5 described above following the guideline for ontology development (Noy and McGuiness 2001). The example is illustrated in Figure 6.

Figure 6. Exemplary ontology with classes (yellow diamonds), class hierarchy and slots (blue arrows).

The resulting ontologies thus serve as basis to formalise design and process rules and guidelines in ontology-typical SWRL-based (Semantic Web Rule Language) rules. SWRL rules (equivalent to IF-THEN expressions) are composed of a conjunction of predicates (subjects, objects, and properties including variable names from the ontology) located in the antecedent, leading to the consequent (Horrocks et al. 2023). Additionally to build up these rules from scratch, existing SWRL-based DfAM rules from literature (Dinar and Rosen 2017; Sanfilippo, Belkadi, and Bernard 2019) can be reused after modification. Due to their dependence of the precisely defined entities in the ontology, SWRL rules reused from existing ontologies need to be rewritten to match the wording within the new ontology.

These gathered information will form the basis for developing the knowledge representation following the Steps 3–7 from Noy and McGuiness (2001) (see Figure 5). The knowledge representation will be subsequently integrated and utilised in the CAD environment for the product and process design synthesis. The next sections will provide specific examples for developing the ontology-based knowledge representation for both product design (Section 3.2.1) and process design (Section 3.2.2) specific to the FLM manufacturing technology and non-assembly mechanisms.

3.2.1. Knowledge representation for FLM product design synthesis

The ontology-based knowledge representation for product design synthesis in this contribution is especially devoted to the design of non-assembly mechanisms. Thereby, specific challenges must be considered when designing a manufacturable and fully functional product. These challenges and associated proposed design solutions are presented in tabular form, according to Lussenburg, Sakes, and Breedveld (2021).

Considering these DfAM solutions from Table 2 and an existing DfAM ontology from own prior work (Schaechtl et al. 2023), in combination with reusing information of existing ontologies from the literature (Kim et al. 2019; Ko et al. 2021; Mayerhofer et al. 2021), the ontology-based knowledge representation for the product design is developed. Therefore, domain-specific knowledge gathered from the sources mentioned in Figure 5 are considered. These sources thus serve for determining classes and defining properties and relationships for the ontology. An excerpt of this is shown in Figure 7.

Figure 7. Extract of the class hierarchy of the FLM product design ontology specific to non-assembly mechanisms.

Table 2. Challenges, general solutions, and design solutions for designing non-assembly mechanisms according to Lussenburg, Sakes, and Breedveld (2021).

Challenge	General solution	DfAM design solution
Support structures	Prevent support structures	Reduce overhang angle
		Decrease clearance
		Change main hinge shape
	Allow support removal	Increase clearance
		Increase spacing between bodies
		Add features such as chamfers
		Use open geometry
		Add drainage/release holes
Surface finish	Reduce surface contact	Add protrusions/markers and dents
	Prevent ‘staircase’ effect	Reduce overhang angle
		Use angular geometric shapes in favour or organic
	Prevent supports	Keep interacting surfaces free from supports
	Consider the build direction	Change the build direction to suit the design
Clearance	Adjust geometry	Change main hinge shape
		Add markers and dents
		Scale entire mechanism up
Strength	Consider the build direction	Change the build direction to suit the design
	Adjust geometry	Optimise cross-section for the direction of movement

The DfAM ontology specific to the FLM technology contains restrictive design guidelines to ensure or improve the general manufacturability (e.g. ‘reduce overhang angle’). Additionally, it incorporates opportunistic guidelines for harnessing the full potential of AM (e.g. ‘use angular geometric or organic shapes in favour’). As an example, it contains a specific guideline for DfAM-compliant design of non-assembly revolute joints:

‘If the part contains revolute joints, the axial displacement should be ensured by inclined surfaces at an angle θ as well as a minimum distance [joint clearance] of $j_{\mathrm{cl}}$ should be fixed and the contact area should be as small as possible to avoid support structures and enhance functionality’ (Schulz, Schlattmann, and Rosenthal 2017).

For the angle θ, a threshold value of ${45}^{\circ }$ can be chosen since it is a general restrictive guideline for most AM technologies, thus omitting the need for support structures. Following this premise, the parameters of this triangular design feature for ensuring axial displacement of the joint can be fixed with: (1) $h=r=\frac{d}{2},$(1) where:

• h  height of the triangular cross-section,

• r  radius of the triangular cross-section,

• d  diameter of the triangular cross-section.

