Search

While Additive Manufacturing promises to reshape the manufacturing landscape, challenges related to part, and process qualification hinder its widespread adoption. The Instance-Qualified approach seeks to qualify individual parts, even for processes with high variability, by leveraging the concept of a digital twin. This work proposes a scalable cyberphysical infrastructure to enable the construction and use of such digital twins. This work also introduces the concept of an Augmented Intelligence Relay, which allows Artificial Intelligence algorithms to predict component performance for a given application even when it is impractical to perform a large number of physical tests.

Because many of the most important defects in Laser Powder Bed Fusion (L-PBF) occur at the size and timescales of the melt pool itself, the development of methodologies for monitoring the melt pool is critical. This works examines the possibility of in-situ detection of keyholing porosity and balling instabilities. Specifically, a visible-light high speed camera with a fixed field of view is used to study the morphology of L-PBF melt pools in the Inconel 718 material system. A scale-invariant description of melt pool morphology is constructed using Computer Vision techniques and unsupervised Machine Learning is used to differentiate between observed melt pools. By observing melt pools produced across process space, in-situ signatures are identified which may indicate flaws such as those observed ex-situ. This linkage of ex-situ and in-situ morphology enabled the use of supervised Machine Learning to classify melt pools observed (with the high speed camera) during fusion of non-bulk geometries such as overhangs.

In-situ detection of processing defects is a critical challenge for Laser Powder Bed Fusion Additive Manufacturing. Many of these defects are related to interactions between the recoater blade, which spreads the powder, and the powder bed. This work leverages Deep Learning, specifically a Convolutional Neural Network (CNN), for autonomous detection and classification of many of these spreading anomalies. Importantly, the input layer of the CNN is modified to enable the algorithm to learn both the appearance of the powder bed anomalies as well as key contextual information at multiple size scales. These modifications to the CNN architecture are shown to improve the flexibility and overall classification accuracy of the algorithm while mitigating many human biases. A case study is used to demonstrate the utility of the presented methodology and the overall performance is shown to be superior to that of methodologies previously reported by the authors.

Solidification microstructure control via melt pool geometry control is investigated in detail for the AlSi10Mg alloy in a laser melting process. Relationships between cell spacing and melting process parameters—beam power and velocity, are investigated. Single beads and multi-layer pad experiments are performed and analyzed for melt pool geometry and microstructure. Information from these experiments is used to achieve location-specific control of cell spacing in a single part using an EOS laser powder bed fusion machine. Along with experiments, thermal modeling is used to map the effect of melting parameters on solidification conditions. Optical microscopy and scanning electron microscopy are used to determine the melt pool dimensions and cell spacing, respectively. Results from the model and experiments show that cell spacing can be varied by controlling the cooling rate of the melt pool via beam power and travel speed for both single bead and multi-layer pad experiments. In addition, the relationship between melt pool area and cooling rate is used to demonstrate integrated melt pool geometry and microstructure control. This is also critical for process qualification using simple observable metrics such as melt pool size. This relationship is then used to demonstrate location-specific control of cell spacing in a single component.

There is a strong desire within the metal additive manufacturing industry to immediately leverage research discoveries to improve process and part quality. Rapid transfer of knowledge from a university setting to a factory floor is greatly enhanced if researchers have direct access to the same type of equipment being used by industry. Challengingly, the operation of much of this equipment requires specially designed facilities—the specifications for which are not well reported in the prior literature. Informed by their recent experience of constructing a new metal additive manufacturing research laboratory, the authors enumerate many of the design challenges they faced and the solutions they implemented.

Despite the rapid adoption of laser powder bed fusion (LPBF) Additive Manufacturing by industry, current processes remain largely open-loop, with limited real-time monitoring capabilities. While some machines offer powder bed visualization during builds, they lack automated analysis capability. This work presents an approach for in-situ monitoring and analysis of powder bed images with the potential to become a component of a real-time control system in an LPBF machine. Specifically, a computer vision algorithm is used to automatically detect and classify anomalies that occur during the powder spreading stage of the process. Anomaly detection and classification are implemented using an unsupervised machine learning algorithm, operating on a moderately-sized training database of image patches. The performance of the final algorithm is evaluated, and its usefulness as a standalone software package is demonstrated with several case studies.

Understanding in-situ melt pool dynamics has become a major research focus for the metals additive manufacturing community. Much of the research on melt pool dynamics in the Laser Powder Bed Fusion processes makes use of a high speed visible-light or infrared camera. Because these cameras are typically installed “after-market” on pre-existing AM machines they are often limited to a fixed field of view. This work presents a method for efficiently converting images captured by a high speed camera with a fixed field of view into a more useful format via a coordinate transform. Specifically, images of the melt pool are transformed such that they appear to have been captured with a camera integrated coaxially with the laser optic train.

Increasing industry acceptance of powder bed metal Additive Manufacturing requires improved real-time detection and classification of anomalies. Many of these anomalies, such as recoater blade impacts, binder deposition issues, spatter generation, and some porosities, are surface-visible at each layer of the building process. In this work, the authors present a novel Convolutional Neural Network architecture for pixel-wise localization (semantic segmentation) of layer-wise powder bed imaging data. Key advantages of the algorithm include its ability to return segmentation results at the native resolution of the imaging sensor, seamlessly transfer learned knowledge between different Additive Manufacturing machines, and provide real-time performance. The algorithm is demonstrated on six different machines spanning three technologies: laser fusion, binder jetting, and electron beam fusion. Finally, the performance of the algorithm is shown to be superior to that of previous algorithms presented by the authors with respect to localization, accuracy, computation time, and generalizability.

Expanding on prior process mapping work by the authors, multiple melt pool cross-sections are measured at multiple process parameter combinations for the Inconel 718 alloy in a Laser Powder Bed Fusion (L-PBF) process. Collection of such data enables the study of the variability of melt pool geometry (e.g. width, depth, and cross-sectional area) across process space. Furthermore, the statistical distribution of the measured melt pool geometries is compared to that of an equivalent normal distribution and intriguing outliers are observed. The cross-sectional morphology of the melt pools are associated with defects such as keyholing porosity and balling and the variability of the defects is quantified. The final product of this work is a robust description of L-PBF In718 melt pool behavior, based on ex-situ observations, which can be linked to in-situ observations of melt pool morphology in future work.