Function-Based Design Principles for Additive Manufacturing

Abstract

The development of additive manufacturing (AM) technologies has brought new design possibilities, and to utilise those possibilities, new sources of AM design knowledge are needed. This paper presents an inductive methodology for extracting AM design knowledge based on the functional analysis of AM products. Extracted AM design knowledge is formalised in 32 AM design principles using the proposed methodology. The AM design principles are organised regarding the functions they solve. Initial validation and intended use are described through a case study. The AM design principles can facilitate systematic design processes and methods, and can be used in early design phases for finding partial solutions for subfunctions of the design problem.

1. Introduction

Additive manufacturing (AM) has brought unique design possibilities and design freedom, manifested in new functionalities and new shapes for our products [1]. Designers can see these possibilities in exciting-looking examples of AM products (e.g., satellite antenna bracket [2] or flow-measurement probe [3]). However, designers, especially those not familiar with AM, struggle to replicate such designs and create products that utilise AM possibilities. To help designers and to support the AM-oriented design process, a new design approach called Design for Additive Manufacturing (DfAM) and focusing on the utilisation of AM capabilities has emerged [4]. While today, numerous approaches exist within the DfAM [5], the majority are focused on later design phases used for optimising design features, and manufacturing process parameters that have little influence on the layout and functionality of the product. To truly utilise all the possibilities AM can offer, it is essential to apply DfAM and AM knowledge (AMK) early in the design process. The importance of early design phases arises from the design activities of establishing the functional and working structure of the product and defining product layout. These activities directly and significantly influence the function and form of the product [6]. Designers working with AM recognise the need to think additively from the beginning of the design process, but feel the lack of DfAM design support and sources of AMK for the early design phases [7].

New DfAM methods and sources of AMK are being developed to support the designers in the early design phases, and different approaches have been proposed in the literature. Probably the first approach developed for the early design phases of DfAM is the design feature database [8], which provides AMK in the form of design features categorised according to the reasons for applying AM in the creation of concepts. Perez et al. [9] proposed the use of crowdsourced design principles (DPs). This approach was further developed into DP Design Cards as a toolset for supporting innovation with AM [10,11]. The DPs are based on the analysis of products made mainly with the fused deposition modelling (FDM) process, and provide broad coverage of AMK, from suggesting design solutions in early design phases to improvements for detail design and the design process itself.

DPs were also proposed in the form of a design catalogue, where AMK is categorised according to functionalities of products that DPs can solve [12]. The approach utilises design knowledge in the form of principle solutions often used to conceptualise products in prescribed systematic design processes [6,13]. The systematisation of AMK according to function was also used in the development of AM DPs based on the multi-material capability of AM [14,15]. However, this approach is limited in scope, as it only focuses on one specific aspect of AM capabilities. None of the three approaches are publicly available for a thorough review.

Another notable example of AMK for early design phases is the AM design heuristics [16,17]. Heuristics provide a process-independent source of AMK for the given design context and enable the utilisation of unique AM possibilities in the conceptualisation of AM products. The final list of 25 design heuristics contains a broad AMK suitable for early design phases and is validated through user studies [18,19]. However, due to their generalisation and lack of relation to product functions, heuristics are not an appropriate source of AMK for function-based design approaches.

While different sources of AMK for early design phases exist in the literature, they vary in the maturity of approaches and scope of AMK they provide. None of the reviewed approaches provides, at the same time, a broad process-independent source of AMK and focus on function-based design. The reason the research is focused on the function-based approach is to facilitate the use of systematic design due to the technical and economic benefits it provides [6,20,21,22,23,24]. Systematic design steers the design process, and among other benefits, increases the chance of finding quality solutions, promotes collaborations, supports management of design activities, helps the automation of design activities, enables reuse of previous design solutions, and can aid designers’ creativity. In addition, functions, as abstract and independent forms of product representation [25,26], enable cross-phase and cross-domain comparison of design, and facilitate the use of different design methods and tools.

Furthermore, to store AMK for early design phases, DPs as a form of knowledge explication are adopted. DP is defined as “A fundamental rule or law, derived inductively from extensive experience and/or empirical evidence, which provides design process guidance to increase the chance of reaching a successful solution” [27]. Compared to design heuristics, DPs are less context-dependent and are based on empirical evidence, rather than tacit knowledge. Meanwhile, in comparison with design guidelines, DPs provide more specific instructions, rather than context-dependent directives, and thus are a suitable form for storing AMK in a function-based approach.