Figure 8 illustrates the connection between the parameters of the design feature and the opportunistic design guideline for this example. The design adaptation enables the revolute joints compliance for non-assembly 3D-printing. Thus, it requires no support structures, and no final post-processing steps depending on the rotation type.

Figure 8. Illustration of the realisation of opportunistic DfAM design guidelines for non-assembly revolute joints.

These guidelines are formalised in standardised SWRL rules for integration into the ontology-based DfAM knowledge representation. Creating SWRL rules requires a careful consideration of different aspects to ensure that the intended knowledge is represented adequately so the rules can be used efficiently. For example, the monotonic behaviour of SWRL rules precludes subsequent editing and makes changes irreversible (Horrocks et al. 2023). Below, an exemplary SWRL rule concerning the revolute joint is presented as a qualitative design recommendation with quantitative values for design adaptation. The basis for this is the general structure of the ontology with its classes, objects and data properties, and relationships (see Figure 7).

SWRL rule:

Part(?p) $\wedge$ hasConstraintName(?p, ‘Cylindrical’$\wedge \wedge$xsd:string) $\wedge$ revoluteJoint(?p) $\to$ hasGuideline(?p, ‘Joint design has inclined walls to prevent axial displacement’$\wedge \wedge$xsd:string) $\wedge$ hasGuideline(?p, ‘Contact area should be small’$\wedge \wedge$xsd:string) $\wedge$ hasGuideline(?p, ‘Inclined walls have a downskin angle of max. ${45}^{\circ }$’$\wedge \wedge$xsd:string)

The above SWRL rule states that if the part has the assembly constraint ‘Cylindrical’, it is a revolute joint. Consequently, two DfAM rules are associated with this type of mechanism. First, the joint design should have inclined walls to prevent axial displacement (with an angle of maximum ${45}^{\circ }$). Second, the contact area should be small to avoid support structures (see Figure 8).

By applying these guidelines, following the general and design solutions from Table 2, the main challenges for non-assembly mechanisms (clearance and support structures) can be resolved. This example is an excerpt from the final knowledge representation. Further examples can be found in Table A1, Appendix.

3.2.2. Knowledge representation for FLM process design synthesis

For further mitigating the challenge of manufacturing non-assembly mechanisms, pre-processing has gained increased attention as it offers possibilities to enhance the quality before printing, thus reducing post-processing effort (e.g. the removal of support structures). Therefore, aspects of pre-processing, e.g. the choice of suitable process parameters and settings and slicing and path planning algorithms can be front-loaded and considered in advance. To achieve this, a process design ontology for the FLM technology is proposed. It primarily comprises a qualitative as well as quantitative representation of the relationship between parameters (process parameter, machine parameter, material parameter, and slicing and path planning algorithms), quality criteria, tolerances, fault causes, material and geometrical features (overhangs, wall thickness, boreholes, etc.) and the benefits and applicability of further pre-processing techniques. The development of the process design ontology followed the identical approach described in the previous section (see Section 3.2), taking into account the requirements for developing the DfAM according to Dinar and Rosen (2017) and the general procedure from Noy and McGuiness (2001). Figure 9 shows an extract of the class hierarchy of the FLM process design ontology. Thereby, the relationship between the FLM process, including specific parameters and fault causes, and tolerances, are illustrated.

Figure 9. Extract of the class hierarchy of the FLM process design ontology.

In the following, two illustrative examples will show the formalisation of guidelines and recommendations in the pre-processing for process design synthesis using SWRL rules for the above-mentioned specific scenarios (see Section 3.2.1) and their subsequent integration into the ontology. These two different examples are intended to show that the application of methods within process design can, on the one hand, take into account requirements from tolerance design (Example 1) and, on the other hand, specifically mitigate manufacturing-related problems originating from product design (Example 2).

In the first example, the cylindricity tolerance and, thus, the clearance of moving parts can be optimised by applying efficient tool path planning in the pre-processing settings. This can be achieved by printing the external perimeter of the cylindrical object (outer diameter and inner diameter) first. In this scenario, the dimensions can be better ensured as the tolerance zone is not violated through the extra material from repressing through the already solidified layers. By complying with the specified tolerance, the pre-defined clearance to ensure the functionality of the non-assembly mechanism can thus be guaranteed. The illustration of this example and the associated SWRL rule are shown below.

