TY - JOUR
AU - Moon, Seung Ki
AB - Abstract The advent of additive manufacturing (AM) has brought about radically new ways of designing and manufacturing of end-use parts and components, by exploiting freedom of design. Due to the unique manufacturing process of AM, both design and process parameters can strongly influence the part properties, thereby enlarging the possible design space. Thus, finding the optimal combination of embodiment design and process parameters can be challenging. A structured and systematic approach is required to effectively search the enlarged design space, to truly exploit the advantages of AM. Due to lowered costs in computing and data collection in the recent years, data-driven strategies have become a viable tool in characterization of process, and researches have starting to exploit data-driven strategies in the design domain. In this paper, a state-of-the-art data-driven design strategy for fused filament fabrication (FFF) is presented. The need for data-driven strategies is explored and discussed from design and process domain, demonstrating the value of such a strategy in designing an FFF part. A comprehensive review of the literature is performed and the research gaps and opportunities are analysed and discussed. The paper concludes with a proposed data-driven framework that addresses the identified research gaps. The proposed framework encompasses knowledge management and concurrent optimization of embodiment design and process parameters to derive optimal FFF part design. Contribution of this paper is twofold: A review of the state-of-the-art is presented, and a framework to achieve optimal FFF part design is proposed. Graphical Abstract Open in new tabDownload slide Graphical Abstract Open in new tabDownload slide additive manufacturing, Bayesian inference, data-driven design, design strategies, fused filament fabrication Highlights: Comprehensive review of state-of-the-art data-driven design methods for fused filament fabrication (FFF). Identify research gaps and opportunities of data-driven design strategies in FFF. Propose a data-driven framework that encompasses knowledge management and concurrent optimization of embodiment design and process parameters. 1. Introduction Additive manufacturing (AM), or 3D printing, is a layered manufacturing technique where digital files are converted into physical parts. AM allows for freedom of design, direct manufacturing of complex geometries and shapes, and has many benefits such as tool-less manufacturing, printing on demand, and reduced material wastage, which are advantageous over traditional manufacturing methods. American Society for Testing and Materials (ASTM) and International Organization for Standardization have classified various AM technologies into seven main categories, namely Powder Bed Fusion, Directed Energy Deposition, Material Extrusion, Vat Photopolymerization, Binder Jetting, Sheet Lamination, and Material Jetting (Standard, 2012; Chua & Leong, 2017). AM traditionally has been used as a rapid prototyping technique that reduces developmental cycle time for product design. In recent years, the popularity of AM has risen exponentially as the technology matures, which resulted in higher adoption for printing direct end-use parts. Complex structure and topology that are near-impossible to be manufactured using conventional manufacturing techniques can now be produced in a cost-effective and efficient manner using the AM technologies (Mineo, Pierce, Nicholson, & Cooper, 2017; Wu, Wei, & Terpenny, 2018; Hu & Wu, 2019; Kim & Yoo, 2020). New paradigms in distributed and sustainable manufacturing environments are enabled by the advent of AM (Ford & Despeisse, 2016). In one of the most notable examples of AM adoption in aviation industry, General Electric designed and produced AM engine fuel nozzles since 2015, which was one of the first flying aircraft parts that is additively manufactured. The suppression of assembly was achieved and the cost of manufacturing of the nozzle is reduced by up to 75% as compared to conventional manufacturing (General Electric Additive, 2018 ). Airframe original equipment manufacturers, such as Boeing and Airbus, have also utilized AM to manufacture spare parts and brackets to meet aircraft delivery schedules (Wagner & Walton, 2016; Ceruti, Marzocca, Liverani, & Bil, 2019). Among the AM technologies, fused filament fabrication (FFF), commonly known as Fused Deposition Modeling™, is one of the most popular and most studied, primarily due to ease of operation and low operating cost of an FFF printer. Typically, material used in FFF process is engineering plastics, such as acrylonitrile butadiene styrene (ABS), polylactic acid, and polycarbonate. Schematic of FFF and the nomenclature used in this paper are shown in Fig. 1. The principle is briefly described as follows. A filament is fed into the liquefier using torque wheels and is heated to above its glass transition temperature and transforms into a semimolten state. The nozzle assembly moves along a predefined path and deposits the semimolten material onto the build platform. Once deposited, the material starts to cool to build chamber temperature. Bonds are formed with adjacent material through polymer chain diffusion. Once a layer is completed, the build platform moves down a predefined distance and the process repeats for the subsequent layers, until the part is completed. As the melted polymer is extruded through a nozzle or an orifice, FFF technique is classified under material extrusion. Figure 1: Open in new tabDownload slide Schematic and nomenclature of FFF process. Figure 1: Open in new tabDownload slide Schematic and nomenclature of FFF process. Freedom of design enabled by AM has vastly increased the design space that AM designers can explore for a particular part with a specific set of requirements. Due to the unique manufacturing process of AM, both design and process parameters can strongly influence part properties. AM designers have the option of optimizing the design and/or process parameters to achieve the functional requirements of a part. In addition, manufacturing constraints associated with traditional manufacturing (e.g. injection moulding) are no longer considered when designing an AM part. Instead, a set of guidelines commonly known as design for additive manufacturing (DfAM) is used to guide AM designers in designing AM parts. This further enlarges the possible design space. A structured and systematic approach is required to effectively search this enlarged design space to exploit the freedom of design that AM brings. To overcome the issue faced when searching for optimal designs, a state-of-the-art review is performed on various data-driven strategies that leading researchers are utilizing. Next, research trends and opportunities are identified. Finally, a data-driven design strategy framework is proposed to address the research opportunities and gaps. Examples from existing literature are amalgamated to demonstrate the feasibility and applicability of the framework. The scope of the paper will be limited to FFF technique, due to its popularity and hence, increasing the potential benefits of this paper. This paper is structured into the following sections: In Section 2, a brief background on data-driven design strategy and the needs of AM designers are discussed. Literature review on the state-of-the-art is performed and discussed in Section 3. Then, the section concludes with an overview of the research trend and opportunities. In Section 4, a data-driven framework for FFF is proposed to address the research gaps and opportunities identified. Lastly, Section 5 provides a summary and some limitations of the proposed framework. 2. The Need for Data-Driven Design Strategy Concurrent exploration of embodiment design and process parameters enables optimal solutions to be derived. The converse is true. Without providing due consideration to both embodiment design and process parameters, FFF parts are manufactured with poor quality. Figure 2 provides an example of such scenario. The specimen shown is an ASTM D638 specimen, which is commonly used for characterization of FFF tensile properties. For the specific parameter settings, voids appear in the curved transition zone, which causes premature failure and interraster delamination. Such a phenomenon is commonly observed in literature (Ahn, Montero, Odell, Roundy, & Wright, 2002; Byberg, Gebisa, & Lemu, 2018). Chang and Huang (2011) termed such errors or voids as “aperture errors” and argued that such errors deteriorate the quality of FFF parts. Adopting concurrent design process in FFF allows for shortfall of the process to be compensated. For example, defining overhang constraints during topology optimization enables structures to be manufactured with little to no support (Gaynor, Meisel, Williams, & Guest, 2014; Wang, Gao, & Kang, 2018; Huang et al., 2020). To achieve lattice structures in FFF parts, there are two possible ways of doing so. One method is that designers can create a model that already has these structures built in. The alternative method is through tuning the toolpath (Ang, Leong, Chua, & Chandrasekaran, 2006; Medellin-Castillo & Zaragoza-Siqueiros, 2019). As a result, both geometry and process parameter can be fine-tuned to achieve desired functional requirements and mechanical properties of an AM part. However, such concurrent design process greatly increases the dimensionality of the design space and exponentially increases the amount of resources required to derive the optimal solution. Figure 2: Open in new tabDownload slide Voids present in an FFF tensile specimen that demonstrates the need for concurrent design. Figure 2: Open in new tabDownload slide Voids present in an FFF tensile specimen that demonstrates the need for concurrent design. Data-driven approaches can efficiently and effectively resolve the issue of enlarged design spaces. Typically, two types of characterization strategies for design spaces, namely data-driven and model-based, are frequently used in literature. A data-driven strategy is defined as the approach of combining data from various sources (through experimental, sensors, past data records, etc.), analysing and processing, and deriving meaningful and actionable insights (An, Kim, & Choi, 2015). The resultant models can be used for optimization and prediction of FFF part properties. Model-based approaches, on the other hand, utilize a priori knowledge of the underlying physical and mathematical knowledge of the process. These first principles are then subsequently used to build a process model for simulation, whose accuracy is then validated by experiments (Yin, Ding, Xie, & Luo, 2014). Through data-driven techniques, cheaper and faster prediction and exploration even within high-dimensional design spaces are made possible. Typical run time for physics model-based approach ranges from hours to days, whereas typical run time for surrogate models is a magnitude faster, ranging from seconds to minutes (Paul et al., 2019; Roy & Wodo, 2020). In addition, surrogate modeling augments results from the expensive physics model-based simulations, which reduces the cost of prediction, and enables data mining to gain insights into relationships between variables and response of interests (Forrester, Sobester, & Keane, 2008). With data-driven techniques, feasible and infeasible design spaces in FFF are identified quickly and effectively. Typical surrogate modeling techniques involve machine learning (ML) techniques such as Gaussian process regression (GRP), support vector regression, Bayesian network (BN), etc. These techniques are trained using available data and are closely related to statistics. Deep learning is a subfield of the ML techniques, which also falls under surrogate modeling. Deep learning techniques utilize multiple layers of representation of data, in contrast with other ML models that represent the data in only a single layer or two. Examples of such deep learning models include neural networks. Advantages of the deep learning techniques are allowing for automation of feature engineering, which has to be performed manually in the other ML methods. However, the deep learning requires significant resources in terms of hardware and data. In addition, the performance of ML models outperforms the deep learning when the dataset is small (Chollet, 2018). Hence, ML models tend to be used in AM, due to the limited availability of the datasets. However, during the model building stage, trade-off between bias and variance needs to be carefully balanced to ensure that the model does not overfit the existing dataset or training data nor having poor predictive accuracy. This trade-off can be alleviated through calibration of model with new empirical data (Olleak & Xi, 2020). For an in-depth review of the existing surrogate modeling approaches, the reader is referred to Wang and Shan (2007) and Forrester et al. (2008). For an overview of surrogate modeling techniques used in AM in general, the readers can refer to Goh, Sing, and Yeong (2020), Wang, Tan, Tor, and Lim (2020), Yu and Jiang (2020), and Meng et al. (2020). In the following sections, design process of FFF parts is segregated into embodiment design and process parameter domain, and the needs of AM designers in each of the phases are articulated and discussed. 2.1 The need for data-driven design strategy from embodiment design perspective Embodiment design is defined as the design stage where a main design concept is developed according to design requirements. During this phase, geometrical structures, material, and requirements are defined, and multiple possible designs can be proposed to meet the defined functional requirements (Xiong et al., 2019). AM designers need to be well versed with DfAM rules and have a good understanding of how the part geometries (e.g. lattice structure) can affect the part properties. AM designers utilize the rules of DfAM to ensure the printability of the designed part, where printability is defined as ability to produce accurate realization of the designed model (Sossou, Demoly, Montavon, & Gomes, 2018; Mycroft et al., 2020). Common DfAM rules for FFF include overhang structure, bridging length (length of unsupported horizontal structure), and thin-walled structures. For AM designers to gain a good comprehension of these rules, they require extensive learning and trial and error to understand the nuances of these rules and their applicability. For example, Fernandez-Vicente, Canyada, and Conejero (2015) concluded that even though the authors have done extensive empirical studies to derive the design rules, the applicability might be limited. These rules only serve as a guideline and may be only applicable to conditions used in the study, and the empirical results can be further optimized by an expert. As such, the resultant steep learning curves and knowledge gaps often act as a barrier to effective AM adoption (Richter, Watschke, Schumacher, & Vietor, 2018). Due to freedom of design, architected material such as lattice structure and honeycomb structure can now be efficiently produced through AM techniques. These structures are often utilized due to their high strength-to-weight ratio and unprecedented mechanical performance (Moon, Tan, Hwang, & Yoon, 2014; Feng, Fu, Lin, Shang, & Li, 2018; Panda, Leite, Biswal, Niu, & Garg, 2018; Valdevit, Bertoldi, Guest, & Spadaccini, 2018). However, the characterization of complex topological structures can be time consuming and tedious due to the non-linear relationship between design and the properties (Liu et al., 2017; Bessa, Glowacki, & Houlder, 2019). The geometry of these structures dominates the overall mechanical performance, and often, finite element analysis (FEA) is required to characterize these structures (Panesar, Abdi, Hickman, & Ashcroft, 2018; Wang et al., 2020). Such simulations can be costly and time consuming, especially considering the limitless permutations of the designs of architected material. 2.2 The need for data-driven design strategy from process perspective The performance and quality of AM parts are highly sensitive to the process parameters and the manufacturing process (Barrrionuevo & Ramos-Grez, 2019). Process parameters and unwanted variations in the manufacturing process (e.g. variations in build chamber thermal field) affect the micro and meso structures of the material, and in turn, these structures affect the final properties of the AM parts. Such interdependencies are commonly referred to as process–structure–property (PSP) linkage and can be hard to characterize due to a large number of parameters and uncertainties involved, and both issues are discussed in detail in the following paragraphs (Xiong et al., 2019; Zhang, Wang, & Gao, 2019; Nguyen & Choi, 2020). Once these interdependencies are characterized and understood, AM designers then would need to perform trade-off analysis to propose an optimal part design. The substantial number of process parameters involved in FFF process can result in unwieldy characterization effort if not done systematically. Figure 3 illustrates the sheer number of process parameters and noise factors along the FFF process chain using an Ishikawa diagram. These process parameters and noise factors originate from sources such as material, environment, printing parameters and machine can influence the final part properties and qualities. Potential confounding between the noise factors and process parameters can further exacerbate the effort to characterize the process. Without a systematic approach to characterize and quantify the effect of these parameters, the quality and final printed part properties may be not able to satisfy the design requirements (Mokhtarianet al., 2019). Figure 3: Open in new tabDownload slide Ishikawa diagram of design and noise factors that can influence part properties in FFF process. Figure 3: Open in new tabDownload slide Ishikawa diagram of design and noise factors that can influence part properties in FFF process. Aleatory and epistemic uncertainty of FFF process can significantly affect the final part properties. Aleatory uncertainties are inherent to the process, for example, variation in the print head movement, chamber temperature, etc. Epistemic uncertainties, on the other hand, arise due to lack of process knowledge (Z. Hu & Mahadevan, 2017). The uncertainties result in process variability. It has been reported by Montazeri, Nassar, Dunbar, and Rao (2020) that during a build of seven parts that are identical in terms of process parameters, except for build orientation, five of the parts failed. The failed parts contain various defects, which makes the parts unusable operationally. Such failures are often costly and prevent AM from being adopted as a manufacturing method for highly critical parts (Mokhtarian et al., 2019; Rizzi et al., 2019). There is a need to manage and characterize the uncertainties to ensure that the process is robust and the quality of FFF parts is consistent. To arrive at an optimal part design, trade-off analysis is often performed by AM designers. Part build orientation (PBO) is a critical process parameter that the AM designers need to have a gasp of to achieve design intent. PBO significantly affects FFF part properties, such as build time, surface roughness, mechanical strength, etc. (Pandey, Thrimurthulu, & Reddy, 2004; Khodaygan & Golmohammadi, 2018; Nguyen & Choi, 2020). Often, conflicting objectives require AM designers to make trade-offs between part properties to arrive at an optimized PBO. For example, it has been found that the strength of FFF parts can be as low as 46% of the injection moulded specimen (Zaldivar et al., 2017), due to the anisotropicity inherent in FFF process as compared to other AM process such as selective laser melting (Simonelli, Tse, & Tuck, 2014). However, to minimize the surface roughness, a different PBO is chosen that can adversely affect the mechanical strength. Common trade-off analysis includes mechanical strength, surface roughness, and dimensional accuracy (Peng, Xiao, & Yue, 2014; Qattawi, Alrawi, & Guzman, 2017; Zhang, Choi, Moon, & Ngo, 2020). Trade-offs that include energy consumption and material consumption against conventional metrics such as mechanical strength are increasingly common, as part of the drive toward green manufacturing (Tian, Ma, & Alizadeh, 2019; Alizadeh, Esfahani, Tian, & Ma, 2020; Qin, Liu, Grosvenor, Lacan, & Jiang, 2020). Without a data-driven approach based on empirical evidence to guide AM designers for the trade-off analysis, the eventual part design may not be an optimal one. In this section, the needs for data-driven design strategy from both embodiment design and process perspective are explored and discussed. The vast design spaces and the knowledge required may impede AM designers from achieving an optimal part design and process parameter selection, without a robust data-driven strategy. In the next section, the state-of-the-art literature that addresses the above issues is reviewed and discussed. 