Search

Data-driven research in Additive Manufacturing (AM) has gained significant success in recent years. This has led to a plethora of scientific literature to emerge. The knowledge in these works consists of AM and Artificial Intelligence (AI) contexts that haven't been mined and formalized in an integrated way. Moreover, no tools or guidelines exist to support data-driven knowledge transfer from one context to another. As a result, data-driven solutions using specific AI techniques are being developed and validated only for specific AM process technologies. There is a potential to exploit the inherent similarities across various AM technologies and adapt the existing solutions from one process or problem to another using AI, such as Transfer Learning. We propose a three-step knowledge transferability analysis framework in AM to support data-driven AM knowledge transfer. As a prerequisite to transferability analysis, AM knowledge is featurized into identified knowledge components. The framework consists of pre-transfer, transfer, and post-transfer steps to accomplish knowledge transfer. A case study is conducted between two flagship metal AM processes: Laser Powder Bed Fusion (LPBF) and Directed Energy Deposition (DED). The relatively mature LPBF is the source while the less developed DED is the target. We show successful transfer at different levels of the data-driven solution, including data representation, model architecture, and model parameters. The pipeline of AM knowledge transfer can be automated in the future to allow efficient cross-context or cross-process knowledge exchange.

Metal additive manufacturing has been rapidly developed for more than 2 decades; however, the nucleation mechanism (or crystal introducing) in MAM processes has been a long-term unsolved problem. This work discusses the importance of nucleation in MAM and investigates the crystal introducing mechanism in laser wire DED fabricated Ti6Al4V. A sample matrix under different fabrication conditions is fabricated under three hypotheses (stress, heat accumulation, and the solid-state transformation). It is approved from different perspectives that the recrystallization is the main reason for the crystal introducing and formation, and it is a comprehensive result of the residual stress and reheating. A more detailed investigation was launched on the early stage of a recrystallized grain at the prior beta grain boundary. The formation tendency of the recrystallized grain indicates its strong dependency on the grain boundary alpha phase. Results also show that the recrystallization does not necessarily take place at the prior beta grain boundary area, and even without grain boundaries, a single beta crystal is still able to recrystallize under the fabrication condition of laser wire DED. Some of these recrystallized grains are able to participate into the following competitive grain growth, thus influence the solidification microstructure in the fabricated part dramatically.

Dispatching strategies for gas turbines (GTs) are changing in modern electricity grids. A growing incorporation of intermittent renewable energy requires GTs to operate more but shorter cycles and more frequently on partial loads. Deep reinforcement learning (DRL) has recently emerged as a tool that can cope with this development and dispatch GTs economically. The key advantages of DRL are a model-free optimization and the ability to handle uncertainties, such as those introduced by varying loads or renewable energy production. In this study, three popular DRL algorithms are implemented for an economic GT dispatch problem on a case study in Alberta, Canada. We highlight the benefits of DRL by incorporating an existing thermodynamic software provided by Siemens Energy into the environment model and by simulating uncertainty via varying electricity prices, loads, and ambient conditions. Among the tested algorithms and baseline methods, Deep Q-Networks (DQN) obtained the highest rewards while Proximal Policy Optimization (PPO) was the most sample efficient. We further propose and implement a method to assign GT operation and maintenance cost dynamically based on operating hours and cycles. Compared to existing methods, our approach better approximates the true cost of modern GT dispatch and hence leads to more realistic policies.

Recent applications of machine learning in metal additive manufacturing (MAM) have shown great potential to resolve critical barriers to MAM's widespread adoption. Recent research on the topic highlights the significance of using melt pool signatures to predict defects on-the-fly. While high-fidelity melt pool image data has the potential to enable accurate predictions, hardly any work exists on the use of state-of-the-art spatiotemporal models to leverage the information embedded in the transient and sequential nature of the additive process. This work introduces and implements some of the major deep spatiotemporal learners that can be adapted to classify melt pool image streams from different materials, systems, and applications. In this regard, two-stream networks with a spatial and temporal stream, a recurrent spatial network, and a factorized 3D convolutional neural network were investigated herein. The generalization abilities of these models to perturbations in the melt pool image data are tested using data perturbation techniques that are grounded in palpable process situations. The implemented architectures exhibit the ability to learn the spatiotemporal features of the melt pool image sequences. However, only the Kinetics400 pre-trained SlowFast network, which belongs to the two-stream category, showed strong generalisation abilities to data perturbations.

