Mapping the design space in function and geometry models supporting redesign for additive manufacturing

ERDSE: efficient reinforcement learning based design space exploration method for CNN accelerator on resource limited platform

The unique tool-less and layer-upon-layer capabilities of additive manufacturing (AM) have largely lessened manufacturing constraints and significantly broadened design freedom. AM offers new opportunities for developing customized products by achieving significant improvements in product performances and lower overall costs. The AM-enabled customization opportunities inevitably challenge traditionally design theories and methodologies (DTMs) that lack design knowledge for AM products. Here, to counteract this limitation, this study presents a novel methodology for constructing knowledge supporting DTMs for AM-enabled customization: customized design for AM (CDfAM). The CDfAM supports two main phases of design for additive manufacturing (DfAM): (1) opportunistic DfAM in a conceptual design phase and (2) restrictive DfAM in an embodiment design phase. In terms of the knowledge construction for the opportunistic DfAM, it is essential to construct knowledge that supports to explore a customized design space enabled by AM’s design freedom from the conceptual design phase. To address the challenge, this study provides a novel method for capturing knowledge of customized AM product behaviors. First, this study proposes an affordance-based DfAM framework by integrating the concept of affordance into three link chain models. Second, an affordance structure is proposed to develop a formal method for representing the knowledge of the customized AM product behaviors utilizing finite state automata. In terms of the knowledge construction for the restrictive DfAM, this study presents a formal method for representing and reasoning customizable modular knowledge of design rules for AM. First, a novel structured modular hierarchy and associated formalisms of AM design rule elements are presented. Second, machine learning is employed to automatically learn design rules from AM data, which leverages the proposed design rule formalisms. A knowledge graph is designed and built to demonstrate the proposed methodology. Case studies with AM products such as a transtibial prosthesis leg and AM data such as ex-situ AM measurement data provide illustrations in this study. The case studies show that the proposed methodology explicitly constructs structured and correlated AM design knowledge of customized AM product behaviors, product geometries, and additive manufacturability constraints. It is also shown that qualified design knowledge in the knowledge graph enhances automation of DfAM decision-making and data-driven knowledge constructions. The proposed CDfAM methodology enables incorporating both customized AM product functionalities and additive manufacturability constraints into product designs. This study provides a foundation for AM knowledge construction and data- and knowledge-driven automation of DfAM decision-making. Validated design rules and representations are expected to contribute to developing international standards on AM design rules.

Research in additive manufacturing (AM) has increased the use of AM in many industries, resulting in a commensurate need for a workforce skilled in AM. In order to meet this need, educational institutions have undertaken different initiatives to integrate design for additive manufacturing (DfAM) into the engineering curriculum. However, limited research has explored the impact of these educational interventions in bringing about changes in the technical goodness of students' design outcomes, particularly through the integration of DfAM concepts in an engineering classroom environment. This study explores this gap using an experimental study with 193 participants recruited from a junior-level course on mechanical engineering design. The participants were split into three educational intervention groups: (1) no DfAM, (2) restrictive DfAM, and (3) restrictive and opportunistic (dual) DfAM. The effects of the educational intervention on the participants' use of DfAM were measured through changes in (1) participants' DfAM self-efficacy, (2) technical goodness of their AM design outcomes, and (3) participants' use of DfAM-related concepts when describing and evaluating their AM designs. The results showed that while all three educational interventions result in similar changes in the participants' opportunistic DfAM self-efficacy, participants who receive only restrictive DfAM inputs show the greatest increase in their restrictive DfAM self-efficacy. Further, we see that despite these differences, all three groups show a similar decrease in the technical goodness of their AM designs, after attending the lectures. A content analysis of the participants' design descriptions and evaluations revealed a simplification of their design geometries, which provides a possible explanation for the decrease in their technical goodness, despite the encouragement to utilize the design freedom of AM to improve functionality or optimize the weight of the structure. These results emphasize the need for more in-depth DfAM education to encourage the use of both opportunistic and restrictive DfAM during student design challenges. The results also highlight the possible influence of how the design problem is stated on the use of DfAM in solving it.

