Romulo Meira-Goes
Pennsylvania State University, PA·Engineering
Data centers run the cloud services and artificial intelligence that much of the economy depends on. They also use over 4% of the world's electricity, much of it to keep computers cool.
- Fields
Engineering
Computer Science
- Topics
fault detection and diagnostics
data centers
data center cooling
- Grants
- 1 known grant, through Sep 2028
What they work on
- 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.
- Identify the attack vectors, or routes for cyber-attack, that pose the greatest threat to data center cooling management, by combining event-driven models with physics-based models of the cooling system.
- Design real-time control responses that adjust cooling operation to keep it running normally during major faults or attacks, using the integrated supervisory control and physics-based framework.
- Evaluate the algorithms in an open-source virtual testbed and in a hardware-in-the-loop testbed that connects real data center cooling equipment to simulations.
More about the research
If a cooling system fails, a data center can shut down entirely and lose data and equipment. Current fault detection and diagnostics (FDD) methods rely on statistics, machine learning or ad-hoc estimates. These methods can miss rare or complex failures, such as cascading faults where one failure triggers others, or deliberate cyber-attacks. This project asks how to detect, explain and respond to such failures with rigorous guarantees.
Why it matters
Failures in data center cooling can shut down systems and destroy data and equipment, a growing risk as demand for AI and cloud computing rises. Fault detection methods that catch cascading failures and cyber-attacks could help data center developers, technology companies, utilities and HVAC manufacturers design cooling systems that are more resilient and harder to attack.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Yuncheng Du
University of Houston, TX·Engineering·Early-career
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.
Zejian Zhou
University of Wyoming, WY·Engineering
Model complex autonomous systems as multi-agent systems in which a disrupted component becomes an irrational agent, either passive (an accidental fault) or active (adversarial, such as a cyber-attack).
Javad Khazaei
Lehigh University, PA·Engineering·Early-career
Learn low-dimensional manifolds from time-series measurements of distributed energy resources and use them for real-time predictive control of voltage, frequency and power sharing in microgrids.
The school
Pennsylvania State University: 60 labs on PI Pages, $1.3B a year on research, 743 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.