This research aims to establish a methodology for extracting DPs for AM, based on the functional analysis of existing products. The methodology utilises the literature recommendations for deriving and defining the DPs [27] and applies them to the domain of AM. Furthermore, this research aims to define a set of DPs derived through the established methodology, and systemise them according to related functions to facilitate their use in the early design phases of the systematic design process. The research builds on the previous proposal of the authors on the conception of an AM DP repository [28] and is the next step in establishing a function-based source of AMK.

The paper presents the developed methodology for deriving DPs in Section 2. Section 3 describes the formulation of DPs, presents the list of 32 derived function-based AM DPs, and categorises them according to the function they are solving. The Section 4 case study describes the intended use of DPs and provides an initial verification of the approach. Section 5 discusses the research outputs and compares them with similar approaches. Finally, the outline of future work and conclusion are presented in Section 6.

2. Methodology

A product is a record of design knowledge. During product development, designers must find an optimal solution to solve the design problem and fulfil the requirements. In doing so, they must use their own design knowledge gathered through experience, intuition, observation of best practices, trial-and-error approach, or some other research and development method. This accumulated knowledge is incorporated into the designed product. This design knowledge can be extracted through a systematic examination of products [29]. Furthermore, because each product has a set of functions, the relations between form and functions can be observed and captured through form-to-function mapping. The form-to-function mapping can be reversed and used in new product development for function-to-form mapping approaches and searches for design solutions for solving product functions [26] The premise used in the methodology in this study was that examples of “good” AM products incorporate the best practice and innovative forms based on AM potentials used for solving the functions of a product. Similar premises were used previously for extracting design knowledge in the work of Singh et al. [30], Yilmaz & Seifert [31] and Blösch-Paidosh & Shea [16]. Therefore, for the extraction of DPs in this study, an inductive approach was used based on the observation and analysis of existing AM products and their key features. The research was focused only on the analysis of products, and did not include other sources of design knowledge, such as experience or observation of designers, because the products were an unbiased source of design knowledge and were the most used source of evidence for extracting design knowledge [27].

The inductive process used for deriving DPs is made of three key phases: collecting the data, analysing the data to extract the patterns, and forming the theory based on the identified patterns [27]. There are multiple examples of inductive approaches being used for deriving various DPs. For example, “Transformation Principles” [30] and “Tolerance design principles” [32] are derived using induction, and according to Fu et al. [27], induction is the most frequently used approach for deriving DPs and other knowledge explications. The inductive approach has also been previously used for deriving AM design knowledge, for instance, in the work of Blösch-Paidosh & Shea [16] or Perez et al. [9]. The methodology used in this study is shown in Figure 1. It consisted of 8 steps divided into three phases. The phases and steps are described below through an explanation of one conducted analysis. While described linearly, the process was highly iterative, requiring numerous comparisons between products, features, and function structures.

The first phase of the methodology consisted of two steps. Firstly, the criteria for selecting AM products to be used for the analysis had to be defined. While analysis of any AM product can reveal a certain amount of knowledge, to have a focused analysis, the emphasis should be placed on products that clearly incorporate unique AM potentials. Thus, to be included in the pool for analysis, the product need to have features that were solely possible with AM, or it needed to gain additional value from AM. Secondly, as one of the research goals was to provide support for design practitioners, to be included in the pool, the product had to be manufacturable on the current commercially available AM equipment. This criterion was used because some AM features are possible only on highly specialised lab equipment. It might be years before they could be made on commercial AM equipment; thus, for this study, such products were excluded from the analysis. Finally, enough data about the product needed to be available to conduct the necessary analysis (e.g., pictures, decryption, physical product, or CAD model).

Using the described criteria, a pool of AM products was created. Products were gathered from three sources: (i) literature on DfAM, (ii) commercial products and demo products of major AM equipment manufacturers, and (iii) from crowdsourced platforms. Literature sources are a valuable source of AM knowledge, as they usually clearly state how and why a product is designed the way it is, and the benefits of using AM. Commercial and demonstration products provide an overview of state-of-the-art AM capabilities. In contrast, crowdsourced repositories of AM designs (e.g., Thingiverse (https://www.thingiverse.com/ (accessed on 5 February 2022)) or Thangs (https://thangs.com/ (accessed on 5 February 2022)) provide many designs that are often made through a trial-and-error approach, thus containing empirical knowledge about AM. The gathered products came from several different domains, with most products being from Aerospace, Mechanical Engineering and Medical domain, which corresponds to industries where AM is mostly used [1]. However, the single biggest portion of the products came from the domain of household products, because such products are common on the crowdsourced repositories that are the biggest source in this study, due to the availability of a great number of products. The initial pool consisted of fifteen products and was incrementally increased until asymptotic convergence was achieved [27]. In total, forty products were analysed, and their distribution across sources and domains is shown in Figure 2.