The cyclic manufacturing processes of Additive Manufacturing (AM) create three-dimensional objects layer by layer, offer a large freedom in design, and suffer from very different restrictions compared to conventional manufacturing processes. Nowadays, AM is a mature manufacturing technology capable of being applied in industrial series production. However, the AM technology enters the industrial series production only slowly. A main reason is the fact that engineers in industry have little to no experience in Design for Additive Manufacturing (DfAM) and an integrated experience transfer in engineering education is not yet widespread. To implement the AM technology, a course for graduate students was developed and established, mainly in the master studies and in the field of mechanical engineering. The goal of the course is that the students gain experience in the possibilities and the restrictions of DfAM. The course is based on the Experience Transfer Model (ETM) and is divided in two parts, a lecture and a team project. In the lecture, the students gain explicit knowledge for DfAM. In the practical team project, the students develop and produce AM prototypes based on a variety of AM processes and gain implicit knowledge. The concept of the team project is based on Problem-Based Learning (PBL). The whole team project follows a product development process, including presentations of milestone. The physical outcomes showed that the students are able to design AM optimized parts without falling back into old design patterns. In a survey at the end of the lecture series, the students gave the authors feedback. The large majority of the feedback was positive.

Laser powder bed fusion (L-PBF) is one of the most novel additive manufacturing methods used for a wide range of industrial grade metallic materials. The process can produce end-use metal parts with desirable qualities and mechanical properties. L-PBF however, remains a complicated and expensive manufacturing method. Design for additive manufacturing (DfAM) is a key aspect leveraging the uptake of advantages and possibilities offered by AM in augmenting its competitiveness against conventional manufacturing (CM) methods. Inconel 718 (IN718) is a nickel-based superalloy boasting high temperature strength, good oxidation, and corrosion resistance at elevated temperatures. IN718 is commonly used for high performance applications, such as power and process industry parts, and gas turbine components. High inherent toughness, hardness, work hardening, and low thermal conductivity properties make the material difficult to manufacture through conventional machining methods. The layer-by-layer building of powder metals via L-PBF makes it possible to build different geometrical intricacy. The offered manufacturing flexibility for complex high-end metal structures for variety of applications makes L-PBF an alternative manufacturing method for high performance metals. This study investigates use of DfAM for a small-scale pressure vessel with predefined geometry, dimensions, design space and load condition. The aim is to introduce and exploit contemporary design optimization methods and their feasibility with AM. Structures, such as lattices and stress field driven geometries based on finite element analysis are investigated in this study. The designs are virtually tested under predefined pressure load of 50 bar. All four design options are manufactured on EOS M290 and IN718 powder. The result of the study shows the different optimizations decrease weight and improve material savings without compromising the linear load capacity. Optimized designs could also be made in such a way that it does not increase the manufacturing duration or add additional steps to it.

Design for manufacturing to design for Additive Manufacturing: Analysis of implications for design optimality and product sustainability

Biomimetics as well as additive manufacturing have prominently produced novel design approaches for parts and products independently from each other. The combination of both has resulted in numerous innovative part designs that were unseen before. However remarkable the marketing impact of individual 3D printed biomimetic parts has been, a widespread industrial application is missing to date. This publication, therefore, takes a closer look at how biomimetic design in additive manufacturing is currently pursued and evaluates the different design approaches based on their suitability for industrial application. The assessment reveals that algorithms and thesaurus tools should be preferred in an industrial biomimetic design process. From the various additive manufacturing methods, laser additive manufacturing today is a dominating industrial application when it comes to metal parts. Thus, several case studies of biomimetic designs produced with laser additive manufacturing are presented. On the basis of the selected examples, the added value through biomimetic design is discussed and reviewed critically, raising the question of when a biomimetic design approach is promising compared to conventional design approaches. Based on the review of current use cases and the potentials that the combination of biomimetics and additive manufacturing offer, recommended fields of research are concluded. Finally, the road to industry for biomimetic additive manufacturing design is outlined, taking into account the findings on existing biomimetic design methodologies and tools.

