Yuncheng Du
University of Houston, TX·Engineering·Early-career
Many systems we depend on are made of many parts that constantly affect one another in nonlinear, hard-to-predict ways. Examples include manufacturing plants, computer networks, automated transportation and medical devices.
- Fields
Engineering
Biomedical Engineering
Computer Science
- Topics
fault-tolerant control
machine learning
fault detection and diagnosis
- Grants
- 1 known grant, through Aug 2027
What they work on
- Develop machine learning methods that find the symptom "fingerprints" of faults directly from data, enabling real-time diagnosis of ventricular assist device problems such as thrombosis and suction.
- Build an online adaptive modeling approach that predicts performance-related variables of a complex system that cannot be measured directly because of cost or technical limits.
- Design a data-driven fault-tolerant controller that automatically adjusts a ventricular assist device after a fault to maintain performance while meeting all operating constraints.
More about the research
When something goes wrong in such a system, such as a failed sensor or a cyber-attack, it can become unsafe or expensive to run. Keeping these systems working is hard when no reliable physics-based model of their behavior exists. This project asks two questions: how can rare, harmful events be predicted from data alone, and how can a system be controlled so it keeps performing well after a fault appears?
Why it matters
Heart failure patients who rely on ventricular assist devices face life-threatening device faults such as clotting and suction. Automatic detection of these faults, followed by automatic adjustment of the pump, could improve their survival and quality of life. The same data-driven control methods could also help protect computer systems from attacks, make manufacturing more efficient, and improve safety in automated transportation. They may reduce the need for costly, expert-built physics models.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Zhongjun Wu
University of Maryland, Baltimore, MD·Biomedical Engineering
Characterize how high mechanical shear stress changes neutrophil structure, rolling and adhesion, and surface markers, and quantify how structural damage scales with stress levels found in VADs and ECMO.
Romulo Meira-Goes
Pennsylvania State University, PA·Engineering
Develop fault detection and diagnostics methods based on supervisory control of discrete event systems, which model a system as a sequence of events, to detect cascading faults and cyber-attacks in data center cooling with rigorous guarantees.
David W Schmidtke
University of Texas at Dallas, TX·Biomedical Engineering
Measure pre-implant leukocyte profiles and plasma biomarkers in advanced heart failure patients and test whether these signs of a compromised inflammatory response predict infection after VAD implantation.
The school
University of Houston: 21 labs on PI Pages, $239M a year on research, 374 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.