Center for Intelligent Metrology & Sensing

We're excited to share that Dr. Jia Xiaodong has been named among the top 2% of scientists globally by Stanford University/Elsevier for the third consecutive year! Ranked 2,011 out of nearly 114,000 in the 'Industrial Engineering and Automation' subfield, this recognition is a testament to our lab's commitment to advancing the capabilities of data-driven measurement. Learn more about the research that earned Dr. Jia this honor on our website: https://lnkd.in/dVChTX9N

Imbalanced data is a major challenge in machine learning for manufacturing quality control. Minority class samples, often representing rare defects, can lead to misclassifications and reduced model accuracy. To tackle this, our COSMOTE resampling approach combines the strengths of undersampling and oversampling to generate synthetic data for minority classes, enhancing both training quality and model reliability. By using COSMOTE within a deep learning model, we create balanced datasets that improve defect detection accuracy. This approach has resulted in 100% true detection rates for critical defects, as demonstrated in our results. For a more comprehensive understanding of our methodology and its practical applications, including case studies, please refer to the full research publication in the link below. paper link: https://lnkd.in/gDsTj6SR

🏆 We are thrilled to announce that Dr. Xiaodong Jia from the Lab for Intelligent Metrology Systems at the University of Cincinnati has been honored with the 2024 CEAS Distinguished Research Award. Join us in congratulating Dr. Jia on this great honor!   #CEAS #ResearchExcellence #UniversityOfCincinnati #Engineering #Innovation

Data quality is critical in prognostics and health management (PHM). However, the rapid growth of big data in the industry poses challenges in assessing data suitability for modeling, often leaving vast amounts of machine-generated data underutilized. To address this, our team has developed a novel framework that systematically evaluates data suitability for PHM modeling. The methodology focuses on three core assessments: detectability, diagnosability, and prognosability. It not only offers visualization tools but also quantitative metrics to pinpoint the most useful datasets for model development. The practical application of this methodology has been validated with real-world case studies, including ball screw and boring tool degradation, proving its effectiveness and practicality in enhancing data-driven solutions. Explore our detailed poster and delve into our research paper for a deeper understanding. #DataQuality #PHM #SmartManufacturing #MachineLearning paper link: https://lnkd.in/gXnA6HqU

Particle contamination during semiconductor manufacturing can lead to significant yield losses. At the Lab for Intelligent Metrology Systems, we have developed a novel assessment method to monitor deposit accumulation in etching chambers. Utilizing virtual metrology systems, our approach improves in-situ particle monitoring and predictive chamber maintenance, leading to better yield and efficiency. The successful application of this method across multiple etching chambers justifies its effectiveness. Explore our innovative techniques by viewing our poster and for an in-depth analysis, refer to our detailed research paper. #SemiconductorManufacturing #DryEtching #ParticleContamination #Innovation #SmartManufacturing paper link: https://lnkd.in/gwwMM8Uj

Thrilled to share that the Lab for Intelligent Metrology Systems was selected to present our research at the P&G Innovation Fair this week. A big thank you to the organizers for hosting the great event!

At the Lab for Intelligent Metrology Systems, we continuously push the boundaries in wafer inspection technology. Our recent work introduced a dynamic clustering approach with Vision Transformers and topological data analysis to advance wafer map analysis. Building on this, we've developed a graph-based defect classification method. Validated on the WM-811K dataset, extraction of rotation-invariant features from wafer maps enables accurate defect classification compared to other methods in the literature. Take a look at our poster for an overview of our methodologies, and walk through our published paper for a deeper understanding of our research. #SemiconductorInspection #WaferMap #SmartManufacturing #DynamicClustering full paper: https://lnkd.in/gDtv2zg8

As semiconductors shrink to unprecedented scales, Virtual Metrology (VM) becomes increasingly crucial to address the complexities of high-mix manufacturing and minimize process variations. Feature selection for VM presents a complex challenge, demanding a high degree of adaptability to effectively tailor the VM process. At the Lab for Intelligent Metrology Systems, we've developed a solution using the GMDH-type polynomial neural network—an algorithm that automates feature selection, enhancing the Virtual Metrology (VM) for the Chemical Mechanical Planarization process. Our method has been validated to be effective through comprehensive benchmarking, demonstrating the lowest Mean Squared Error (MSE) among various VM models. 🏆 Proudly, our innovative approach led us to victory in the PHM conference data competition in 2016, outperforming 24 teams. Check out our poster to discover how we're refining semiconductor manufacturing with precision. #SemiconductorManufacturing #VirtualMetrology #R2RControl #SmartManufacturing paper link: https://lnkd.in/ezwUcQD7

At the Lab for Intelligent Metrology Systems, University of Cincinnati, we're advancing automated production with smart, precise quality control solutions. Check out our poster featuring an intelligent system that pinpoints bottle capping failures by analyzing motor current signals. With a robust IoT system paired with advanced machine learning algorithms, multiple capping failure patterns are accurately identified, enhancing the reliability of manufacturing lines. Take a moment to explore the insights of our innovative approach in the poster and delve deeper into our research with the full paper linked below! #SmartManufacturing #QualityControl #FaultDetection #Innovation #MachineLearning   Full paper link: https://lnkd.in/gY__zRvs