Figure 10. Illustration of the pre-processing setting ‘External perimeter first’ for optimising cylindricity tolerance.

Figure 11. Illustration of the pre-processing setting ‘Detect bridging perimeters’ to omit the need for support structures.

SWRL rule example 1:

Part(?p) $\wedge$ Cylinder(?cyl) $\wedge$ hasFeature(?p, ?cyl) $\wedge$ Cylindricity(?cyl_tol) $\wedge$ hasTolerance(?cyl, ?cyl_tol) $\to$ hasGuideline(?p, ‘external_perimeters_first= 1’$\wedge \wedge$xsd:string)

The second example illustrates an approach to overcome the challenge of the need for support structures for overhangs that exceed a certain threshold if design actions (e.g. including chamfers or self-supporting overhangs) are no longer feasible. Therefore, a FLM-specific pre-processing setting called ‘Bridging’ can be enabled. Thereby, specific process parameters, such as the printing velocity and the cooling through the layer fan, are adjusted depending on the used material, causing the deposited layer to solidify mid-air. Thus, the need for support structures and subsequent post-processing is omitted. Figure 11 illustrates this example and the associated SWRL rule.

SWRL rule example 2:

Part(?p) $\wedge$ unsupportedOverhang(?uOv) $\wedge$ hasFeatureName(?p, ?uOv) $\wedge$ hasMaxValue(?uOv, ?len_uOv) $\wedge$ swrlb:greaterThan(?len_uOv, ‘2.5’$\wedge \wedge$xsd:decimal) len_uOv $\wedge$ swrlb:lessThan(?len_uOv, ‘60’$\wedge \wedge$xsd:decimal) $\to$ hasGuideline(?p, ‘detect_ bridging_perimeters’$\wedge \wedge$xsd:string)

These examples are an excerpt from the final knowledge representation. Further examples can be found in Table A1, Appendix.

3.3. Combination of product and process design knowledge representation

Since the ontology-based knowledge representation for FLM product and process design were developed individually for reasons of reusability (Noy and McGuiness 2001), the main objective is now their combination (Klein 2001). Through this combination, their complementarity can be exploited to obtain unified knowledge about the process from design to manufacturing (Babalou and König-Ries 2020). Therefore, cross-domain interactions for simultaneous application and provision within product and process design can be considered. To achieve the combination of existing ontologies, different concepts are available. Thereby, a distinction can be drawn between the concepts of Aligning, Mapping and Merging. Alignment is ‘a set of matches between two (or more) ontologies in the same domain or in related domains’ (Amrouch and Mostefai 2012). The actual process to align two ontologies is called Mapping (Amrouch and Mostefai 2012). In contrast, Merging is the ‘creation of a new ontology from two, possibly overlapping, input ontologies’ (Euzenat and Shvaiko 2013). The resulting ontology contains the knowledge of both ontologies whereby the original ontologies remain the same. The state-of-the-art literature provides various approaches and tools for automatising this process. Within this contribution, the actual merging of the ontologies is directly performed within the software tool Protégé (Musen 2015) where the knowledge representations were initially built up using the built-in function ‘Refactor $\to$ Merge Ontologies’. Merging the ontologies in Protégé includes some prerequisites. The ontologies to be merged need the same ontology IRIs (Internationalised Resource Identifiers). Further prerequisites involve resolving class and property mappings by overlapping or equivalent classes and properties. The first prerequisite is already automatically fulfilled here, as both ontologies have been built within the same environment. The second prerequisite requires additional effort and expertise and is a more time consuming task since entities are copied and now result in duplicates within the consolidated knowledge representation. The merged knowledge representation can be automatically checked for consistency, coherency and non-redundancy (McGuinness et al. 2000) within the ontology using mechanisms of reasoning. In this way, overlaps between both knowledge representations can be identified. The challenge lies in mapping semantically close concepts through equivalence and subsumption relations. By investigating the automatically identified overlaps, the duplicates can be resolved by merging them to maintain the connected entities. This task requires experience and background knowledge in the respective domains but finally ensures a consistent, coherent and non-redundant consolidated knowledge representation of both domains. Although various approaches and tools for automatising this process exist in literature, within this contribution this task is performed manually. This can be justified by the fact that both ontologies to be merged were developed by the authors themselves and thus enough experience and background knowledge is given to perform this task more efficiently than to implement and apply sophisticated mapping and merging tools.

