skip to main content


Title: Deriving Metamodels to Relate Machine Learning Quality to Repository Characteristics in the Context of Additive Manufacturing
The widespread growth of additive manufacturing, a field with a complex informatic “digital thread”, has helped fuel the creation of design repositories, where multiple users can upload distribute, and download a variety of candidate designs for a variety of situations. Additionally, advancements in additive manufacturing process development, design frameworks, and simulation are increasing what is possible to fabricate with AM, further growing the richness of such repositories. Machine learning offers new opportunities to combine these design repository components’ rich geometric data with their associated process and performance data to train predictive models capable of automatically assessing build metrics related to AM part manufacturability. Although design repositories that can be used to train these machine learning constructs are expanding, our understanding of what makes a particular design repository useful as a machine learning training dataset is minimal. In this study we use a metamodel to predict the extent to which individual design repositories can train accurate convolutional neural networks. To facilitate the creation and refinement of this metamodel, we constructed a large artificial design repository, and subsequently split it into sub-repositories. We then analyzed metadata regarding the size, complexity, and diversity of the sub-repositories for use as independent variables predicting accuracy and the required training computational effort for training convolutional neural networks. The networks each predict one of three additive manufacturing build metrics: (1) part mass, (2) support material mass, and (3) build time. Our results suggest that metamodels predicting the convolutional neural network coefficient of determination, as opposed to computational effort, were most accurate. Moreover, the size of a design repository, the average complexity of its constituent designs, and the average and spread of design spatial diversity were the best predictors of convolutional neural network accuracy.  more » « less
Award ID(s):
1825535
NSF-PAR ID:
10171192
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
ASME 2020 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract Machine learning can be used to automate common or time-consuming engineering tasks for which sufficient data already exist. For instance, design repositories can be used to train deep learning algorithms to assess component manufacturability; however, methods to determine the suitability of a design repository for use with machine learning do not exist. We provide an initial investigation toward identifying such a method using “artificial” design repositories to experimentally test the extent to which altering properties of the dataset impacts the assessment precision and generalizability of neural networks trained on the data. For this experiment, we use a 3D convolutional neural network to estimate quantitative manufacturing metrics directly from voxel-based component geometries. Additive manufacturing (AM) is used as a case study because of the recent growth of AM-focused design repositories such as GrabCAD and Thingiverse that are readily accessible online. In this study, we focus only on material extrusion, the dominant consumer AM process, and investigate three AM build metrics: (1) part mass, (2) support material mass, and (3) build time. Additionally, we compare the convolutional neural network accuracy to that of a baseline multiple linear regression model. Our results suggest that training on design repositories with less standardized orientation and position resulted in more accurate trained neural networks and that orientation-dependent metrics were harder to estimate than orientation-independent metrics. Furthermore, the convolutional neural network was more accurate than the baseline linear regression model for all build metrics. 
    more » « less
  2. The data generated during additive manufacturing (AM) practice can be used to train machine learning (ML) tools to reduce defects, optimize mechanical properties, or increase efficiency. In addition to the size of the repository, emerging research shows that other characteristics of the data also impact suitability of the data for AM-ML application. What should be done in cases for which the data in too small, too homogeneous, or otherwise insufficient? Data augmentation techniques present a solution, offering automated methods for increasing the quality of data. However, many of these techniques were developed for machine vision tasks, and hence their suitability for AM data has not been verified. In this study, several data augmentation techniques are applied to synthetic design repositories to characterize if and to what degree they enhance their performance as ML training sets. We discuss the comparative advantage of these data augmentation techniques across several canonical AM-ML tasks. 
    more » « less
  3. null (Ed.)
    Modern digital manufacturing processes, such as additive manufacturing, are cyber-physical in nature and utilize complex, process-specific simulations for both design and manufacturing. Although computational simulations can be used to optimize these complex processes, they can take hours or days--an unreasonable cost for engineering teams leveraging iterative design processes. Hence, more rapid computational methods are necessary in areas where computation time presents a limiting factor. When existing data from historical examples is plentiful and reliable, supervised machine learning can be used to create surrogate models that can be evaluated orders of magnitude more rapidly than comparable finite element approaches. However, for applications that necessitate computationally- intensive simulations, even generating the training data necessary to train a supervised machine learning model can pose a significant barrier. Unsupervised methods, such as physics- informed neural networks, offer a shortcut in cases where training data is scarce or prohibitive. These novel neural networks are trained without the use of potentially expensive labels. Instead, physical principles are encoded directly into the loss function. This method substantially reduces the time required to develop a training dataset, while still achieving the evaluation speed that is typical of supervised machine learning surrogate models. We propose a new method for stochastically training and testing a convolutional physics-informed neural network using the transient 3D heat equation- to model temperature throughout a solid object over time. We demonstrate this approach by applying it to a transient thermal analysis model of the powder bed fusion manufacturing process. 
    more » « less
  4. The goal of this work to mitigate flaws in metal parts produced from laser powder bed fusion (LPBF) additive manufacturing (AM) process. As a step towards this goal, the objective of this work is to predict the build quality of a part as it is being printed via deep learning of in-situ layer-wise images obtained from an optical camera instrumented in the LPBF machine. To realize this objective, we designed a set of thin-wall features (fins) from Titanium alloy (Ti-6Al-4V) material with varying length-to-thickness ratio. These thin-wall test parts were printed under three different build orientations and in-situ images of their top surface were acquired during the process. The parts were examined offline using X-ray computed tomography (XCT), and their build quality was quantified in terms of statistical features, such as the thickness and consistency of its edges. Subsequently, a deep learning convolutional neural network (CNN) was trained to predict the XCT-derived statistical quality features using the layer-wise optical images of the thin-wall part as inputs. The statistical correlation between CNN-based predictions and XCT-observed quality measurements exceeds 85%. This work has two outcomes consequential to the sustainability of additive manufacturing: (1) It provides practitioners with a guideline for building thin-wall features with minimal defects, and (2) the high correlation between the offline XCT measurements and in-situ sensor-based quality metrics substantiates the potential for applying deep learning approaches for the real-time prediction of build flaws in LPBF. 
    more » « less
  5. Abstract Background

