This paper provides a comprehensive analysis of recent advancements in additive manufacturing, a transformative approach to industrial production that allows for the layer-by-layer construction of complex parts directly from digital models. Focusing specifically on Directed Energy Deposition, it begins by clarifying the fundamental principles of metal additive manufacturing as defined by International Organization of Standardization and American Society for Testing and Materials standards, with an emphasis on laser- and powder-based methods that are pivotal to Directed Energy Deposition. It explores the critical process mechanisms that can lead to defect formation in the manufactured parts, offering in-depth insights into the factors that influence these outcomes. Additionally, the unique mechanisms of defect formation inherent to Directed Energy Deposition are examined in detail. The review also covers the current landscape of process evaluation and non-destructive testing methods essential for quality assurance, including both traditional and contemporary in situ monitoring techniques, with a particular focus given to advanced machine-vision-based methods for geometric analysis. Furthermore, the integration of process monitoring, multiphysics simulation models, and data analytics is discussed, charting a forward-looking roadmap for the development of Digital Twins in Laser–Powder-based Directed Energy Deposition. Finally, this review highlights critical research gaps and proposes directions for future research to enhance the accuracy and efficiency of Directed Energy Deposition systems.

The 3D printing industry faces challenges in ensuring reliable and repeatable processes, as increasing slopes can lead to defects and faults forming layer by layer in fabricated parts. Detecting these defects through post-processing quality inspections is time-consuming and laborious. To address this, a new method is proposed that incorporates a scoring scheme to quantitatively evaluate process performance using clad height data during printing. This approach aims to save time and cost while preserving the structural integrity of the part. The study develops a layer-wise point cloud processing technique to convert the incoming unordered data streams into rasterized points. By transforming raw signals into spatially equidistant points representing clad height in a 2-D Cartesian plane, a heatmap tomography of each layer is generated. Subsequently, a novel Defects-Finder algorithm is developed to locate and cluster surface defects based on the assigned scores. The findings demonstrate the algorithm’s ability to identify the root cause of propagated faults, which can result in severe defects. Additionally, the study employs various statistical measures in the layer-level analysis to evaluate miniature process faults or shifts, which may get overshadowed, including cases of under and over-deposition. Through comparison and validation with physical artifacts, the proposed method proves effective in identifying and assessing process faults. Ultimately, this method enhances big data visualization for cost-effective quality control and boosts the overall productivity of additive manufacturing processes. By streamlining defect detection and performance evaluation, it addresses the challenges faced by the industry in ensuring reliable 3D printing outcomes.

The process of laser metal deposition (LMD) is becoming increasingly popular due to its ability to manufacture intricate components for various industrial applications. However, ensuring the reliability of the deposition process has proven to be a challenging task. Despite the control mechanisms in place during the LMD process, process shifts may occur and negatively impact the component as deviations propagate along the layer growth direction. Relying solely on previous time steps to predict depositions on subsequent time steps may not be sufficient. To achieve robust real-time geometry control, this paper introduces a spatiotemporal adjacency-based predictive model that incorporates two types of features: (1) inter-neighbouring (\(\zeta\)) and (2) intra-neighbouring (\(\xi\)) features. The former samples the melt pool height of previous layers, whilst the latter includes the neighbouring heights of the predicted value within the same layer. To validate the proposed method, nine samples of bulk structures were fabricated. Initially, five different machine learning algorithms were compared without the adjacency features, from which long short-term memory was selected for downstream analysis. Three different sets of experiments were cross-validated to check the robustness of the selected predictive model. For the cuboid structure, in all three cases, with \(\zeta = 3\) and \(\xi = 1\), the root-mean-square error (RMSE), mean absolute error (MAE), and R² score of the model is the highest at 21.040 ± 0.251 µm, 16.339 ± 0.144 µm, and 0.719 ± 0.011, respectively. Nevertheless, experimental results have shown a significant improvement when the adjacent features are incorporated as additional input variables. Comparatively, when neither adjacent features are included, the RMSE, MAE, and R² values are 29.984 ± 0.238 µm, 23.028 ± 0.100 µm, and 0.430 ± 0.032. These findings verify the effectiveness of the proposed method to increase the performance of the prediction model of the melt pool height.

Microsoft Kinect camera can capture depth images of the subject during surveillance of Human Activity Recognition (HAR) and subsequently obtain the skeletal data. Several studies have attempted to use and analyse human actions based on skeletal data and other complex feature representation extraction methods. Most authors have proposed obtaining Spatio-temporal information as one of the extraction methods. Therefore, this study automatically extracts the Spatio-temporal information from the skeletal data by using an Imaging time series (ITS) method called Recurrence Plots (RP) to transform the skeleton joint coordinates into 2D images. The raw data are preprocessed and partitioned into three-channel matrices (R, G, B) before applying the principal component analysis (PCA). The generated RP images are used as input to Convolutional Neural Network (CNN) to distinguish between different activities. The proposed method uses the UTD-MHAD dataset for benchmarking and shows that our approach outperforms previous studies with a maximum accuracy of 92.6%.