Figure 12. Exemplary excerpt of the combined ontology with classes (yellow diamonds) and slots (blue arrows).

Figure 12 shows an exemplary excerpt for merging and mapping the DfAM ontology and the FLM process ontology, including process design and tolerancing and their interactions.

Finally, this combined knowledge representation can be used for SWRL-based formalisation of guidelines in the interaction of product and process design. In this process, the rules already existing in the respective ontologies are examined for overlap and, if necessary, supplemented with necessary information from the respective other representation. Therefore, again, expert knowledge is required to adapt or rewrite the SWRL-based rules. The guideline for the opportunistic design adaptation of revolute joints in Section 3.2.1 can be used as an example for showing the benefits of the merged knowledge representation. Using the DfAM ontology, the existing product is identified as a revoluteJoint through the constraint type Cylindrical and, therefore the guidelines for design adaptation are proposed (see Section 3.2.1). Currently, the information regarding the achievable clearance (in this case denoted as $j_{\mathrm{cl}}$) is missing for achieving the final product design. The combination shows that the value of this design parameter is limited (lim) by process parameters like layer height and machine-specific parameters like nozzle diameter and extrusion width. This information can thereby be derived from the FLM process design ontology. By combining both knowledge representations, the guideline for adapting revolute joints to make them compliant for 3D-printing can be enriched for the specific manufacturing technology and machine as precise data from prior experiments, included in the overall guidelines, can be extracted. This joint information from product and process design can finally be provided to the designer to assist in finding a suitable trade-off between product design and process design decisions. Furthermore, the combination additionally enables a tolerance transfer (Hong and Chang 2002; Weill 1988) as process parameters and settings (extPerimeterFirst) can be derived directly from initially defined PMI like the cylindricity. Therefore, this information can be automatically transferred from design to manufacturing stage, without further user interaction.

Through this combination of different domains, an enhanced ready-for-manufacturing design for a specific product and manufacturing environment can be finally achieved. The above-mentioned example illustrates the general approach, which is applicable and transferable for further scenarios as well (see Table A1 in Appendix).

4. Implementation and application

This section first describes the prototypical implementation of the basic framework from Figure 3 (see Section 3) in Section 4.1. Subsequently, this prototypical implementation is exemplarily applied to non-assembly mechanisms to illustrate the general procedure and its applicability in Section 4.2. The overall goal is to support the systematic product and process design synthesis for making an existing product design compliant for 3D-printing as a non-assembly mechanism. Finally, the overall framework is assessed based on the evaluation of the resulting manufactured case study concerning manufacturability, functionality and overall quality in Section 4.3.

4.1. Prototypical implementation

A prototypical implementation is proposed to test the framework for applicability and highlight the benefits and limitations. The general software architecture of the prototypical implementation and the procedure for product and process design synthesis of non-assembly mechanisms is shown in Figure 13.

Figure 13. Software architecture and procedure of the proposed prototypical implementation for product and process design synthesis of non-assembly mechanisms.

For the prototypical implementation, an interface was developed in Python (Python version 3.10). This interface is mainly responsible for the processing and transmission of information, the communication between the ontology-based knowledge representation and the CAD environment, and the connection to the pre-processing tool. For the communication between CAD and ontology, the python-specific library owlready2 (Lamy 2017) was used, which allows to access, edit, and add to ontologies easily. Additionally, it facilitates the link to other applications, like in this case, the CAD tool, finally achieving interoperability.