The next phase of the methodology was the data analysis. In this phase, the used approach differed from other approaches in literature, as it utilised the function model of a product for facilitating the analysis and extraction of DPs [33]. The function model describes the overall function of a system through its subfunctions in a solution neutral way [25]. The function model is represented through function structure, a graphical layout of product subfunctions connected with energy, material, and signal flows that represent the overall product function [6,21]. As a solution-neutral representation of the product, function structure facilitates the comparison of products, but also a comparison of forms and solutions used to solve the product subfunctions needed for extracting DPs.

The first step of the data analysis was the creation of a function structure for each product in the pool. To have a formal function representation, rather than using natural language to define functions and flows, a standardised vocabulary called “Functional Basis” was adopted to achieve repeatable and meaningful results from function structures [34]. However, because the Functional Basis has three hierarchy levels, only the second level was used, as it provided an optimum between abstraction and detailed description of product functions [35]. The predefined vocabulary was combined with the physics-based reasoning approach for modelling flows in a function structure [36]. When the function structure of a product was created, in the second step of data analysis, it was observed how individual product functions were solved. Here, key features and AM solutions were observed and compared with the created function structure. The focus of observations was to understand how the functions of a product are solved using AM. The AM form used was mapped onto product functions and extracted.

To illustrate the procedure described above, the example of analysis conducted on the AM milireactor is explained and shown in Figure 3. The AM milireactor is a piece of laboratory equipment used to synthesise liquid chemicals [37]. The purpose of the milireactor is to quickly synthesise two liquid chemicals by inducing the turbulent flow in small channels using internal chambers and barriers. Furthermore, as it is used for experimental synthesis, it must provide a means for visual observation of the synthesis. The product analysis started with the creation of the function structure (top of Figure 3). The function structure consisted of ten functions operating on the flows of Liquid, Chemical Energy (CE), and Status (visual information). In the second step of data analysis, it was observed how functions of a product (bottom of Figure 3) were solved using AM solutions. Four observations were documented. The functions Import Liquid and Export Liquid were solved by integrating the threaded channel opening (marked red). The function Guide Liquid was solved using winding channels integrated into the body of milireactor (marked green). The function Mix Liquid and Liquid was solved with internal chambers and geometry that increased the turbulent flow and enabled better mixing of liquids (marked purple). Finally, one of the requirements of the milireactor is to allow visual observation of chemical reaction and mixing of liquids, thus, a set of functions related to this requirement (Sense Liquid, Indicate Status, Export Status) were solved using semi-transparent material and material distribution to allow visual observation through the entire length of the milireactor channel (marked orange).

The exact process was applied to other products from the pool of AM products. When multiple products were analysed, different AM solutions used for solving product functions were extracted. In the third step of data analysis, the extracted solutions were grouped into patterns for consolidation through iterative analysis. In the final phase of the proposed methodology, the grouped observations are formalised into DPs using the predefined syntax and categorised (described in Section 3). For the presented example, the red observation was combined with other similar solutions and generalised into principle #DP6 (Enable interaction with environment by integrating standard geometry), the green observation became #DP9 (Enhance fluid performance by using integrated internal channels), the purple became #DP10 (Enhance material/energy conversion by shaping internal chamber for the use case), and the orange became #DP27 (Convey information and/or change permutability of light by applying custom material distribution).

After conducting the above-described analysis, the number of analysed AM products was compared with the number of newly derived DPs in chronological order (Figure 4). For the initial set of 15 products, 23 DPs were extracted (green dashed line). Observation revealed the convergence of DPs derived with each new analysed product, suggesting the finite set of DPs could be derived. The convergence analysis, where a set of data is observed until the number of extracted observations starts to converge to a horizontal asymptote, is often used to determine the size of the data set in inductive approaches for knowledge extractions [27]. Therefore, the pool of AM products was incrementally increased, and the analysis of AM solutions was repeated until the number of derived DPs started to asymptote toward a finite number (blue dotted line). The convergence occurred after the 25th analysed product, in which 31 DPs were found (purple dashed line). Only one more DP was extracted in an additional analysis of 10 products. Hence, the process was stopped after 40 products were analysed and the number of DPs converged.