3. Data-Driven Design Strategy in FFF In this section, literature review is performed to understand the state-of-the-art data-driven design strategies in both design domain and process domain, as presented in Tables 1 and 2, respectively. In both tables, the proposed methodology and the response of interest are presented, along with a brief description of the parameters that were investigated. Lastly, the literature source is cited to provide reference to readers for further reading. The scope of this paper is limited to FFF and includes generic approaches that are applicable to FFF as discussed in this section. Table 1: Overview of the literature that utilizes data-driven strategy in design domain. . Response of interest . Investigated parameters . Strategies and methodology used to achieve design goals . Reference . Design space exploration Dimensional accuracy Process and material-related domain, design-related domain, part-related domain BN for merging knowledge from different sources and enabling inferencing with uncertainty (Wang, Blache, Zheng, & Xu, 2018) Stiffness of response for functionally graded structure Design space Framework for design space exploration and exploitation (Xiong et al., 2019) Designing of a platform variant AM production costs, AM process settings, customer-perceived utility Fuzzy time-driven activity-based costing and adaptive neuro-fuzzy inference system (Yao, Moon, & Bi, 2016) Personalized design of parts to be manufactured by AM Affordance, effectivity, preferences Finite state automata (Ko, Moon, & Otto, 2015) Architected material Mechanical properties of cellular structure Dimensions of structs and unit cell Particle swarm theory (Chu, Graf, & Rosen, 2008) Compression response Top and base diameter, height, cross section of the longerons Bayesian ML (Bessa et al., 2019) Yield strength, modulus of elasticity Wall thickness and cell size of honeycomb structure Genetic programming and automated neural network models, response surface regression comparison (Panda et al., 2018) Compressive strength, modulus Geometry (strut orientation, strut angle, wall thickness, strut spacing) Mixed fractional factorial design with two and three levels of factors for rectangular prism, surrogate modeling (Levasseur, Ploeg, & Petit, 2012) Design rules Printable threshold overhang angle Extrusion temperature, fan speed, print speed Design of experiment (Jiang, Stringer, Xu, & Zhong, 2018) Printable bridge length Printing speed, nozzle temperature, cooling fan speed L32 orthogonal array with BPNN (Jiang et al., 2019) . Response of interest . Investigated parameters . Strategies and methodology used to achieve design goals . Reference . Design space exploration Dimensional accuracy Process and material-related domain, design-related domain, part-related domain BN for merging knowledge from different sources and enabling inferencing with uncertainty (Wang, Blache, Zheng, & Xu, 2018) Stiffness of response for functionally graded structure Design space Framework for design space exploration and exploitation (Xiong et al., 2019) Designing of a platform variant AM production costs, AM process settings, customer-perceived utility Fuzzy time-driven activity-based costing and adaptive neuro-fuzzy inference system (Yao, Moon, & Bi, 2016) Personalized design of parts to be manufactured by AM Affordance, effectivity, preferences Finite state automata (Ko, Moon, & Otto, 2015) Architected material Mechanical properties of cellular structure Dimensions of structs and unit cell Particle swarm theory (Chu, Graf, & Rosen, 2008) Compression response Top and base diameter, height, cross section of the longerons Bayesian ML (Bessa et al., 2019) Yield strength, modulus of elasticity Wall thickness and cell size of honeycomb structure Genetic programming and automated neural network models, response surface regression comparison (Panda et al., 2018) Compressive strength, modulus Geometry (strut orientation, strut angle, wall thickness, strut spacing) Mixed fractional factorial design with two and three levels of factors for rectangular prism, surrogate modeling (Levasseur, Ploeg, & Petit, 2012) Design rules Printable threshold overhang angle Extrusion temperature, fan speed, print speed Design of experiment (Jiang, Stringer, Xu, & Zhong, 2018) Printable bridge length Printing speed, nozzle temperature, cooling fan speed L32 orthogonal array with BPNN (Jiang et al., 2019) Open in new tab Table 1: Overview of the literature that utilizes data-driven strategy in design domain. . Response of interest . Investigated parameters . Strategies and methodology used to achieve design goals . Reference . Design space exploration Dimensional accuracy Process and material-related domain, design-related domain, part-related domain BN for merging knowledge from different sources and enabling inferencing with uncertainty (Wang, Blache, Zheng, & Xu, 2018) Stiffness of response for functionally graded structure Design space Framework for design space exploration and exploitation (Xiong et al., 2019) Designing of a platform variant AM production costs, AM process settings, customer-perceived utility Fuzzy time-driven activity-based costing and adaptive neuro-fuzzy inference system (Yao, Moon, & Bi, 2016) Personalized design of parts to be manufactured by AM Affordance, effectivity, preferences Finite state automata (Ko, Moon, & Otto, 2015) Architected material Mechanical properties of cellular structure Dimensions of structs and unit cell Particle swarm theory (Chu, Graf, & Rosen, 2008) Compression response Top and base diameter, height, cross section of the longerons Bayesian ML (Bessa et al., 2019) Yield strength, modulus of elasticity Wall thickness and cell size of honeycomb structure Genetic programming and automated neural network models, response surface regression comparison (Panda et al., 2018) Compressive strength, modulus Geometry (strut orientation, strut angle, wall thickness, strut spacing) Mixed fractional factorial design with two and three levels of factors for rectangular prism, surrogate modeling (Levasseur, Ploeg, & Petit, 2012) Design rules Printable threshold overhang angle Extrusion temperature, fan speed, print speed Design of experiment (Jiang, Stringer, Xu, & Zhong, 2018) Printable bridge length Printing speed, nozzle temperature, cooling fan speed L32 orthogonal array with BPNN (Jiang et al., 2019) . Response of interest . Investigated parameters . Strategies and methodology used to achieve design goals . Reference . Design space exploration Dimensional accuracy Process and material-related domain, design-related domain, part-related domain BN for merging knowledge from different sources and enabling inferencing with uncertainty (Wang, Blache, Zheng, & Xu, 2018) Stiffness of response for functionally graded structure Design space Framework for design space exploration and exploitation (Xiong et al., 2019) Designing of a platform variant AM production costs, AM process settings, customer-perceived utility Fuzzy time-driven activity-based costing and adaptive neuro-fuzzy inference system (Yao, Moon, & Bi, 2016) Personalized design of parts to be manufactured by AM Affordance, effectivity, preferences Finite state automata (Ko, Moon, & Otto, 2015) Architected material Mechanical properties of cellular structure Dimensions of structs and unit cell Particle swarm theory (Chu, Graf, & Rosen, 2008) Compression response Top and base diameter, height, cross section of the longerons Bayesian ML (Bessa et al., 2019) Yield strength, modulus of elasticity Wall thickness and cell size of honeycomb structure Genetic programming and automated neural network models, response surface regression comparison (Panda et al., 2018) Compressive strength, modulus Geometry (strut orientation, strut angle, wall thickness, strut spacing) Mixed fractional factorial design with two and three levels of factors for rectangular prism, surrogate modeling (Levasseur, Ploeg, & Petit, 2012) Design rules Printable threshold overhang angle Extrusion temperature, fan speed, print speed Design of experiment (Jiang, Stringer, Xu, & Zhong, 2018) Printable bridge length Printing speed, nozzle temperature, cooling fan speed L32 orthogonal array with BPNN (Jiang et al., 2019) Open in new tab Table 2: Overview of the literature that utilizes data-driven strategy in process domain. . Response of Interest . Parameters investigated . Strategies and methodology used to achieve design goals . Reference . Toolpath planning Reducing geometric errors (voids/aperture errors), improve printability of features Adaptive contour parallel toolpath, bead width Level-set method for path planning and GRP for determination of bead geometry (Xiong et al., 2019) Dimensional accuracy, tensile properties Infill density, infill pattern, extrusion temperature, LT L9 orthogonal array with Taguchi method (Qattawi, 2018) Tensile strength Infill density, extrusion temperature GA-ANN (genetic algorithm artificial neural network) (Yadav, Chhabra, Garg, Ahlawat, & Phogat, 2020) Strength-to-weight ratio Infill pattern, infill density Numerical modeling and ANN (Zhou, Hsieh, & Ting, 2018) Optimization Surface roughness Build time, PBO NSGA-II (Pandey et al., 2004) PBO PBO NSGA-II, Kriging, TOPSIS (Khodaygan & Golmohammadi, 2018) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature NSGA-II, TOPSIS, (Alizadeh et al., 2020) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature Karush–Kulh–Tucker (KKT)-based non-linear programming optimization (Tian et al., 2019) Flexural strength, impact strength PBO, LT, RA, RW, RRAG Gray relation, Taguchi method (Liu et al., 2017) Tensile strength, flexural strength, impact strength LT, PBO, RA, RW, RRAG Response surface methodology using FCCD design matrix (Sood, Ohdar, & Mahapatra, 2010) Compressive Strength LT, PBO, RA, RW, RRAG Quantum-behaved particle swarm optimization with FCCD design matrix (Sood, Ohdar, & Mahapatra, 2012) Geometrical accuracy (length, width, thickness) LT, RRAG, RA, PBO, RW, number of contours RSM with I-optimality design matrix (Mohamed, Masood, & Bhowmik 2016b) Compressive strength LT, PBO, RA Multigene genetic programming (MGGP) and general regression neural network (GRNN) (Panda, Bahubalendruni, & Biswal, 2015) Creep compliance and recoverable compliance LT, RRAG, RA, PBO, RW Definitive screening design, multilayer perceptron neural network (Mohamed, Masood, & Bhowmik, 2017) Dimensional accuracy, surface precision Nozzle temperature, printing speed, print spacing, filament feed rate IPSO-BPNN (improved particle swarm optimization with back propagation neural network) (Yan, 2020) Tensile strength LT, infill density, printing speed, RA ANOVA and adaptive neuro-fuzzy inference system (ANFIS) (Barrrionuevo & Ramos-Grez, 2019) Tensile strength Nozzle temperature, LT, RA Artificial neural network – Levenberg Marquardt (ANN) (Bayraktar, Uzun, Çakiroğlu, & Guldas, 2017) Dimensional accuracy (length, width and thickness) LT, PBO, infill density, number of contours RSM – ANN, ANN, ANN-GA (Deswal, Narang, & Chhabra, 2019) Surface roughness PBO Radial basis function neural network (RBFNNs), imperialist competitive algorithm (ICA), compare against (Vahabli & Rahmati, 2016) Dimensional errors, warp deformation, build time Print speed, LT, feed rate, RW compensation factor RSM + fuzzy inference system (FIS), using uniform 17 array (Peng et al., 2014) Dimensional accuracy (length, width, thickness) LT, PBO, RA, RW, AG L27 Taguchi method + fuzzy inference system (Padhi et al., 2017) Dynamic mechanical properties (loss compliance and storage compliance) LT, RRAG, RA, PBO, RW, number of contours IV-optimality RSM, multilayer feed-forward neural networks (MFNN) (Mohamed, Masood, & Bhowmik, 2016a) Compressive strength LT, PBO, RA, RW, RRAG M5′-genetic programming (Garg, Tai, Lee, & Savalani, 2014) Compared against SVR, ANFIS Cost of optimization of parameters Mechanical strength, surface roughness, etc. Fully connected neural network (FCNN) (Zhou, Hsieh, & Wang, 2019) Surface roughness LT, print speed, extruder temperature, outer shell print speed RSM, particle swarm optimization, symbiotic organism search (Saad, Nor, Baharudin, Zakaria, & Aiman, 2019) Prediction Porosity Bonding between adjacent toolpath ANN (Haghighi & Li, 2020) Geometrical accuracy, surface roughness Offline process settings (feed/flow ratio of nozzle), in situ process variables (acceleration) Quantitative and qualitative (QQ) models (Sun, Rao, Kong, Deng, & Jin, 2017) Geometrical accuracy Process parameters, geometry of AM parts Bayesian inference (Sabbaghi & Huang, 2018) Geometrical accuracy Process parameters, geometry of AM parts Adaptive Bayesian (Sabbaghi, Huang, & Dasgupta, 2018) Thermal profile Direction of print, number of layers Neural network (Roy & Wodo, 2020) In plane geometric accuracy Geometry Multitask learning, GRP (Zhu, Anwer, Huang, & Mathieu, 2018) Thermal field, geometric accuracy Geometry GRP, Bayesian calibration (Li, Jin, & Hang, 2018) Structural reliability, mechanical properties RRAG Kriging surrogate model combined with modified classical laminate theory, Monte Carlo simulation (Liu & Wang, 2016) Tensile strength LT, RA, infill density Latin Hypercube Sampling with Kriging, Polynomial Response Surface and Radial Basis Function. Sobol's method and Global Sensitivity analysis (Jothibabu & Kumar, 2018) Geometrical accuracy Shape-independent error, shape-dependent error Linear regression (Cheng, Tsung, & Wang, 2017) Surface roughness Offline (LT, extrusion temperature, feed/flow ratio of nozzle), online (max. build plate and extruder vibration, skewness of extruder vibration, standard deviation of build plate temperature, mini melt pool temperature) Classification and regression trees (CART), random vector functional link network (RVFL), ridge regression (RR), support vector regression (SVR), random forest (RF), AdaBoost (Li, Zhang, Shi, & Wu, 2019) . Response of Interest . Parameters investigated . Strategies and methodology used to achieve design goals . Reference . Toolpath planning Reducing geometric errors (voids/aperture errors), improve printability of features Adaptive contour parallel toolpath, bead width Level-set method for path planning and GRP for determination of bead geometry (Xiong et al., 2019) Dimensional accuracy, tensile properties Infill density, infill pattern, extrusion temperature, LT L9 orthogonal array with Taguchi method (Qattawi, 2018) Tensile strength Infill density, extrusion temperature GA-ANN (genetic algorithm artificial neural network) (Yadav, Chhabra, Garg, Ahlawat, & Phogat, 2020) Strength-to-weight ratio Infill pattern, infill density Numerical modeling and ANN (Zhou, Hsieh, & Ting, 2018) Optimization Surface roughness Build time, PBO NSGA-II (Pandey et al., 2004) PBO PBO NSGA-II, Kriging, TOPSIS (Khodaygan & Golmohammadi, 2018) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature NSGA-II, TOPSIS, (Alizadeh et al., 2020) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature Karush–Kulh–Tucker (KKT)-based non-linear programming optimization (Tian et al., 2019) Flexural strength, impact strength PBO, LT, RA, RW, RRAG Gray relation, Taguchi method (Liu et al., 2017) Tensile strength, flexural strength, impact strength LT, PBO, RA, RW, RRAG Response surface methodology using FCCD design matrix (Sood, Ohdar, & Mahapatra, 2010) Compressive Strength LT, PBO, RA, RW, RRAG Quantum-behaved particle swarm optimization with FCCD design matrix (Sood, Ohdar, & Mahapatra, 2012) Geometrical accuracy (length, width, thickness) LT, RRAG, RA, PBO, RW, number of contours RSM with I-optimality design matrix (Mohamed, Masood, & Bhowmik 2016b) Compressive strength LT, PBO, RA Multigene genetic programming (MGGP) and general regression neural network (GRNN) (Panda, Bahubalendruni, & Biswal, 2015) Creep compliance and recoverable compliance LT, RRAG, RA, PBO, RW Definitive screening design, multilayer perceptron neural network (Mohamed, Masood, & Bhowmik, 2017) Dimensional accuracy, surface precision Nozzle temperature, printing speed, print spacing, filament feed rate IPSO-BPNN (improved particle swarm optimization with back propagation neural network) (Yan, 2020) Tensile strength LT, infill density, printing speed, RA ANOVA and adaptive neuro-fuzzy inference system (ANFIS) (Barrrionuevo & Ramos-Grez, 2019) Tensile strength Nozzle temperature, LT, RA Artificial neural network – Levenberg Marquardt (ANN) (Bayraktar, Uzun, Çakiroğlu, & Guldas, 2017) Dimensional accuracy (length, width and thickness) LT, PBO, infill density, number of contours RSM – ANN, ANN, ANN-GA (Deswal, Narang, & Chhabra, 2019) Surface roughness PBO Radial basis function neural network (RBFNNs), imperialist competitive algorithm (ICA), compare against (Vahabli & Rahmati, 2016) Dimensional errors, warp deformation, build time Print speed, LT, feed rate, RW compensation factor RSM + fuzzy inference system (FIS), using uniform 17 array (Peng et al., 2014) Dimensional accuracy (length, width, thickness) LT, PBO, RA, RW, AG L27 Taguchi method + fuzzy inference system (Padhi et al., 2017) Dynamic