Felipe Lopez

Identifying uncertainty in laser powder bed fusion additive manufacturing models

As additive manufacturing (AM) matures, models are beginning to take a more prominent stage in design and process planning. A limitation frequently encountered in AM models is a lack of indication about their precision and accuracy. Often overlooked, model uncertainty is required for validation of AM models, qualification of AM-produced parts, and uncertainty management. This paper presents a discussion on the origin and propagation of uncertainty in laser powder bed fusion (L-PBF) models. Four sources of uncertainty are identified: modeling assumptions, unknown simulation parameters, numerical approximations, and measurement error in calibration data. Techniques to quantify uncertainty in each source are presented briefly, along with estimation algorithms to diminish prediction uncertainty with the incorporation of online measurements. The methods are illustrated with a case study based on a thermal model designed for melt pool width predictions. Model uncertainty is quantified for single track experiments, and the effect of online estimation in overhanging structures is studied via simulation. [DOI: 10.1115/1.4034103]

Particle filtering on GPU architectures for manufacturing applications

HighlightsWe implemented a particle filter and an auxiliary particle filter on a GeForce GTX TITAN GPU.Both estimators are able to meet real-time operating constraints in a remelting process.The fully adapted auxiliary particle filter is more efficient in this peaked-likelihood case.The auxiliary particle filter is 40 times faster than the particle filter for the same accuracy. Particle filters are nonlinear estimators that can be used to detect anomalies in manufacturing processes. Although promising, their high computational cost often prevents their implementation in real-time applications. Recently, the introduction of graphics processing units (GPUs) has enabled the acceleration of computationally intensive processes with their massive parallel capabilities. This article presents the acceleration of the particle filter and the auxiliary particle filter, two of the most important particle methods, on a GPU using NVIDIA CUDA technology. This is illustrated via simulation for a remelting process where the accelerated algorithms return accurate estimates while still being two orders of magnitude faster than the physical process even for calculations that involve millions of particles.

A Method for Characterizing Model Fidelity in Laser Powder Bed Fusion Additive Manufacturing

As Additive Manufacturing (AM) matures as a technology, modeling methods have become increasingly sought after as a means for improving process planning, monitoring and control. For many, modeling offers the potential to complement, and in some cases perhaps ultimately supplant, tedious part qualification processes. Models are tailored for specific applications, focusing on specific predictions of interest. Such predictions are obtained with different degrees of fidelity. Limited knowledge of model fidelity hinders the user’s ability to make informed decisions on the selection, use, and reuse of models. A detailed study of the assumptions and approximations adopted in the development of models could be used to identify their predictive capabilities. This could then be used to estimate the level of fidelity to be expected from the models. This paper conceptualizes the modeling process and proposes a method to characterize AM models and ease the identification and communication of their capabilities, as determined by assumptions and approximations. An ontology is leveraged to provide structure to the identified characteristics. The resulting ontological framework enables the sharing of knowledge about indicators of model fidelity, through semantic query and knowledge browsing capabilities.

Identifying uncertainty in Laser Powder Bed Fusion models

A limitation frequently encountered in additive manufacturing (AM) models is a lack of indication about their precision and accuracy. Often overlooked, information on model uncertainty is required for validation of AM models, qualification of AM-produced parts, and uncertainty management. This paper presents a discussion on the origin and propagation of uncertainty in Laser Powder Bed Fusion (L-PBF) models. Four sources of uncertainty are identified: modeling assumptions, unknown simulation parameters, numerical approximations, and measurement error in calibration data. Techniques to quantify uncertainty in each source are presented briefly, along with estimation algorithms to diminish prediction uncertainty with the incorporation of online measurements. The methods are illustrated with a case study based on a transient, stochastic thermal model designed for melt pool width predictions. Model uncertainty is quantified for single track experiments and the effect of online estimation in overhanging structures is studied via simulation. The application of these concepts to estimation and control of the L-PBF process is suggested.© 2016 ASME

Implementation of a particle filter on a GPU for nonlinear estimation in a manufacturing remelting process

This paper discusses the use of modern methods for estimation in Vacuum Arc Remelting, a manufacturing process used in the production of specialty metals for aerospace applications. Accurate estimation in this process is challenging because the system is nonlinear and all available measurements are corrupted with noise. Particle filters are nonlinear estimators that sample a set of points, called particles, in the state space to construct discrete approximations of probability density functions. Real-time issues arise when using these methods in systems with low signal-to-noise ratios because of the large number of particles required to reach acceptable accuracy. In these cases, the throughput of the particle filter becomes critical, and parallelization becomes a necessity. This paper presents the implementation of a particle filter using a GPU with NVIDIAs CUDA technology, whose large number of processor cores allows massive parallelization.