Geometric modelling has been a crucial component of the design process ever since the introduction of the first computer-aided design systems. Additive Manufacturing (AM) pushes design freedom to previously unachievable limits. AM allows the manufacturing of lattice structures which are otherwise close to impossible to be manufactured conventionally. Yet, the geometric modelling of heterogeneous lattice structures is still greatly limited. Thus, the AM industry is now in a situation where the manufacturing capabilities exceed the geometric modelling capabilities. While there have been advancements in the modelling of heterogeneous lattice structures, the review of relevant literature revealed critical limitations of the existing approaches. These limitations include their inability to model non-linear variation of geometric parameters, as well as the limited amount of controllable geometric parameters. This work presents a novel geometric modelling methodology based on function representation as an attempt to bridge this gap. The proposed approach avoids the manual definition of geometric parameters and provides a method to control them with mathematical functions instead. A software prototype implementing the proposed approach is presented, and several use-cases are analysed.Graphical Abstract

The design dataset is the backbone of data-driven design. Ideally, the dataset should be fairly distributed in both shape and property spaces to efficiently explore the underlying relationship. However, the classical experimental design focuses on shape diversity and thus yields biased exploration in the property space. Recently developed methods either conduct subset selection from a large dataset or employ assumptions with severe limitations. In this paper, fairness- and uncertainty-aware data generation (FairGen) is proposed to actively detect and generate missing properties starting from a small dataset. At each iteration, its coverage module computes the data coverage to guide the selection of the target properties. The uncertainty module ensures that the generative model can make certain and thus accurate shape predictions. Integrating the two modules, Bayesian optimization determines the target properties, which are thereafter fed into the generative model to predict the associated shapes. The new designs, whose properties are analyzed by simulation, are added to the design dataset. This constructs an active learning mechanism that iteratively samples new data to improve data representativeness and machine learning model performance. An S-slot design dataset case study was implemented to demonstrate the efficiency of FairGen in auxetic structural design. Compared with grid and randomized sampling, FairGen increased the coverage score at twice the speed and significantly expanded the sampled region in the property space. As a result, the generative models trained with FairGen-generated datasets showed consistent and significant reductions in mean absolute errors.

With the rapid advancement of the multimaterial additive manufacturing (AM) technology, the heterogeneous lattice structures (HLSs) comprising the multiphase materials with gradual variations have become feasible and accessible to the industry. However, the multimaterial AM capabilities have far outpaced the modeling capability of design systems to model and thus design novel HLSs. To further expand the design space for the utilization of AM technology, this paper proposes a method for modeling HLS with complex geometries and smooth material transitions. The geometric modeling and material modeling problems are formulated in a rigorous and computationally effective manner. The geometric complexity of HLS is significantly reduced by a semi-analytical unit cell decomposition strategy that is applied to split HLS into material units: struts and connectors. The smooth material transitions of the connector associated with multimaterial struts are realized by interpolating the discrete material property values defined at control points using a multiquadric radial basis function network.Graphical Abstract

The lattice structure is a type of cellular material that can achieve a variety of promising physical properties. Additive Manufacturing (AM) has relieved the difficulty of fabricating lattice structures with complex geometries. However, the quality of the AM fabricated lattice structure still needs improvement. In this paper, the influence of parameters of the Fused Deposition Modeling (FDM) process on lattice structures was investigated by the Taguchi method. S/N ratio analysis was used to find the optimal process parameters that improve the printing quality, and ANOVA provided a significance ranking of the various factors analyzed in this paper. It was found that the optimum level and significance of each process parameter vary for horizontal and inclined struts. In addition, compression tests investigate the influence of process parameters on the mechanical properties of lattice structures. The results show that process parameters optimized by print quality can also improve the elastic modulus and the ultimate strength of these lattice structures.

Multitasking or task switching has been a topic of interest and research in the field of operations management. There has been a little yet no full proof as to how increasing multitasking levels affect the performance of a whole engineering system. The goal of this paper is to introduce the concept of task switching into a network of engineers which handles and executes the quality complaints of a major aerospace firm and to observe the trends of performance measures such as average lead time, system utilisation and queuing time with increasing multitasking levels. The paper develops a mathematical model based on queuing theory and Jackson networks which is then applied to a discrete events-based simulation model.

As additive manufacturing (AM) evolves from Rapid Prototyping (RP) to the end-of-use product manufacturing process, manufacturing constraints have been largely alleviated and design freedom for part consolidation is extremely broadened. AM enabled part consolidation method promises a more effective way to achieve part count reduction and the ease of assembly compared with traditional Design for Manufacture and Assembly (DFMA) method. However, how to achieve AM enabled part consolidation is not well developed. In this paper, a new part consolidation method comprehensively considering function integration and structure optimization is proposed. This presented method is characterized by two main modules. The first one is to achieve better functionality through surface-level function integration and sequential part-level function integration based on design specifications with an initial CAD model which is designed for conventional manufacturing process. The other module is to realize better performance through the introduction and optimization of heterogeneous lattice structures according to performance requirements. The proposed part consolidation method highlights itself from the perspective of functionality achievement and performance improvement. An example of a triple clamp is studied to verify the effectiveness of the proposed model. The optimized results show that the part count has been reduced from 19 to 7 with a less weight by 20% and demonstrates better performance.

