Stress-strain curves are essential for understanding the mechanical behaviour of materials, particularly in additive manufacturing. However, achieving accurate predictions for material extrusion (MEX) printed components often demands extensive experimental data, while traditional machine learning models struggle to generalise across diverse printing parameters. Here, we introduce a novel long short-term memory and temporal convolutional network with transfer learning (LSTM-TCN-TL) framework that enables accurate stress-strain predictions under varying process conditions using minimal data. Our approach employs transfer learning, adapting models trained on stress-strain data for samples printed at specific angles to predict behaviour across different infill densities. By leveraging transfer learning, the number of required specimens per process parameter is reduced from five to one, achieving an 80 % reduction in experimental workload while maintaining high predictive accuracy. Experimental results demonstrate that the LSTM-TCN model achieves high predictive accuracy on the original dataset (R2 = 0.94, RMSE = 0.07) and maintains robust performance following transfer learning (R2 = 0.81, RMSE = 0.09), even when dataset size is significantly reduced, and input variables change from angles to densities. The predicted stress-strain curves closely align with experimental results, highlighting the framework's generalisability and efficiency. This study provides a scalable and data-efficient solution for stress-strain prediction, reducing data requirements while enhancing model applicability across varying MEX process parameters, thereby advancing predictive modelling capabilities in additive manufacturing.

3D printing has become widely-applied for manufacturing sacrificial moulds. However, conventional mould designs often involve simple geometries like cube, leading to excessive material usage and high production costs. The moulds will need to be removed after casting, resulting in wasted mould material. Currently, with the development of additive manufacturing technologies, complex and light-weight mould structures can be fabricated easily with a low cost. Therefore, this study proposes a stress-guided lightweight design methodology for 3D printing sacrificial moulds. The approach begins by applying an outward offset to the target model, generating a uniform-thickness shell with the inner surface aligning with the target model's geometry. The shell thickness is then optimised based on stress distribution to ensure that the mould can withstand fluid pressure during injection moulding, forming a non-uniform thickness shell model. The potential leakage problem due to 3D printing interlayer gaps is also considered and optimised in our proposed method. Experimental validation demonstrates that the optimised shell model achieves accurate casting of the target geometry. Compared to traditional cube-shaped sacrificial moulds, it achieves up to 94.7% reduction in volume, 95.01% material saving, and 83% improvement in fabrication time, all while maintaining structural stability. This method offers a practical solution for cost-effective and efficient 3D printing sacrificial mould design.

Abstract The current boom in soft robotics development has spurred extensive research into these flexible, deformable, and adaptive robotic systems. However, the unique characteristics of soft materials, such as non-linearity and hysteresis, present challenges in modeling, calibration, and control, laying the foundation for a compelling exploration based on finite element analysis (FEA), machine learning (ML), and digital twins (DT). Therefore, in this review paper, we present a comprehensive exploration of the evolving field of soft robots, tracing their historical origins and current status. We explore the transformative potential of FEA and ML in the field of soft robotics, covering material selection, structural design, sensing, control, and actuation. In addition, we introduce the concept of DT for soft robots and discuss its technical approaches and integration in remote operation, training, predictive maintenance, and health monitoring. We address the challenges facing the field, map out future directions, and finally conclude the important role that FEA, ML, and DT play in shaping the future of soft robots.

Deployable energy absorption structures are widely utilized in aircraft landing gear, seismic support systems, and transport vessels due to their unique designs that significantly improve energy absorption capacity. However, current studies encounter challenges related to insufficient connection strength and suboptimal energy absorption performance. To address these issues, this paper proposed an interlocking negative Poisson's ratio connector (INPR-Connector) with expansion capabilities and geometric interlocking functions, aimed at enhancing both connectivity and energy absorption. We developed two types of structures: complete structure filling (CSF) and intermediate part filling (IPF), and experimentally validated the superior connection performance and energy absorption capabilities of unit cell-generated structures under various geometric configurations. Moreover, the proposed connection structure was integrated with a rigid plate to create an expandable, bistable origami structure embedded INPR-Connector. When the load is applied, the hinge can store energy through deformation, converting the applied load into tensile forces within the horizontal flexible hinges. This structure can also recover its original shape after multiple cycles of compression, demonstrating excellent load-bearing capacity. Both numerical simulations and physical experiments confirm the effectiveness and feasibility of the designed connection structure within expandable configurations. The results indicate that this structure not only possesses adjustable energy absorption capabilities but also significantly enhances impact resistance.

