Ongoing Project III: Statistical Learning for Interdisciplinary Scientific/Engineering Applications
[NSF 1904165, 1929091, 19201920]
Health Prognostics of Complex Engineered Systems
[National Science Foundation (NSF), CMMI 1904165]
A designed engineering system is never a black box. Sensor signals are not only statistically correlated but also physically connected through unequivocal system working principles.
The overall goal of this project is to develop and test (based on real data ) a set of novel methodologies which enables the integration of fundamental system physics and sensor monitoring data for the modeling and health prognostics of engineering systems.
Ensemble Tree-Based Method for Recurrence Data
[National Science Foundation (NSF), OIA 1929091]
The goal of this project is to develop new statistical learning methodologies for the modeling, prediction and optimization of large-scale recurrent event processes with heterogeneous (both static and dynamic) feature information.
Gradient Boosted Trees for Medical Image Data
[partially supported by National Science Foundation (NSF), OIA 1920920]
The first project aims to develop gradient boosting trees for spatially correlated data (Boost.S). The method helps us model and predict FDG-PET (fluorodeoxyglucose-positron emission tomography) imaging data from cancer chemoradiotherapy (*the figure is from the collaborator, Dr. Stephen Bowen's group, from University of Washington).
The second project aims to develop Structured Adaptive Boosting Trees algorithm (AdaBoost.S) to surmount the edge detection problem associated with medical images. The method has been applied to detect the multicellular aggregates in fluorescence intravital microscopy (*the figure is from the collaborator, Dr. Margaret Bennewitz's group, from West Virginia University).
Stochastic Degradation, Optimal Planning of Reliability Testing, and Condition-Based Maintenance