The procedure of the workflow starts with the automatic extraction of product-specific information like Features, Constraints and PMI (see Section 3.1) ($\text{192}$). Thereby, the retrieved keywords (stored in .txt files) are used for the query within the ontology-based knowledge representation via the interface $\left(\text{193}\right)$. The knowledge representation was built using the software tool Protégé 5.5.0 (Musen 2015) using SPARQL 1.1 (SPARQL 1.1 Query Language 2023) as query language. Subsequently, these queried guidelines, including both qualitative and quantitative information (cf. examples from Section 3.2) are processed through the interface and integrated into the CAD environment and visualised $\left(\text{194}\right)$ using Siemens NX 2212 and NXOpen as an API (Application Programming Interface). The illustration and visualisation of the product and process design guidelines thus serve as proposals and recommendations for the designer for manual and semi-automatic design adaptation through user interaction and input $\left(\text{195}\right)$. This framework explicitly uses Siemens NX, but can be adapted to other CAD programs, if they offer an adequate programming interface. As a last step, the final product design is exported as an input for the pre-processing tool for slicing and path planning via a file format specific for 3D-printing (*.3mf – 3D Manufacturing Format) (3MF Consortium 2023) $\left(\text{196}\right)$. The relevant process information, extracted from the knowledge representation in Protégé as strings, is (automatically) annotated to the configuration files $\left(\text{197}\right)$. These configuration files (short: config), ‘config.ini’, are files used to define parameters and initial settings for computer programs. In this case, the file contains all relevant information for slicing, path planning and parameter settings, like layer height, material, and printer settings. Both files can subsequently be loaded into the pre-processing tool (in this case, PrusaSlicer 2.6.0) to generate the machine code (*.gcode) and finally manufacture the part $\left(\text{198}\right)$. A case study will illustrate the proposed workflow in detail in the following section.

4.2. Case study

The case study is a conventionally designed, manufactured and assembled 2D robot gripper for simple pick-and-place tasks (see Figure 14 top). It is composed of 24 parts in total (parts needed for mounting the coupling and the stepper motor are not counted). The objective is to make it compliant for 3D-printing as a non-assembly mechanism for its direct usage after manufacturing without post-processing steps (no need for removing support structures) and assembly. Thus, several AM benefits like part consolidation, individualisation and lightweight design can be exploited. For making the product compliant for FLM, certain trade-offs will have to be considered in the design regarding the following aspects, proposed in Jansen, Doubrovski, and Verlinden (2014), in line with the initial design:

• minimal joint clearance for accurate movement vs. sufficient joint clearance to facilitate manufacturability (prevent fusion of parts and provide accessibility for the removal of support structures if necessary)

• small dimensions vs. space for movement of numerous mechanical parts

• minimised mechanical friction vs. inferior surface quality and high friction due to AM characteristics (e.g. staircase effect, material properties)

• minimal need for support structures to prevent failure and minimise post-processing effort

Figure 14 illustrates the application of the proposed framework specific to the use case using the software architecture and following the stepwise procedure from Figure 13. The initial design concept for the robot gripper is illustrated in Figure 14 top.

Following the procedure, specific keywords are retrieved from the product-specific information (Features, Constraints and PMI, see Figure 13) stored in the automatically extracted text files from the CAD model using Python scripts $\left(\text{192}\right)$. Within this use case, the exemplary keywords are the constraint name ‘Cylindrical’ and the PMI ‘&4’ (cylindricity, see Section 3.1). These are used in the next step $\left(\text{193}\right)$ to query the consolidated ontology-based knowledge representation in Protégé using Python scripts. As a result, SWRL-based DfAM guidelines specific to revolute joints (product design) and process design recommendations specific to cylindricity tolerances are retrieved.

For product design synthesis, the SWRL-based DfAM guidelines including quantitative data for the specification of the minimal achievable joint clearance ($j_{\mathrm{cl}}$) can subsequently be visualised within the API in the CAD environment to support the designer $\left(\text{194}\right)$. Through design adaptation via user input, the initial design can finally be adapted to be compliant for 3D-printing as non-assembly mechanisms $\left(\text{195}\right)$. By realising this proposal, the above-mentioned trade-offs regarding the design of non-assembly mechanisms like need for support structures and minimal joint clearance are considered due to the support-structure-free design and the definition of an ideal $j_{\mathrm{cl}}$ value. This specific guideline and SWRL rule is explained in detail in Section 3.2.1 and Figure 8. Finally, a ready-for-manufacturing product design is achieved and can be exported in the AM-specific ‘.3mf’ format $\left(\text{196}\right)$.

For product design synthesis, the SWRL-based process design rules ‘External perimeter first’ (Example 1, Figure 10) and ‘Bridging’ (Example 2, Figure 11) from Section 3.2.2 are exemplarily retrieved from the consolidated ontology-based knowledge representation based on the extracted keywords. These guidelines can be visualised within the API of the CAD to assist the designer and be automatically written to the ‘config.ini’ file used for manufacturing (see Section 4.1). Figure 14 $\left(\text{197}\right)$ shows this file with the highlighted changes to enable ‘Bridging’ and ‘External perimeter first’. Thus, through this product design synthesis, process-related shortcomings like shape deviations, potentially affecting the minimal joint clearance and restrictions like unsupported overhangs and the need for support structure can be resolved without further adapting the design.