Material extrusion additive manufacturing enables the efficient fabrication of polymer components, yet modelling structures with internal complexity remains challenging. This study develops a deep neural network (DNN) model to predict printing parameters, specifically infill density and printing angles, based on pointwise stress-strain data from material extrusion printed polylactic acid (PLA) specimens. Unlike conventional methods applying a single angle per specimen, this work introduces the combination of three distinct printing angles within the same specimen, achieving in-sample angle compounding and greater structural complexity. Twelve experimental groups were designed using a Taguchi L12 orthogonal array, varying infill levels of 30, 50, 70, and 100 percent, and angle sequences. Tensile tests were conducted to obtain stress-strain curves, with extracted pointwise data used as model inputs. The DNN model achieved a coefficient of determination of 0.9481 and an MSE of 0.0518, demonstrating strong prediction performance. These results demonstrate that the proposed model not only captures the mechanical behaviour of compounded-structure material extrusion prints but also enables performance-driven design by recommending printing configurations that meet specific mechanical targets, reducing trial-and-error and expanding design efficiency.

Complex components can now be fabricated in innovative ways thanks to additive manufacturing (AM) technology, but it also presents a severe challenge in the detection of defects, primarily due to extensive labeling efforts of defect samples. In order to address the labeling problem in AM defect identification, this work suggests a unique strategy called the Dual-Classifier Semi-supervised Learning (DCSL) method. DCSL reduces the requirement for intensive tagging and improves detection accuracy by utilizing both labeled and unlabeled data. Two distinct classifiers, namely the one-hot classifier and the semantic classifier, are designed to train the defect class labels from different perspectives. The one-hot classifier adopts a one-hot encoding of labels, treating each class independently from the visual perspective, while the semantic classifier employs a distributed representation of labels, grouping potentially similar classes from the natural language view. The incorporation of dual classifiers transforms the proposed DCSL into a vision-language model, leveraging the semantic classifier to improve pseudo-labeling quality by identifying semantic relationships among class labels. Extensive tests on the publicly accessible AM defect dataset confirm that DCSL is more efficacious than state-of-the-art techniques for AM defect identification at the moment. The findings show that our DCSL is designed to enhance image discrimination skills towards label-free defect identification by training classifiers with different data perspectives in a synergistic manner. Given its innovative approach to defect detection, which can improve the effectiveness and precision of quality control and ultimately aid in the broad adoption of intelligent manufacturing across numerous industries, this study has the potential to herald in a new era of embodied intelligence in the current manufacturing system.

High energy dissipation materials are crucial for impact protection gear. Additionally, if these materials also have shape memory property, they can offer a better body fit and increase comfort feeling. Herein, we present a novel auxetic composite foam with ultrahigh specific energy dissipation (SED) and shape memory property, which was prepared by directly foaming with low-melting-point alloy (LMPA) in polyurethane (PU) followed by thermal compression process. Due to the synergetic action of LMPA and auxetic PU foam (APU), APU/LMPA foam showed better energy dissipation than pristine PU foam. The compression test showed the energy dissipation and SED of the APU/LMPA foam were 13.4 times and 4.8 times higher than non-APU foam, respectively. Furthermore, the SED improvement of APU/LMPA foam was much higher than other reported auxetic foams. The impact test demonstrated that APU/LMPA foam with 30% thinner thickness could reduce transmitted peak force by 62.1% compared to non-APU foam. Additionally, APU/LMPA foam exhibited shape memory effect due to the phase transition of LMPA, allowing it to adapt to different body shapes through thermal process. With its outstanding energy dissipation and shape memory properties, this composite foam is highly promising for personal safety protection, offering excellent user experience.

The growing demand for sustainable manufacturing practices has catalysed interest in integrating recycled materials into additive manufacturing (AM) processes. This review provides a comprehensive overview of recent advances, challenges, and opportunities in the use of recycled polymers, metals, composites, and glass/ceramics for AM. It examines key material sources, preprocessing techniques, and the influence of recycling on printability, mechanical performance, and part quality across various AM technologies. This review highlights current research gaps and outlines future directions for advancing the reliable and scalable use of recycled materials in AM, paving the way toward greener manufacturing ecosystems.

