Additive Manufacturing Quality Control Research Articles

Federated learning enables privacy-preserving and data-efficient dimension prediction and part qualification across additive manufacturing factories

A crucial part of quality control in additive manufacturing (AM) is the decision to accept or reject parts based on their dimensional accuracy. Machine learning (ML) models can learn nuanced relationships between process parameters and resulting geometries; however, obtaining large quantities of supervised learning data to train ML models incurs non-trivial costs. Relevant measurement data may be available at other factories or stations within the same factory but cannot be pooled together for conventional centralized learning (CL) due to privacy constraints. Here, we propose federated learning (FL) that uses private data and information from distributed sources to train ML models that predict part dimensions and inform part qualification, thereby allowing collaborative model building without compromising privacy. We manufacture and measure 405 parts having three different designs and distribute them across three to 45 factories. The performance of FL is evaluated when factories produce similar parts, parts of different designs, and parts of different qualities. When factories produce similar parts or parts of different designs, FL predicts part geometry within 10 µm of the CL benchmark and within 15 µm of the process capability limit, with as few as nine parts at each factory. FL also outperforms individual learning where factories train on their own data by up to 96% in geometry prediction, thereby providing a win-win paradigm for privacy-preserving collaborative learning. When factories produce parts of different qualities, FL outperforms individual factories only when they produce less than 15 parts locally. We show that FL performance is influenced by two factors: the quantity of data in the federation and the statistical homogeneity of data across participating factories. Overall, this research demonstrates the promise of FL for privacy-preserving, data-efficient quality control and offers key insights to design data federations in scalable AM production.

Physics-Informed Approximation of Internal Thermal History for Surface Deformation Predictions in Wire Arc Directed Energy Deposition

Abstract This work presents a physics-informed fusion methodology for deformation detection using multi-sensor thermal data. A challenge with additive manufacturing (AM) is that abnormalities commonly occur due to rapid changes in the thermal gradient. Different non-destructive in-situ thermal sensors capture parts of the thermal history but are limited by the visible temperature spectrum and sensor field of view of the fabrication process. Various sensors mitigate problems with the loss of thermal history information; however, it brings forth challenges with integrating different data streams and the need to interpolate the internal thermal history. This study develops a thermal data-informed heat flux methodology that fills the gap in fusing numerical temperature approximation with data-driven knowledge of the surface of additive manufactured (AM) components. First, this study fuses infrared (IR) thermal data complexities during the AM process with the Goldak double ellipsoidal heat flux to model the energy input into the component. Second, a thermal physics-informed model input (PIMI) is created with thermal data-informed heat flux to capture internal thermal history. Lastly, regression convolutional neural network (CNN) captures the relationship between the three-dimensional thermal gradient and surface deformation. The rapid thermal gradient formation and identification of deformation is a key step toward using thermal history data and machine learning to improve quality control in AM. The proposed surface deformation detection model achieved an MSE of 1.14 mm and an R2 of 0.89.

Deep learning based reconstruction of transient 3D melt pool geometries in laser powder bed fusion from coaxial melt pool images

Melt pool monitoring can provide guidance for quality control in additive manufacturing. Existing techniques focus on 2D melt pool characteristics. In this work, we explore in situ prediction and control of 3D melt pool geometries within a simulation environment. To reconstruct the transient 3D melt pool geometry with only the 2D coaxial image as an input, we adopt a U-net model trained with a synthetic image dataset generated from simulations. The results show that the root-mean-square error of the pixel-wise predicted melt depth is 1.14μm, demonstrating the potential of the model in enabling precise laser path-wise 3D melt pool control.

