skip to main content


Search for: All records

Award ID contains: 1849085

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. Abstract

    The research reported in this article is concerned with the question of detecting and subsequently determining the endpoint in a long-stretch, ultraprecision surface polishing process. While polishing endpoint detection has attracted much attention for several decades in the chemical-mechanical planarization of semiconductor wafer polishing processes, the uniqueness of the surface polishing process under our investigation calls for novel solutions. To tackle the research challenges, we develop both an offline model and an online detection method. The offline model is a functional regression that relates the vibration signals to the surface roughness, whereas the online procedure is a change-point detection method that detects the energy turning points in the vibration signals. Our study reveals a number of insights. The offline functional regression model shows clearly that the polishing process progresses in three states, including a saturation phase, over which the polishing action could be substantially shortened. The online detection method signals in real-time when to break a polishing cycle and to institute a follow-up inspection, rather than letting the machine engage in an overpolishing cycle for too long. When implemented properly, both sets of insights and the corresponding methods could lead to substantial savings in polishing time and energy and significantly improve the throughput of such polishing processes without inadvertently affecting the quality of the final polish.

     
    more » « less
  2. We present HiRA-Pro, a novel procedure to align, at high spatio-temporal resolutions, multimodal signals from real-world processes and systems that exhibit diverse transient, nonlinear stochastic dynamics, such as manufacturing machines. It is based on discerning and synchronizing the process signatures of salient kinematic and dynamic events in these disparate signals. HiRA-Pro addresses the challenge of aligning data with sub-millisecond phenomena, where traditional timestamp, external trigger, or clock-based alignment methods fall short. The effectiveness of HiRA-Pro is demonstrated in a smart manufacturing context, where it aligns data from 13+ channels acquired during 3D-printing and milling operations on an Optomec-LENS® MTS 500 hybrid machine. The aligned data is then voxelized to generate 0.25 second aligned data chunks that correspond to physical voxels on the produced part. The superiority of HiRA-Pro is further showcased through case studies in additive manufacturing, demonstrating improved machine learning-based predictive performance due to precise multimodal data alignment. Specifically, testing classification accuracies improved by almost 35% with the application of HiRA-Pro, even with limited data, allowing for precise localization of artifacts. The paper also provides a comprehensive discussion on the proposed method, its applications, and comparative qualitative analysis with a few other alignment methods. HiRA-Pro achieves temporal-spatial resolutions of 10-1000 𝜇s and 100 𝜇m in order to generate datasets that register with physical voxels on the 3D-printed and milled part. These resolutions are at least an order of magnitude finer than the existing alignment methods that employ individual timestamps, statistical correlations, or common clocks, which achieve precision of hundreds of milliseconds. 
    more » « less
    Free, publicly-accessible full text available March 13, 2025
  3. This study presents an overview and a few case studies to explicate the transformative power of diverse imaging techniques for smart manufacturing, focusing largely on variousin-situandex-situimaging methods for monitoring fusion-based metal additive manufacturing (AM) processes such as directed energy deposition (DED), selective laser melting (SLM), electron beam melting (EBM).In-situimaging techniques, encompassing high-speed cameras, thermal cameras, and digital cameras, are becoming increasingly affordable, complementary, and are emerging as vital for real-time monitoring, enabling continuous assessment of build quality. For example, high-speed cameras capture dynamic laser-material interaction, swiftly detecting defects, while thermal cameras identify thermal distribution of the melt pool and potential anomalies. The data gathered fromin-situimaging are then utilized to extract pertinent features that facilitate effective control of process parameters, thereby optimizing the AM processes and minimizing defects. On the other hand,ex-situimaging techniques play a critical role in comprehensive component analysis. Scanning electron microscopy (SEM), optical microscopy, and 3D-profilometry enable detailed characterization of microstructural features, surface roughness, porosity, and dimensional accuracy. Employing a battery of Artificial Intelligence (AI) algorithms, information from diverse imaging and other multi-modal data sources can be fused, and thereby achieve a more comprehensive understanding of a manufacturing process. This integration enables informed decision-making for process optimization and quality assurance, as AI algorithms analyze the combined data to extract relevant insights and patterns. Ultimately, the power of imaging in additive manufacturing lies in its ability to deliver real-time monitoring, precise control, and comprehensive analysis, empowering manufacturers to achieve supreme levels of precision, reliability, and productivity in the production of components.