3. Design Principles for Additive Manufacturing

The final phase of the methodology was the formalisation of the DPs. It consisted of the two-step formulation of DPs and the organisation of DPs. To support designers, DPs must clearly articulate the intended meaning of the AM design knowledge they contain. Therefore, the DPs needed to be carefully formalised. The two main formats for expressing DPs are the descriptive and prescriptive formats. The descriptive format is stated in grammatical declarative form. The descriptive format informs about a concept, fact, or knowledge that can be applied to a given context [27]. On the other hand, the prescriptive format is stated in imperative grammatical form, through which an action is prescribed. Thus, the prescriptive format provides clear instructions for designers on what action must be conducted to apply the design knowledge contained in the DP. Up to now, both formats have been used in the formalisation of AMK. For example, Lauff et al. [10] and Hwang et al. [11] used prescriptive format with formal syntax to express their AM DPs, while Blösch-Paidosh & Shea [16] use the descriptive format in their formalisation of AMK through design heuristics.

As the intention of DPs in this research is to provide a means for solving product functions, a prescriptive format expressed in imperative grammatical form was used. A syntax proposed by Lauff et al. [10] was adopted with smaller modifications to better reflect the function-oriented nature of the derived DPs. The syntax firstly states a design problem or requirement, followed by the conjunctive “by” and a generalised action to address the problem. The syntax can be stated as: (solve) DESIGN PROBLEM by USING “X” AM CHARACTERISTIC. In total, 32 DPs were derived and shown in

Table 1. List of derived DPs for AM.