mechanical properties (loss compliance and storage compliance) LT, RRAG, RA, PBO, RW, number of contours IV-optimality RSM, multilayer feed-forward neural networks (MFNN) (Mohamed, Masood, & Bhowmik, 2016a) Compressive strength LT, PBO, RA, RW, RRAG M5′-genetic programming (Garg, Tai, Lee, & Savalani, 2014) Compared against SVR, ANFIS Cost of optimization of parameters Mechanical strength, surface roughness, etc. Fully connected neural network (FCNN) (Zhou, Hsieh, & Wang, 2019) Surface roughness LT, print speed, extruder temperature, outer shell print speed RSM, particle swarm optimization, symbiotic organism search (Saad, Nor, Baharudin, Zakaria, & Aiman, 2019) Prediction Porosity Bonding between adjacent toolpath ANN (Haghighi & Li, 2020) Geometrical accuracy, surface roughness Offline process settings (feed/flow ratio of nozzle), in situ process variables (acceleration) Quantitative and qualitative (QQ) models (Sun, Rao, Kong, Deng, & Jin, 2017) Geometrical accuracy Process parameters, geometry of AM parts Bayesian inference (Sabbaghi & Huang, 2018) Geometrical accuracy Process parameters, geometry of AM parts Adaptive Bayesian (Sabbaghi, Huang, & Dasgupta, 2018) Thermal profile Direction of print, number of layers Neural network (Roy & Wodo, 2020) In plane geometric accuracy Geometry Multitask learning, GRP (Zhu, Anwer, Huang, & Mathieu, 2018) Thermal field, geometric accuracy Geometry GRP, Bayesian calibration (Li, Jin, & Hang, 2018) Structural reliability, mechanical properties RRAG Kriging surrogate model combined with modified classical laminate theory, Monte Carlo simulation (Liu & Wang, 2016) Tensile strength LT, RA, infill density Latin Hypercube Sampling with Kriging, Polynomial Response Surface and Radial Basis Function. Sobol's method and Global Sensitivity analysis (Jothibabu & Kumar, 2018) Geometrical accuracy Shape-independent error, shape-dependent error Linear regression (Cheng, Tsung, & Wang, 2017) Surface roughness Offline (LT, extrusion temperature, feed/flow ratio of nozzle), online (max. build plate and extruder vibration, skewness of extruder vibration, standard deviation of build plate temperature, mini melt pool temperature) Classification and regression trees (CART), random vector functional link network (RVFL), ridge regression (RR), support vector regression (SVR), random forest (RF), AdaBoost (Li, Zhang, Shi, & Wu, 2019) Open in new tab Table 2: Overview of the literature that utilizes data-driven strategy in process domain. . Response of Interest . Parameters investigated . Strategies and methodology used to achieve design goals . Reference . Toolpath planning Reducing geometric errors (voids/aperture errors), improve printability of features Adaptive contour parallel toolpath, bead width Level-set method for path planning and GRP for determination of bead geometry (Xiong et al., 2019) Dimensional accuracy, tensile properties Infill density, infill pattern, extrusion temperature, LT L9 orthogonal array with Taguchi method (Qattawi, 2018) Tensile strength Infill density, extrusion temperature GA-ANN (genetic algorithm artificial neural network) (Yadav, Chhabra, Garg, Ahlawat, & Phogat, 2020) Strength-to-weight ratio Infill pattern, infill density Numerical modeling and ANN (Zhou, Hsieh, & Ting, 2018) Optimization Surface roughness Build time, PBO NSGA-II (Pandey et al., 2004) PBO PBO NSGA-II, Kriging, TOPSIS (Khodaygan & Golmohammadi, 2018) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature NSGA-II, TOPSIS, (Alizadeh et al., 2020) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature Karush–Kulh–Tucker (KKT)-based non-linear programming optimization (Tian et al., 2019) Flexural strength, impact strength PBO, LT, RA, RW, RRAG Gray relation, Taguchi method (Liu et al., 2017) Tensile strength, flexural strength, impact strength LT, PBO, RA, RW, RRAG Response surface methodology using FCCD design matrix (Sood, Ohdar, & Mahapatra, 2010) Compressive Strength LT, PBO, RA, RW, RRAG Quantum-behaved particle swarm optimization with FCCD design matrix (Sood, Ohdar, & Mahapatra, 2012) Geometrical accuracy (length, width, thickness) LT, RRAG, RA, PBO, RW, number of contours RSM with I-optimality design matrix (Mohamed, Masood, & Bhowmik 2016b) Compressive strength LT, PBO, RA Multigene genetic programming (MGGP) and general regression neural network (GRNN) (Panda, Bahubalendruni, & Biswal, 2015) Creep compliance and recoverable compliance LT, RRAG, RA, PBO, RW Definitive screening design, multilayer perceptron neural network (Mohamed, Masood, & Bhowmik, 2017) Dimensional accuracy, surface precision Nozzle temperature, printing speed, print spacing, filament feed rate IPSO-BPNN (improved particle swarm optimization with back propagation neural network) (Yan, 2020) Tensile strength LT, infill density, printing speed, RA ANOVA and adaptive neuro-fuzzy inference system (ANFIS) (Barrrionuevo & Ramos-Grez, 2019) Tensile strength Nozzle temperature, LT, RA Artificial neural network – Levenberg Marquardt (ANN) (Bayraktar, Uzun, Çakiroğlu, & Guldas, 2017) Dimensional accuracy (length, width and thickness) LT, PBO, infill density, number of contours RSM – ANN, ANN, ANN-GA (Deswal, Narang, & Chhabra, 2019) Surface roughness PBO Radial basis function neural network (RBFNNs), imperialist competitive algorithm (ICA), compare against (Vahabli & Rahmati, 2016) Dimensional errors, warp deformation, build time Print speed, LT, feed rate, RW compensation factor RSM + fuzzy inference system (FIS), using uniform 17 array (Peng et al., 2014) Dimensional accuracy (length, width, thickness) LT, PBO, RA, RW, AG L27 Taguchi method + fuzzy inference system (Padhi et al., 2017) Dynamic mechanical properties (loss compliance and storage compliance) LT, RRAG, RA, PBO, RW, number of contours IV-optimality RSM, multilayer feed-forward neural networks (MFNN) (Mohamed, Masood, & Bhowmik, 2016a) Compressive strength LT, PBO, RA, RW, RRAG M5′-genetic programming (Garg, Tai, Lee, & Savalani, 2014) Compared against SVR, ANFIS Cost of optimization of parameters Mechanical strength, surface roughness, etc. Fully connected neural network (FCNN) (Zhou, Hsieh, & Wang, 2019) Surface roughness LT, print speed, extruder temperature, outer shell print speed RSM, particle swarm optimization, symbiotic organism search (Saad, Nor, Baharudin, Zakaria, & Aiman, 2019) Prediction Porosity Bonding between adjacent toolpath ANN (Haghighi & Li, 2020) Geometrical accuracy, surface roughness Offline process settings (feed/flow ratio of nozzle), in situ process variables (acceleration) Quantitative and qualitative (QQ) models (Sun, Rao, Kong, Deng, & Jin, 2017) Geometrical accuracy Process parameters, geometry of AM parts Bayesian inference (Sabbaghi & Huang, 2018) Geometrical accuracy Process parameters, geometry of AM parts Adaptive Bayesian (Sabbaghi, Huang, & Dasgupta, 2018) Thermal profile Direction of print, number of layers Neural network (Roy & Wodo, 2020) In plane geometric accuracy Geometry Multitask learning, GRP (Zhu, Anwer, Huang, & Mathieu, 2018) Thermal field, geometric accuracy Geometry GRP, Bayesian calibration (Li, Jin, & Hang, 2018) Structural reliability, mechanical properties RRAG Kriging surrogate model combined with modified classical laminate theory, Monte Carlo simulation (Liu & Wang, 2016) Tensile strength LT, RA, infill density Latin Hypercube Sampling with Kriging, Polynomial Response Surface and Radial Basis Function. Sobol's method and Global Sensitivity analysis (Jothibabu & Kumar, 2018) Geometrical accuracy Shape-independent error, shape-dependent error Linear regression (Cheng, Tsung, & Wang, 2017) Surface roughness Offline (LT, extrusion temperature, feed/flow ratio of nozzle), online (max. build plate and extruder vibration, skewness of extruder vibration, standard deviation of build plate temperature, mini melt pool temperature) Classification and regression trees (CART), random vector functional link network (RVFL), ridge regression (RR), support vector regression (SVR), random forest (RF), AdaBoost (Li, Zhang, Shi, & Wu, 2019) . Response of Interest . Parameters investigated . Strategies and methodology used to achieve design goals . Reference . Toolpath planning Reducing geometric errors (voids/aperture errors), improve printability of features Adaptive contour parallel toolpath, bead width Level-set method for path planning and GRP for determination of bead geometry (Xiong et al., 2019) Dimensional accuracy, tensile properties Infill density, infill pattern, extrusion temperature, LT L9 orthogonal array with Taguchi method (Qattawi, 2018) Tensile strength Infill density, extrusion temperature GA-ANN (genetic algorithm artificial neural network) (Yadav, Chhabra, Garg, Ahlawat, & Phogat, 2020) Strength-to-weight ratio Infill pattern, infill density Numerical modeling and ANN (Zhou, Hsieh, & Ting, 2018) Optimization Surface roughness Build time, PBO NSGA-II (Pandey et al., 2004) PBO PBO NSGA-II, Kriging, TOPSIS (Khodaygan & Golmohammadi, 2018) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature NSGA-II, TOPSIS, (Alizadeh et al., 2020) Energy consumption, part geometrical accuracy LT, printing speed, extrusion temperature Karush–Kulh–Tucker (KKT)-based non-linear programming optimization (Tian et al., 2019) Flexural strength, impact strength PBO, LT, RA, RW, RRAG Gray relation, Taguchi method (Liu et al., 2017) Tensile strength, flexural strength, impact strength LT, PBO, RA, RW, RRAG Response surface methodology using FCCD design matrix (Sood, Ohdar, & Mahapatra, 2010) Compressive Strength LT, PBO, RA, RW, RRAG Quantum-behaved particle swarm optimization with FCCD design matrix (Sood, Ohdar, & Mahapatra, 2012) Geometrical accuracy (length, width, thickness) LT, RRAG, RA, PBO, RW, number of contours RSM with I-optimality design matrix (Mohamed, Masood, & Bhowmik 2016b) Compressive strength LT, PBO, RA Multigene genetic programming (MGGP) and general regression neural network (GRNN) (Panda, Bahubalendruni, & Biswal, 2015) Creep compliance and recoverable compliance LT, RRAG, RA, PBO, RW Definitive screening design, multilayer perceptron neural network (Mohamed, Masood, & Bhowmik, 2017) Dimensional accuracy, surface precision Nozzle temperature, printing speed, print spacing, filament feed rate IPSO-BPNN (improved particle swarm optimization with back propagation neural network) (Yan, 2020) Tensile strength LT, infill density, printing speed, RA ANOVA and adaptive neuro-fuzzy inference system (ANFIS) (Barrrionuevo & Ramos-Grez, 2019) Tensile strength Nozzle temperature, LT, RA Artificial neural network – Levenberg Marquardt (ANN) (Bayraktar, Uzun, Çakiroğlu, & Guldas, 2017) Dimensional accuracy (length, width and thickness) LT, PBO, infill density, number of contours RSM – ANN, ANN, ANN-GA (Deswal, Narang, & Chhabra, 2019) Surface roughness PBO Radial basis function neural network (RBFNNs), imperialist competitive algorithm (ICA), compare against (Vahabli & Rahmati, 2016) Dimensional errors, warp deformation, build time Print speed, LT, feed rate, RW compensation factor RSM + fuzzy inference system (FIS), using uniform 17 array (Peng et al., 2014) Dimensional accuracy (length, width, thickness) LT, PBO, RA, RW, AG L27 Taguchi method + fuzzy inference system (Padhi et al., 2017) Dynamic mechanical properties (loss compliance and storage compliance) LT, RRAG, RA, PBO, RW, number of contours IV-optimality RSM, multilayer feed-forward neural networks (MFNN) (Mohamed, Masood, & Bhowmik, 2016a) Compressive strength LT, PBO, RA, RW, RRAG M5′-genetic programming (Garg, Tai, Lee, & Savalani, 2014) Compared against SVR, ANFIS Cost of optimization of parameters Mechanical strength, surface roughness, etc. Fully connected neural network (FCNN) (Zhou, Hsieh, & Wang, 2019) Surface roughness LT, print speed, extruder temperature, outer shell print speed RSM, particle swarm optimization, symbiotic organism search (Saad, Nor, Baharudin, Zakaria, & Aiman, 2019) Prediction Porosity Bonding between adjacent toolpath ANN (Haghighi & Li, 2020) Geometrical accuracy, surface roughness Offline process settings (feed/flow ratio of nozzle), in situ process variables (acceleration) Quantitative and qualitative (QQ) models (Sun, Rao, Kong, Deng, & Jin, 2017) Geometrical accuracy Process parameters, geometry of AM parts Bayesian inference (Sabbaghi & Huang, 2018) Geometrical accuracy Process parameters, geometry of AM parts Adaptive Bayesian (Sabbaghi, Huang, & Dasgupta, 2018) Thermal profile Direction of print, number of layers Neural network (Roy & Wodo, 2020) In plane geometric accuracy Geometry Multitask learning, GRP (Zhu, Anwer, Huang, & Mathieu, 2018) Thermal field, geometric accuracy Geometry GRP, Bayesian calibration (Li, Jin, & Hang, 2018) Structural reliability, mechanical properties RRAG Kriging surrogate model combined with modified classical laminate theory, Monte Carlo simulation (Liu & Wang, 2016) Tensile strength LT, RA, infill density Latin Hypercube Sampling with Kriging, Polynomial Response Surface and Radial Basis Function. Sobol's method and Global Sensitivity analysis (Jothibabu & Kumar, 2018) Geometrical accuracy Shape-independent error, shape-dependent error Linear regression (Cheng, Tsung, & Wang, 2017) Surface roughness Offline (LT, extrusion temperature, feed/flow ratio of nozzle), online (max. build plate and extruder vibration, skewness of extruder vibration, standard deviation of build plate temperature, mini melt pool temperature) Classification and regression trees (CART), random vector functional link network (RVFL), ridge regression (RR), support vector regression (SVR), random forest (RF), AdaBoost (Li, Zhang, Shi, & Wu, 2019) Open in new tab To source for literature, the following procedure was adopted. Keywords such as “data-driven,” “additive manufacturing,” “surrogate modeling,” “Fused Filament Fabrication,” “FFF embodiment design,” and “FFF process optimization” were used. Search for the literature was carried out in major database such as Scopus®. In total, approximately 100 articles consisting of journal articles and review papers were found, and the final list was narrowed down into 46, after removing articles with repeated methodology. The papers in the final list were then further classified into the two main categories, namely embodiment design (10 articles) and process parameter (36 articles). To ensure that the literature is state-of-the-art, Fig. 4 provides a plot of the number of articles over the different time periods. Most of the literature used (>80%) is less than 5 years old. This could also signify that research interest drastically increased over the recent years. The literature is summarized in Tables 1 and 2 to provide readers with an overview of the strategies used, respectively, in two categories. Datasets in most of literature cited can be directly obtained from the journal articles or requested from the authors for future work and verification. Figure 4: Open in new tabDownload slide The number of literature used for review, plotted against publication time for (a) embodiment design domain and (b) process domain. Figure 4: Open in new tabDownload slide The number of literature used for review, plotted against publication time for (a) embodiment design domain and (b) process domain. As the nomenclature for various parameters is not standardized, in the following paragraphs, nomenclature used is defined as follows: Layer thickness (LT): The layer height of each of the slices. Also known as slice thickness. PBO: Defines the orientation of a part in three-dimensional Euclidean space, and how the part is being built in the AM process. Printing speed: The speed with which the print head moves when depositing material onto the build chamber. Feed rate: Defines the amount of filament per unit time that is being pushed into the nozzle. Infill: The bulk material that makes up the part. Primarily comprises rasters. Extrusion temperature: Temperature of the nozzle that the FFF material is extruded at. Also known as nozzle temperature. For the following list of process parameters, refer to Fig. 1 for the nomenclature. Raster-to-raster air gap (RRAG): Defines the amount of spacing between adjacent rasters. Negative value indicates that there is overlapping between rasters, and positive value indicates separation. Raster width (RW): Defines the width of the raster that is extruded. Raster angle (RA): Defines the direction of extruded raster with respect to an axis. Contour width: Defines the width of perimeter surrounding the part. 3.1 Embodiment design domain In the design phase, data-driven strategies are often utilized to aid in the design of AM parts. Functional requirements of the end-use parts are defined, and the AM designers have to ensure that the requirements are met through the design. Moreover, the designers need to have knowledge of DfAM for FFF to ensure that the designed parts are printable. 3.1.1 Design space exploration To effectively explore the design space, BN, a type of probabilistic graphical models, can be utilized by AM designers. BN is commonly used to represent and provide solutions to highly complex problems that use probability as a representation of uncertainty and aid users make trade-off decisions. A directed acyclic graph is used to show the joint probability distribution and conditional dependency of variables (Larrañaga & Moral, 2011; Larrañaga, Karshenas, Bielza, & Santana, 2013). Due to the aforementioned dependencies and complexities observed in AM, BN has been frequently utilized in embodiment design of AM parts. Wang, Chen, Houqi, and Li (2018) proposed a knowledge management system that utilized discrete BN to model process and material domain, design-related and part-related domain at different design stages, and was able to cater to designers with different levels of experience with AM. Similarly, Xiong et al. (2019) introduced a BN classifier-based approach that allowed for preliminary evaluation of the printability and parameters required. Process parameters were classified into feasible and infeasible design region according to the design requirements. Subsequently, for detailed refinement and design, GRP model was used, and further refinement was performed using Markov chain Monte Carlo resampling method. Researchers have also proposed novel approaches for product design to exploit the freedom of design for AM. Ko et al. (2015) proposed a framework for design for AM-facilitated and personalized products that captured design requirements. Yao et al. (2016) and Yao, Moon, and Bi (2017) introduced a framework for trade-off analysis to aid designers into making informed decisions, such as maximizing customers’ perceived utilities in different market segments of a variable platform design for a given cost constraint. 