EAS PhD Proposal Defense by Richard Bellizzi
When: Thursday,
May 20, 2021
2:00 AM
-
3:00 AM
Where: Online
Description: EAS PhD Program (CSE Option/Mechanical Engineering)
PhD PROPOSAL DEFENSE by Richard Bellizzi
Date: May 20, 2021
Time: 2:00 pm
Topic: Strategic Implementation of Computational Methodologies in Lubricant Testing Analysis
Zoom Teleconference: https://umassd.zoom.us/j/5903522937?pwd=VnNoT2dUOHVBVHdJSTg1MGFHV3l5QT09
Abstract:
The Lubricant industry is prosperous with historical data and methodologies that establish opportunities to leverage Machine Learning and Computer Vision methods to advance lubricant testing analysis procedures. An approach like this requires digitized data to allow for a computer to interpret the inputs and provide a satisfactory output. The bearing corrosion analysis procedure from the ASTM EMCOR method is a perfect proof of concept for this type of analysis approach. The standard specifically calls for the inspection method to use only the visual acuity of a human technician. Techniques like this leave a lot of room for interpretation and variability that this project solves through modern imaging and analysis techniques. While this may lead to results that differ from technicians, the overall system can repeatably provide a meaningful result that allows for improved collaborative testing on a global platform by removing the variability technician to technician.
Mask R-CNN models are representative of some improvements that modern methods provide that expand on human capability. With models like this used in medical imaging analysis and autonomous driving, it intuitively makes sense that they start migrating into other industrial realms, like the lubricant realm. The first project explores the feasibility of Convolutional Neural Networks (CNN) and Transfer Learning (TL) methods applied to bearing lubricant defect detection. After demonstrating this initial methodology, the second project progresses into more advanced methods, like the Mask R-CNN framework, yielding a model with bearing lubricant defect recognition, classification, and segmentation features. Overall, improving approaches in the lubricants industry that rely on technicians' visual interpretation shows one adoption of Machine Learning methods that provides immediate value. This type of analysis allows for increasing the depth of evaluation performed on specimens providing new ways for quantifying products. Automated segmentation presents various advantages since corrosion and other surface defects contain differences leaving room for further categorization. Differentiating these defects promotes research into surface interactions yielding different defect instances on bearings, further aiding product development.
In addition to the model, methods for digitizing the data and establishing automated analyses contribute to the overall transformation occurring through the lubricant industry, like most other industries.
ADVISOR(S): Dr. Alfa Heryudono, Department of Mathematics
(aheryudono@umassd.edu, 508-999-8516)
Dr. Yanlai Chen, Department of Mathematics
(yanlai.chen@umassd.edu, 508-999-8438)
COMMITTEE MEMBERS: Dr. Wenzhen Huang, Department of Mechanical Engineering
Dr. Scott Field, Department of Mathematics
Dr. Jason Galary, Nye Lubricants Research & Development
NOTE: All MNE and EAS Students are ENCOURAGED to attend.
PhD PROPOSAL DEFENSE by Richard Bellizzi
Date: May 20, 2021
Time: 2:00 pm
Topic: Strategic Implementation of Computational Methodologies in Lubricant Testing Analysis
Zoom Teleconference: https://umassd.zoom.us/j/5903522937?pwd=VnNoT2dUOHVBVHdJSTg1MGFHV3l5QT09
Abstract:
The Lubricant industry is prosperous with historical data and methodologies that establish opportunities to leverage Machine Learning and Computer Vision methods to advance lubricant testing analysis procedures. An approach like this requires digitized data to allow for a computer to interpret the inputs and provide a satisfactory output. The bearing corrosion analysis procedure from the ASTM EMCOR method is a perfect proof of concept for this type of analysis approach. The standard specifically calls for the inspection method to use only the visual acuity of a human technician. Techniques like this leave a lot of room for interpretation and variability that this project solves through modern imaging and analysis techniques. While this may lead to results that differ from technicians, the overall system can repeatably provide a meaningful result that allows for improved collaborative testing on a global platform by removing the variability technician to technician.
Mask R-CNN models are representative of some improvements that modern methods provide that expand on human capability. With models like this used in medical imaging analysis and autonomous driving, it intuitively makes sense that they start migrating into other industrial realms, like the lubricant realm. The first project explores the feasibility of Convolutional Neural Networks (CNN) and Transfer Learning (TL) methods applied to bearing lubricant defect detection. After demonstrating this initial methodology, the second project progresses into more advanced methods, like the Mask R-CNN framework, yielding a model with bearing lubricant defect recognition, classification, and segmentation features. Overall, improving approaches in the lubricants industry that rely on technicians' visual interpretation shows one adoption of Machine Learning methods that provides immediate value. This type of analysis allows for increasing the depth of evaluation performed on specimens providing new ways for quantifying products. Automated segmentation presents various advantages since corrosion and other surface defects contain differences leaving room for further categorization. Differentiating these defects promotes research into surface interactions yielding different defect instances on bearings, further aiding product development.
In addition to the model, methods for digitizing the data and establishing automated analyses contribute to the overall transformation occurring through the lubricant industry, like most other industries.
ADVISOR(S): Dr. Alfa Heryudono, Department of Mathematics
(aheryudono@umassd.edu, 508-999-8516)
Dr. Yanlai Chen, Department of Mathematics
(yanlai.chen@umassd.edu, 508-999-8438)
COMMITTEE MEMBERS: Dr. Wenzhen Huang, Department of Mechanical Engineering
Dr. Scott Field, Department of Mathematics
Dr. Jason Galary, Nye Lubricants Research & Development
NOTE: All MNE and EAS Students are ENCOURAGED to attend.
Contact: > See Description for contact information
Topical Areas: Faculty, Students, Students, Graduate, Students, Undergraduate, Bioengineering, Civil and Environmental Engineering, College of Engineering, Computer and Information Science, Co-op Program, Electrical and Computer Engineering, Mechanical Engineering, Physics