The forward prediction and inverse design of 4D printing have primarily focused on 2D rectangular surfaces or plates, leaving the challenge of 4D printing parts with arbitrary shapes underexplored. This gap arises from the difficulty of handling varying input sizes in machine learning paradigms. To address this, we propose a novel machine learning-driven approach for forward prediction and inverse design tailored to 4D printed hierarchical architectures with arbitrary shapes. Our method encodes non-rectangular shapes with special identifiers, transforming the design domain into a format suitable for machine learning analysis. Using Residual Networks (ResNet) for forward prediction and evolutionary algorithms (EA) for inverse design, our approach achieves accurate and efficient predictions and designs. The results validate the effectiveness of our proposed method, with the forward prediction model achieving a loss below 10 -2 mm, and the inverse optimization model maintaining an error near 1 mm, which is low relative to the entire shape of the optimized model. These outcomes demonstrate the capability of our approach to accurately predict and design complex hierarchical structures in 4D printing applications.

Herein, we report a novel hybrid auxetic foam (HAF) with high energy dissipation and self-healing properties prepared by integrating shear thickening gel (STG) with auxetic polyurethane foam (APF). Due to the synergetic action of shear thickening property of STG and the negative Poisson's ratio of APF, HAF shows better impact protection performance than APF and PU foam. The quasi-static compression test shows the energy dissipation ability of HAF is around 4 times that of APF. The dynamic impact test demonstrates that the force reduction of HAF increases by as high as 64 %, compared to APF. Notably, the force reduction improvement of the HAF is much higher than other hybrid auxetic materials. It is also found that the peak force of HAF is reduced as the amount of STG increases. Additionally, the peak force difference between HAF and APF becomes larger when they are subjected to higher impact energies, due to the rate-dependent effect of STG inside the foam. The Poisson's ratio results for HAF with different STG content under low and high compression strain rates reveal that the dimension of auxetic cell structures and STG content are required to be carefully designed to maximize the synergistic effect of auxetic property and shear thickening property. Besides, HAF demonstrates self-healing ability, allowing it to repair damage sustained during use and can be assembled like Lego blocks to make structures with any irregular shapes. Our work provides ideas for the development of advanced auxetic materials, with the potential to revolutionize a wide range of applications.

This paper introduces a methodology for optimizing 4D printing design through the integration of Residual Neural Network (ResNet) and Genetic Algorithms (GA). Departing from traditional forward design approaches, our inverse design methodology addresses both the forward prediction and inverse optimization problems. ResNet efficiently predicts the performance of 4D-printed parts given their design, while GA optimizes material allocation and stimuli distribution to achieve desired configurations. The ResNet model exhibits high accuracy, converging to a small error (10-3), as validated across diverse cases. The GA demonstrates effectiveness in achieving optimal or near-optimal solutions, illustrated through case studies shaping parts into a parabola and a sinusoid. Experimental results align with optimized and simulated outcomes, showcasing the practical applicability of our approach in 4D printing design optimization.

Additive manufacturing (AM) has undergone significant development over the past decades, resulting in vast amounts of data that carry valuable information. Numerous research studies have been conducted to extract insights from AM data and utilize it for optimizing various aspects such as the manufacturing process, supply chain, and real-time monitoring. Data integration into proposed digital twin frameworks and the application of machine learning techniques is expected to play pivotal roles in advancing AM in the future. In this paper, we provide an overview of machine learning and digital twin-assisted AM. On one hand, we discuss the research domain and highlight the machine-learning methods utilized in this field, including material analysis, design optimization, process parameter optimization, defect detection and monitoring, and sustainability. On the other hand, we examine the status of digital twin-assisted AM from the current research status to the technical approach and offer insights into future developments and perspectives in this area. This review paper aims to examine present research and development in the convergence of big data, machine learning, and digital twin-assisted AM. Although there are numerous review papers on machine learning for additive manufacturing and others on digital twins for AM, no existing paper has considered how these concepts are intrinsically connected and interrelated. Our paper is the first to integrate the three concepts big data, machine learning, and digital twins and propose a cohesive framework for how they can work together to improve the efficiency, accuracy, and sustainability of AM processes. By exploring latest advancements and applications within these domains, our objective is to emphasize the potential advantages and future possibilities associated with integration of these technologies in AM.

