An official website of the United States government
Extruded aluminum supply chains are materially inefficient with around 40% of the billet likely to be scrapped before the profile is embedded in a product. One of the largest sources of scrap is the removal due to weld integrity concerns of the tongue-shaped transverse weld(s) that forms between consecutively extruded billets. Process setting and die geometry optimization can decrease the weld length (and hence scrapped material) by modest amounts. We explore a process for significant scrap savings using profiled dummy blocks to generate shorter welds by compensating for the differential metal flow velocities across the billet cross-section as it flows through the die ports. We develop a design process for defining the profiled dummy block shape. For a given part and press, we first define an ideal dummy block shape by extracting the velocity field from finite element simulations of the conventional process and assuming perfectly rigid tooling. Next, we rationalize the tool shape using stress and deflection limits (preventing plastic deformation and interference with the container wall) and ductile damage limits for the billet to prevent cracking. We then simulate the likely effect of the rationalized dummy block design on back-end defect removal. The methodology is demonstrated for four profiles of increasing complexity. The process’ potential is evaluated experimentally using billets machined to match the ideal dummy block shape. The results show that profiled billets can achieve weld length reductions >50% for simple shapes. We demonstrate that multi-profile tooling can deliver scrap savings across a family of similar profiles.
more »
« less
Process-BERT: A Framework for Representation Learning on Educational Process Data
- Award ID(s):
- 1917713
- PAR ID:
- 10339507
- Date Published:
- Journal Name:
- International Conference on Educational Data Mining
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Modern data acquisition routinely produce massive amounts of event sequence data in various domains, such as social media, healthcare, and financial markets. These data often ex- hibit complicated short-term and long-term temporal dependencies. However, most of the ex- isting recurrent neural network-based point process models fail to capture such dependencies, and yield unreliable prediction performance. To address this issue, we propose a Transformer Hawkes Process (THP) model, which leverages the self-attention mechanism to capture long- term dependencies and meanwhile enjoys computational efficiency. Numerical experiments on various datasets show that THP outperforms existing models in terms of both likelihood and event prediction accuracy by a notable margin. Moreover, THP is quite general and can incorpo- rate additional structural knowledge. We provide a concrete example, where THP achieves im- proved prediction performance for learning multiple point processes when incorporating their relational information.more » « less
-
Abstract Polishing of additively manufactured products is a multi-stage process, and a different combination of polishing pad and process parameters is employed at each stage. Pad change decisions and endpoint determination currently rely on practitioners’ experience and subjective visual inspection of surface quality. An automated and objective decision process is more desired for delivering consistency and reducing variability. Toward that objective, a model-guided decision-making scheme is developed in this article for the polishing process of a titanium alloy workpiece. The model used is a series of Gaussian process models, each established for a polishing stage at which surface data are gathered. The series of Gaussian process models appear capable of capturing surface changes and variation over the polishing process, resulting in a decision protocol informed by the correlation characteristics over the sample surface. It is found that low correlations reveal the existence of extreme roughness that may be deemed surface defects. Making judicious use of the change pattern in surface correlation provides insights enabling timely actions. Physical polishing of titanium alloy samples and a simulation of this process are used together to demonstrate the merit of the proposed method.more » « less
-
Educational process data, i.e., logs of detailed student activities in computerized or online learning platforms, has the potential to offer deep insights into how students learn. One can use process data for many downstream tasks such as learning outcome prediction and automatically delivering personalized intervention. However, analyzing process data is challenging since the specific format of process data varies a lot depending on different learning/testing scenarios. In this paper, we propose a framework for learning representations of educational process data that is applicable across many different learning scenarios. Our framework consists of a pre-training step that uses BERT-type objectives to learn representations from sequential process data and a fine-tuning step that further adjusts these representations on downstream prediction tasks. We apply our framework to the 2019 nation’s report card data mining competition dataset that consists of student problem-solving process data and detail the specific models we use in this scenario. We conduct both quantitative and qualitative experiments to show that our framework results in process data representations that are both predictive and informative.1more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