Auxiliary particle filter-model predictive control of the vacuum arc remelting process

Solidification control is required for the suppression of segregation defects in vacuum arc remelting of superalloys. In recent years, process controllers for the VAR process have been proposed based on linear models, which are known to be inaccurate in highly-dynamic conditions, e.g. start-up, hot-top and melt rate perturbations. A novel controller is proposed using auxiliary particle filter-model predictive control based on a nonlinear stochastic model. The auxiliary particle filter approximates the probability of the state, which is fed to a model predictive controller that returns an optimal control signal. For simplicity, the estimation and control problems are solved using Sequential Monte Carlo (SMC) methods. The validity of this approach is verified for a 430 mm (17 in) diameter Alloy 718 electrode melted into a 510 mm (20 in) diameter ingot. Simulation shows a more accurate and smoother performance than the one obtained with an earlier version of the controller.

Controlling pool depth during VAR of Alloy 718

A longtime goal of superalloy producers has been to control the geometry of the liquid pool in solidifying ingots. Accurate pool depth control at appropriate values is expected to result in ingots free of segregation defects. This article describes an industrial VAR experiment in which a 430mm (17 in) diameter Alloy 718 electrode was melted into a 510mm (20 in) ingot. In the experiment, the depth of the liquid pool at the mid-radius was controlled to three different set-points: 137 mm (nominal), 193 mm (deep) and 118 mm (shallow). At each level, the pool depth was marked by a power cutback of several minutes. The ingot was sectioned and longitudinal slices were cut out. Analysis of the photographed ingot revealed that accurate control was obtained for both the nominal and deep pool cases, while the third one was not conclusive.

DIGITAL SOLUTIONS FOR INTEGRATED AND COLLABORATIVE ADDITIVE MANUFACTURING

Software tools, knowledge of materials and processes, and data provide three pillars on which Additive Manufacturing (AM) lifecycles and value chains can be supported. These pillars leverage efforts dedicated to the development of AM databases, high-fidelity models, and design and planning support tools. However, as of today, it remains a challenge to integrate distributed AM data and heterogeneous predictive models in software tools to drive a more collaborative AM development environment. In this paper, we describe the development of an analytical framework for integrated and collaborative AM development. Information correlating material, product design, process planning and manufacturing operations are captured and managed in the analytical framework. A layered structure is adopted to support the composability of data, models and knowledge bases. The key technologies to enable composability are discussed along with a suite of tools that assist designers in the management of data, models and knowledge components. A proof-of-concept case study demonstrates the potential of the AM analytical framework.Copyright

Investigating Predictive Metamodeling for Additive Manufacturing

Additive manufacturing (AM) is a new and disruptive technology that comes with a set of unique challenges. One of them is the lack of understanding of the complex relationships between the numerous physical phenomena occurring in these processes. Metamodels can be used to provide a simplified mathematical framework for capturing the behavior of such complex systems. At the same time, they offer a reusable and composable paradigm to study, analyze, diagnose, forecast, and design AM parts and process plans. Training a metamodel requires a large number of experiments and even more so in AM due to the various process parameters involved. To address this challenge, this work analyzes and prescribes metamodeling techniques to select optimal sample points, construct and update metamodels, and test them for specific and isolated physical phenomena. A simplified case study of two different laser welding process experiments is presented to illustrate the potential use of these concepts. We conclude with a discussion on potential future directions, such as data and model integration while also accounting for sources of uncertainty.Copyright

Solidification Mapping of a Nickel Alloy 718 Laboratory VAR Ingot

The solidification microstructure of a laboratory-scale Nickel alloy 718 vacuum arc remelted (VAR) ingot was analyzed. The cylindrical, 210-mm-diameter ingot was sectioned along a plane bisecting it length-wise, and this mid-plane surface was ground and etched using Canada’s reagent to reveal segregation contrast. Over 350 photographs were taken of the etched mid-plane surface and stitched together to form a single mosaic image. Image data in the resulting mosaic were processed using a variety of algorithms to extract quantities such as primary dendrite orientation, primary dendrite arm spacing (PDAS), and secondary dendrite arm spacing (SDAS) as a function of location. These quantities were used to calculate pool shape and solidification rate during solidification using existing empirical relationships for Nickel Alloy 718. The details and outcomes of this approach, along with the resulting comparison to experimental processing conditions and computational models, are presented.