Nondestructive Quantitative Measurement for Precision Quality Control in Additive Manufacturing Using Hyperspectral Imagery and Machine Learning

Nondestructive Quantitative Measurement for Precision Quality Control in Additive Manufacturing Using Hyperspectral Imagery and Machine Learning

In-process monitoring and direct simulation of Argon shielding gas and vapour dynamics to control laser-matter interaction in laser powder bed fusion additive manufacturing

Laser powder bed fusion (L-PBF) additive manufacturing (AM) enables the fabrication of parts with precise dimensional control, freedom of design and material properties similar to or better than those fabricated using traditional manufacturing approaches. AM quality control depends upon the fundamental of the laser-matter interaction during metal AM using L-PBF to exploit the potential use of the materials and process control. In this work, thermal-fluid dynamics in gas chamber experimentally and computationally is used to elucidate the interplay between vapour, liquid, and solid phases in L-PBF. It is revealed that the argon (Ar) shielding gas flow with varied inlet velocities by different nozzles has a pronounced effect to minimise the laser-fume interaction, resulting in the reduction in unstable metal vapour flow and enhancing laser absorptivity. In-process monitoring via high-speed visualisation has been used to understand the simultaneous gas plume dynamics as a result of vapourisation and subsequent laser-fume interaction, backed up by thermal-fluid flow simulation. Unfavourable process dynamics associated with unwanted defects such as lack of fusion can be avoided to improve process design and enhance process stability.

A lattice structure coupon sample for build quality control in metal additive manufacturing

Quality control in Additive Manufacturing (AM) is vital, asprocess parameters can lead to defects, impacting mechanical properties. The present study proposes the use of alattice structure coupon sample as a proxy for build quality control. Lattices aresensitive to variations in process parameters,impacting the resulting mechanical properties, more so than bulk mechanical test specimens. Thesensitivity of lattice structures allows for the detection ofsubtle changes in process parameters, which can be valuable for quality control efforts, i.e. maintaining constant manufacturing quality. In this study, a series of lattice coupon samples were manufactured with varying factors as a proof of concept to demonstrate this concept. Non-destructive micro-computed tomography (CT) wasused to quantify the manufacturing variations/defects. The lattice coupons are subjected to quasi-static compression testing to evaluate theirmechanical properties, with differences shown indicative of changes in the manufacturing process. The work reported here acts as a proof of concept for a low-cost build quality evaluation methodology using lattice coupons and compression testing alone. When the build quality deteriorates or parameters change without the knowledge of this by the operator, this will be observed in changes in the compressive test results for lattice coupons.

Recurrence Network-Based 3D Geometry Representation Learning for Quality Control in Additive Manufacturing of Metamaterials

Abstract Metamaterials are designed with intricate geometries to deliver unique properties, and recent years have witnessed an upsurge in leveraging additive manufacturing (AM) to produce metamaterials. However, the frequent occurrence of geometric defects in AM poses a critical obstacle to realizing the desired properties of fabricated metamaterials. Advances in three-dimensional (3D) scanning technologies enable the capture of fine-grained 3D geometric patterns, thereby providing a great opportunity for detecting geometric defects in fabricated metamaterials for property-oriented quality assurance. Realizing the full potential of 3D scanning-based quality control hinges largely on devising effective approaches to process scanned point clouds and extract geometric-pertinent information. In this study, a novel framework is developed to integrate recurrence network-based 3D geometry profiling with deep one-class learning for geometric defect detection in AM of metamaterials. First, we extend existing recurrence network models that focus on image data to represent 3D point clouds, by designing a new mechanism that characterizes points’ geometric pattern affinities and spatial proximities. Then, a one-class graph neural network (GNN) approach is tailored to uncover topological variations of the recurrence network and detect anomalies associated with geometric defects in the fabricated metamaterial. The developed methodology is evaluated through comprehensive simulated and real-world case studies. Experimental results have highlighted the efficacy of the developed methodology in identifying both global and local geometric defects in AM-fabricated metamaterials.