#	Design Principles for AM
#DP1	Fit user by using custom ergonomic geometry AM enables manufacturing of complex and curved geometry. Furthermore, each product manufactured with AM can have different geometry. Therefore, the geometry in interaction with the user can be easily customised for an individual user or different groups of users to provide an optimum ergonomic.
#DP2	Absorb energy by using lattice structures AM enables easy manufacturing of lattice structures, on multiple levels of hierarchy, through the entire geometry or only in part of the geometry. The lattice structure can absorb energy through elastic or plastic movement, deformation, or breakage.
#DP3	Absorb energy by using elastic material AM can process a variety of materials with different material properties. Create structures using a material with adequate elasticity to absorb energy.
#DP4	Enhance interaction with environment by customising surface texture With AM, it is relatively easy to manufacture different patterns and textures embedded in the surface of the product. Create customised textures to ensure good grip and interaction with the environment.
#DP5	Enhance interaction with environment by using customised surface features With AM, it relatively easy to manufacture small 3D features on the surface of the products. Create custom features to provide an adequate interaction with the environment.
#DP6	Enable interaction with environment by integrating standard geometry AM can be used to create simple shapes, replicate existing geometry, and integrate multiple features in a single design. Utilise this possibility of AM and integrate a standard geometry (e.g., threads, bores) into a design to enable interaction with the environment and connection with other products.
#DP7	Enable interaction with environment by using custom geometry to fit the use case AM enables easy customisation of geometry. Customise geometry for the particular use case to enable interaction with the environment and direct fit with the geometry of other components.
#DP8	Enhance fluid performance by using customised surfaces AM can manufacture complex shapes. Use this possibility of AM when interacting with fluids to create surfaces whose shape will be in accordance with fluid dynamics to increase the overall performance.
#DP9	Enhance fluid performance by using integrated internal channels AM enables the creation of complex internal geometry. Use the custom internal channels whose shape will be in accordance with fluid dynamics for guiding and distributing fluids to increase the overall performance.
#DP10	Enhance material/energy conversion by shaping internal chamber for the use case AM enables the creation of complex internal geometry. Create custom internal chambers with geometry that will be adjusted for a particular physical process happening inside the chamber, such as energy or material conversion.
#DP11	Enhance energy magnitude by shaping internal chamber for the use case AM enables the creation of complex internal geometry. Create custom internal chambers with geometry that will help change the magnitude of the energy (e.g., increase/decrease pressure, change acoustic or kinetic energy).
#DP12	Conduct mechanical energy and forces by applying lattice structures AM enables easy manufacturing of lattice structures, on multiple levels of hierarchy, through the entire geometry or only in part of the geometry. Use the lattice structures to conduct mechanical energy through the product and create a lightweight but stiff product.
#DP13	Conduct mechanical energy and forces by applying topologically optimised geometry AM enables manufacturing of complex geometries. Use this possibility to create topologically optimised geometry to create optimised designs with reduced mass and/or increased performance that will conduct mechanical energy or forces.
#DP14	Conduct mechanical energy and forces by using void structures AM enables manufacturing of complex geometries. Use this possibility to create void structures to conduct mechanical energy or forces with reduced weight.
#DP15	Conduct energy by using material with appropriate properties AM processes can utilise a variety of different materials. Use appropriate material for the use case.
#DP16	Interact with object or material by customising geometry for the use case AM enables easy customisation of geometry. Customise geometry for the particular use case to enable interaction with the object of interest to ensure an adequate fit.
#DP17	Import/export fluid by applying appropriate openings With AM, it is easy to create custom openings on the surface. Customise the openings according to fluid dynamics to increase the overall performance during import or export of fluids from the system.
#DP18	Conduct energy by embracing anisotropy and layered structure One of the characteristics of most AM processes is the anisotropic properties in a final product. Utilise the product orientation a layer like structure to increase the strength of the part in the direction of energy flow.
#DP19	Enable movement of the system by using compliant mechanism Compliant mechanisms enable desired motions through relative flexibility of the mechanism shape. The AM capability of manufacturing complex shapes with varying wall thickness provides a means to create custom 2D and 3D single-body compliant mechanisms.
#DP20	Achieve degree of freedom/desired behaviour by applying custom material distribution AM enables manufacturing of shapes with varying wall thicknesses. Use the material distribution to achieve movement and desired behaviour with the relative movement of a single body part.
#DP21	Solve function with non-AM component by using integrated attachment point/geometry Sometimes AM cannot solve the product function, or the function can be solved more easily or cheaply with a non-AM component. Use the customisability of AM to create an attachment point or geometry for easy integration of non-AM components into a product.
#DP22	Detect and indicate temperature change by using thermal sensitive material AM can process a variety of materials, including thermal sensitive polymers. Use thermal sensitive polymer to indicate a change in temperature to the user without the need for a dedicated sensor.
#DP23	Connect another part or flexible end of a part by using integrated attachment point/geometry When a multi-part design or part opening is needed to solve the function, integrate the attachments into a product to ease the assembly and disassembly without the need for additional fastening elements.
#DP24	Allow pass through of fluid by using lattice structures AM enables easy manufacturing of lattice structures on multiple levels of hierarchy, through the entire geometry or only in part of the geometry. Use the lattice structures to allow the pass through of fluid over a larger surface area.
#DP25	Convey information with colour by using multicolour AM Some AM processes are capable of building multi-material structures. Use the multi-material capability to embed and convey information through colour.
#DP26	Convey information by customising geometry AM enables easy manufacturing of complex geometry on multiple levels of hierarchy. Use custom geometry to convey information to the user by embedding it directly to the part.
#DP27	Convey information and/or change permutability of light by applying custom material distribution In AM, there is no need for uniform wall thickness. Use material distribution to control permutability of light to embed the information that will be conveyed to the user when the part will be in front of the light source.
#DP28	Conduct light by using transparent material AM can process a variety of materials, including transparent and semi-transparent materials. Use transparent materials to conduct and distribute light customised for the use case.
#DP29	Store energy by using material elasticity With AM, it is possible to create custom structures that will act as a spring due to their shape and elasticity. Use such structures to store mechanical energy.
#DP30	Provide movement by using single build assemblies AM can build entire assemblies in a single build without the need for additional assembly operations. Use these capabilities to connect multiple parts and enable movement of the product.
#DP31	Change motion or force by using single build mechanisms AM can build entire mechanisms in a single build without the need for additional assembly operations. Use these capabilities to create entire mechanisms in a single build that will control the motion or change the force/energy in a predetermined manner.
#DP32	Achieve desired behaviour by using multi-material AM Some AM processes are capable of building multi-material structures. Use the multi-material capability to design different properties in different areas of the part to achieve desired behaviour (e.g., flexible area vs. stiff area to achieve controlled flexibility of the part)

together with the short description for each DP.

