Investigating the impact of additive manufacturing data exchange standards for re-distributed manufacturing

The ISO/ASTM 52910:2018 Additive manufacturing—design—requirements, guidelines and recommendations document categorises AM processes into (1) Binder Jetting, in which a liquid bonding agent is selectively deposited to join powder materials; (2) Directed Energy Deposition, in which focused thermal energy is used to fuse materials by melting as they are being deposited; (3) Material Extrusion, in which material is dispensed through a moving nozzle or orifice; (4) Material Jetting, in which droplets of build material are selectively deposited; (5) Powder Bed Fusion, in which thermal energy selectively fuses regions of a powder bed; (6) Sheet Lamination, in which sheets of material are bonded to form an object; and (7) Vat Photopolymerization, in which liquid photopolymer in a vat is selectively cured by light-activated polymerization. Compared to traditional manufacturing processes, such as milling, grinding, turning, EDM, etc., AM technology defines a process of adding the print material according to the layer-upon-layer accumulation.

However, it makes AM technology has a number of advantages [6]. It allows almost complete 3D control over artefact geometries and material properties, thus dramatically increasing the range of feasible products made without the need for tooling in directly print process. In many cases, it also produces less waste material. Manufacture of interlocking parts using AM eliminates the need for additional assembly processes and because it does not require specialist tooling for each part, AM also enables local, specialised and on-demand manufacturing. This means that parts can be produced at short notice, close to the location of consumption, and based on exact customer specifications. Since some raw materials are more easily procured in feeder or powder form, AM also simplifies supply chains by layering the required material [7]. As a result of these advantages, AM offers opportunities to extend conventional manufacturing methods as the integration of hybrid manufacturing, thus enabling the development of newer products, innovative business and information models, and flexible digital chains. For these reasons, AM has been referred to as the next industrial revolution (reference needed) although many practical barriers still exist. Commercial and industrial adoption of AM is challenged in several areas, such as quality control issues, artefact reliability, information interoperability, process repeatability, automatically data integration, etc.

In an AM process, an artefact is first created as a computer-aided design (CAD) model that contains the geometry of the part before being produced using the specific AM methods. There are a number of stages and processes that need to be completed to realise this physical artefact. Several scholars have attempted to define key stages of the AM process. For example, Nassar and Reutzel [8] distinguish four stages. In the designing stage, a digital model of the artefact is created through the use of solid-modelling CAD software or by 3D Scanning an existing object and importing the data into the CAD environment. Next, in the process planning stage, the 3D model data is translated into a set of instructions for the operation of the AM machine. In the execution stage, the AM system creates the physical artefact in a layer-wise process. In the finishing stage, the support material is removed and the parts are dealt with according to the post-processing requirements, such as heat treatment. Finally, in the verification stage, the finished artefact is compared against the original CAD model for inspection and validation. The definition of the four stages are considered to be general and as a result, they have been deeply extended by Kim et al. [9] into further eight sub-stages linked by seven activities in Fig. 1.

While the specific stages described above cover most practices observed in the industry, some steps are argued by other researchers. For example, [10] proposed skipping the creation of a tessellated model, i.e. an approximate geometric description by discrete surfaces such as triangles. Instead, slicing, i.e. the generation of horizontal layers which combined form the physical model could be done directly from the CAD model with more calculations and handling. This would increase the precision of slices and unify the data sources for execution and verification [2]. For this paper, we propose a view of the AM process that combines those described by Nassar and Reutzel [8] and Kim et al. [9] as shown in Table 1.

To achieve an effective print, the production requirements must be contained in the manufacturing file. Most efforts for AM file standards appear to be focused on standardising the definition of CAD geometry such as improving the STL file through AMF and 3MF initiatives [11]. In addition to the CAD geometry, other important build information and data requirements that are needed include object orientation, support structures, slice structures, machine paths, object packing information, and tolerance data [10]. According to Hiller and Lipson [12], a robust file format should consider aspects of technology independence, simplicity, scalability, and future compatibility. For example, information such as geometrical data should be independent of manufacturing processes, while other information such as tool paths should be dependent on the manufacturing process [2]. This means that information can be considered to be machine or device dependent. This is particularly important for laser- and electron-beam AM processes where relevant data includes the sequence and timing of deposition paths that influence the overall quality and composition of eventual product with regards to stress and microstructure [8]. Other researchers also reported that some information can be lost during the stages of manufacture. For example, the original geometry information (the native CAD file) and the tessellated features can be displaced after the slicing process. Slicing is used to convert a 3D CAD model into a set of instructions for the AM machine in the form of a machine G-code. As there is no standard framework for the exchange of data for the AM production stages, Nassar and Reutzel [8] and Kim et al. [9], proposed four additional file formats in addition to the existing AMF format. They suggested an “AMSF” format that could contain information about the “slices”, whereby the “AMPF” extension would capture data on path planning and process parameters; “AMQF” would be used for sensor data and as a qualification record; while “AMVF” would be used in the last stages of verification and validation phase.