In Situ Analysis of Curling Defects in Powder Bed Fusion of Polyamide by Simultaneous Application of Laser Profilometry and Thermal Imaging

Additive manufacturing (AM) is one of the key technologies in the global manufacturing market within various application sectors. The unique capabilities of AM enable high structural and part complexity, low material waste, and benefits in productivity by reducing design cycles and time to market. Efficient real-time quality control is still an important challenge in AM. In this paper, a real-time and in situ approach for monitoring the process in powder bed fusion of polyamide (PBF-LB/P/PA12) is proposed using the simultaneous application of two individual sensors, enabling the overlay and direct comparison of independent output data. An industrial grade laser profilometer and a thermal infrared (IR) camera were successfully integrated into a commercial system for PBF-LB/P. Artificially created curling defects were induced in a reproducible way by the manipulation of process parameters. The radiometric data was evaluated and processed into 3D topology and profile measurements to highlight peaks and curling progression. The results measured using different powder bed conditions were contrasted with corresponding thermographic data to prove the thermal visibility of curling and the influence of inhomogeneous temperature distribution on geometrical powder surface defects. The experimental setup enables the measuring of the entire powder bed surface inside the machine, with no limitations to sub-areas. Results indicate the measurable presence of curling and related temperature influences. When curling reached maximum values, inverted warpage into the negative z-direction was detected at part center as a further effect. These results can be used for improving real-time quality control in AM.

AI-Based Support System for Monitoring the Quality of a Product within Industry 4.0 Paradigm

Three-dimensional (3D) printing, also known as additive manufacturing (AM), has already shown its potential in the fourth technological revolution (Industry 4.0), demonstrating remarkable applications in manufacturing, including of medical devices. The aim of this publication is to present the novel concept of support by artificial intelligence (AI) for quality control of AM of medical devices made of polymeric materials, based on the example of our own elbow exoskeleton. The methodology of the above-mentioned inspection process differs depending on the intended application of 3D printing as well as 3D scanning or reverse engineering. The use of artificial intelligence increases the versatility of this process, allowing it to be adapted to specific needs. This brings not only innovative scientific and technological solutions, but also a significant economic and social impact through faster operation, greater efficiency, and cost savings. The article also indicates the limitations and directions for the further development of the proposed solution.

A deep learning framework for layer-wise porosity prediction in metal powder bed fusion using thermal signatures

Part quality manufactured by the laser powder bed fusion process is significantly affected by porosity. Existing works of process–property relationships for porosity prediction require many experiments or computationally expensive simulations without considering environmental variations. While efforts that adopt real-time monitoring sensors can only detect porosity after its occurrence rather than predicting it ahead of time. In this study, a novel porosity detection-prediction framework is proposed based on deep learning that predicts porosity in the next layer based on thermal signatures of the previous layers. The proposed framework is validated in terms of its ability to accurately predict lack of fusion porosity using computerized tomography (CT) scans, which achieves a F1-score of 0.75. The framework presented in this work can be effectively applied to quality control in additive manufacturing. As a function of the predicted porosity positions, laser process parameters in the next layer can be adjusted to avoid more part porosity in the future or the existing porosity could be filled. If the predicted part porosity is not acceptable regardless of laser parameters, the building process can be stopped to minimize the loss.

Numerical approach for a strongly coupled thermomechanical process with multi-phase transition and applications in additive manufacturing

<p indent="0mm">We propose a monolithic incremental Lagrangian framework based on the hot optimal transportation meshfree (HOTM) method for a strongly coupled thermomechanical process with multiphase transition. The HOTM method combines the optimal transportation meshfree method and the variational thermomechanical constitutive updates. This numerical approach is applied to conduct meshfree direct numerical simulations (DNS) for metal additive manufacturing. We present two kinds of simulations, the powder bed selective laser melting (SLM) and the laser material deposition technology referred to as extreme high-speed laser material deposition (EHLA) at the powder scale. In the simulations, the powder particles are modeled explicitly using the size distribution measured in experiments and discretized using nodes and material points. The governing equations, including the linear momentum conservation and energy conservation equations, are solved simultaneously in the meshfree framework to predict the deformation, temperature, and local state of the powder particles. A full-field constitutive model is developed to simulate multiphase flow with melting and solidification. The Lagrangian feature of the HOTM method overcomes various challenges in the DNS of melt pool thermodynamics in additive manufacturing. The proposed approach is used to quantify the influence of the main processing parameters, such as the powder size distribution, laser power, laser radius, deposition speed, powder mass flow, and axial feed on the layer thickness, surface roughness, and porosity of the bonding zone. It further enables an in-depth understanding of quality control in additive manufacturing.