Since the purpose of derived DPs is to support designers in solving product functions, for easier search and implementation, DPs are organised in the “Function-Flow-Design Principle” (FFDP) matrix (Figure 5). This type of systematisation was previously used for similar purposes. An example can be found in the work of Schumacher et al. [14]. The FFDP matrix lists functions and flows in the left column and top rows, while other cells are populated with DPs that can solve associated function-flow relations. While in the analysis and derivation of the DPs, the secondary level of Functional Basis was used, the matrix also contains a primary level. This has two purposes. Firstly, it enables easier searching, and secondly, when there is no solution, or the solution is not satisfying, the designer can look for solutions on a higher level of abstractions. For example, instead of looking for a solution for Import Solid, the designer can look for solutions in all the cells under the higher-level function and flow Channel Material. This expands the pool of suggested solutions by including DPs for similar purposes, such as for Import Liquid or Export Solid.

4. Case Study

To verify the validity of using DPs for the conceptualisation of AM products, a case study was conducted [38]. The case study was a redesign of a gear to be manufactured with AM. As the purpose of the case study was to verify the use of DPs in conceptual design, the only design requirements were that the gear must be made with AM, it must have a lightweight structure, and it must be cooled with a liquid coolant. Other design requirements, such as operating conditions, life span and strength, were ignored in the conceptual design, as they were part of the later design phase.

Firstly, the function structure of the gear with a function of cooling the gear with liquid was developed (Figure 6). The function structure consisted of 14 functions connected with material flows of liquid and solid, and energy flows of mechanical energy (ME) and thermal energy (TE). Next, DPs that could be used as solutions were found using the FFDP matrix and mapped onto the function structure for each function in the function structure.

Three different concepts were created by combining multiple combinations of DPs [39] (Figure 7). Due to the design requirements, all three concepts utilised #DP6 (standard geometry) for functions related to interaction with the shaft and the second gear, as this part of that geometry needed to remain as it was in the original design. Hence functions related to guiding mechanical energy and cooling of the gear were solved using various AM DPs.

Concept A uses a lattice structure (#DP12) to guide ME from shaft to gear teeth. The lattice structure is also used to increase the surface and allow the passthrough of coolant (#DP24). The coolant is scooped using custom blades (#DP7). Concept B utilises topologically optimised geometry (#DP13) for guiding ME. The coolant is pumped through the shaft using openings on the inner surface of the bore (#DP17). The coolant is guided through internal channels (#DP9) to cool the gear. Concept C guides ME using void structures (#DP14). It also incorporates internal channels (#DP9) to cool the gear, as in concept B, but the coolant is scooped using custom blades (#DP7).

The variety of suggested DPs for a simple product such as a gear clearly shows the benefit of using function-based DPs in a systematic design process, as multiple concepts are created that utilise the unique possibilities of AM. For further evaluation of the DPs and their role in the design process, concept C was chosen using the Pugh matrix and a variety of selection criteria [23] for further development. Due to the need for brevity, the description of the embodiment design and detail design are omitted. Once the final 3D model was created, concept C was manufactured using the FDM (Fused Deposition Modelling) process in polymer and using the DMLS (Direct Metal Laser Sintering) process in stainless steel (Figure 8). The metal gear also included markings, thus embodying additional principle #DP26. The conceptual design and embodiment of gear provide an initial validation of DPs. The fabrication of gear in two different materials and using two different AM processes show the universality and process independence of derived DPs.

5. Discussion

The methodology for deriving the AM DPs described in Section 2 enabled a systematic analysis of existing AM products to extract and formalise AMK. While similar methodologies were used previously (e.g., [16,30,31]), the novelty of the proposed methodology is the utilisation of functional analysis to facilitate the process of extracting the design knowledge. The functional analysis was based on the creation of function structures for each product included in the analysis. The use of function structures enabled a rigorous analysis and assisted the observations of relations between the product functions and AM forms used for solving these functions, as shown in the example of AM milireactor analysis. The focus on the functions of AM solutions used in a product rather than their embodiment and dimensions aided in extracting AMK for the early design phases of the design process.

Furthermore, the common representation of products through predefined vocabulary facilitated the systematic approach to analysis and comparison of AM solutions. It aided in reviewing and documenting the analysis process. Furthermore, while the proposed methodology focuses on the analysis of AM products and extractions of the AMK, it is general enough to be used, with necessary modifications, for other purposes. By changing the inclusion criteria for establishing the data pool, the methodology can be applied to a broader set of AM products or to the different domains of the design area to develop new sources of function-oriented design knowledge.