The recycling of Carbon Fibers Reinforced Plastics (CFRP) wastes is becoming increasingly important in the aerospace industry. For most of the technologies, the recycled CF (rCF) are discontinuous, misaligned and unsized. Compared to thermal treatments, the orientation, the length and the brittleness of the rCF are better preserved with the SuperCritical Water Solvolysis (SCWS). The effect of the sizing removal on the recycled CFRP behavior is studied by conducting static characterizations. Realigned virgin CF were used to manufacture sample plates which underwent short beam shear tests. An environmental assessment was carried out on the CFRP end-of-life with a focus on SCWS.

Manufacturing, where great amount of energy and materials are being consumed, should take response to have cleaner production and to improve its sustainability. Additive manufacturing (AM) technology shows potential to reduce environment impact as a more sustainable manufacturing method; however, the lack of well documented energy consumption and material flow data limits the development of Life-Cycle Inventory (LCI) analysis of AM technology. This paper presents an energy and material consumption model of Binder-Jetting (BJ) process. A Unit-Process (UP) level model is created and validated by experimental data to provide LCI data for further Life-Cycle Analysis (LCA) of BJ additive manufacturing processes. The accurate process model provides a tool to industry to understand the energy consumption and material efficiency aspect of the binder-jetting process and to allow comparisons with traditional processes.

Fully automated machining would require automatic inspection process planning and real–time inspection result feedback. As inspection process planning is still based on G&M codes containing low level information or vendor–specific bespoke routines, inspection process planning is mostly isolated from machining process planning. With the development of new data model standards such as STEP and STEP–NC providing high level product information for the entire manufacturing chain, it is conceivable that both machining and inspection process planning are considered hand–in–hand to generate optimal machining and inspection sequences with real–time measurement result feedback. This paper introduces an integrated process planning system architecture for combined machining and inspection. On–machine inspection (OMI) is chosen to carry out inspection operations in–between machining operations and to provide real–time feedback to process planning. The system is able to automatically insert OMI operations for online measurement. A case study testified the feasibility of the proposed architecture.

Laser powder bed fusion (LPBF) is an additive manufacturing (AM) process widely adopted in multiple industries for various purposes. When LPBF is used for part fabrication, determining the manufacturability of a specific design is a challenge. Therefore, this study aimed to identify a printable design using a novel approach to predict the potential printing failures of a given design via the LPBF process. A voxel-based convolutional neural network (CNN) model is developed for analyzing the design aspect, and a neural network (NN) model is applied to the process aspect. The two models are then combined to predict the manufacturability of the given design in the selected LPBF process settings. The validation samples were selected randomly, and the results verified that the developed model can accurately predict the manufacturability of the specific design. However, the proposed model is restricted by the computational power and the number of training datasets and therefore requires further investigation in this regard.

Metal additive manufacturing (MAM) technology is now widely applied in the manufacturing and remanufacturing industry, among which microstructure simulation gradually shows its importance. Traditional solidification microstructure simulation methods all have their merits as well as drawbacks when applied in MAM cases. In this work, a deterministic solidification microstructure model, named “invasion model”, is established to avoid the essential defects of traditional methods. This model focuses on the interaction between the neighboring bi-crystals instead of simulating the growth kinetics of each columnar grain or deriving the field form of variables. Within a bi-crystal system, the grain boundary tilt from thermal gradient direction is understood as a transient invasion behavior of one grain to another, and the competitive grain growth behavior along the buildup process of MAM is a summary of all the invasions in bi-crystal systems. To fill the gaps in the rapid solidification theory, a database recording the anisotropic growth effect under a rapid directional solidification condition was established with the help of an artificial neural network (ANN). Wire feed directed energy deposition (DED) fabricated Ti6Al4V thin-wall samples with full dendritic columnar grains (prior beta) are used as a benchmark to test the validation of the novel simulation model. The grain geometrical structure of reconstructed prior β grains along the build-up direction has a good agreement with the simulation result. The model can also be applied to other cases of MAM or combine with various models to achieve real-time as-solidified crystallographic feature prediction when it meets the scope of application.

This is an editorial. There should not be an abstract published.