Today, the most commonly used file format for transferring the data model for AM is the STL file format. STL is a digital format that is used to store tessellated surface information. It has been recognised that the STL format has several shortcomings that reduce the suitability of using this format for newer AM machines with multi-nozzle and functionally graded materials capabilities. STL files are prone to redundant information and geometrical defects such as missing, overlapping and degenerate facets, guarding against and repairing of which is computationally and procedurally expensive. The STL file also lacks the provision to store material, texture, colour, measurement or structural information [6]. The surface triangle information model for STL does not provide inherent mechanisms to retain manifold surface information. As a result, a number of proprietary alternatives for 3D model data storage have been proposed and efforts are underway to introduce standards to comprehensively fulfil the design requirements of AM. An overview of the advantages and disadvantages of the file formats frequently associated with AM use is presented in Table 2 [10]; ISO TC184 SC4 WG3 [12, 13]; 3MF [5, 14, 15], p. 261; [2]; ISO TC184 SC1 WG7 [16].

The most prominent effort to date is the AMF format which is an official ISO/ASTM standard for AM. Another file format for AM is 3MF that arose as an effort to bridge the gap between hardware and software systems. Both are based on the extensible mark-up language (XML), which means that the formats follow a standardised, text-based, human-readable encoding format following the open XML specification [17]. Both AMF and 3MF are open and evolving standards that are intended to handle large amounts of design and geometric data required by AM processes. It remains to be seen whether AM hardware and software vendors will refrain from proliferating their own controlled, closed-source formats or to adopt a single industry-wide framework that can cope with the complexity of different AM processes and machines. In the manufacturing spectrum, an alternative approach to the standardisation of AM specific file formats are STEP and STEP-NC machining standards. STEP standards differ from other standards, which is not only product model, but also print format. It provides a broad range of product information descriptions, such as geometry information, material information, and other related specifications and requirements. However, STEP-NC can be viewed as an extended STEP standard. Therefore, STEP-compliant product model can involve the entire product lifecycle. They are efforts from organisations including ISO and ASTM to standardise product and production related information across a number of manufacturing processes. To date, both standards aim to provide extensions that can support AM infrastructure and services. Unlike STL, AMF and 3MF formats, the STEP format includes related geometric and tessellated model data. STEP-NC has a more ambitious aim by building on STEP to include more processing information, but avoiding the tessellated model and to only contain the geometric model.

In the previous section, we provided an overview of the AM process and file formats for data exchange and sharing. From a review of the literature, we identified notable advantages and disadvantages of existing file formats. While we did not find any literature on context-dependency in data requirements of AM data transfer, there is a very clear distinction that separates the use of AM for rapid prototyping and AM as an industrial manufacturing process. For rapid prototyping, features such as exact tolerance adherence and material gradation are perceived to be of minor importance, and therefore the STL format appears to continue to be seen as sufficient for single-material prints. The current de-facto standard of using STL to describe surfaces has some shortcomings due to its inability to describe the properties of the object such as material gradation and colour. First, as there is an increased demand for such features to be used by AM, the use of STL is less capable to meet the demands of the next generation of AM systems. Second, while some of these issues have been addressed by newer file formats such as AMF and STEP-NC, the formats are usually software or hardware-dependent and the build files are still sometimes difficult to be translated across different machine systems. STEP-compliant NC physical file not only describe CAD geometry and design information but also represent the definitions of tools and manufacturing process. Third, AM as an industrial process should be capable of going beyond the mere volumetric and geometric description of an artefact to be manufactured. Some production parameters have relevance to artefact integrity and need to be contained in the file for production such as the built orientation or the melt pool size. Finally, the majority of current models favour tessellated descriptions of volumes. A model with originally smooth surfaces will be represented as a number of edges and vertices. Through the tessellation process, it is inevitable that some precision will be lost. When it was originally conceived with limited computing power, tessellation was seen as an efficient method that could simplify the necessary calculations for slicing. However, as processing power in modern computers has increased, data formats should now enable direct processing of geometric models. The authors believe that future requirements for an AM file format based on a hypothetical RDM scenario should include support for Intellectual Property, quality assurance, and product liability. An RDM scenario where the end user can modify parts might also take into account limited 3D modelling and engineering skill as well as capturing the knowledge between end users and conventional artefact modellers. As such, an RDM-compatible AM data transfer standard should include features that can be modified and other manufacturing related features such as the minimum and maximum wall thicknesses in artefacts should be locked and not editable.