Figure 14. Product and process design synthesis of the case study.

Finally, both files (‘config.ini’ and ‘*.3mf’) can be directly used for manufacturing the final product (see Figure 14 bottom). This use case shows the general working principle for extracting guidelines and recommendations and their subsequent application for product and process design synthesis exemplarily for certain guidelines. Further SWRL-based guidelines, used for the product and process synthesis of this case study, can be found in Table A1 in Appendix. It therefore demonstrates the feasibility of integrating the consolidated knowledge representation in CAD tools and providing these information on a product-specific basis theoretically.

4.3. Evaluation of manufacturability and functionality

The manufacturability, the final functionality and thus the quality of the product are practically evaluated based on the initially defined requirements (see Figure 13 and Section 4.2) to demonstrate the applicability of the framework. Figure 15 shows the final gripper right after printing, with the motor mounted, and in operation. Through its print-first-time-right manufacturing, the overall manufacturability of the product after its product and process design synthesis, by applying the proposed framework, is verified.

Figure 15. Final 2D robot gripper right after printing and in operation.

The application of the knowledge-based product design synthesis saved 23 parts in total as a result of part consolidation (parts needed for mounting the coupling and the stepper motor are not counted), comparing the initial design (24 parts) with the adapted design (1 part). During this process, a significant weight reduction of over $78\mathrm{\%}$ was accomplished as well. This statement is based on the analysis of the weight of the initially virtually designed gripper (approx. 404 grams) derived from the CAD model. The bolts were assumed to be standardised steel components and the linkages of the gripper were assumed to be made out of polymer (Acrylonitrile butadiene styrene, ABS). The components necessary for attaching the motor and the motor itself were neglected in this weight analysis. According to the calculation of the pre-processing tool, the adjusted design of the robot gripper was approx. 90 grams for an infill density of $50\mathrm{\%}$. The material used for 3D-printing was ABS and thermoplastic polyurethane (TPU) for the compliant gripper fingers, resulting in an estimated total price of approx. $8.80\text{€}$ (calculated by the pre-processing tool). Through the process design synthesis by adapting the printing settings in the config file, a print-first-time-right, non-assembly product was achieved, whereby any post-processing steps were avoided. Finally, only the stepper motor, including the coupling, has to be mounted. The SWRL-based guidelines and recommendations for the product and process synthesis of this case study can be found in Table A1 in Appendix.

5. Discussion

In summary, the initially defined overall goal (see Section 2.3) was accomplished through the proposed framework in combination with the obtained results from its application to a case study. Thus, by applying the implemented proposed framework while harnessing an ontology-based knowledge representation based on formalised guidelines in combination with experimental data and the subsequent integration into product development and realisation, a simultaneous product and process design synthesis specific to an existing product can be accomplished. During this process, the additionally defined aspects were also addressed as follows. Process parameters, like the layer height and orientation as well as FLM-specific settings have a decisive impact on the design and final quality and functionality and also can enhance them, if applied adequately. Therefore, knowledge about these restrictions and opportunities has to be front-loaded to achieve an optimal product (RQ1). The representation of this knowledge in an ontology-based manner has proven the potential to merge different domains like product and process design as qualitative and quantitative information can be combined in SWRL rules (RQ2). The integration and visualisation within design environments (CAD) and pre-processing tools for FLM technology therefore allow an efficient application of these product-specific guidelines for product and process design synthesis of existing products for manufacturing as non-assembly mechanisms (RQ3).

As demonstrated through the use case, the framework is capable of achieving a ready-for-production design for non-assembly mechanisms, taking into account process-specific information in advance and incorporating them into a virtual design environment. Thereby, a certain degree of automation of the design adaptation is achievable for specific and mostly restrictive guidelines. Hence, the presented framework provides high potential for reasonable assistance for novice and inexperienced designers without an in-depth background in AM or DfAM to rethink their existing design and eventually redesign it for manufacturing as a non-assembly mechanism. Highlighting the relationship between product and process design through integrating illustrations of product-specific guidelines and their formalisation in human-interpretable SWRL-based rules within the CAD environment also raises further awareness of the benefits and restrictions in this area. In addition, the framework offers the possibility for its simple extension for a company-specific use beyond robot gripper. The ontology-based knowledge representation and merging of the different domains also enables the elicitation of new correlations through the use of inference mechanisms and, thus, the continuous evolution within this area.