3-D printing, or additive manufacturing (AM), leverages 3-D computer-aided design models and numerical control to produce objects layer-by-layer, playing a key role in Industry 4.0 and Industry 5.0. Despite its potential to revolutionize manufacturing by creating complex structures more efficiently and cost-effectively, 3-D printing still faces quality issues due to a lack of sufficient data, resulting in improper process parameter settings and poor analyzability. This work introduces a data-driven and physics-assisted machine learning (DP-ML) approach for a 3-D-printed BeltClip object, integrating finite element analysis (FEA) and physics-informed machine learning (PIML). The proposed DP-ML framework provides a cost-effective and time-efficient data collection method using Digimat-AM and a warpage classification algorithm. The data collection begins with obtaining the STereoLithography (STL) file of the BeltClip object from Thingiverse and slicing it in Ultimaker (c), Cura, considering process parameters such as infill amount, toolpath pattern, layer height, print speed, and extrusion temperature. The resulting G-code file is then input into Digimat-AM for further parameter setting and analysis. In Digimat-AM, glass fiber-filled and unfilled material types are set, undergoing the virtual 3-D printing process, followed by a warpage analysis of the printed BeltClip. The collected 3-D printing data is used to build ML models-deep neural network (DNN), decision tree (DT), support vector machine (SVM), logistic regression (LR), and random forest. The DNN contains three architectures-DNN-1, DNN-2, and DNN-3. Based on the metrics of precision, recall, F1-score, and accuracy, DNN-3 outperforms the others and is chosen for the warpage classification algorithm. The presented DP-ML approach is compared with the state-of-the-art methods and shows a promising capability to predicting warpage, optimizing process parameters, and improving the overall quality and efficiency of a 3-D-printed BeltClip.

Soft robots are developed and applied in aspects such as grasping delicate objects. Their inherent flexibility also enables applications that are unattainable by humans, especially those in life-threatening environments. However, the object grasping performed by most pneumatic soft robotics during transportation requires continuous external power/force, a highly energy-consuming process, particularly for long-distance transportation. In this paper, we propose a low-melting-point alloy (LMPA)-integrated soft robot, manufactured by material extrusion additive manufacturing, requiring no power/force for holding objects during the moving process and thus presenting energy-saving characteristics. The working principles of the LMPA-integrated soft robot are as follows: (1) The LMPA is injected inside the soft robot using material extrusion. (2) The LMPA is heated to above its melting temperature so that the soft robot can change its shape. (3) At this stage, the soft robot is able to grasp an object. (4) While the soft robot is holding or grasping the object, the LMPA is cooled down to room temperature so that it turns into a solid state, and from this point onward, the soft robot can hold the object without relying on extra power for object grasping. (5) Once the soft robot arrives at the destination, the LMPA will be melted again to change the shape of the soft robot for releasing the grip and/or getting ready for another object grasping. In summary, this paper presents a case study of soft grippers, using 3D printing, specifically material extrusion, for fabricating an LMPA-integrated soft robot.

Additive manufacturing is a commonly used manufacturing method in complex part fabrication, instant assemblies, part consolidation, mass customization and personalization, on-demand manufacturing, lightweight, and topological optimization due to its advantage of lower costs, flexibility to learn and use, reduced raw material wastage, digital design integration, high efficiency, environmental-friendliness. However, the current metal 3D printing, which is mainly fabricated layer by layer using laser, is expensive to manufacture metal parts. Therefore, in this paper, a low-cost high-quality metal manufacturing process called low-melting-point alloys (LMPAs) integrated extrusion additive manufacturing will be examined. This manufacturing process can fabricate complex metal structures and integrated circuits with simple fused deposition modeling, which is a cost-effective method for producing these objects. LMPAs with different melting points are used for performance comparison to find out the optimized mechanical behavior, energy absorption properties, and electrical conductivity. Our investigation into LMPAs integrated extrusion additive manufacturing has revealed significant findings. Tensile tests conducted on LMPAs with varying melting points have illuminated distinct mechanical behaviors. Notably, lower melting points contribute to increased ductility but reduced stiffness, while higher melting points result in greater stiffness but diminished ductility. These results emphasize the importance of tailoring LMPA selection based on specific application requirements. Furthermore, our examination of lattice and triply periodic minimal surface structures has demonstrated consistent energy absorption properties across different manufacturing temperatures, highlighting the adaptability and versatility of LMPAs for energy absorption applications. Additionally, our electrical conductivity assessments have shown that LMPAs with melting points of 47^∘C and 120^∘C exhibit higher electrical conductivity, making them suitable for applications requiring good electrical conduction properties. These findings collectively underscore the significance of LMPAs in additive manufacturing, offering valuable insights for material selection and applications in various domains.