ABSTRACTMinimising total production time in the additive or layered manufacturing is a critical concern, and in this respect, the idea of balancing assembly time and build time is rapidly gaining research attention. The proposed work intends to provide benefit in terms of reduced lead time to customers in a collaborative environment with simultaneous part printing. This paper formulates a mixed-integer non-linear programming (MINLP) model to evaluate the near optimal threshold area and support material allocation while segregating parts for a single material additive manufacturing set-up. The resulting time minimisation model is finitely bounded with respect to support material volume, total production time and total assembly cost constraints. A novel swarm intelligence-based part segregation procedure is proposed to determine the number of part assemblies and part division scheme that adheres to cross-sectional shape, cross-sectional area, and height restrictions. The proposed approach is illustrated and evaluated for objects with regular as well as free-form surfaces using two different hypothetically simulated real size 3D models. Results indicate that the proposed approach is able to reduce the total amount of manufacturing time in comparison with single part build time for all the tested cases.

The design methodologies and part shape algorithms for additive manufacturing (AM) are rapidly growing fields, proven to be of critical importance for the uptake of additive manufacturing of parts with enhanced performance in all major industrial sectors. The current trend for part design is a computationally driven approach where the parts are algorithmically morphed to meet the functional requirements with optimized performance in terms of material distribution. However, the manufacturability restrictions of AM processes are not considered at the primary design phases but at a later post-morphed stage of the part’s design. This paper proposes an AM design method to ensure: (1) optimized material distribution based on the load case and (2) the part’s manufacturability. The buildability restrictions from the direct energy deposition (DED) AM technology were used as input to the AM shaping algorithm to grant high AM manufacturability. The first step of this work was to define the term of AM manufacturability, its effect on AM production, and to propose a framework to estimate the quantified value of AM manufacturability for the given part design. Moreover, an AM design method is proposed, based on the developed internal stresses of the build volume for the load case. Stress tensors are used for the determination of the build orientation and as input for the part morphing. A top-down mesoscale geometric optimization is used to realize the AM part design. The DED Design for Additive Manufacturing (DfAM) rules are used to delimitate the morphing of the part, representing at the same time the freeform mindset of the AM technology. The morphed shape of the part is optimized in terms of topology and AM manufacturability. The topology optimization and AM manufacturability indicator (TMI) is introduced to screen the percentage of design elements that serve topology optimization and the ones that serve AM manufacturability. In the end, a case study for proof of concept is realized.

Design for manufacture and assembly (DfMA) in the architectural, engineering, and construction (AEC) industry is attracting the attention of designers, practitioners, and construction project stakeholders. Digital fabrication (Dfab) and design for additive manufacturing (DfAM) practices are found in current need of further research and development. The DfMA’s conceptual function is to maximize the process efficiency of Dfab and AM building projects. This work reviewed 171 relevant research articles over the past few decades. The concepts and the fundamentals of DfMA in building and construction were explored. In addition, DfMA procedures for Dfab, DfAM, and AM assembly processes were discussed. Lastly, the current machine learning research on DfMA in construction was also highlighted. As Dfab and DFAM are innovated, practical DFMA techniques begin to develop to a great extent. Large research gaps in the DfMA for Dfab and DfAM can be filled in terms of integrating them with product structural performance, management, studied cases, building information modeling (BIM), and machine learning to increase operational efficiency and sustainable practices.

Additive manufacturing (AM) technologies have seen fast growth in the last few decades. AM needs the implementation of new methods in design, fabrication, and delivery to end-users. Hence, AM techniques have given great flexibility to designers as the design of complex components and highly customized products are no longer binding from a manufacturability point of view. In addition to high material variety, this allows multi-material and variable mechanical characteristics of product manufacturing. This review paper addresses the design for additive manufacturing (DfAM) rules, guidelines, and tools to guide the designer to take advantage of the opportunities provided by AM whether in the early design stages (EDS) or in the later phase using computer-aided design (CAD) tools. It discusses issues related to the design for AM and proposes a DfAM framework applied in the design for the additive manufacturing process.