The limitations of this framework lie in its degree of automation. Since the presented guidelines are mostly opportunistic, their application may drastically change the product's appearance and behaviour. Since this is not always desirable, a fully automatic design adaptation is only advisable to a certain limit to keep the designer in charge of the product and is therefore neglected within this contribution. Consequently, it may be advisable to speak of knowledge– and data-informed product and process synthesis rather than knowledge–driven since the overall goal is no fully automated design adaptation. However, the design and process decisions need to be inspired and subsequently verified through this knowledge-informed approach. Another shortcoming lies in the comparatively high effort of the initial development of this consolidated ontology-based knowledge representation. But, this is only a one-time investment, which will be amortised through its continuous enhancement and application and can certainly be automated as its development progresses.

6. Conclusion and outlook

The high shape complexity, as well as the possibilities in process control, hold the potential to restructure traditional product development through Additive Manufacturing. However, this requires mechanisms to incorporate the associated advantages and, most importantly, the associated challenges appropriately into the newly conceived workflow. This article, therefore, focuses on the quality assurance of additively manufactured non-assembly mechanisms through a knowledge-based product and process design synthesis. Integrating specific knowledge into the development procedure thus helps to achieve the vision of print-first-time-right. It also creates increased awareness of the restrictions and opportunities of the AM technology for less experienced designers. Adapting the initial design, combined with the front-loaded consideration of aspects of process design and its impact on the product quality, will ensure its overall manufacturability while exploiting the high potentials of AM. A case study of a FLM manufactured 2D non-assembly robot gripper demonstrated the workflow, and thereby proved the general applicability of the framework. Despite its benefits, the limitations and restrictions mentioned in the previous chapter (high initial effort to develop knowledge base and degree of automation) offer opportunities for further enhancement of the proposed framework.

Further research activities within this framework will focus on enhancing the proposed methods' applicability and the integration of approaches for quantitatively assessing the resulting quality and functionality of the final product employing tolerance analyses. The influence of the operating behaviour will also be investigated and taken into account for making a reliable statement about the long-term applicability of the presented use case. Furthermore, the degree of automation, taking into account the responsibility of the designer for decisive design decisions, as well as the maintenance and the continuous expansion of the ontology-based knowledge representation could be improved. Therefore, considering further approaches for automating this process like integrating techniques of Data Mining, and Machine Learning, e.g. Natural Language Processing and Large Language Models, would be helpful. In order to test the applicability, transferability, and the overall benefits of the presented framework to other products and in the general product development process, a user testing will be conducted following this work.

Even though the presented framework was developed and applied for FLM exemplarily, its application is also possible for other AM technologies, like metal-based laser powder bed fusion (L-PBF). For this purpose, the general approach can be maintained, only the process and product design guidelines specific for the respective AM technology have to be identified and integrated into the ontology after their formalisation. For this purpose, the methods proposed in Section 3.2 (cf. Figure 5) can be used as a guide. In addition, ontologies have the advantage of being easier to adapt to another AM technology than hard-coded rules. Within this framework, they can even be exchanged for L-PBF-specific ontologies with little rework.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

The data supporting this study's findings are available from the corresponding author, Paul Schaechtl, upon request.

Additional information

Funding

The authors thank the German Research Foundation (DFG) for supporting the research project ‘Pro${}^2$AMech: Computer-aided PROduct and PROcess design of Additively manufactured MECHanisms’ under the grant numbers WA 2913/27-2 and SCHL 2233/4-2.

Appendix

Table A1. Overview of the product and process design guidelines used for the illustrated case study used within this contribution in SWRL syntax.