3.1.2 Architected material Architected materials are structures whose behavior is governed by the structure rather than the composition of the material. Common examples of architected materials are cellular structures, lattice structures, and honeycomb structures. Research into parts with the architected materials is popularized due to advent of AM, which allows for freedom of design and material to be placed at specific locations where it is required. The architected material has the potential of greatly reducing the weight of an AM part and material wastage, and imbues structures with unique properties such as supercompressibility and anisotropic stiffness (Jackson, Liu, Patrikalakis, Sachs, & Cima, 1999; Chu et al., 2008; Coulais, Teomy, De Reus, Shokef, & Van Hecke, 2016; Overvelde, Weaver, Hoberman, & Bertoldi, 2017; Panesar et al., 2018; Nguyen & Choi, 2020). Chu et al. (2008) proposed a framework to design parts with cellular structure that considered objective functions such as minimization of weight, compliance distribution, and motion. To effectively explore the design space for determining the topology and dimensions of the structure, particle swarm optimization was utilized, based on physics-based surrogate modeling. Bessa et al. (2019) utilized Bayesian regression and classification to design and tune the geometry of the structure-dominated material to explore structures that are capable of unique properties and functionality, such as high compressibility structures without permanent deformation. Sensitivity analysis was carried out to identify the printing parameters that affected the performance, and the design space was segregated into feasible and infeasible design regions, with the ability for prediction of response. Panda et al. (2018) proposed a process flow for the optimization of FFF honeycomb structures for compressive properties (more specifically, yield strength and modulus of elasticity). Various numerical modeling techniques, such as genetic programming and response surface regression, were used to model and predict the compressive properties. Levasseur et al. (2012) explored printing of FFF prisms for bone surrogate. In their study, a mixed experimental design is used to investigate the effect of various structural parameters for the embodiment design such as wall thickness and strut orientation, and the effect of compressive properties of these prisms. 3.1.3 Design rules As established, DfAM is key in ensuring the printability of the FFF part. Jiang et al. (2019) explored the printable bridge length to reduce the need for support materials for overhangs. Process parameters, such as print speed and print temperature, were explored using an L32 orthogonal array, and a back propagation neural network was used to estimate the longest bridge length printable with acceptable deformation. Jiang et al. (2018) investigate the limit of the printability limit of overhang structures, by varying process parameters such as extrusion temperature, fan speed, and printing speed. The images of resultant overhang structures are digitally analysed to determine the degree of deformation and the surface quality to determine the acceptable printability limit for FFF overhang structures. In summary, Bayes theorem is commonly utilized by researchers to support AM designers in designing space exploration and finalizing the embodiment design. Benefits of adopting BN allow designers to represent the information and the causal relationships between design variables, which combines experts’ knowledge and empirical data. Subsequently, forward inference can be conducted with prior knowledge, which enables prediction of properties of the printed parts, such as the part mechanical strength. These predictions often have a degree of uncertainty built into it. Lastly, BN enables learning, whereby prior information is updated once there is additional information or observation presented (Koller & Friedman, 2009; Zhang & Thai, 2016). To search for optimal design of cellular structures and to explore DfAM rules, surrogate modeling is commonly utilized. 3.2 Process domain At this phase, data-driven strategies are often utilized to dial in the required process parameters to commence the printing process. The strategies include planning for the toolpath, optimization, managing the uncertainty, and prediction of the properties of a final part. As discussed in prior paragraphs, surrogate modeling is one of the key data-driven design strategies to resolve issues related to enlarged design space. Computationally and time-consuming multiphysics simulation can be approximated by simpler surrogate models that require less resources and time to run but still retain sufficient fidelity to predict an accurate response (Tapia, Elwany, & Sang, 2016; Tapia, Khairallah, Matthews, King, & Elwany, 2018; Wang et al., 2019; Xiong et al., 2019; Wang, Zhu, Fuh, Zhang, & Yan, 2020). Other benefits of using the meta modeling include ability for quick exploration of design space and allowing for integration of domain-dependent data from different simulation software or codes (Simpson, Poplinski, Koch, & Allen, 2001). Typically, the processing of the surrogate modeling encompasses the three steps. A sampling plan is mapped out to define experiments with investigating process variables. Subsequently, the experiments are conducted according to the defined sampling plan and observed. Lastly, the experiment data are analysed, and manipulations such as dimension reduction can be performed to reduce the complexity of the analysis (Wang & Shan, 2007; Forrester et al., 2008; Roy & Wodo, 2020; Johnson & Wichern 2020). Review papers, such as the ones presented by A Garg, Tai, and Savalani (2014) and Sheoran and Kumar (2020), provide a comprehensive review of the various optimizations of FFF process parameters through empirical modeling. This section will not go into details to discuss the results of each of the modeling, as shown in the two papers. Instead, the following paragraphs will dwell on the strategies various researchers used to achieve their goal of exploring the design space. 3.2.1 Toolpath planning Toolpath planning is an important aspect of FFF process as it dictates how the rasters and contour will be laid, which have a significant impact on the mechanical properties of the part. Due to the process characteristics, voids are significantly more prevalent and can adversely impact the mechanical performance (Eiliat & Urbanic, 2018, 2019; Medellin-Castillo & Zaragoza-Siqueiros, 2019). Many investigations have been performed to exploit the toolpath to improve mechanical properties, such as surface roughness and tensile and flexural strength (Avdeev et al., 2019; Kubalak, Wicks, & Williams, 2019; Wang et al., 2018). Xiong et al. (2019) proposed a process planning approach for adaptive contour parallel toolpath with variable bead width using level-set based algorithm with implicit approach, for reduction in the voids due to geometric errors and improving printability of features. GRP was utilized to determine the bead geometry, whereas level-set based algorithm was utilized to determine the optimal toolpath for a given part geometry. Eiliat and Urbanic (2019) introduced a novel method to predict and reduce the occurrence of contiguous voids. Such voids can adversely impact the strength and performance of FFF parts. An exhaustive search optimization is used to determine an optimal set of parameters, including PBO, to reduce the presence of contiguous voids within the FFF parts. Through the toolpath planning, cellular structures can be obtained through manipulating the infill pattern and density. However, cellular structures created through toolpath planning are often restricted to within the slicing plane. Zhou et al. (2018) explored the effect of infill pattern and density on the ultimate tensile strength (UTS) of the part. Several infill patterns, such as rectilinear, honeycomb, and triangular, were investigated in the study. Artificial neural network (ANN) was used to model and predict the UTS-to-weight ratio. A mesostructure numerical model was proposed and a knowledge-based library was constructed for the prediction of elastic modulus and infill patterns. Qattawi (2018) examined the relationship between infill patterns and process parameters on the mechanical properties and dimensional accuracy. Three different infill patterns, namely rectilinear, triangular, and hexagonal, were used. A design of experiment was performed to determine the optimal infill and process parameters for strength and dimensional accuracy. Yadav et al. (2020) investigated the effect of infill parameters and extrusion temperature on across the different types of material. Hybrid genetic algorithm and ANN were utilized to search for the optimal process parameters for tensile strength optimization. 3.2.2 Optimization Due to many conflicting objectives for optimization of FFF part properties, AM designers often are required to perform trade-off analysis to select optimal parameters for a particular design. This section explores various methodologies proposed by researchers to achieve the optimal parameters for a set of given objective functions. To guide AM designers to choose the optimal PBO, Pandey et al. (2004) proposed a methodology for determining optimal build orientation by minimizing two contradicting objective functions. In this early work, the authors modeled build time and part surface roughness as a function of part geometry and build orientation, and using non-dominated sorting algorithm (NSGA-II), a Pareto optimal front was determined for a decision maker to select the final PBO. Khodaygan and Golmohammadi (2018) improved on the surrogate modeling of the surface roughness and the build time by using Kriging method. NSGA-II was then utilized to identify the Pareto-optimum solutions to minimize both the build time and surface roughness, and a technique for order of preference by similarity to ideal solution (TOPSIS) was utilized to distinguish the optimal build orientation. Vahabli and Rahmati (2016) utilized radial basis function neural network (RBFNN) for the prediction of surface quality of FFF parts by varying the PBO. The prediction accuracy of the proposed method was compared against existing models in literature, and the RBFNN model is superior in terms of prediction accuracy. Researchers have also investigated minimization of energy consumption of AM process for sustainable manufacturing, while maximizing part properties such as geometrical accuracy. Alizadeh et al. (2020) developed a methodology for optimization between energy consumption and part geometry accuracy. Regression analysis was used to determine the effect of printing parameters on the responses. Principal component analysis was used to reduce the dimensions of the response function, and NSGA-II was then used to determine the Pareto front of the minimizing of the response variables and TOPSIS to select the final optimal parameter set for the best geometrical accuracy with the least energy consumed from the Pareto front. Tian et al. (2019) also examined the energy consumption and its relationship between the process parameters and part quality. In the proposed approach, Karush–Kulh–Tucker (KKT)-based non-linear programming optimization was used to determine the optimized solutions between the three process parameters. Regression analysis was first performed to formulate the regression coefficients and KKT was then used to optimize the solution, and the response of geometrical accuracy was then categorized into regions achievable as a function of studied parameters. Due to the large number of variable process parameters, investigation into parameter optimization for a given mechanical property is commonly conducted by researchers. These approaches can be divided into two categories, namely statistical approaches and artificial intelligence approaches (An et al., 2015). Statistical approach, such as robust parameter design and response surface, is discussed first, followed by an artificial intelligence approach. A robust parameter design methodology, or more commonly known as Taguchi method, is to optimize a response and reduce the variance through adjusting process parameters (Kackar, 1985; Vining & Myers, 1990). Due to the efficiency of the methodology, robust parameter design is popular among literature that seeks to optimize the process parameters for a given response (Huynh, Nguyen, Ha, & Thai, 2017; Srivastava & Rathee, 2018; Vishwas, Basavaraj, & Vinyas, 2018). As there are many pieces of literature that utilized Taguchi method, as such, only selected literature will be discussed in this section. Liu et al. (2017) utilized a combination of Gray relation analysis and Taguchi method to study the influence of parameters such as deposition orientation and LT on the effect on tensile, flexural, and impact strength performance. From the Taguchi method, it was determined that the three different performance indexes were optimized by different parameter sets, and Gray relation analysis was performed to identify the parameter set that optimized the three indexes simultaneously. Response surface methodology is another popular statistical approach for process parameter optimization. Sood et al. (2010) studied the effect of five process parameters on tensile, flexural, and impact strength using central composite design (CCD). Similar to Liu et al. (2017), the authors proposed a desirability function for optimizing three indexes simultaneously. Mohamed et al. (2016b) studied the process parameters such as LT, air gap, and the number of contours on the geometrical accuracy in the X, Y, and Z axes using I-optimality optimal design, which minimized the integrated prediction variance over the region of interest. Desirability index was used to optimize the geometrical accuracy in all three different axes simultaneously. Neural network is a type of artificial intelligence approach that is also commonly utilized in a surrogate modeling technique for the optimization of process parameters. Sood et al. (2012) used a Face Centred Composite Design (FCCD) for gathering of experimental data and particle swarm optimization to determine optimal process parameters for compressive strength. To aid in the convergence of the neural network, Schrodinger equation replaced the position and velocity in conventional PSO algorithms. Panda et al. (2015) maximized the compressive strength utilizing multigene genetic programming and general regression neural network. Mohamed et al. (2017) utilized definitive screening design to explore the effect of process parameters on the part creep and recoverable compliance properties. Subsequently, multilayer perceptron neural network and desirability function were utilized to optimize both compliance responses. Yan (2020) proposed an optimization algorithm based on neural network for the optimization of FFF process parameters to achieve geometrical accuracy and surface accuracy. Barrrionuevo and Ramos-Grez (2019) introduced adaptive neuro-fuzzy inference system (ANFIS) for optimizing process parameters for tensile properties. Bayraktar et al. (2017) determined the process parameters required to maximize the tensile strength of FFF parts using ANN. Deswal et al. (2019) identified the significant process parameters, and utilized methodology such as ANN-GA, RSM-GA, and ANN and compared the performance between the three different models for the optimization of geometric accuracy. Zhou et al. (2019) introduced a multifidelity framework that utilizes low-fidelity simulation model, in conjunction with high-fidelity experimental model to iteratively perform optimization on the AM processes. The proposed framework was to reduce the cost of performing additional experimentation and is sufficiently robust against noise. Saad et al. (2019) propose an optimization methodology for surface roughness for FFF part using process parameters. Regression model is derived using RSM method, and subsequently, the optimization of surface roughness is carried out using PSO, SOS, and Response Surface Methodology (RSM), respectively, and their accuracy is compared against the experimental values. 3.2.3 Prediction The prediction of part properties with a given set of parameters is critical in ensuring that a part is able to meet functional requirements, without having to print the physical part. Sun et al. (2017) proposed a framework whereby part quality variables and process variables were modeled by quantitative and qualitative (QQ) models. Scalar offline process setting variables and in situ process variables were used as inputs, and then an initial estimator was used in combination with hierarchical nonnegative garrote for the QQ modeling. The predictive model was then validated for the prediction of part dimensional accuracy (quantitative response) and surface roughness (quantitative response) for FFF. Wu et al. (2018) developed a predictive approach to surface roughness using random forest algorithm by processing and analysing data from accelerometers and infrared sensors mounted on the printer. Similarly, Li et al. (2019) utilized both online and offline methods to predict part surface roughness. Offline parameters consist of the print process parameters such as extrusion temperature, whereas online parameters consist of vibration and temperature measurements using sensors. These data are then used to train various ML algorithms and their accuracies are compared. Liu and Wang (2016) proposed a stochastic multilevel modeling to predict mechanical properties for FFF parts. In the proposed framework, uncertainties from material, structural, and modeling are combined with a surrogate model of enhanced laminate theory to predict the structural reliability and mechanical strength. Jothibabu and Kumar (2018) modeled the tensile performance of the printed parts by varying the infill density and process parameters such as LT and RA. The performances of surrogate models using Kriging, polynomial response surface and radial basis function were evaluated and compared. Zhu et al. (2018) proposed a model to predict in-plane geometric deviations. The proposed model identified three transformation types that might cause geometric deviations: translation, rotation, and material shrinkage. GRP was used to model the variations in the geometric deviations by incorporating knowledge transferred from other shapes. Similar knowledge transfer approaches were suggested by Sabbaghi and Huang (2018). They proposed a framework that allowed for model transfers across to a new AM process with limited experimental runs through characterizing unobserved lurking variable’s effect on the process mean and applied a compensation plan to the new process, which was derived by utilizing Bayesian inference method instead of using existing methods such as ANOVA and TrAdaBoost. Sabbaghi et al. (2018) developed an adaptive Bayesian methodology to predict the in-plane geometrical deviation. The proposed methodology involved Bayesian discrepancy measure to identify the new features as compared to sampled shapes. Trends were then subsequently identified and modeled from the measure, and lastly, a hierarchical structure was specified to account for these trends and deviations. Similarly, Cheng et al. (2017) utilized transfer learning to predict the geometry accuracy for new shapes based on past generalized datasets. This is accomplished through modeling of shape-independent error and shape-specific error separately. Due to the influence of thermal profiles on the mechanical properties, researchers have also investigated predicting and modeling of the build chamber temperature profiles for FFF process. Roy and Wodo (2020) proposed a surrogate modeling of thermal profiles for FFF using ANN. The trained model was able to predict the thermal profile of regular cuboids directly from the GCode data. Li et al. (2018) introduced a framework whereby a Gaussian surrogate model was constructed using data from high-fidelity simulations to model and predict layer-to-layer thermal field for a printed part. Bayesian calibration was used to correct for model discrepancies and account for potential unknown parameters from the simulation. Haghighi and Li (2020) developed a hybrid physics-based and data-driven approach to model the porosity variation and bonding between filaments. The proposed methodology improved the prediction accuracy of the bonding and porosity in FFF parts, by calibrating the physics-based model with empirical data, to account for multiphysics phenomenon. To sum up, due to the costly empirical studies and simulations, many researchers have turned to the strategy of using surrogate modeling for the prediction of response equivalent to an original model at a fraction of the cost without sacrificing accuracy (Roy & Wodo, 2020). Surrogate models allow for a quick validation of the printing parameter and reduction of the experimental runs required to fine-tune the parameters (Murphy, Imediegwu, Hewson, Santer, & Muir, 2019; Zhang, Moon, Ngo, Tou, & Yusoff, 2019; Olleak & Xi, 2020). Moreover, surrogate model often only requires relatively small and limited sample sizes to generate a prediction (Olleak & Xi, 2020). 