To illustrate the features of AM in a RDM scenario, we define AM as a production process and RDM as a hypothetical manufacturing scenario that takes place in the future. The manufacturing process transforms raw materials into specifically designed physical artefacts. A number of manufacturing techniques, business schemas, and global supply chains have been established in today’s landscape of modern production. Depending on the component, traditional manufacturing is still usually defined by geographically concentrated production centres that are aligned with an already established supply chain network. In such manufacturing centres, large amounts of generally identical items are produced for mass consumption and shipped to remote locations. The position of manufacturing centres is also subject to the availability of technology and manpower. In modern times, the aspect of shipping has become a key factor in the face of the relative efficiency of economies of scale, meaning that it is cheaper to ship products than to set up additional production plants. Industrial development has been predicted to move away from cheap mass production towards customised and personalised products. Manufacturing industry will move towards industrial capacity being re-allocated into reconfigurable factories for continually changing supply chains, super factories for complex products, and even domestic production [18]. Accordingly, five strategic themes have been identified for the future of manufacturing by InnovateUK which is a UK non-departmental public body set up to accelerate UK economic growth by stimulating and supporting business-led innovation, focusing on resource efficiency, manufacturing systems, materials integration, manufacturing processes, and business models. Manufacturing systems and processes, in particular, refer to the creation of more efficient and effective manufacturing models by developing new, agile, and increasingly cost-effective manufacturing processes. There is potential for manufacturing to become even more flexible and adaptive, specifically by applying Additive Manufacturing and to emphasise on high-value manufacturing (HVM). HVM refers to “the application of leading-edge technical knowledge and expertise for the creation of products, production services and associated services” [19], as well as “the reduction of material and energy use in production, and the repatriation or on-shoring of production” (ibid). A part of this is a trend towards Distributed Manufacturing, which is an “on-demand, local manufacturing, made possible by combining digital technologies with new production processes” [20]. Distributed Manufacturing does away with manufacturing centres. In this scenario, manufacturing is de-centralised and the final product is manufactured close to the eventual customer. This leads to the concept of Re-Distributed Manufacturing (RDM). RDM is defined as technology, systems and strategies that have the potential to change the economics and organisation of manufacturing, particularly with regard to location and scale [21]. RDM represents the production of evolved, smart and sustainable products, designed and produced locally using customer input and collaboration. As such, RDM poses future social, technological and economic challenges. One of the key challenges for RDM is how to enable competitive local production, presumably through the consumer or small local manufacturers. In this context, AM has been identified as a technology that might lead to new business models [22] and disrupt the current dominance of cheap global shipping and economies of scale since the principle of AM as a system that can, in principle, produce anything eliminates the need for specialised manufacturing facilities and reduces the efficiency advantage of scale economies. There are limitations of AM that cast doubt on the notion that it might replace mass-production technologies for serial manufacturing of cheap items [23]. However, it is still conceivable that AM would feature prominently in a combined approach where modern and smart products are modularised, with complex or consumable parts produced in traditional manufacturing, and personalised parts produced by local manufacturing or consumers. Based on this, the role of AM in an RDM environment can be defined as the manufacturing of adapted, customer-configured or individualised artefacts close to the customers’ location. This would lead to a number of conceivable use cases for implementing AM in an RDM scenario. Thiesse et al. [24] referred to this scenario as: “A company constructs functional product parts whereas the customer contributes towards the product design. The production is then carried out by a service provider.” On one end of the user spectrum, a user might download the desired 3D model for free or for a fee. On the other end, a user might completely model the desired artefact autonomously. In between exists a scenario where a downloaded model is adapted to fit the user’s needs. The final model is printed either by the user on his or her own AM machine or through an AM service, depending on product requirements such as the material, resolution or quality. For example, material extrusion systems might be suitable for some users and sufficient for a number of parts, whereas powder-based materials might be more suitable for high-quality or high-performance parts. Systems using this technology are less likely to be at the disposal of typical end users. However, hybrid solutions are also conceivable where only some parts, specifically those that are customisable, are modified and printed by the end user, whereas the core part is manufactured remotely [25]. Commercial and non-commercial model repositories such as thingyverse (www.​thingyverse.​com), Shapeways (www.​shapeways.​com), Ponoko (www.​ponoko.​com), and Sculpteo (www.​sculpteo.​com) and other AM vendors are some examples.