Research on Morphology Detection of Metal Additive Manufacturing Process Based on Fringe Projection and Binocular Vision

This paper considers the three-dimensional (3D) shape measurement of metal parts during an additive manufacturing process in a direct energy deposition (DED) printing system with high temperature and strong light; a binocular measurement system based on ultraviolet light source projection is built using fringe projection and Fourier analysis. Firstly, ultraviolet light projection and an optical filter are used to obtain high-quality fringe patterns in an environment with thermal radiation. Then, Fourier analysis is carried out by using a single deformed fringe, and a spatial phase unwrapping algorithm is applied to obtain an unambiguous unwrapping phase, which is used as the guiding basis for the binocular matching process and 3D shape reconstruction. Finally, the accuracy of the measuring system is evaluated using a standard ball-bar gauge and the measurement error of this system is within 0.05 mm @ 100 × 100 mm. The results show that the system can measure 3D shape changes of metal parts in the additive manufacturing process. The proposed method and system have the potential to be used for online inspection and quality control of additive manufacturing.

Gaussian Process Monitoring of Layerwise-Dependent Imaging Data

Additive Manufacturing (AM) enables the direct production of complex geometries from computer-aided designs (CAD). The AM fabrication process is often executed in a layer-by-layer manner, whereby minute printing errors in one layer can manifest significant defects in the final part. In-situ quality monitoring and control are currently limited for AM processes and cause low repeatability. Recently. advanced imaging is increasingly invested in AM and leads to the proliferation of layerwise imaging data, which provides an opportunity to transform quality control of AM from post-build inspection to in-situ quality monitoring. However, existing methodologies for in-situ inspection primarily focus on key characteristics of image profiles that tend to be limited in the ability to analyze the variance components, as well as root causes and failure patterns that are critical to process improvement. This letter presents an Additive Gaussian Process with dependent layerwise correlation (AGP-D) to model the spatio-temporal correlation of layerwise imaging data for AM quality monitoring. The AGP-D consists of three independent GP modules. The first GP approximates the base profile, whereas the second and third GP capture the correlation within the same layer and among layers, respectively. Based on posterior predictions of new layers, Hotelling <b xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">T</b> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and generalized likelihood ratio (GLR) control tests are formulated to detect process shifts in the newly fabricated layer and analyze root causes. The proposed methodology is evaluated and validated using both simulation data and real-world case study of a cylinder build fabricated by a laser powder bed fusion (LPBF) machine. Experimental results show the proposed AGP-D is effective for real-time modeling and monitoring of layerwise-correlated imaging data.

Recurrence network analysis of design-quality interactions in additive manufacturing

Powder bed fusion (PBF) additive manufacturing (AM) provides a great level of flexibility in the design-driven build of metal products. However, the more complex the design, the more difficult it becomes to control the quality of AM builds. The quality challenge persistently hampers the widespread application of AM technology. Advanced imaging (e.g., X-ray computed tomography scans and high-resolution optical images) has been increasingly explored to enhance the visibility of information and improve the AM quality control. Realizing the full potential of imaging data depends on the advent of information processing methodologies for the analysis of design-quality interactions. This paper presents a design of AM experiment to investigate how design parameters (e.g., build orientation, thin-wall width, thin-wall height, and contour space) interact with quality characteristics in thin-wall builds. Note that the build orientation refers to the position of thin-walls in relation to the recoating direction on the plate, and the contour space indicates the width between rectangle hatches. First, we develop a novel generalized recurrence network (GRN) to represent the AM spatial image data. Then, GRN quantifiers, namely degree, betweenness, pagerank, closeness, and eigenvector centralities, are extracted to characterize the quality of layerwise builds. Further, we establish a regression model to predict how the design complexity impacts GRN behaviors in each layer of thin-wall builds. Experimental results show that network features are sensitive to build orientations, width, height, and contour space under the significant level α = 0.05. Thin-walls with the width bigger than 0.1 mm printed under orientation 0° are found to yield better quality compared to 60° and 90°. Also, thin-walls build with orientation 60° are more sensitive to the changes in contour space compare to the other two orientations. As a result, the orientation 60° should be avoided while printing thin-wall structures. The proposed design-quality analysis shows great potential to optimize engineering design and enhance the quality of PBF-AM builds.