The proposed methodology was applied to the pool of 40 AM products, from which 32 DPs were derived. DPs are the repository of AMK to be used in early design phases, especially in function-based systematic design methods. During the formalisation phase, a predefined syntax was used, with the aim of achieving balance between the generalisation of DPs and their specificity regarding particular functions. Because some AM features and DPs can solve multiple different functions, the syntax avoided mentioning functions in DP definitions. Formulating DPs to include mention of a particular function would lead to the extensive expansion of the number of DPs, which would be exhausting to comprehend with little additional value to the stored AMK. Similarly, lack of reference to specific materials and AM technologies gives universality to the derived DPs and emphasises their focus on early design phases.

This logic can be seen in the group of principles referring to the functionality of interaction with the environment (#DP4, #DP5, #DP6, #DP7). If the expression “Enhance interaction with environment” were replaced with a particular function, in this case with a total of 15 different functions (e.g., Import Gas, Import Liquid, Secure Solid, etc.), instead of four DPs, the list would have contained 34 DPs (the number of appearances of the four DPs in the FFDP matrix) without providing additional benefit, as these four DPs were already categorised in appropriate cells in the FFDP matrix referring to necessary functions. On the other hand, referring to AM capabilities through only a generalised description could hinder some subtle AM capabilities. The AM feature mentioned in #DP7 (use custom geometry to fit the use case) is a broad characteristic of AM, and features referred to in #DP4, #DP5, and #DP6 could be considered sub-features of custom geometry. However, by referring to surface texture (microscale geometry), surface features (macroscale geometry), and standard geometry that could be incorporated directly into a part, a customisable geometry characteristic of AM is explained with the specificity for a particular use case. For example, #DP7 is applicable for a broad set of functions (found in 13 cells in FFDP matrix), while #DP4 and #DP4 are more specific and refer to only four functions, thus providing specific information for these use cases. The same logic is applied in the definition of the rest of DPs as well. These subtle differences could be emphasised with additional modalities of AM DP representation, such as examples, pictures, or 3D models [28], but this is outside of the scope of this paper.

The derived DPs are categorised in the FFDP matrix, where one DP could be found in multiple cells for every function it solves. Due to one of the characteristics of AM products (a great number of functions related to flows of material and mechanical energy), for some functions and flows (Channel Mechanical Energy and Channel Material) there are a number of suggested DPs, while for others, there are none. Hence, the use of the FFDP matrix in its current form is limited to the design of products whose main functions operate on the flows of material or mechanical energy. This can be primarily attributed to the criteria of product selection. As we focused on the current state of the AM, most products had common AM features, such as lattice structures, informal cooling, and internal geometry, or topologically optimised geometry. Likewise, the focus of the analysis was on the AM parts, rather than entire assemblies that may include parts not manufactured with AM, such as sensors or electric motors; thus, there is a notable lack of electrical energy and control signals often found in electromechanical products.

If the design process as prescribed by Otto & Wood [21] is observed, the DPs can be used for the conceptual design of AM products. After designers create the function structure of a product, based on previously defined design requirements, the FFDP matrix can be used for finding the DPs that could be used as partial solutions for product subfunctions. The broadness of DPs and the multiple suggestions from the FFDP matrix enable the exploration of the AM-enabled design space and conceptualisation of the design. By combining different DPs, a designer can create one or more concepts of a product, as shown in the case study above. The DPs are a helpful tool for concept generation, but are not suitable for embodiment and detail design phases and should be complemented with specific design rules and guidelines appropriate for a design case and AM technology. In the presented case study, DPs enabled the conceptualisation of structures for conducting mechanical energy and cooling in a gear, showing their suitability for that phase of the design process. However, later design phases, where calculations regarding strength, heat exchange, and necessary dimensions are needed, require different AMK that is not part of this study.

The derived DPs are similar to the exiting crowdsourced AM DPs of Perez et al. [9,11] and AM design heuristics [16,18]. However, there are notable differences in three areas: the formulation, intended use, and scope of the AMK they contain. The comparison was conducted to verify and validate our approach. The authors acknowledge the proposal of other similar approaches, notably multi-material AM DPs, which have similar systematisation [14]. However, they are not publicly disclosed, so a proper comparison is not possible. When the formulation of the three knowledge explications was compared, the derived function-based AM DPs were similar to crowdsourced DPs, as a similar syntax is used. Hence, both sets of DPs provide clear instructions for the designer on which action needs to be performed in a particular use case. The design heuristics are presented through more generalised descriptive formulation and do not provide specific guidance for solving the product functions, but rather provide a description of directions the designer could take to explore the AM design space.