     
    more » « less
    Free, publicly-accessible full text available December 21, 2024
  4. As the model size grows rapidly, fine-tuning the large pre-trained language model has become increasingly difficult due to its extensive memory usage. Previous works usually focus on reducing the number of trainable parameters in the network. While the model parameters do contribute to memory usage, the primary memory bottleneck during training arises from storing feature maps, also known as activations, as they are crucial for gradient calculation. Notably, machine learning models are typically trained using stochastic gradient descent. We argue that in stochastic optimization, models can handle noisy gradients as long as the gradient estimator is unbiased with reasonable variance. Following this motivation, we propose a new family of unbiased estimators called WTA-CRS , for matrix production with reduced variance, which only requires storing the sub-sampled activations for calculating the gradient. Our work provides both theoretical and experimental evidence that, in the context of tuning transformers, our proposed estimators exhibit lower variance compared to existing ones. By replacing the linear operation with our approximated one in transformers, we can achieve up to 2.7× peak memory reduction with almost no accuracy drop and enables up to 6.4× larger batch size. Under the same hardware, WTA-CRS enables better down-streaming task performance by applying larger models and/or faster training speed with larger batch sizes. The code is available at https://github.com/zirui-ray-liu/WTACRS/. 
    more » « less
    Free, publicly-accessible full text available December 16, 2024
  5. Embedding is widely used in recommendation models to learn feature representations. However, the traditional embedding technique that assigns a fixed size to all categorical features may be suboptimal due to the following reasons. In recommendation domain, the majority of categorical features' embeddings can be trained with less capacity without impacting model performance, thereby storing embeddings with equal length may incur unnecessary memory usage. Existing work that tries to allocate customized sizes for each feature usually either simply scales the embedding size with feature's popularity or formulates this size allocation problem as an architecture selection problem. Unfortunately, most of these methods either have large performance drop or incur significant extra time cost for searching proper embedding sizes. In this article, instead of formulating the size allocation problem as an architecture selection problem, we approach the problem from a pruning perspective and proposePruning-basedMulti-sizeEmbedding (PME) framework. During the search phase, we prune the dimensions that have the least impact on model performance in the embedding to reduce its capacity. Then, we show that the customized size of each token can be obtained by transferring the capacity of its pruned embedding with significant less search cost. Experimental results validate that PME can efficiently find proper sizes and hence achieve strong performance while significantly reducing the number of parameters in the embedding layer.

     
    more » « less
  6. Designing alloys for additive manufacturing (AM) presents significant opportunities. Still, the chemical composition and processing conditions required for printability (ie., their suitability for fabrication via AM) are challenging to explore using solely experimental means. In this work, we develop a high-throughput (HTP) computational framework to guide the search for highly printable alloys and appropriate processing parameters. The framework uses material properties from stateof- the-art databases, processing parameters, and simulated melt pool profiles to predict processinduced defects, such as lack-of-fusion, keyholing, and balling. We accelerate the printability assessment using a deep learning surrogate for a thermal model, enabling a 1,000-fold acceleration in assessing the printability of a given alloy at no loss in accuracy when compared with conventional physics-based thermal models. We verify and validate the framework by constructing printability maps for the CoCrFeMnNi Cantor alloy system and comparing our predictions to an exhaustive ’in-house’ database. The framework enables the systematic investigation of the printability of a wide range of alloys in the broader Co-Cr-Fe-Mn-Ni HEA system. We identified the most promising alloys that were suitable for high-temperature applications and had the narrowest solidification ranges, and that was the least susceptible to balling, hot-cracking, and the formation of macroscopic printing defects. A new metric for the global printability of an alloy is constructed and is further used for the ranking of candidate alloys. The proposed framework is expected to be integrated into ICME approaches to accelerate the discovery and optimization of novel high-performance, printable alloys. 
    more » « less
  7. This article presents an overview of the emerging themes in Autonomous Materials Discovery andManufacturing (AMDM). This interdisciplinary field is garnering a growing interest among the sci-entists and engineers in the materials and manufacturing domains as well as those in the ArtificialIntelligence (AI) and data sciences domains, and it offers immense research potential for the indus-trial systems engineering (ISE) and manufacturing fields. Although there are a few reviews relatedto this topic, they had focused exclusively on sequential experimentation techniques, AI/machinelearning applications, or materials synthesis processes. In contrast, this review treats AMDM as acyberphysical system, comprising an intelligent softwarebrainthat incorporates various computa-tional models and sequential experimentation strategies, and a hardwarebodythat integratesequipment platforms for materials synthesis with measurement and testing capabilities. Thisreview offers a balanced perspective of the software and the hardware components of an AMDMsystem, and discusses the current state-of-the-art and the emerging challenges at the nexus ofmanufacturing/materials sciences and AI/data sciences in this nascent, exciting area 
    more » « less