Uncertainty quantification and reduction in metal additive manufacturing

Uncertainty quantification (UQ) in metal additive manufacturing (AM) has attracted tremendous interest in order to dramatically improve product reliability. Model-based UQ, which relies on the validity of a computational model, has been widely explored as a potential substitute for the time-consuming and expensive UQ solely based on experiments. However, its adoption in the practical AM process requires overcoming two main challenges: (1) the inaccurate knowledge of uncertainty sources and (2) the intrinsic uncertainty associated with the computational model. Here, we propose a data-driven framework to tackle these two challenges by combining high throughput physical/surrogate model simulations and the AM-Bench experimental data from the National Institute of Standards and Technology (NIST). We first construct a surrogate model, based on high throughput physical simulations, for predicting the three-dimensional (3D) melt pool geometry and its uncertainty with respect to AM parameters and uncertainty sources. We then employ a sequential Bayesian calibration method to perform experimental parameter calibration and model correction to significantly improve the validity of the 3D melt pool surrogate model. The application of the calibrated melt pool model to UQ of the porosity level, an important quality factor, of AM parts, demonstrates its potential use in AM quality control. The proposed UQ framework can be generally applicable to different AM processes, representing a significant advance toward physics-based quality control of AM products.

Thermal tomography 3D imaging of additively manufactured metallic structures

Thermal tomography is a computational method for heat diffusion-based imaging of solids, which provides 3D visualization of data from flash thermography measurements. We investigate thermal tomography imaging and nondestructive evaluation of stainless steel and nickel super alloy metallic structures produced with the laser powder bed fusion (LPBF) additive manufacturing (AM) process. Metallic structures produced with LPBF contain defects, and there are limited capabilities to evaluate these structures non-destructively. Thermal tomography reconstruction of 3D apparent spatial effusivity provides information about AM structure geometry and internal material flaws. We study performance of thermal tomography in imaging of metallic structures through COMSOL computer simulations of transient heat transfer and through reconstruction of data obtained from experimental measurements. Thermal tomography reconstructions of structure shape and dimensions are shown for the Inconel 718 AM structure which has variations in the horizontal plane but is uniform along the depth dimension. Reconstruction of internal defects is investigated using a stainless steel 316L specimen with flat bottom hole (FBH) indentations, and the Inconel 718 plate is produced with the LPBF method, which contains imprinted hemispherical shape low density regions containing non-sintered metallic powder. The FBHs have the same sizes as the imprinted defects in the LPBF specimens but offer better imaging contrast. Thermal tomography reconstructions provide visualizations of internal defects and allow for estimation of their sizes and locations. Results of this study demonstrate that thermal tomography can be used for visualization and quality control in AM.

Defect detection in additive manufacturing via a toolpath overlaid melt-pool-temperature tomography

Additive manufacturing of metals has emerged as a potential technology for companies to create highly integrated and individualized products. In particular, powder-based laser metal deposition has advantages such as flexibility and multimaterial capabilities. It is possible to mix powders and create alloys inside the melt-pool during the build process. Consequently, purpose made material combinations with set or even varying thermal properties can be realized. Inherently, the process becomes increasingly challenging because of the great number of variables. Analyzation of the manufactured part ensures top quality and detects errors and defects. To accomplish this, specimens have to be x-rayed or ground and cut into microsections. In order to save time and keep the parts’ integrity, a new method uses temperature data from the process to determine irregularities. During the additive manufacturing process, a 680 W diode laser melts the substrate and the powder locally. The powder is composed of 42% nickel and 58% iron. A pyrometer samples the temperature of the molten pool at a spectral range from 1.45 to 1.85 μm. The recorded data are mapped onto the toolpath of the process head. A script converts the time dependent signal to spatially resolved temperature points. The feedrate and the laser status aid to synchronize the data throughout. As a result, the overlaid melt-pool temperature visualizes the process and creates a tomography for the produced part. Initial experiments show that errors and defects like porosities and cavities are identifiable inside the manufactured structure. Furthermore, correlations between the visualization and errors detected with microsections are possible. Overall, this technique is an addition to the repertoire of data visualization and quality control in additive manufacturing and can be transferred to other machines and laser processes.