SWRL rule	Note
Part(?p) $\wedge$ hasConstraintName(?p, ‘Cylindrical’$\wedge \wedge$xsd:string) $\wedge$ revoluteJoint(?p) $\to$ hasGuideline(?p, ‘Joint design has inclined walls to prevent axial displacement’$\wedge \wedge$xsd:string) $\wedge$ hasGuideline(?p, ‘Contact area should be small’$\wedge \wedge$xsd:string) $\wedge$ hasGuideline(?p, ‘Inclined walls have a downskin angle of max. ${45}^{\circ }$’$\wedge \wedge$xsd:string)
Part(?p) $\wedge$ hasConstraintName(?p, ‘Cylindrical’$\wedge \wedge$xsd:string) $\wedge$ revoluteJoint(?p) $\to$ hasGuideline(?p, ‘requires clearance to prevent bonding’$\wedge \wedge$xsd:string)
Part(?p) $\wedge$ Cylinder(?cyl) $\wedge$ hasFeature(?p, ?cyl) $\wedge$ Cylindricity(?cyl_tol) $\wedge$ hasTolerance(?cyl, ?cyl_tol) $\to$ hasGuideline(?p, ‘external_perimeters_first=1’$\wedge \wedge$xsd:string)
Part(?p) $\wedge$ unsupportedOverhang(?uOv) $\wedge$ hasFeatureName(?p, ?uOv) $\wedge$ hasMaxValue(?uOv, ?len_uOv) $\wedge$ swrlb:greaterThan(?len_uOv, ‘2.5’$\wedge \wedge$xsd:decimal) len_uOv $\wedge$ swrlb:lessThan(?len_uOv, ‘60’$\wedge \wedge$xsd:decimal) $\to$ hasGuideline(?p, ‘enable Bridging’$\wedge \wedge$xsd:string)
Part(?p) $\wedge$ Thread(?th) $\wedge$ hasFeature(?p, ?th) $\to$ hasGuideline(?p, ‘Threads should be aligned with printing direction’$\wedge \wedge$xsd:string)	Adapted from Schulz, Schlattmann, and Rosenthal (2017) and Schaechtl et al. (2023)
Part(?p) $\wedge$ hasConstraintName(?p, ‘Ball’$\wedge \wedge$xsd:string) $\wedge$ ballJoint(?p) $\to$ hasGuideline(?p, ‘Ball socket and ball shank should be placed perpendicular to build plate’$\wedge \wedge$xsd:string) $\wedge$ hasGuideline(?p, ‘Adapt geometry ball shank’$\wedge \wedge$xsd:string)	Adapted from Schulz, Schlattmann, and Rosenthal (2017) and Schaechtl et al. (2023)
Part(?p) $\wedge$ hasConstraintName(?p, ‘Ball’$\wedge \wedge$xsd:string) $\wedge$ ballJoint(?p) $\wedge$ hasOrientation(?p, ‘XY’$\wedge \wedge$xsd:string) $\to$ hasGuideline(?p, ‘Adapt geometry ball’$\wedge \wedge$xsd:string)
Part(?p) $\wedge$ thinFeature(?thF) $\wedge$ hasFeature(?p, ?thF) $\wedge$ tallFeature(?tF) $\wedge$ hasFeature(?p, ?tF) $\wedge$ LargeSurface(?lS) $\wedge$ hasFeature(?p, ?lS) $\to$ hasGuideline(?lS, ‘Orient planar on build plate’$\wedge \wedge$xsd:string)
Part(?part) $\wedge$ inclinedSurface(?incS) $\wedge$ Angularity(?angTol) $\to$ hasGuideline(?part, ‘apply Non-planar Printing’$\wedge \wedge$xsd:string)	Adapted from Schaechtl, Schleich, and Wartzack (2022)
Wall(?wall) $\wedge$ wallThickness(?wt) $\wedge$ parametrizes(?wt, ?wall) $\wedge$ hasValue(?wt, ?th) $\wedge$ lessThan(?nozzleDia, 0.4) $\to$ NotManufacturable(?wall)
Part(?p) $\wedge$ hasFeature(?part, ?edge) $\wedge$ Edge(?edge) $\to$ hasGuideline(?part, ‘Blend Edge’$\wedge \wedge$xsd:string) $\wedge$ greaterThan(?edge, 1.0)
LargeSurface(?lS) $\wedge$ Part(?p) $\wedge$ hasFeature(?p, ?lS) $\to$ hasGuideline(?lS, ‘Specify as primary datum’$\wedge \wedge$xsd:string)	Adapted from Vakouftsis et al. (2020)