    Identifying splice site regions is an important step in the genomic DNA sequencing pipelines of biomedical and pharmaceutical research. Within this research purview, efficient and accurate splice site detection is highly desirable, and a variety of computational models have been developed toward this end. Neural network architectures have recently been shown to outperform classical machine learning approaches for the task of splice site prediction. Despite these advances, there is still considerable potential for improvement, especially regarding model prediction accuracy, and error rate.

    Results

    Given these deficits, we propose EnsembleSplice, an ensemble learning architecture made up of four (4) distinct convolutional neural networks (CNN) model architecture combination that outperform existing splice site detection methods in the experimental evaluation metrics considered including the accuracies and error rates. We trained and tested a variety of ensembles made up of CNNs and DNNs using the five-fold cross-validation method to identify the model that performed the best across the evaluation and diversity metrics. As a result, we developed our diverse and highly effective splice site (SS) detection model, which we evaluated using two (2) genomicHomo sapiensdatasets and theArabidopsis thalianadataset. The results showed that for of theHomo sapiensEnsembleSplice achieved accuracies of 94.16% for one of the acceptor splice sites and 95.97% for donor splice sites, with an error rate for the sameHomo sapiensdataset, 4.03% for the donor splice sites and 5.84% for theacceptor splice sites datasets.

    Conclusions

    Our five-fold cross validation ensured the prediction accuracy of our models are consistent. For reproducibility, all the datasets used, models generated, and results in our work are publicly available in our GitHub repository here:https://github.com/OluwadareLab/EnsembleSplice

     
    more » « less