Decision-Level Data Fusion in Quality Control and Predictive Maintenance

Data fusion integrates data from multiple sources to improve prediction performance. While significant research has been conducted to develop data-level and feature-level fusion methods, very few studies are performed to develop more effective decision-level data fusion methods. This research aims at developing a decision-level data fusion approach that transforms low-dimensional decisions (i.e., predictions) made based on individual sensor data such as temperature and vibration to high-dimensional decisions. Integration of these high-dimensional decisions is formulated as a convex optimization problem rather than a traditional multivariate linear regression problem. The proposed decision-level data fusion approach is demonstrated in two cases: 1) quality control in additive manufacturing and 2) predictive maintenance in aircraft engines. Experimental results have shown that the proposed decision-level fusion method can reduce prediction variance by at least 30% as well as increase prediction accuracy by 45%.

Development of 18 Quality Control Gates for Additive Manufacturing of Error Free Patient-Specific Implants.

Unlike subtractive manufacturing technologies, additive manufacturing (AM) can fabricate complex shapes from the macro to the micro scale, thereby allowing the design of patient-specific implants following a biomimetic approach for the reconstruction of complex bone configurations. Nevertheless, factors such as high design variability and changeable customer needs are re-shaping current medical standards and quality control strategies in this sector. Such factors necessitate the urgent formulation of comprehensive AM quality control procedures. To address this need, this study explored and reported on a variety of aspects related to the production and the quality control of additively manufactured patient-specific implants in three different AM companies. The research goal was to develop an integrated quality control procedure based on the synthesis and the adaptation of the best quality control practices with the three examined companies and/or reported in literature. The study resulted in the development of an integrated quality control procedure consisting of 18 distinct gates based on the best identified industry practices and reported literature such as the Food and Drug Administration (FDA) guideline for AM medical devices and American Society for Testing and Materials (ASTM) standards, to name a few. This integrated quality control procedure for patient-specific implants seeks to prepare the AM industry for the inevitable future tightening in related medical regulations. Moreover, this study revealed some critical success factors for companies developing additively manufactured patient-specific implants, including ongoing research and development (R&D) investment, investment in advanced technologies for controlling quality, and fostering a quality improvement organizational culture.

Quality assurance in metal powder bed fusion via deep-learning-based image classification

Purpose This paper aims to establish a standardized, quick, reliable and cost-efficient method of quality control (QC) in metal powder bed fusion (PBFM) based on process monitoring data. Design/methodology/approach Based on destructive testing results that emerged from a statistical investigation on powder bed fusion process exceeding reproducibility of mechanical properties, it was investigated if the generated monitoring data from a concept laser machine allows reliable deductions on resulting mechanical properties of the manufactured specimens. Findings The application of machine learning on generated melt pool images, under-recognition of destructive testing results, enables enhanced pattern recognition. The generated computational model successfully classified 9,280 unseen layer images by 98.9 per cent accuracy. This finding offers an automated approach to quality control within PBFM. Originality/value To the authors knowledge, it is the first time that machine learning has been applied for the purpose of QC in additive manufacturing. The ability of deep convolutional neural networks to recognize patterns, which are imperceptible to the human eye, shows high potential to facilitate activities of QC and to minimize QC-related costs and throughput times. The achieved processing speed for image analyses also points a way for future developments of self-corrective PBFM systems.