Engineered lattice materials with pre-designed crack paths can provide critical component protection and tunable fracture energy. However, most existing crack path design methods require a change in the local topology of the material, which induces inconsistency in the material density and additional topological design complexity. In this study, the authors propose an optical programing method via digital light processing (DLP) to engineer the crack path in lattice materials without altering the material’s topological design. In this method, the effect of curing light intensity on the mechanical properties of the material is studied to develop an effective crack path engineering method, followed by the generation of a bio-inspired crack path deflection on an octet-truss lattice structure. Based on a derived theoretical model of material properties, a crack path engineering method is proposed by applying a grayscale on the desired crack path locations in DLP printing layer images. To mimic the weak crack deflection interfaces of natural materials, vertical crack deflection zones are designed with different relative positions on the octet-truss lattice structures. Consequently, a 15%–152% increase in quantitively controllable fracture energy is observed.

Laser-powder-based directed energy deposition (DED) process has been used extensively and becomes more and more popular in part manufacturing, repairing and prototype fabrication. It offers an effective way to deposit material via simultaneous delivery of the powders of desired material and laser beam. Complex physical events take place during the process and have great impacts on the success of the process, in which powder stream and its interaction with laser beam play a crucial role. In this paper, a new comprehensive numerical model for powder stream dynamics and heating process is developed to describe the coaxial powder stream and its interaction with laser beam. Reynolds averaged Navier–Stokes (RANS) approach is used for turbulent continuum gas flow while a discrete phase model (DPM) is applied to describe the dynamic behavior of the powder. Two-way coupling approach is adopted to account for the momentum transfer between gas and powder. A new thermal model is developed for powder stream, in which the interaction between powder stream and Gaussian laser beam is modelled through a novel numerical submodel. The obtained results of powder stream are compared with the experimental results from published literature and show good agreement.

Additive Manufacturing (AM) has largely relieved the design freedom of functional parts. Topology optimization has been widely used to design lightweight structures fabricated by AM. In this paper, a general design method is proposed to design solid lattice hybrid structures. An optimization algorithm is used in this method that can generate a functionally graded heterogeneous lattice structure connecting the solid part. The manufacturability can be improved due to the lattice structure supporting the overhangs. The selection of the lattice topology and the generation of the solid-lattice space are explained in detail. A hybrid element model is used to simulate the mechanical performance and optimize the material distribution of the lattice structure. To validate the design theory and the advantage of the hybrid structure, a three-point bending beam is designed by the proposed method and the existing methods. Both the simulation result and the experimental result show that the hybrid structure has a higher stiffness, yield strength, and critical buckling load than the pure solid structure and the pure lattice structure.

•A review of the microstructure characteristics and simulation methods in metal additive manufacturing.•The reviewed microstructure simulation methods all have their merits as well the essential defects.•Great progresses have been made in microstructure simulation techniques, however, many problems remain to be solved.

Easily segregated Cu-15Ni-8Sn alloy bulk material was fabricated using a selective laser melting (SLM) process. The microstructure of SLM-manufactured Cu-15Ni-8Sn alloy was investigated using optical microscopy (OM), scanning electron microscopy (SEM), X-ray diffraction (XRD), and transmission electron microscopy (TEM). Differences in the microstructures and elemental segregation of gas-atomized alloy powder, cast ingots, and SLM-manufactured samples were analyzed. The statistical average grain size of the SLM-manufactured Cu-15Ni-8Sn alloy was 4.03 μm. Microstructures of the SLM-manufactured sample were mainly composed of epitaxially grown slender cellular structures with submicron widths. Microsegregation was detected by TEM, and 80- to 200-nm Sn-enriched precipitates were dispersed between cellullar structures. Many dislocations and dislocation tangles appeared around the precipitates. An EBSD test revealed that most local misorientations within 3 degrees were concentrated in fusion line regions. Compared with cast ingots, the yield strength Rp0.2, ultimate tensile strength Rm, elongation A, and elastic modulus Eof the SLM-manufactured sample increased by 67%, 24.6%, 360%, and 7%, respectively. Moreover, the SLM-manufactured Cu-15Ni-8Sn alloy could be directly aged at 350℃ for 12 h, reaching Rm = 991.1 MPa and A =3 %, with no need for solid solution treatment or cold working.

Cellular architectures are promising in a variety of engineering applications due to attractive material properties. Additive manufacturing has reduced the difficulty in the fabrication of three-dimensional (3D) cellular materials. In this paper, the numerical homogenization method for 3D cellular materials is provided based on a short, self-contained matlab code. It is an educational description that shows how the homogenized constitutive matrix is computed by a voxel model with one material to be void and another material to be solid. A voxel generation algorithm is proposed to generate the voxel model easily by the wireframe scripts of unit cell topologies. The format of the wireframe script is defined so that the topology can be customized. The homogenization code is then extended to multimaterial cellular structures and thermal conductivity problems. The result of the numerical homogenization shows that different topologies exhibit anisotropic elastic properties to a different extent. It is also found that the anisotropy of cellular materials can be controlled by adjusting the combination of materials.