Opportunities are offered by multiple Additive Manufacturing (AM) processes nowadays. Design rules are evolving to lead to lighter and stiffer parts with really more complex shapes than those obtained by conventional processes. Worldwide, new methodologies/tools of assistance for the design are developed such as Design for Additive Manufacturing (DfAM). Additive manufacturing can allow the development of new metamaterials and health-matter evaluation based on energy flow evaluation. In this paper, the objective is to generate a new methodology with DfAM based on mesoscale knowledge. It is generated with open lab bench and simple object characterization. A methodology is presented to formalize and quantify information at multilayer dimension. A database is also generated following Design of Experiments (DoE) to obtain metamodels. They are developed for specific features representative of AM geometric class such as overhanging, holes or walls for instance. Mereotopological primitives with their AM definitions are used to define features in term of space and time variables. This theory enables the formalization of knowledge at the mesoscopic scale taken into consideration layer by layer build-up. It is then possible to use it to integrate data and information to the different feature juxtapositions using recognition algorithm. Information for each feature can then be included and explicitly used to help the designer during detailed design phase. A global 4-steps DfAM methodology maximizing the potential of AM is presented and validated through a part from the space industry use case. It includes the definition of skeleton/skin entities, pattern decomposition, information associated based on material evaluation and decision for AM part.

Additive Manufacturing (AM) technologies enable the fabrication of parts and devices that are geometrically complex, have graded material compositions, and can be customized. To take advantage of these capabilities, it is important to assist designers in exploring unexplored regions of design spaces. We present a Design for Additive Manufacturing (DFAM) method that encompasses conceptual design, process selection, later design stages, and design for manufacturing. The method is based on the process-structure-property-behavior model that is common in the materials design literature. A prototype CAD system is presented that embodies the method. Manufacturable ELements (MELs) are proposed as an intermediate representation for supporting the manufacturing related aspects of the method. Examples of cellular materials are used to illustrate the DFAM method.

A large body of work can be found in literature on Design Space Exploration (DSE) methods for distributed embedded system architecting (DESA). However, almost none of these methods is successfully adopted in automotive industry. To clarify the reasons, this paper 1) analyzes the current state of the art (SOTA) on DSE methods for DESA through a systematic literature study, focusing on the assumed architecting process and concerns; 2) investigates the state of practice (SOP) on DESA in the automotive industry through a literature study and interviews with experienced system architects from five different automotive manufacturers; and 3) analyzes the gap between SOTA and SOP, and thereby discusses potential improvements of DSE methods.

Additive manufacturing (AM) is a “game-changing” digital fabrication technology that can fundamentally alter the way products are designed and manufactured. In nuclear engineering, the application of additive manufacturing can enable advanced reactor designs with better performance, lower cost, and quicker deployment. To exploit the maximum potential of additive manufacturing, design for additive manufacturing (DfAM) is needed, DfAM can leverage the complexity freedom offered by additive manufacturing technology to break traditional design constraints and achieve enhanced performance. Generative design, one of the key technologies in DfAM, is an iterative exploration method using AI-driven algorithms to generate a series of design solutions that meet specified constraints. In this work, a generative design platform for micro-reactors is developed based on DASSULT SIMULIA Isight. To simulate and analyze the multi-physics performance of the reactor design, various multi-physics simulation codes are integrated on this generative design platform, including RMC for neutron physics analysis, ANSYS Fluent for computational fluid dynamics analysis, and ANSYS Mechanical for structural mechanics analysis. A multi-objective optimization method is used to explore geometric parameters and fuel configurations to achieve better neutronics, thermal-fluidics, and structural mechanics performance. The generative design solution is a 4 MW helium-cooled nuclear reactor which includes 3D-printed silicon carbide (SiC) matrix embedded with tri-structural isotropic (TRISO) particle fuel. Helium at 3 MPa is chosen as the coolant, and the cross-shaped coolant channels are distributed within the matrix to remove the heat generated by fission. According to the multiphysics analysis, the reactor has a initial keff of 1.0571, power peaking factor FQ of 1.50, and operating lifetime of 1200 EFPD (Effective Full Power Days), with a maximum temperature of 1331 K and a peak thermal stress of 141.3 MPa. The generative design method and platform provide a systematic and efficient approach for the design of additively manufactured micro-reactors, and the generative design solution demonstrates the feasibility and effectiveness of the proposed method.