If the intended use is observed, the derived DPs are similar to AM design heuristics, as both are focused on the early design phases. Crowdsourced DPs, due to the broader AMK they contain, can be used across the design process. The notable difference is in the way these sources of AMK are used. Neither heuristics nor crowdsourced DPs provide a means for systematic application of AMK, but rather designers must go through each heuristic or DP and evaluate if they can be applied in their context. On the other hand, the presented DPs provide a systematic search and application of AMK through the FFDP matrix. When a designer is looking at how to solve a particular function, the FFDP matrix narrows down the entire list of DPs and provides only DPs applicable for the case.

Finally, the three approaches involve different scopes of AMK. The crowdsourced DPs provide the broadest AMK, which, besides design solutions, provide guidance for conducting and improving the design process and guidance for detail design to ensure the printability of created designs. On the other hand, AM design heuristics only contain knowledge about conceiving and improving design solutions, and in that view, the derived DPs are very similar to heuristics. However, they contain only design knowledge that can be related to functions, and do not refer to design requirements (e.g., recyclability, weight, or aesthetics) in the manner of heuristics.

The comparison of DPs with similar AMK sources provides an initial validation of the derived DPs. However, this research has two main limitations that must be discussed: the scope of the gathered AMK and the verification and validation. DPs were derived through an inductive approach and the analysis of existing AM products. The analysis was carried out until the number of derived DPs converged to a finite number. In our case, this occurred after analysing 40 products. According to Fu et al. [27], inductive approaches based on the analysis of products and asymptotic convergence have a sample size that goes from as low as 3 products up to 190 products. The conducted analysis falls into that scope and is similar to Hwang et al. [11], who reported 80 observations needed for deriving 90% of their crowdsourced AM DPs. These similarities and asymptotic convergence provide confidence that the derived set of function-based AM DPs covers a broad scope of AMK. However, it must be emphasised this is not a closed set of DPs, as the analysis included a few specific AM domains. If new products and AM domains are included in the analysis, new DPs could emerge. Furthermore, as DPs cover the current state of AM, with further development of AM technologies, but also with the new and creative application of AM to solve product functions, it is probable that new DPs will emerge.

Another limitation of the study is the verification and validation of the DPs. If the four-step procedure for validation of early-phase DfAM design supports is followed [40] the presented approach is only validated through two steps. Firstly, a case study needed for verifying the feasibility of the DPs was conducted. It showed the DPs were a suitable tool for searching for AM-based solutions through the function structure of a product, and enabled the creation of concepts that utilised the unique design possibilities of AM. Secondly, the DPs were compared with other similar validated DfAM approaches. The observed similarities provide an initial validation of the DPs. However, for complete validation, controlled validation and workshop validation are needed. They are planned for future research, but at the moment, they are outside of the scope of this paper.

6. Conclusions & Future Work

The work presented in this paper addresses the growing need for design support for the early phases of AM-oriented design processes [7]. The literature review identified the lack of function-orientated design supports and sources of AMK for conducting design activities of conceptual design. The two main contributions of this paper are the methodology for extraction of AM design knowledge and the initial list of AM design principles (DPs). The proposed methodology for deriving AM DPs is based on existing literature recommendations [27]. The uniqueness and novelty of the inductive methodology is the use of functional analysis of existing AM products to extract AMK. The functional analysis was based on the use of function structures, created through reverse engineering, to extract and formulate DPs. The derived DPs were carefully formulated in the prescriptive format using imperative grammatical form [10,27,41]. The second contribution is the list of 32 DPs with a short description of each DP. The DPs are organised in the “Function-Flow-Design Principle” (FFDP) matrix, which enables the search for and application of DPs according to the function they solve. The conducted case study showed the applicability of DPs for creating concepts using a systematic function-based approach. The paper discusses the validity of derived DPs and compares them with other existing DfAM approaches for early design phases. The research will continue by increasing the number of analysed AM products, and conducting further work on the modality of DP representation and the establishment of an online repository of AM DPs, as envisioned in previous work [28].