3.3 Research gaps and opportunities AM enables creation of unique structures and structures with unprecedented mechanical properties through optimizing geometry (architected materials) and process parameters. However, most literature considers embodiment design and process parameters separately, without providing due consideration to both aspects. As a result, the conundrum shown in Fig. 2 may occur. Voids present in the part can cause premature failure and interraster delamination, decreasing the observed mechanical strength and providing inaccurate characterization results. As a result, characterization results may not be representative of the final part performance. Lack of AM specific standards further exacerbates the problem identified in the previous paragraph. Standard which homogenize the characterization process of unique structures, such as the lattice structures and supercompressible metamaterial in Bessa et al. (2019), has yet to exist. As a result, the reliability and repeatability of such characterization efforts may be put in question. Another issue with current characterization methodology is the specificity of the results. An optimized model can be specific to a geometrical model or parameter set, which may not be transferrable to other geometries, which limit their usefulness. For example, optimized parameters are studied for tensile strength on ABS using a desktop printer that may not be directly applicable to other printers or materials. As such, once the printer is changed or a new material is utilized, the designers have to repeat the characterization process. Current research performed on transfer learning, such as an approach proposed by Sabbaghi & Huang (2018), has the potential to reduce the repetitive work. However, the research explored only geometrical accuracy. The research on this topic is still in an early stage with limited literature. Alternatively, Aboutaleb et al. (2017) proposed a novel method that leveraged on existing empirical data to reduce the experimentation runs required for characterization. However, there is room for development for this idea, such as utilizing a classifier to determine outlier datasets from literature, and utilizing an inference system to predict how variations between systems (e.g. beam distribution) can affect the response of interest. In addition, implementation has yet to be carried out in FFF. Current state-of-the-art DfAM rules for FFF still rely heavily on the knowledge that is learnt through trial and error, and there is a lack of a data-driven strategy to help designers implement DfAM rules to improve printability. As reviewed under Section 3.1, Jiang et al. (2018, 2019) established that design rules are based firmly on empirical data. On the other hand, literature such as Urbanic and Hedrick (2016) and Ahn et al. (2002) provides generic FFF design rules that may be based on experience and expert advices. These generic FFF design rules are lacking in terms of applicability, where changes in process parameters or embodiment design may nullify these recommendations. Moreover, as discussed by Huang et al. (2020), there are a wide variety of DfAM rules to be considered, whose current literature only covers a small subset. 4. Data-Driven Design Strategy for FFF In this section, a framework model is proposed to develop a data-driven design strategy that addresses the shortcoming aforementioned for FFF technique. Much of the knowledge is commonly resided within the experienced AM designer, as a form of tacit knowledge. The knowledge is difficult to access and will be no longer accessible, if the AM designer leaves and the knowledge not externalized and captured within a database. As such, there is a need for systematic handling of the knowledge to enable designers to tap on and exploit knowledge for AM part design and printing (Heisig, 2009). Ontologies have been proposed as a possible way for managing knowledge and data in AM, and providing decision making capabilities (Witherell et al., 2014; Roh et al., 2016; Sanfilippo, Belkadi, & Bernard, 2019). Commonly, in knowledge management frameworks, there are three key elements (Lin & Ha, 2015): Identification: define what is required and knowledge to collect and/or create. Management: arrangement of knowledge objects for ease of access, and enable AM designers to understand what they need and what they have in hand. Implementation: using of knowledge to support the completion of task, and knowledge can be revised or updated upon the completion of task. Although there are multiple pieces of literature that discuss and propose similar knowledge management frameworks, the frameworks are often limited in application or do not fully encompass the whole design process with data-driven methods. Majeed, Lv, and Peng (2019) proposed a big data-based analytics framework, which included a data mining and knowledge management framework for AM. However, the proposed framework needed details on how the mined data were utilized effectively to improve the quality and reduce the amount of experimentation and trial and error that AM designers required during process parameter selection or exploration of design space during an embodiment design phase. Zhou et al. (2018) implemented a knowledge-based library for predicting the elastic modulus for infill patterns utilizing ANN. However, the application was limited to predict the strength of the part with varying infill density. Nagarajan et al. (2019) proposed a hybrid learning network, using both theoretical knowledge and empirical data for modeling and exploration of manufacturing space, requiring fewer experimentation runs for characterization. However, the learning network is primarily based on the prediction of part performance by varying the process parameters, and does not discuss on the embodiment design aspect. Li et al. (2018) introduced a hybrid physics-based and data-driven-based modeling approach and utilize Bayesian calibration to calibrate the surrogate model. However, the framework was currently still applicable to the prediction of thermal properties. The common theme between the frameworks presented in various pieces of literature is focused on a knowledge management system, which capture and store information for future inference and usage. In this paper, the proposed framework seeks to incorporate elements such as simultaneous exploration of embodiment design and process parameter spaces, enabling transferring and application of learnt knowledge and calibration of models once new data are available. Knowledge management is key to ensuring that the AM parts are built according to design intent. As shown in Fig. 5, the proposed knowledge management network encompasses the whole AM process, from embodiment design to process parameter selection, and finally manufacturing of the AM part. Database can be contained within a BN as discussed in Wang et al. (2018) or a standalone database such as the SQL database. To ensure that the database is robust and able to encompass various design scenarios, a large amount of data is required. It may not be feasible to build a comprehensive model from scratch. To counteract this drawback, three strategies are proposed as follows. Figure 5: Open in new tabDownload slide Information flow for enabling data-driven design strategy in a knowledge management network. Figure 5: Open in new tabDownload slide Information flow for enabling data-driven design strategy in a knowledge management network. First, information from sources, such as from literature, past experiences with printed parts, etc., is organized and collated into a data pool. For example, Goh et al. (2020) provided a comprehensive overview of the different process parameter setting on the final part properties, which can be imported into database for future reference. However, data obtained from such method need to be sorted through and spurious data were to be discarded before further data processing and cleaning, of which Hertlein, Deshpande, Venugopal, Kumar, and Anand (2020) provided an example. Second, transfer learning can also be utilized for building up of the database. The transfer learning is defined as applying relevant knowledge from source tasks and improving learning in related target tasks (Torrey & Shavlik, 2010). For example, characterization for a new FFF system or a new material can be considered as a target task, and knowledge from a past system and material is considered as source. Through the transfer learning, improvement in learning for the target tasks can be achieved through three scenarios: better initial performance achieved, faster learning, and better final performance achievable. More specifically, inductive transfer learning can be utilized to reduce the hypothesis space. Lastly, a hybrid method of physics-based data-driven approaches (i.e. combination of empirical data and numerical simulations) enables reduction in the amount of data required to build an accurate model and can be achieved via two ways. First, numerical simulation of the process or phenomenon can be conducted with physics-based numerical simulation, and the simulation can be complemented by low fidelity or limited empirical data while building the model (Lu & Wang, 2018; Haghighi & Li, 2020). The second method would be to use the empirical data for calibration of surrogate models obtained, which may be built from only numerical simulations. Through physical observations, unknown parameters about the system can be learnt, to correct the model discrepancies and improve on the predictability of the model (Kennedy & O’Hagan, 2001; Li et al., 2018; Olleak & Xi, 2020). To ensure the printability of designed AM parts, the database includes specific DfAM rules, formed by amalgamating knowledge from published literature, such as described in Ahn et al. (2002) and Urbanic and Hedrick (2016), expert opinions, and AM designers’ experience. The collated information from various sources forms the a priori knowledge for the next design development and is leveraged to improve printability of the part. This can reduce the resources incurred in the embodiment design phase for experimentation and can even guide AM novices to “get it right the first time.” The database is updated whenever new characterization experiments have been conducted or when new expert knowledge becomes available. To achieve the goal of “get it right the first time,” a probable approach combining a voxel-based discretization technique for analysis of the AM part geometry as described in Guo, Lu, and Fuh (2020) in combination with a BN, such as described in Wang et al. (2018), can be utilized for the printability analysis and recommendation. Figure 6 shows the proposed data-driven design strategy framework that further develops on the knowledge management network described in Fig. 5. The following paragraphs provide a step-by-step description of the steps involved in the framework, citing relevant literature and research that can be utilized to fulfill the steps. Table 3 provides a compressive summary including the deliverables, baseline, and alternatives/optional methods of the steps. Figure 6: Open in new tabDownload slide Data-driven design strategy framework. Figure 6: Open in new tabDownload slide Data-driven design strategy framework. Table 3: Descriptions of architecting steps, deliverables, and available methods in the proposed framework. Architecting steps . Deliverables . Baseline . Alternatives/optional . Step 1: Defining functional requirements Product requirements, such as surface roughness, mechanical properties are clearly defined and identified • List of functional requirements - • Product specifications • FEA Step 2: Creating embodiment design and assigning process parameters Part designs with defined process parameters and toolpaths that are able to meet the defined product requirements • Identification of feasible design space using BN—(Wang et al., 2018) • Topological optimization, with functionally graded lattice structure—(Nguyen et al., 2019) • Adaptive toolpath planning with variable bead width—(Xiong et al., 2019) • Reduction in voids by selecting optimal PBO—(Eiliat & Urbanic, 2019) • Using cost of part to drive design—(Yao et al., 2016) • Prediction of geometrical accuracy with selected process parameter using neural network—(Yan, 2020) Step 3: Validating design and process parameters Validated final design in terms of printability and ability to meet the defined product requirements • Narrowing down the design space to optimize point-wise design using ANN and TOPSIS—(Khodaygan & Golmohammadi, 2018) • Narrowing down the design space to optimize point-wise design using GPR—(Xiong et al., 2019) • Evaluation of printability using ML classifiers—(Mycroft et al., 2020) • Visualization of PSP relationship using self-organizing map (Gan et al., 2019) • Manual evaluation of printability using knowledge from literature • Bayesian inference to reduce the number of experimentations—(Sabbaghi et al., 2018) • Prediction of geometrical accuracy and distortions—(Cattenone et al., 2019) • Prediction of surface roughness using in situ process variables and offline process parameters—(Sun et al., 2017) Step 4: Manufacturing of parts Manufactured parts as per design and process specifications • FFF - Step 5: Measuring and validating printed parts, calibrating Bayesian model Collection of data pertaining to the build and part, incorporating this information into the data base for future use and reference. Calibration of surrogate models pertaining to the build • Measuring and validation of predicted properties—(Zhu et al., 2018) - • Bayesian calibration of surrogate model—(Li et al., 2018) Architecting steps . Deliverables . Baseline . Alternatives/optional . Step 1: Defining functional requirements Product requirements, such as surface roughness, mechanical properties are clearly defined and identified • List of functional requirements - • Product specifications • FEA Step 2: Creating embodiment design and assigning process parameters Part designs with defined process parameters and toolpaths that are able to meet the defined product requirements • Identification of feasible design space using BN—(Wang et al., 2018) • Topological optimization, with functionally graded lattice structure—(Nguyen et al., 2019) • Adaptive toolpath planning with variable bead width—(Xiong et al., 2019) • Reduction in voids by selecting optimal PBO—(Eiliat & Urbanic, 2019) • Using cost of part to drive design—(Yao et al., 2016) • Prediction of geometrical accuracy with selected process parameter using neural network—(Yan, 2020) Step 3: Validating design and process parameters Validated final design in terms of printability and ability to meet the defined product requirements • Narrowing down the design space to optimize point-wise design using ANN and TOPSIS—(Khodaygan & Golmohammadi, 2018) • Narrowing down the design space to optimize point-wise design using GPR—(Xiong et al., 2019) • Evaluation of printability using ML classifiers—(Mycroft et al., 2020) • Visualization of PSP relationship using self-organizing map (Gan et al., 2019) • Manual evaluation of printability using knowledge from literature • Bayesian inference to reduce the number of experimentations—(Sabbaghi et al., 2018) • Prediction of geometrical accuracy and distortions—(Cattenone et al., 2019) • Prediction of surface roughness using in situ process variables and offline process parameters—(Sun et al., 2017) Step 4: Manufacturing of parts Manufactured parts as per design and process specifications • FFF - Step 5: Measuring and validating printed parts, calibrating Bayesian model Collection of data pertaining to the build and part, incorporating this information into the data base for future use and reference. Calibration of surrogate models pertaining to the build • Measuring and validation of predicted properties—(Zhu et al., 2018) - • Bayesian calibration of surrogate model—(Li et al., 2018) Open in new tab Table 3: Descriptions of architecting steps, deliverables, and available methods in the proposed framework. Architecting steps . Deliverables . Baseline . Alternatives/optional . Step 1: Defining functional requirements Product requirements, such as surface roughness, mechanical properties are clearly defined and identified • List of functional requirements - • Product specifications • FEA Step 2: Creating embodiment design and assigning process parameters Part designs with defined process parameters and toolpaths that are able to meet the defined product requirements • Identification of feasible design space using BN—(Wang et al., 2018) • Topological optimization, with functionally graded lattice structure—(Nguyen et al., 2019) • Adaptive toolpath planning with variable bead width—(Xiong et al., 2019) • Reduction in voids by selecting optimal PBO—(Eiliat & Urbanic, 2019) • Using cost of part to drive design—(Yao et al., 2016) • Prediction of geometrical accuracy with selected process parameter using neural network—(Yan, 2020) Step 3: Validating design and process parameters Validated final design in terms of printability and ability to meet the defined product requirements • Narrowing down the design space to optimize point-wise design using ANN and TOPSIS—(Khodaygan & Golmohammadi, 2018) • Narrowing down the design space to optimize point-wise design using GPR—(Xiong et al., 2019) • Evaluation of printability using ML classifiers—(Mycroft et al., 2020) • Visualization of PSP relationship using self-organizing map (Gan et al., 2019) • Manual evaluation of printability using knowledge from literature • Bayesian inference to reduce the number of experimentations—(Sabbaghi et al., 2018) • Prediction of geometrical accuracy and distortions—(Cattenone et al., 2019) • Prediction of surface roughness using in situ process variables and offline process parameters—(Sun et al., 2017) Step 4: Manufacturing of parts Manufactured parts as per design and process specifications • FFF - Step 5: Measuring and validating printed parts, calibrating Bayesian model Collection of data pertaining to the build and part, incorporating this information into the data base for future use and reference. Calibration of surrogate models pertaining to the build • Measuring and validation of predicted properties—(Zhu et al., 2018) - • Bayesian calibration of surrogate model—(Li et al., 2018) Architecting steps . Deliverables . Baseline . Alternatives/optional . Step 1: Defining functional requirements Product requirements, such as surface roughness, mechanical properties are clearly defined and identified • List of functional requirements - • Product specifications • FEA Step 2: Creating embodiment design and assigning process parameters Part designs with defined process parameters and toolpaths that are able to meet the defined product requirements • Identification of feasible design space using BN—(Wang et al., 2018) • Topological optimization, with functionally graded lattice structure—(Nguyen et al., 2019) • Adaptive toolpath planning with variable bead width—(Xiong et al., 2019) • Reduction in voids by selecting optimal PBO—(Eiliat & Urbanic, 2019) • Using cost of part to drive design—(Yao et al., 2016) • Prediction of geometrical accuracy with selected process parameter using neural network—(Yan, 2020) Step 3: Validating design and process parameters Validated final design in terms of printability and ability to meet the defined product requirements • Narrowing down the design space to optimize point-wise design using ANN and TOPSIS—(Khodaygan & Golmohammadi, 2018) • Narrowing down the design space to optimize point-wise design using GPR—(Xiong et al., 2019) • Evaluation of printability using ML classifiers—(Mycroft et al., 2020) • Visualization of PSP relationship using self-organizing map (Gan et al., 2019) • Manual evaluation of printability using knowledge from literature • Bayesian inference to reduce the number of experimentations—(Sabbaghi et al., 2018) • Prediction of geometrical accuracy and distortions—(Cattenone et al., 2019) • Prediction of surface roughness using in situ process variables and offline process parameters—(Sun et al., 2017) Step 4: Manufacturing of parts Manufactured parts as per design and process specifications • FFF - Step 5: Measuring and validating printed parts, calibrating Bayesian model Collection of data pertaining to the build and part, incorporating this information into the data base for future use and reference. Calibration of surrogate models pertaining to the build • Measuring and validation of predicted properties—(Zhu et al., 2018) - • Bayesian calibration of surrogate model—(Li et al., 2018) Open in new tab The proposed framework starts with the identification of functional and performance requirements of a part. Examples of these requirements include surface roughness, mechanical strength, or fatigue performance. Subsequently, FEA is performed to understand the stress distribution, which corresponds to the load conditions and scenarios experienced by the part within its product lifecycle. After the stress analysis is concluded, the embodiment design phase commences. The second step of the framework involves creation of the embodiment design, assigning process parameters and defining the toolpath. The goal is to achieve the required part properties by tuning the process parameter and the geometries of the part, which is similar to the backward induction mapping methodology discussed in Rosen (2015). AM designers may incorporate cellular structures, and topological optimization can be carried out on functional volumes that have less critical stress loadings (Yang, Tang, & Zhao, 2015; Nguyen, Kim, & Choi, 2019). These cellular structures and optimized design can provide performance improvement and significant weight reduction. Next, assignment of process parameters is carried out.Part requirements and stress distribution information are fed into surrogate models built from empirical data. These surrogate models and classifiers compute and define the feasible design space for embodiment design and process parameter, such as material, RA, and PBO. BN classifiers as described in Wang et al. (2018) provide a probabilistic view of the feasible and infeasible design space, and help AM designers determine a range of process parameters that can be used to achieve the required properties. Toolpath planning is another critical aspect that is necessary to be performed to reduce voids and strengthen critical regions in this stage. Through adaptive toolpath planning as proposed by Xiong et al. (2019), reduction in the voids present in part and improvement in geometrical accuracy can be achieved. Alternatively, Eiliat and Urbanic (2019) proposed using optimization of PBO to minimize the voids in the parts. However, the proposed method can be utilized on closed systems (e.g. Stratasys FDM™ printers), even though the method can be cumbersome and further constrains the choice of build orientation. Further, depending on the requirements of the part design, the design and process parameter selection can be driven by cost, or if tolerancing is critical (e.g. an interference fit is required), the design space can be further constrained by methodologies as discussed in Yao et al. (2016) and Yan (2020), respectively. In the previous preliminary design step, multiple possible designs can exist that can potentially fulfill the requirements. In the next step, however, the number of possible designs is to be reduced to only one and iterations may be performed to fine-tune the selected detailed design. In this framework, point-wise selection of design is based on two criteria: (1) ability to meet the functional requirements and (2) printability evaluation. To search for the design that best meets the requirements, a combination of ANN and TOPSIS described in Khodaygan & Golmohammadi (2018), which provides a Pareto front of the optimized design(s) can be utilised. If required, further optimization (e.g. using desirability function) can be carried out to select the optimal design points or manual selection by AM designers can be performed. An alternative method to search for optimal design includes the combination of Monte Carlo-based resampling and GRP can be utilized as described in Xiong et al. (2019). However, the performance of GPR in high-dimensional space or multimodal problems can be poorer as compared to ANN. If necessary or desired, further fine-tuning of design and process parameters can be performed by AM designers. To aid in this fine-tuning process, visualization in the form of maps can be used. These maps enable AM designers to visually perceive the PSP linkage, showing the feasible design spaces and mapping the design and process parameter changes in real time between design spaces. Examples of such visualization include design and process parameter maps shown in Fig. 7(a) and (b), respectively, and self-organizing maps as described in Gan et al. (2019). Figure 7: Open in new tabDownload slide Fine-tuning the detailed design using (a) design space mappings between design and process parameters and (b) process parameter mapping. Figure 7: Open in new tabDownload slide Fine-tuning the detailed design using (a) design space mappings between design and process parameters and (b) process parameter mapping. Printability evaluation is conducted to identify areas with possible printability issues. The information on structures’ printability can be based on data from past experiments, such as the characterization methodology proposed by Jiang et al. (2019), or from open literature sources such as Bikas, Lianos, and Stavropoulos (2019). Evaluation can be carried out via ML classifiers, similar to the methodology as proposed in Mycroft et al. (2020), by analysing vertices of the geometry and describing the printability using probability. Alternatively, manual evaluation can be performed; however, this comes as a relatively higher cost and time consuming. Evaluation criteria include self-supporting angle, bridging length, and printability of features (e.g. small features, thin wall structures, and cellular structure). Areas with potential printability issues are highlighted to AM designers with possible rectification solutions. Once a detailed design is finalized, a final round of validation is performed. Such validation can be performed by feeding the selected design point back into the surrogate models. AM designers can also perform a check on other FFF part performance indicators that are not part of the design requirement, such as geometrical accuracy and distortions (Cattenone, Morganti, Alaimo, & Auricchio, 2019). Further, critical requirements can be estimated during the printing process using an online prediction methodology as discussed in Sun et al. (2017). When existing solutions are not able to meet any of the functional requirements, further simulation and empirical testing are required. To reduce the resources and effort consumed, techniques such as Bayesian inference as detailed in Sabbaghi et al. (2018) can be utilized. Such Bayesian inference methods maximize the value of information obtained from each individual experiment point. As a result, Bayesian methods take significantly lesser experimental points to characterize new design spaces to obtain an optimized solution as compared to conventional design of experiment or uncertainty sampling (Pandita, Bilionis, & Panchal, 2019). These experiments and simulation results are then stored in the database for future references. The finalized detailed design, together with corresponding printing parameters, is sent to the FFF printer for manufacturing. Once the parts are printed, key properties and functional requirement such as surface roughness and geometrical accuracies are measured and compared against the surrogate models’ prediction, similar to the verification approach adopted in Zhu et al. (2018). For the evaluation of mechanical properties, coprinted specimens or sacrificial parts are tested. Results from the tests are collated and verified against the simulation results. Collated information is then feedback and stored in database for calibration of surrogate models to improve on the prediction accuracy, using methods such as Bayesian calibration as discussed in Li et al. (2018). Through the method of direct characterization of the printed parts, accurate mechanical performance of the parts can be obtained, circumventing the issue of having no standardized and representative characterization methods for FFF parts. An example of multilevel surrogate modeling techniques used at the embodiment design phase is shown in Fig. 8. Two major benefits can be achieved utilizing such strategy. First, by adopting a multilevel surrogate modeling strategy, errors can be reduced for surrogate modeling. Three different stages of the hierarchical model are presented in the example shown in Fig. 8. Validation can then be conducted at each level of the surrogate model and ensures that the fidelity of each stage is sufficiently high to reduce compounding errors, which reduces the uncertainty that is cascaded down each level (Wang et al., 2019). Second, aleatory uncertainty can be characterized. In this example, the effect of fluctuating extruder temperature and build chamber is determined as aleatory uncertainties, with unknown distributions, which can affect the mechanical properties for the printed parts (Dinwiddie, Love, & Rowe, 2013; Nelson et al., 2015). Figure 8: Open in new tabDownload slide Example of a multilevel Bayesian surrogate model for FFF process for the prediction of mechanical properties. Figure 8: Open in new tabDownload slide Example of a multilevel Bayesian surrogate model for FFF process for the prediction of mechanical properties. Aleatory sources of uncertainty have to be characterized for the prediction of part mechanical properties. For this example, it is determined that the thermal history affects the bonding characteristics of the adjacent rasters, which affects the mechanical performance of the final part. First stage contains the surrogate models for the thermal field and the extruder temperature, which can be approximated by multivariate distribution |${\theta _1}\sim \ P( {{\boldsymbol{x}},{\boldsymbol{\ }}t} )$| as a function of the Euclidean space and time, and extruder temperature as a function of time |${\theta _2}\sim \ P( t )$| that has thermal feedback control, and they are assumed to be independent. The degree bond formation between adjacent rasters, |$\lambda $|⁠, which can be modeled by Arrhenius kinetics theory, is a function of time and temperature (Malekmotiei, Voyiadjis, Samadi-Dooki, Lu, & Zhou, 2017; Prajapati, Chalise, Ravoori, Taylor, & Jain, 2019). As such, the two thermal surrogates are treated as prior hyperparameters and are used to model the likelihood of the degree of bond formation |$\lambda |{\theta _1},\ {\theta _2}\sim P(\lambda |{\theta _1},\ {\theta _2})$|⁠. The tuning of hyperparameter used can be performed in multifarious manner, including maximum likelihood estimation based on existing database as discussed in Gramacy (2016), gradient descent, and hill climb as discussed in Toal, Bressloff, and Keane (2008). Lastly, the mechanical performance, |$\ {Y_i}$|⁠, is dependent on the degree of bond formation between the adjacent raster, |${Y_i}|\lambda ,\ {\theta _1},\ {\theta _2}\sim P({Y_i}|\lambda ,\ {\theta _1},\ {\theta _2})$|⁠. With observed mechanical performance, |${Y_i},$|⁠, the marginal maximum likelihoods of parameters and hyperparameters |${\theta _1}$|⁠, |${\theta _2}$|⁠, and |$\lambda $| can be updated. Once data from printed parts are available, calibration can be performed on the model to achieve higher accuracies, using methods such as Bayesian calibration as proposed by Kennedy and O’Hagan (2001). In this section, a data-driven design strategy framework for designing of FFF parts is proposed. The framework is driven primarily by empirical data and evidence from experimentations, which are subsumed within a knowledge database. The proposed framework enables AM designers to navigate the high-dimensional design space rapidly and efficiently using surrogate models and classifiers, and the feasibility of the proposed framework is illustrated using examples from literature. 5. Conclusion This paper provided an overview of the various data-driven processes that were implemented in AM, from the perspective of the embodiment design and process parameters for FFF parts. State-of-the-art review was performed, and the existing gaps in the literature were identified. Due to the unique characteristics of AM, embodiment design and process parameter design space should be simultaneously explored and exploited to achieve design intent and meet the functional requirements of an AM part. However, a need for a systematic, data-driven approach was required to explore and exploit the enlarged design space. Current literature often focused solely on either embodiment design or process domain. A data-driven design framework with embedded knowledge management system was proposed to help AM designers exploit the advantages of AM technologies. Through exploiting data-driven design strategies, FFF can be characterized more effectively and at a significantly reduced cost, which in turn reduces the barrier for users to adopt FFF as a direct manufacturing method. The following are the limitations of the current work and further opportunities in FFF: The review is constrained only to the state-of-the-art data-driven design strategy for FFF. Future work will include data-driven strategies for other AM technologies, such as selective laser melting, which has a wealth of literature discussing on data-driven design strategy. The proposed framework can be expanded to include data collected from sensors or in-process monitoring for feedback of the part performance in real time, which can serve for improving the design in the future and allowing for live tuning of parameters to achieve the desired part properties. The proposed framework can be expanded to include data-driven strategy for quality assurance of AM part, such as ensuring that parts printed are defect free. ACKNOWLEDGEMENTS This research was supported by a grant from ST Engineering Aerospace, EDB-IPP, Singapore Centre for 3D Printing (SC3DP), the National Research Foundation, Prime Minister's Office, Singapore under its Medium-Sized Centre funding scheme. Conflict of interest statement None declared. References Aboutaleb A. M. , Bian L., Elwany A., Shamsaei N., Thompson S. M., Tapia G. ( 2017 ). Accelerated process optimization for laser-based additive manufacturing by leveraging similar prior studies . IISE Transactions , 49 ( 1 ), 31 – 44 . Google Scholar OpenURL Placeholder Text WorldCat Ahn S.-H. , Montero M., Odell D., Roundy S., Wright P. K. ( 2002 ). Anisotropic material properties of fused deposition modeling ABS . Rapid Prototyping Journal , 8 ( 4 ), 248 – 257 . Google Scholar OpenURL Placeholder Text WorldCat Alizadeh M. , Esfahani M. N., Tian W., Ma J. ( 2020 ). Data-driven energy efficiency and part geometric accuracy modeling and optimization of green fused filament fabrication processes . Journal of Mechanical Design , 142 ( 4) , 041701 . Google Scholar OpenURL Placeholder Text WorldCat An D. , Kim N. H., Choi J.-H. ( 2015 ). Practical options for selecting data-driven or physics-based prognostics algorithms with reviews . Reliability Engineering & System Safety , 133 , 223 – 236 . Google Scholar OpenURL Placeholder Text WorldCat Ang K. C. , Leong K. F., Chua C. K., Chandrasekaran M. ( 2006 ). Investigation of the mechanical properties and porosity relationships in fused deposition modelling-fabricated porous structures . Rapid Prototyping Journal , 12 ( 2 ), 100 – 105 . Google Scholar OpenURL Placeholder Text WorldCat Avdeev A. , Shvets A., Gushchin I., Torubarov I., Drobotov A., Makarov A., Serdobintsev Y. ( 2019 ). Strength increasing additive manufacturing fused filament fabrication technology, based on spiral toolpath material deposition . Machines , 7 ( 3 ), 57 . Google Scholar OpenURL Placeholder Text WorldCat Barrrionuevo G. O. , Ramos-Grez J. A. ( 2019 ). Machine learning for optimizing technological properties of wood composite filament-Timberfill fabricated by fused deposition modeling . In International Conference on Applied Technologies (ICAT 2019) . (pp. 119 – 132) . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bayraktar Ö. , Uzun G., Çakiroğlu R., Guldas A. ( 2017 ). Experimental study on the 3D-printed plastic parts and predicting the mechanical properties using artificial neural networks . Polymers for Advanced Technologies , 28 ( 8 ), 1044 – 1051 . Google Scholar OpenURL Placeholder Text WorldCat Bessa M. A. , Glowacki P., Houlder M. ( 2019 ). Bayesian machine learning in metamaterial design: Fragile becomes supercompressible . Advanced Materials , 31 ( 48 ), 1904845 . Google Scholar OpenURL Placeholder Text WorldCat Bikas H. , Lianos A., Stavropoulos P. ( 2019 ). A design framework for additive manufacturing . The International Journal of Advanced Manufacturing Technology , 103 ( 9–12 ), 3769 – 3783 . Google Scholar OpenURL Placeholder Text WorldCat Byberg K. I. , Gebisa A. W., Lemu H. G. ( 2018 ). Mechanical properties of ULTEM 9085 material processed by fused deposition modeling . Polymer Testing , 72 , 335 – 347 . Google Scholar OpenURL Placeholder Text WorldCat Cattenone A. , Morganti S., Alaimo G., Auricchio F. ( 2019 ). Finite element analysis of additive manufacturing based on fused deposition modeling: distortions prediction and comparison with experimental data . Journal of Manufacturing Science and Engineering , 141 ( 1) , 011010 . Google Scholar OpenURL Placeholder Text WorldCat Ceruti A. , Marzocca P., Liverani A., Bil C. ( 2019 ). Maintenance in aeronautics in an Industry 4.0 context: The role of augmented reality and additive manufacturing . Journal of Computational Design and Engineering , 6 ( 4 ), 516 – 526 . Google Scholar OpenURL Placeholder Text WorldCat Chang D.-Y. , Huang B.-H. ( 2011 ). Studies on profile error and extruding aperture for the RP parts using the fused deposition modeling process . The International Journal of Advanced Manufacturing Technology , 53 ( 9–12 ), 1027 – 1037 . Google Scholar OpenURL Placeholder Text WorldCat Cheng L. , Tsung F., Wang A. ( 2017 ). A statistical transfer learning perspective for modeling shape deviations in additive manufacturing . IEEE Robotics and Automation Letters , 2 ( 4 ), 1988 – 1993 . Google Scholar OpenURL Placeholder Text WorldCat Chollet F. ( 2018 ). Deep learning with R. Shelter Island, NY , USA: Manning Publications . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Chu C. , Graf G., Rosen D. W. ( 2008 ). Design for additive manufacturing of cellular structures . Computer-Aided Design and Applications , 5 ( 5 ), 686 – 696 . Google Scholar OpenURL Placeholder Text WorldCat Chua C. K. , Leong K. F. ( 2017 ). 3D printing and additive manufacturing: Principles and applications . (5th Ed.). Singapore : World Scientific . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Coulais C. , Teomy E., De Reus K., Shokef Y., Van Hecke M. ( 2016 ). Combinatorial design of textured mechanical metamaterials . Nature , 535 ( 7613 ), 529 – 532 . Google Scholar OpenURL Placeholder Text WorldCat Deswal S. , Narang R., Chhabra D. ( 2019 ). Modeling and parametric optimization of FDM 3D printing process using hybrid techniques for enhancing dimensional preciseness . International Journal on Interactive Design and Manufacturing (IJIDeM) , 13 ( 3 ), 1197 – 1214 . Google Scholar OpenURL Placeholder Text WorldCat Dinwiddie R. B. , Love L. J., Rowe J. C. ( 2013 ). Real-time process monitoring and temperature mapping of a 3D polymer printing process . In Proceedings of SPIE, Thermosense: Thermal Infrared Applications XXXV (Vol. 8705 , p. 87050L ). SPIE . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Eiliat H. , Urbanic J. ( 2018 ). Visualizing, analyzing, and managing voids in the material extrusion process . The International Journal of Advanced Manufacturing Technology , 96 ( 9–12 ), 4095 – 4109 . Google Scholar OpenURL Placeholder Text WorldCat Eiliat H. , Urbanic J. ( 2019 ). Determining the relationships between the build orientation, process parameters and voids in additive manufacturing material extrusion processes . The International Journal of Advanced Manufacturing Technology , 100 ( 1–4 ), 683 – 705 . Google Scholar OpenURL Placeholder Text WorldCat Feng J. , Fu J., Lin Z., Shang C., Li B. ( 2018 ). A review of the design methods of complex topology structures for 3D printing . Visual Computing for Industry, Biomedicine, and Art , 1 ( 1 ), 1 – 16 . Google Scholar OpenURL Placeholder Text WorldCat Fernandez-Vicente M. , Canyada M., Conejero A. ( 2015 ). Identifying limitations for design for manufacturing with desktop FFF 3D printers . International Journal of Rapid Manufacturing , 5 ( 1 ), 116 – 128 . Google Scholar OpenURL Placeholder Text WorldCat Ford S. , Despeisse M. ( 2016 ). Additive manufacturing and sustainability: an exploratory study of the advantages and challenges . Journal of Cleaner Production , 137 , 1573 – 1587 . Google Scholar OpenURL Placeholder Text WorldCat Forrester A. , Sobester A., Keane A. ( 2008 ). Engineering design via surrogate modelling: A practical guide . John Wiley & Sons . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Gan Z. , Li H., Wolff S. J., Bennett J. L., Hyatt G., Wagner G. J., Liu W. K. ( 2019 ). Data-driven microstructure and microhardness design in additive manufacturing using a self-organizing map . Engineering , 5 ( 4 ), 730 – 735 . Google Scholar OpenURL Placeholder Text WorldCat Garg A. , Tai K., Lee C., Savalani M. ( 2014 ). A hybrid M5′-genetic programming approach for ensuring greater trustworthiness of prediction ability in modelling of FDM process . Journal of Intelligent Manufacturing , 25 ( 6 ), 1349 – 1365 . Google Scholar OpenURL Placeholder Text WorldCat Garg A. , Tai K., Savalani M. ( 2014 ). State-of-the-art in empirical modelling of rapid prototyping processes . Rapid Prototyping Journal , 20 ( 2 ), 164 – 178 . Google Scholar OpenURL Placeholder Text WorldCat Gaynor A. T. , Meisel N. A., Williams C. B., Guest J. K. ( 2014 ), Topology optimization for additive manufacturing: considering maximum overhang constraint . In 15th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC General Electric Additive . ( 2018 ). New manufacturing milestone: 30,000 additive fuel nozzles . Retrieved from https://www.ge.com/additive/blog/new-manufacturing-milestone-30000-additive-fuel-nozzles, Accessed date: 08 March 2020 . OpenURL Placeholder Text WorldCat Goh G. D. , Sing S. L., Yeong W. Y. ( 2020 ). A review on machine learning in 3D printing: applications, potential, and challenges . Artificial Intelligence Review , 1 – 32 ., In Press, DOI: https://doi.org/10.1007/s10462-020-09876-9 . Google Scholar OpenURL Placeholder Text WorldCat Goh G. D. , Yap Y. L., Tan H., Sing S. L., Goh G. L., Yeong W. Y. ( 2020 ). Process–structure–properties in polymer additive manufacturing via material extrusion: A review . Critical Reviews in Solid State and Materials Sciences , 45 ( 2 ), 113 – 133 . Google Scholar OpenURL Placeholder Text WorldCat Gramacy R. B. ( 2016 ). laGP: large-scale spatial modeling via local approximate Gaussian processes in R . Journal of Statistical Software , 72 ( 1 ), 1 – 46 . Google Scholar OpenURL Placeholder Text WorldCat Guo Y. , Lu W. F., Fuh J. Y. H. ( 2020 ). Semi-supervised deep learning based framework for assessing manufacturability of cellular structures in direct metal laser sintering process . Journal of Intelligent Manufacturing , 1 – 13 .., In Press, DOI: https://doi.org/10.1007/s10845-020-01575-0. Google Scholar OpenURL Placeholder Text WorldCat Haghighi A. , Li L. ( 2020 ). A hybrid physics-based and data-driven approach for characterizing porosity variation and filament bonding in extrusion-based additive manufacturing . Additive Manufacturing , 36, 101399 . Google Scholar OpenURL Placeholder Text WorldCat Heisig P. ( 2009 ). Harmonisation of knowledge management-comparing 160 KM frameworks around the globe . Journal of Knowledge Management , 13 ( 4 ), 4 – 31 . Google Scholar OpenURL Placeholder Text WorldCat Hertlein N. , Deshpande S., Venugopal V., Kumar M., Anand S. ( 2020 ). Prediction of selective laser melting part quality using hybrid Bayesian network . Additive Manufacturing , 32 , 101089 . Google Scholar OpenURL Placeholder Text WorldCat Hu Z. , Mahadevan S. ( 2017 ). Uncertainty quantification and management in additive manufacturing: current status, needs, and opportunities . The International Journal of Advanced Manufacturing Technology , 93 ( 5–8 ), 2855 – 2874 . Google Scholar OpenURL Placeholder Text WorldCat Hu F. , Wu D. ( 2019 ). Cellular structures design for wrist rehabilitation considering 3d printability and mechanics lightweight . In 2019 WRC Symposium on Advanced Robotics and Automation (WRC SARA) . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Huang J. , Chen Q., Jiang H., Zou B., Li L., Liu J., Yu H. ( 2020 ). A survey of design methods for material extrusion polymer 3D printing . Virtual and Physical Prototyping , 15 ( 2 ), 148 – 162 . Google Scholar OpenURL Placeholder Text WorldCat Huynh H. N. , Nguyen A. T., Ha N. L., Thai T. T. H. ( 2017 ). Application of fuzzy Taguchi method to improve the dimensional accuracy of fused deposition modeling processed product . In 2017 International Conference on System Science and Engineering (ICSSE) . IEEE . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jackson T. , Liu H., Patrikalakis N., Sachs E., Cima M. ( 1999 ). Modeling and designing functionally graded material components for fabrication with local composition control . Materials & Design , 20 ( 2–3 ), 63 – 75 . Google Scholar OpenURL Placeholder Text WorldCat Jiang J. , Stringer J., Xu X., Zhong R. Y. ( 2018 ). Investigation of printable threshold overhang angle in extrusion-based additive manufacturing for reducing support waste . International Journal of Computer Integrated Manufacturing , 31 ( 10 ), 961 – 969 . Google Scholar OpenURL Placeholder Text WorldCat Jiang J. , Hu G., Li X., Xu X., Zheng P., Stringer J. ( 2019 ). Analysis and prediction of printable bridge length in fused deposition modelling based on back propagation neural network . Virtual and Physical Prototyping , 14 ( 3 ), 253 – 266 . Google Scholar OpenURL Placeholder Text WorldCat Johnson R. A. , Wichern D. W., (2020). ; Applied multivariate statistical analysis . 5th Editions , Upper Saddle River , NJ: Prentice hall . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Jothibabu G. , Kumar S. ( 2018 ). Surrogate based sensitivity analysis of part strength due to process parameters in fused deposition modelling . Procedia Computer Science , 133 , 772 – 778 . Google Scholar OpenURL Placeholder Text WorldCat Kackar R. N. ( 1985 ). Off-line quality control, parameter design, and the Taguchi method . Journal of Quality Technology , 17 ( 4 ), 176 – 188 . Google Scholar OpenURL Placeholder Text WorldCat Kennedy M. C. , O’Hagan A. ( 2001 ). Bayesian calibration of computer models . Journal of the Royal Statistical Society: Series B (Statistical Methodology) , 63 ( 3 ), 425 – 464 . Google Scholar OpenURL Placeholder Text WorldCat Khodaygan S. , Golmohammadi A. ( 2018 ). Multi-criteria optimization of the part build orientation (PBO) through a combined meta-modeling/NSGAII/TOPSIS method for additive manufacturing processes . International Journal on Interactive Design and Manufacturing (IJIDeM) , 12 ( 3 ), 1071 – 1085 . Google Scholar OpenURL Placeholder Text WorldCat Kim J. , Yoo D.-J. ( 2020 ). 3D printed compact heat exchangers with mathematically defined core structures . Journal of Computational Design and Engineering , 7 ( 4 ), 527 – 550 . Google Scholar OpenURL Placeholder Text WorldCat Ko H. , Moon S. K., Otto K. N. ( 2015 ). Design knowledge representation to support personalised additive manufacturing . Virtual and Physical Prototyping , 10 ( 4 ), 217 – 226 . Google Scholar OpenURL Placeholder Text WorldCat Koller D. , Friedman N. ( 2009 ). Probabilistic graphical models: Principles and techniques . MIT press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kubalak J. R. , Wicks A. L., Williams C. B. ( 2019 ). Exploring multi-axis material extrusion additive manufacturing for improving mechanical properties of printed parts . Rapid Prototyping Journal , 25 ( 2 ), 356 – 362 . Google Scholar OpenURL Placeholder Text WorldCat Larrañaga P. , Moral S. ( 2011 ). Probabilistic graphical models in artificial intelligence . Applied Soft Computing , 11 ( 2 ), 1511 – 1528 . Google Scholar OpenURL Placeholder Text WorldCat Larrañaga P. , Karshenas H., Bielza C., Santana R. ( 2013 ). A review on evolutionary algorithms in Bayesian network learning and inference tasks . Information Sciences , 233 , 109 – 125 . Google Scholar OpenURL Placeholder Text WorldCat Levasseur A. , Ploeg H.-L., Petit Y. ( 2012 ). Comparison of the influences of structural characteristics on bulk mechanical behaviour: Experimental study using a bone surrogate . Medical & Biological Engineering & Computing , 50 ( 1 ), 61 – 67 . Google Scholar OpenURL Placeholder Text WorldCat Li J. , Jin R., Hang Z. Y. ( 2018 ). Integration of physically-based and data-driven approaches for thermal field prediction in additive manufacturing . Materials & Design , 139 , 473 – 485 . Google Scholar OpenURL Placeholder Text WorldCat Li Z. , Zhang Z., Shi J., Wu D. ( 2019 ). Prediction of surface roughness in extrusion-based additive manufacturing with machine learning . Robotics and Computer-Integrated Manufacturing , 57 , 488 – 495 . Google Scholar OpenURL Placeholder Text WorldCat Lin Y. C. , Ha N.-H. ( 2015 ). The framework for KM implementation in product and service oriented SMEs: Evidence from field studies in Taiwan . Sustainability , 7 ( 3 ), 2980 – 3000 . Google Scholar OpenURL Placeholder Text WorldCat Liu Y. , Wang P. ( 2016 ). Probabilistic modeling and analysis of fused deposition modeling process using surrogate models . In ASME 2016 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Liu X. , Zhang M., Li S., Si L., Peng J., Hu Y. ( 2017 ). Mechanical property parametric appraisal of fused deposition modeling parts based on the gray Taguchi method . The International Journal of Advanced Manufacturing Technology , 89 ( 5–8 ), 2387 – 2397 . Google Scholar OpenURL Placeholder Text WorldCat Lu Y. , Wang Y. ( 2018 ). Monitoring temperature in additive manufacturing with physics-based compressive sensing . Journal of Manufacturing Systems , 48 , 60 – 70 . Google Scholar OpenURL Placeholder Text WorldCat Majeed A. , Lv J., Peng T. ( 2019 ). A framework for big data-driven process analysis and optimization for additive manufacturing . Rapid Prototyping Journal , 25 ( 2 ), 308 – 321 . Google Scholar OpenURL Placeholder Text WorldCat Malekmotiei L. , Voyiadjis G. Z., Samadi-Dooki A., Lu F., Zhou J. ( 2017 ). Effect of annealing temperature on interrelation between the microstructural evolution and plastic deformation in polymers . Journal of Polymer Science Part B: Polymer Physics , 55 ( 17 ), 1286 – 1297 . Google Scholar OpenURL Placeholder Text WorldCat Medellin-Castillo H. I. , Zaragoza-Siqueiros J. ( 2019 ). Design and manufacturing strategies for fused deposition modelling in additive manufacturing: A review . Chinese Journal of Mechanical Engineering , 32 ( 1 ), 53 . Google Scholar OpenURL Placeholder Text WorldCat Meng L. , McWilliams B., Jarosinski W., Park H.-Y., Jung Y.-G., Lee J., Zhang J. ( 2020 ). Machine learning in additive manufacturing: A review . Jom Journal of the Minerals Metals and Materials Society , 27 : (6) , 2363 – 2377 . Google Scholar OpenURL Placeholder Text WorldCat Mineo C. , Pierce S. G., Nicholson P. I., Cooper I. ( 2017 ). Journal of Computational Design and Engineering , 4 ( 3 ), 192 – 202 . Mohamed O. A. , Masood S. H., Bhowmik J. L. ( 2016a ). Analytical modelling and optimization of the temperature-dependent dynamic mechanical properties of fused deposition fabricated parts made of PC-ABS . Materials , 9 ( 11 ), 895 . Google Scholar OpenURL Placeholder Text WorldCat Mohamed O. A. , Masood S. H., Bhowmik J. L. ( 2016b ). Optimization of fused deposition modeling process parameters for dimensional accuracy using I-optimality criterion . Measurement , 81 , 174 – 196 . Google Scholar OpenURL Placeholder Text WorldCat Mohamed O. A. , Masood S. H., Bhowmik J. L. ( 2017 ). Influence of processing parameters on creep and recovery behavior of FDM manufactured part using definitive screening design and ANN . Rapid Prototyping Journal , 23 ( 6 ), 998 – 1010 . Google Scholar OpenURL Placeholder Text WorldCat Mokhtarian H. , Hamedi A., Nagarajan H. P., Panicker S., Coatanéa E., Haapala K. ( 2019 ). Probabilistic modelling of defects in additive manufacturing: A case study in powder bed fusion technology . Procedia CIRP , 81 , 956 – 961 . Google Scholar OpenURL Placeholder Text WorldCat Montazeri M. , Nassar A. R., Dunbar A. J., Rao P. ( 2020 ). In-process monitoring of porosity in additive manufacturing using optical emission spectroscopy . IISE Transactions , 52 ( 5 ), 500 – 515 . Google Scholar OpenURL Placeholder Text WorldCat Moon S. K. , Tan Y. E., Hwang J., Yoon Y.-J. ( 2014 ). Application of 3D printing technology for designing light-weight unmanned aerial vehicle wing structures . International Journal of Precision Engineering and Manufacturing-Green Technology , 1 ( 3 ), 223 – 228 . Google Scholar OpenURL Placeholder Text WorldCat Murphy R. D. , Imediegwu C., Hewson R., Santer M. J., Muir M. ( 2019 ). Multiscale concurrent optimization towards additively manufactured structures . In AIAA Scitech 2019 Forum . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Mycroft W. , Katzman M., Tammas-Williams S., Hernandez-Nava E., Panoutsos G., Todd I., Kadirkamanathan V. ( 2020 ). A data-driven approach for predicting printability in metal additive manufacturing processes . Journal of Intelligent Manufacturing , 31 , 1769 – 1781 . Google Scholar OpenURL Placeholder Text WorldCat Nagarajan H. P. , Mokhtarian H., Jafarian H., Dimassi S., Bakrani-Balani S., Hamedi A., Haapala K. R. ( 2019 ). Knowledge-based design of artificial neural network topology for additive manufacturing process modeling: A new approach and case study for fused deposition modeling . Journal of Mechanical Design , 141 ( 2) , 021705 . Google Scholar OpenURL Placeholder Text WorldCat Nelson J. , Rennie A., Abram T., Adiele A., Wood M., Tripp M., Galloway G. ( 2015 ). Effect of process conditions on temperature distribution in the powder bed during laser sintering of polyamide-12 . Journal of Thermal Engineering , 1 ( 3 ), 159 – 165 . Google Scholar OpenURL Placeholder Text WorldCat Nguyen C. H. P. , Choi Y. ( 2020 ). Concurrent density distribution and build orientation optimization of additively manufactured functionally graded lattice structures . Computer-Aided Design , 127 , 102884 . Google Scholar OpenURL Placeholder Text WorldCat Nguyen C. H. P. , Kim Y., Choi Y. ( 2019 ). Design for additive manufacturing of functionally graded lattice structures: A design method with process induced anisotropy consideration . International Journal of Precision Engineering and Manufacturing-Green Technology , 1 – 17 ., In Press, DOI: https://doi.org/10.1007/s40684-019-00173-7. Google Scholar OpenURL Placeholder Text WorldCat Olleak A. , Xi Z. ( 2020 ). Calibration and validation framework for selective laser melting process based on multi-fidelity models and limited experiment data . Journal of Mechanical Design , 142 (8) , 081801 ). Google Scholar OpenURL Placeholder Text WorldCat Overvelde J. T. , Weaver J. C., Hoberman C., Bertoldi K. ( 2017 ). Rational design of reconfigurable prismatic architected materials . Nature , 541 ( 7637 ), 347 – 352 . Google Scholar OpenURL Placeholder Text WorldCat Padhi S. K. , Sahu R. K., Mahapatra S., Das H. C., Sood A. K., Patro B., Mondal A. ( 2017 ). Optimization of fused deposition modeling process parameters using a fuzzy inference system coupled with Taguchi philosophy . Advances in Manufacturing , 5 ( 3 ), 231 – 242 . Google Scholar OpenURL Placeholder Text WorldCat Panda B. N. , Bahubalendruni M. R., Biswal B. B. ( 2015 ). A general regression neural network approach for the evaluation of compressive strength of FDM prototypes . Neural Computing and Applications , 26 ( 5 ), 1129 – 1136 . Google Scholar OpenURL Placeholder Text WorldCat Panda B. , Leite M., Biswal B. B., Niu X., Garg A. ( 2018 ). Experimental and numerical modelling of mechanical properties of 3D printed honeycomb structures . Measurement , 116 , 495 – 506 . Google Scholar OpenURL Placeholder Text WorldCat Pandey P. M. , Thrimurthulu K., Reddy N. V. ( 2004 ). Optimal part deposition orientation in FDM by using a multicriteria genetic algorithm . International Journal of Production Research , 42 ( 19 ), 4069 – 4089 . Google Scholar OpenURL Placeholder Text WorldCat Pandita P. , Bilionis I., Panchal J. ( 2019 ). Bayesian optimal design of experiments for inferring the statistical expectation of expensive black-box functions . Journal of Mechanical Design , 141 ( 10 ), 101404 . Google Scholar OpenURL Placeholder Text WorldCat Panesar A. , Abdi M., Hickman D., Ashcroft I. ( 2018 ). Strategies for functionally graded lattice structures derived using topology optimisation for additive manufacturing . Additive Manufacturing , 19 , 81 – 94 . Google Scholar OpenURL Placeholder Text WorldCat Paul A. , Mozaffar M., Yang Z., Liao W.-k., Choudhary A., Cao J., Agrawal A. ( 2019 ). A Real-Time Iterative Machine Learning Approach for Temperature Profile Prediction in Additive Manufacturing Processes , IEEE International Conference on Data Science and Advanced Analytics (DSAA) , Washington, DC , USA , pp. 541 – 550 ., doi: 10.1109/DSAA.2019.00069 . Google Scholar OpenURL Placeholder Text WorldCat Peng A. , Xiao X., Yue R. ( 2014 ). Process parameter optimization for fused deposition modeling using response surface methodology combined with fuzzy inference system . The International Journal of Advanced Manufacturing Technology , 73 ( 1–4 ), 87 – 100 . Google Scholar OpenURL Placeholder Text WorldCat Prajapati H. , Chalise D., Ravoori D., Taylor R. M., Jain A. ( 2019 ). Improvement in build-direction thermal conductivity in extrusion-based polymer additive manufacturing through thermal annealing . Additive Manufacturing , 26 , 242 – 249 . Google Scholar OpenURL Placeholder Text WorldCat Qattawi A. ( 2018 ). Investigating the effect of fused deposition modeling processing parameters using Taguchi design of experiment method . Journal of Manufacturing Processes , 36 , 164 – 174 . Google Scholar OpenURL Placeholder Text WorldCat Qattawi A. , Alrawi B., Guzman A. ( 2017 ). Experimental optimization of fused deposition modelling processing parameters: A design-for-manufacturing approach . Procedia Manufacturing , 10 , 791 – 803 . Google Scholar OpenURL Placeholder Text WorldCat Qin J. , Liu Y., Grosvenor R., Lacan F., Jiang Z. ( 2020 ). Deep learning-driven particle swarm optimisation for additive manufacturing energy optimisation . Journal of Cleaner Production , 245 , 118702 . Google Scholar OpenURL Placeholder Text WorldCat Richter T. , Watschke H., Schumacher F., Vietor T. ( 2018 ). Exploitation of potentials of additive manufacturing in ideation workshops . In DS 89: Proceedings of the Fifth International Conference on Design Creativity (ICDC 2018) . Bath, UK : University of Bath . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rizzi F. , Khalil M., Jones R. E., Templeton J. A., Ostien J. T., Boyce B. L. ( 2019 ). Bayesian modeling of inconsistent plastic response due to material variability . Computer Methods in Applied Mechanics and Engineering , 353 , 183 – 200 . Google Scholar OpenURL Placeholder Text WorldCat Roh B.-M. , Kumara S. R., Simpson T. W., Michaleris P., Witherell P., Assouroko I. ( 2016 ). Ontology-based laser and thermal metamodels for metal-based additive manufacturing . In International Design Engineering Technical Conferences and Computers and Information in Engineering Conference . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rosen D. W. ( 2015 ). A set-based design method for material-geometry structures by design space mapping . In International Design Engineering Technical Conferences and Computers and Information in Engineering Conference . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Roy M. , Wodo O. ( 2020 ). Data-driven modeling of thermal history in additive manufacturing . Additive Manufacturing , 32 , 101017 . Google Scholar OpenURL Placeholder Text WorldCat Saad M. S. , Nor A. M., Baharudin M. E., Zakaria M. Z., Aiman A. ( 2019 ). Optimization of surface roughness in FDM 3D printer using response surface methodology, particle swarm optimization, and symbiotic organism search algorithms . The International Journal of Advanced Manufacturing Technology , 105 ( 12 ), 5121 – 5137 . Google Scholar OpenURL Placeholder Text WorldCat Sabbaghi A. , Huang Q. ( 2018 ). Model transfer across additive manufacturing processes via mean effect equivalence of lurking variables . The Annals of Applied Statistics , 12 ( 4 ), 2409 – 2429 . Google Scholar OpenURL Placeholder Text WorldCat Sabbaghi A. , Huang Q., Dasgupta T. ( 2018 ). Bayesian model building from small samples of disparate data for capturing in-plane deviation in additive manufacturing . Technometrics , 60 ( 4 ), 532 – 544 . Google Scholar OpenURL Placeholder Text WorldCat Sanfilippo E. M. , Belkadi F., Bernard A. ( 2019 ). Ontology-based knowledge representation for additive manufacturing . Computers in Industry , 109 , 182 – 194 . Google Scholar OpenURL Placeholder Text WorldCat Sheoran A. J. , Kumar H. ( 2020 ). Fused Deposition modeling process parameters optimization and effect on mechanical properties and part quality: Review and reflection on present research . Materials Today: Proceedings , 21 , 1659 – 1672 . Google Scholar OpenURL Placeholder Text WorldCat Simonelli M. , Tse Y. Y., Tuck C. ( 2014 ). Effect of the build orientation on the mechanical properties and fracture modes of SLM Ti–6Al–4V . Materials Science and Engineering: A , 616 , 1 – 11 . Google Scholar OpenURL Placeholder Text WorldCat Simpson T. W. , Poplinski J., Koch P. N., Allen J. K. ( 2001 ). Metamodels for computer-based engineering design: Survey and recommendations . Engineering with Computers , 17 ( 2 ), 129 – 150 . Google Scholar OpenURL Placeholder Text WorldCat Sood A. K. , Ohdar R. K., Mahapatra S. S. ( 2010 ). Parametric appraisal of mechanical property of fused deposition modelling processed parts . Materials & Design , 31 ( 1 ), 287 – 295 . Google Scholar OpenURL Placeholder Text WorldCat Sood A. K. , Ohdar R. K., Mahapatra S. S. ( 2012 ). Experimental investigation and empirical modelling of FDM process for compressive strength improvement . Journal of Advanced Research , 3 ( 1 ), 81 – 90 . Google Scholar OpenURL Placeholder Text WorldCat Sossou G. , Demoly F., Montavon G., Gomes S. ( 2018 ). An additive manufacturing oriented design approach to mechanical assemblies . Journal of Computational Design and Engineering , 5 ( 1 ), 3 – 18 . Google Scholar OpenURL Placeholder Text WorldCat Srivastava M. , Rathee S. ( 2018 ). Optimisation of FDM process parameters by Taguchi method for imparting customised properties to components . Virtual and Physical Prototyping , 13 ( 3 ), 203 – 210 . Google Scholar OpenURL Placeholder Text WorldCat Standard A. ( 2012 ). ISO/ASTM 52900: 2015 additive manufacturing-general principles-terminology . ASTM F2792-10e1 . Google Scholar OpenURL Placeholder Text WorldCat Sun H. , Rao P. K., Kong Z. J., Deng X., Jin R. ( 2017 ). Functional quantitative and qualitative models for quality modeling in a fused deposition modeling process . IEEE Transactions on Automation Science and Engineering , 15 ( 1 ), 393 – 403 . Google Scholar OpenURL Placeholder Text WorldCat Tapia G. , Elwany A., Sang H. ( 2016 ). Prediction of porosity in metal-based additive manufacturing using spatial Gaussian process models . Additive Manufacturing , 12 , 282 – 290 . Google Scholar OpenURL Placeholder Text WorldCat Tapia G. , Khairallah S., Matthews M., King W. E., Elwany A. ( 2018 ). Gaussian process-based surrogate modeling framework for process planning in laser powder-bed fusion additive manufacturing of 316L stainless steel . The International Journal of Advanced Manufacturing Technology , 94 ( 9–12 ), 3591 – 3603 . Google Scholar OpenURL Placeholder Text WorldCat Tian W. , Ma J., Alizadeh M. ( 2019 ). Energy consumption optimization with geometric accuracy consideration for fused filament fabrication processes . The International Journal of Advanced Manufacturing Technology , 103 ( 5–8 ), 3223 – 3233 . Google Scholar OpenURL Placeholder Text WorldCat Toal D. J. , Bressloff N. W., Keane A. J. ( 2008 ). Kriging hyperparameter tuning strategies . AIAA Journal , 46 ( 5 ), 1240 – 1252 . Google Scholar OpenURL Placeholder Text WorldCat Torrey L. , Shavlik J. ( 2010 ). Transfer learning . In Handbook of research on machine learning applications and trends: Algorithms, methods, and techniques , (pp. 242 – 264 .). IGI global . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Urbanic R. , Hedrick R. ( 2016 ). Fused deposition modeling design rules for building large, complex components . Computer-Aided Design and Applications , 13 ( 3 ), 348 – 368 . Google Scholar OpenURL Placeholder Text WorldCat Vahabli E. , Rahmati S. ( 2016 ). Application of an RBF neural network for FDM parts’ surface roughness prediction for enhancing surface quality . International Journal of Precision Engineering and Manufacturing , 17 ( 12 ), 1589 – 1603 . Google Scholar OpenURL Placeholder Text WorldCat Valdevit L. , Bertoldi K., Guest J., Spadaccini C. ( 2018 ). Architected materials: synthesis, characterization, modeling, and optimal design . Journal of Materials Research , 33 ( 3 ), 241 – 246 . Google Scholar OpenURL Placeholder Text WorldCat Vining G. G. , Myers R. H. ( 1990 ). Combining Taguchi and response surface philosophies: a dual response approach . Journal of Quality Technology , 22 ( 1 ), 38 – 45 . Google Scholar OpenURL Placeholder Text WorldCat Vishwas M. , Basavaraj C., Vinyas M. ( 2018 ). Experimental investigation using Taguchi method to optimize process parameters of fused deposition modeling for abs and nylon materials . Materials Today: Proceedings , 5 ( 2 ), 7106 – 7114 . Google Scholar OpenURL Placeholder Text WorldCat Wagner S. M. , Walton R. O. ( 2016 ). Additive manufacturing's impact and future in the aviation industry . Production Planning & Control , 27 ( 13 ), 1124 – 1130 . Google Scholar OpenURL Placeholder Text WorldCat Wang G. G. , Shan S. ( 2007 ). Review of metamodeling techniques in support of engineering design optimization . Journal of Mechanical Design , 129 ( 4 ), 370 – 380 . Google Scholar OpenURL Placeholder Text WorldCat Wang Y. , Gao J., Kang Z. ( 2018 ). Level set-based topology optimization with overhang constraint: Towards support-free additive manufacturing . Computer Methods in Applied Mechanics and Engineering , 339 , 591 – 614 . Google Scholar OpenURL Placeholder Text WorldCat Wang Y. , Blache R., Zheng P., Xu X. ( 2018 ). A knowledge management system to support design for additive manufacturing using Bayesian networks . Journal of Mechanical Design , 140 ( 5) , 051701 . Google Scholar OpenURL Placeholder Text WorldCat Wang Y. , Chen Z., Houqi L., Li S. ( 2018 ). Theory and methodology for high-performance material-extrusion additive manufacturing under the guidance of force-flow . In Proceedings of the 29th Annual International Solid Freeform Fabrication Symposium – An Additive Manufacturing Conference Reviewed Paper , , Austin, Texas , USA . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Wang Z. , Liu P., Xiao Y., Cui X., Hu Z., Chen L. ( 2019 ). A data-driven approach for process optimization of metallic additive manufacturing under uncertainty . Journal of Manufacturing Science and Engineering , 141 ( 8 ), 081004 . Google Scholar OpenURL Placeholder Text WorldCat Wang C. , Tan X., Tor S. B., Lim C. ( 2020 ). Machine learning in additive manufacturing: State-of-the-art and perspectives . Additive Manufacturing , ( 36 ), 101538 . Google Scholar OpenURL Placeholder Text WorldCat Wang C. , Gu X., Zhu J., Zhou H., Li S., Zhang W. ( 2020 ). Concurrent design of hierarchical structures with three-dimensional parameterized lattice microstructures for additive manufacturing . Structural and Multidisciplinary Optimization , 61 ( 3 ), 869 – 894 . Google Scholar OpenURL Placeholder Text WorldCat Wang S. , Zhu L., Fuh J. Y. H., Zhang H., Yan W. ( 2020 ). Multi-physics modeling and Gaussian process regression analysis of cladding track geometry for direct energy deposition . Optics and Lasers in Engineering , 127 , 105950 . Google Scholar OpenURL Placeholder Text WorldCat Witherell P. , Feng S., Simpson T. W., Saint John D. B., Michaleris P., Liu Z.-K., Martukanitz R. ( 2014 ). Toward metamodels for composable and reusable additive manufacturing process models . Journal of Manufacturing Science and Engineering , 136 ( 6 ), 061025 . Google Scholar OpenURL Placeholder Text WorldCat Wu D. , Wei Y., Terpenny J. ( 2018 ). Surface roughness prediction in additive manufacturing using machine learning . In ASME 2018 13th International Manufacturing Science and Engineering Conference . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Xiong Y. , Duong P. L. T., Wang D., Park S.-I., Ge Q., Raghavan N., Rosen D. W. ( 2019 ). Data-driven design space exploration and exploitation for design for additive manufacturing . Journal of Mechanical Design , 141 ( 10 ), 101101 . Google Scholar OpenURL Placeholder Text WorldCat Xiong Y. , Park S.-I., Padmanathan S., Dharmawan A. G., Foong S., Rosen D. W., Soh G. S. ( 2019 ). Process planning for adaptive contour parallel toolpath in additive manufacturing with variable bead width . The International Journal of Advanced Manufacturing Technology , 105 ( 10 ), 4159 – 4170 . Google Scholar OpenURL Placeholder Text WorldCat Yadav D. , Chhabra D., Garg R. K., Ahlawat A., Phogat A. ( 2020 ). Optimization of FDM 3D printing process parameters for multi-material using artificial neural network . Materials Today: Proceedings , 21 , 1583 – 1591 . Google Scholar OpenURL Placeholder Text WorldCat Yan J. ( 2020 ). 3D printing optimization algorithm based on back-propagation neural network . Journal of Engineering, Design and Technology , 18 ( 5 ), 1223 – 1230 . Google Scholar OpenURL Placeholder Text WorldCat Yang S. , Tang Y., Zhao Y. F. ( 2015 ). A new part consolidation method to embrace the design freedom of additive manufacturing . Journal of Manufacturing Processes , 20 , 444 – 449 . Google Scholar OpenURL Placeholder Text WorldCat Yao X. , Moon S. K., Bi G. ( 2016 ). A cost-driven design methodology for additive manufactured variable platforms in product families . Journal of Mechanical Design , 138 ( 4 ), 041701 . Google Scholar OpenURL Placeholder Text WorldCat Yao X. , Moon S. K., Bi G. ( 2017 ). Multidisciplinary design optimization to identify additive manufacturing resources in customized product development . Journal of Computational Design and Engineering , 4 ( 2 ), 131 – 142 . Google Scholar OpenURL Placeholder Text WorldCat Yin S. , Ding S. X., Xie X., Luo H. ( 2014 ). A review on basic data-driven approaches for industrial process monitoring . IEEE Transactions on Industrial Electronics , 61 ( 11 ), 6418 – 6428 . Google Scholar OpenURL Placeholder Text WorldCat Yu C. , Jiang J. ( 2020 ). A perspective on using machine learning in 3D bioprinting . International Journal of Bioprinting , 6 , 95 . Google Scholar OpenURL Placeholder Text WorldCat Zaldivar R. , Witkin D., McLouth T., Patel D., Schmitt K., Nokes J. ( 2017 ). Influence of processing and orientation print effects on the mechanical and thermal behavior of 3D-Printed ULTEM® 9085 Material . Additive Manufacturing , 13 , 71 – 80 . Google Scholar OpenURL Placeholder Text WorldCat Zhang G. , Thai V. V. ( 2016 ). Expert elicitation and Bayesian network modeling for shipping accidents: A literature review . Safety Science , 87 , 53 – 62 . Google Scholar OpenURL Placeholder Text WorldCat Zhang J. , Wang P., Gao R. X. ( 2019 ). Deep learning-based tensile strength prediction in fused deposition modeling . Computers in industry , 107 , 11 – 21 . Google Scholar OpenURL Placeholder Text WorldCat Zhang H. , Moon S. K., Ngo T. H., Tou J., Yusoff M. A. B. M. ( 2019 ). Rapid process modeling of the aerosol jet printing based on Gaussian process regression with Latin hypercube sampling . International Journal of Precision Engineering and Manufacturing , 21 , 127 – 136 . Google Scholar OpenURL Placeholder Text WorldCat Zhang H. , Choi J. P., Moon S. K., Ngo T. H. ( 2020 ). A hybrid multi-objective optimization of aerosol jet printing process via response surface methodology . Additive Manufacturing , 33 , 101096 . Google Scholar OpenURL Placeholder Text WorldCat Zhou X. , Hsieh S.-J., Ting C.-C. ( 2018 ). Modelling and estimation of tensile behaviour of polylactic acid parts manufactured by fused deposition modelling using finite element analysis and knowledge-based library . Virtual and Physical Prototyping , 13 ( 3 ), 177 – 190 . Google Scholar OpenURL Placeholder Text WorldCat Zhou X. , Hsieh S.-J., Wang J.-C. ( 2019 ). Accelerating extrusion-based additive manufacturing optimization processes with surrogate-based multi-fidelity models . The International Journal of Advanced Manufacturing Technology , 103 ( 9–12 ), 4071 – 4083 . Google Scholar OpenURL Placeholder Text WorldCat Zhu Z. , Anwer N., Huang Q., Mathieu L. ( 2018 ). Machine learning in tolerancing for additive manufacturing . CIRP Annals , 67 ( 1 ), 157 – 160 . Google Scholar OpenURL Placeholder Text WorldCat © The Author(s) 2021. Published by Oxford University Press on behalf of the Society for Computational Design and Engineering. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com
TI - Data-driven design strategy in fused filament fabrication: status and opportunities
JO - Journal of Computational Design and Engineering
DO - 10.1093/jcde/qwaa094
DA - 2021-01-08
UR - https://www.deepdyve.com/lp/oxford-university-press/data-driven-design-strategy-in-fused-filament-fabrication-status-and-08XrawE4eT
SP - 1
EP - 1
VL - Advance Article
IS - 
DP - DeepDyve
ER -