Additive manufacturing facilitates the production of parts with tailored mechanical properties, yet achieving specific stress-strain responses remains a significant challenge due to the intricate relationship between printing parameters and material behaviour. This study introduces a novel approach utilising long short-term memory (LSTM) models to predict parameter configurations for material extrusion additive manufacturing (MEX) parts, aiming to meet specific stress-strain requirements. The proposed framework transforms raw tensile test data for LSTM compatibility, yielding a coefficient of determination of 0.8648 and a mean square error of 0.1348 in inverse prediction tasks. Additionally, validation with new data resulted in an error percentage below 2%. Our approach enables the efficient design of customised parts with accurate mechanical properties, reducing the need for extensive experimentation and allowing adaptation to various additive manufacturing processes.

In conventional additive manufacturing, the layer-by-layer approach leads to mechanical weaknesses, particularly in the vertical tensile strength (Z-axis) and the shear resistance between layers. The unique mechanism of mechanical enhancement found in natural materials has served as inspiration for solving the above problems. Here this study introduces two novel Interlaced Printing strategies for 3-axis printers inspired by nature. The proposed strategies involve moving the deposition head in the XY plane while periodically adjusting its height in the Z-axis, enhancing interlayer bonding and shear resistance. These strategies were closely examined to understand their impact on toolpath width and layer thickness, considering various parameters. Both strategies resulted in “dumbbell”-shaped toolpath geometries, a characteristic that can be lessened by reducing print speed. Mechanical tests revealed that objects printed using these strategies significantly outperform traditional planar toolpath methods in terms of mechanical strength, showing improvements of 31.9% and 67.5% in interlayer shear resistance. Notably, these new strategies can be combined with each other or with conventional methods, broadening their potential applications.

Owing to the rapid development in additive manufacturing, the potential to fabricate intricate structures has become a reality, emphasizing the importance of designing structures conducive to additive manufacturing processes. A crucial consideration is the ability to design structures requiring no additional support during manufacturing. This paper employs implicit B-spline representations for self-supporting structure design by integrating a topology optimization model with self-supporting constraints derived analytically from the implicit representation. This analytical derivation for detecting overhang regions enables accurate and efficient calculation of constraints, outperforming other B-spline-based methods. Compared to the traditional voxel-based methods, the implicit B-spline representation significantly expedites the optimization process by reducing the number of design variables. Additionally, several acceleration techniques are implemented to enhance the efficiency of our method, allowing simulations of 3D models with millions of finite elements to be completed within one and half an hour, excelling other B-spline-based methods and voxel-based methods. Various numerical experiments validate its excellent performance, confirming the effectiveness and efficiency of the proposed algorithm.

This paper explores the methodology for the utilization of spider web-inspired additive manufacturing to enhance overhang support structures in 3D printing. Inspired by the strength and flexibility of spider silk, we propose an approach that reduces material consumption and postprocessing efforts. The methodology includes 3D printing spider webs, addressing key questions on silk production, web strength, and printing path generation. Experimental results demonstrate substantial weight reduction in printed objects, showcasing the efficiency of spider web-inspired support compared to traditional methods. The potential applications extend to hollow shell printing and efficient mass production.

Thirty years into its development, additive manufacturing has become a mainstream manufacturing process. Additive manufacturing fabricates products by adding materials layer-by-layer directly based on a 3D model. It is able to manufacture complex parts and allows more freedom of design optimization compared with traditional manufacturing techniques. Machine learning is now a hot technology that has been used in medical diagnosis, image processing, prediction, classification, learning association, regression, etc. Currently, focuses are increasingly given to using machine learning in the manufacturing industry, including additive manufacturing. Due to the rapid development of machine learning in additive manufacturing, a special issue ‘Machine Learning in Additive Manufacturing’ in International Journal of Computer Integrated Manufacturing is organized. This paper gives a comprehensive understanding of the current status of machine learning enhanced additive manufacturing technologies for this special issue. Discussions and future perspectives are also provided.