This question was answered by a qualitative text analysis of survey questions Q4 and Q5. Half of the participants (n = 13) stated that an AM data exchange standard would improve the manufacturing process mostly through overall simplification, being able to automate the tessellation step, specifically the step from CAD to a manufactured model (n = 5), and a decrease of variation among parts from identical data (n = 4). Four participants stated that the AM data transfer standard would aid in general manufacturing improvements such as the homogenisation of interfaces for manufacturing machines or bridging software and hardware gaps, speeding up the manufacturing process and increasing the process flexibility. Software and hardware compatibility were seen by five participants as a transformative outcome for the RDM landscape. Some participants predicted the standardisation of AM data transfer formats would support future improvements of AM in areas such as model optimisation and part analysis (n = 5). Further, the impact on the industry was seen by some participants through a future adoption of AM in additional areas of manufacturing (n = 5), an active involvement of end users in manufacturing (n = 3), and enabling even more possible hardware geometries (n = 2). An improvement in AM model representation, possibly through alternatives to tessellation, was expected by three participants, and a better coverage of information required for an AM manufacturing process by two participants. Further expected and predicted benefits of an AM data standard include promoting location-independent manufacturing, improved collaboration across disciplines, increased competition in the sector of AM service vendors, and improved reputation of the AM industry.

This question was answered by identifying the users and beneficiaries from Q4 and Q5 in the participants’ responses. A clear majority of responses suggested that manufacturers, i.e. the producers of AM parts, are the main beneficiaries of AM data transfer standards (n = 20). These were followed by end users (n = 8) and tool providers (n = 6). Tool providers such as developers of mesh optimisers would profit from easy access to an official standard as well as the possibility of reaching the entire market by supporting a single, comprehensive standard supporting the whole range of AM systems. End users, i.e. the eventual benefactors of AM products, would profit from a number of effects, including more arbitrary artefact shapes and consideration of exclusive end user concerns such as privacy, in the case of AM medical prostheses. Industrial customers were identified as beneficiaries (n = 9) through an improved access to AM and an increased ability of AM to produce arbitrarily shaped products. A few participants identified the general public (n = 3) as benefiting, through a reduced impact of AM production, and also developers of AM machines (n = 2), through further improvement of AM data transfer formats enabled through standards.

The five features deemed most important by the participants were regular internal structures/lattices, manufacturing tolerances, geometric representation, curvature representation, and surface structures. The full rating result, ordered by average rating, is presented in Table 7. Some participants used the free text fields to express their expectation that a future AM data transfer standard would come with features such as enabling copyright and privacy protection, in particular for bespoke medical artefacts such as prostheses.

Table 7 lists the rating results ordered by average and the supported features per standard. To determine which data exchange standard has the greatest competitive advantage, the standards themselves needed to be ranked. We adopted the gold-first ranking, whereby the standard supporting the most important feature for AM data transfer formats is ranked with the highest score. In the event of a tie, support for the next-most-important feature determines the ranking, until there is a clear order. However, such a ranking would ignore the number of features actually supported—it is conceivable that a standard supporting all but one feature might lose out to a standard supporting only one feature, which by chance could have been ranked with the highest score. Alternatively, the highest score could be given to the standard with the most supported features. Yet another alternative was to apply a system of weighed ranking, where each feature would be assigned a weight, e.g. according to its rank, its average importance, or its importance mode. The last four measures are shown in Table 7. The most important feature, regular internal structures, is only supported by AMF. Interestingly, STEP-NC covers the same number features when compared to AMF, 3MF or STEP but STEP-NC includes “toolpaths” at the expense of “internal structures”. STEP-NC is also ranked above AMF when weighed by average feature importance, modal feature importance, or feature rank. The research suggests that the AM data exchange standard most suitable for an RDM scenario is STEP-NC. This is paradoxical as the STEP-NC standard is still in the process of standardising interfaces for machining and industrial scale production for AM. Beyond the question is whether the STEP-NC standard should have been eligible for comparison in this research to begin with, as the research results can be seen as a tentative suggestion that STEP-NC already supports a wide range of features that are rated desirable. AMF and the fairly new 3MF format were expected to come out top in the competitiveness ranking. While AMF came second to STEP-NC only marginally, 3MF did not perform well, although some of its features are supported neither by AMF nor STEP-NC. It should also be argued that internal structures which are contained in AMF are more important than toolpath features for AM because geometric data is a key inherent feature for representation. Overall, it appears that in their current versions, AMF is more sophisticated than 3MF.

In the focus group sessions, it was acknowledged that commercial interests could drift towards proprietary standards. At the same time, focus group participants emphasised that standards needed to be accessible across disciplines for interoperability in the context of RDM. “Open architecture” was rated 2.92 on average, somewhat below the overall average “important” rating. Only one participant did not rate the feature. The mode, i.e. most common rating, for “open architecture” was four (“very important”), and it was the feature with the biggest gap between average and mode. When compared to all other features, participants were unusually split in their opinion on having an “open architecture”. Altogether, “open architecture” received ten ratings of “very important”, and seven ratings each for “important” or “somewhat important”, but also three ratings for “unimportant”. The topic of open architecture was also mentioned in the open text answers of Q3 and Q4. Two participants predicted that open standards could reduce the barrier of entry for the developers of software tools, leading to more competition among existing software tools such as mesh optimisers, and also to develop other new improvements for AM data exchange formats.