PhD-BPD Dissertation Defense Presentation: Jinzhao Tian

Title: Scalable Fault Impact Analysis for Improving Building Performance

Name: Jinzhao Tian, Ph.D. candidate in Building Performance and Diagnostics (PhD-BPD)

Date: Wednesday, April 29, 2026
Time: 10:00am-12:00pm ET
Location: Intelligent Workplace (IW) Conference Room, MMCH 415 & Zoom

Advisory Committee:

Prof. Vivian Loftness, FAIA, LEED AP (Chair)
University Professor 
School of Architecture
Carnegie Mellon University

Mario Berges, Ph.D.
Professor
Civil and Environmental Engineering
Carnegie Mellon University

Susan Finger, Ph.D.
Professor Emeritus
Civil and Environmental Engineering
Carnegie Mellon University

Kun Zhang, Ph.D.
Professor
Department of Philosophy
Carnegie Mellon University

Abstract:
Effective HVAC management requires not only detecting faults but also understanding how they affect building performance under real operating conditions, enabling prioritization of corrective actions. However, fault impact analysis (FIA) has received far less attention than fault detection, and prior FIA studies have relied primarily on experiments and simulations, which provide valuable insights but are difficult to scale across large, diverse building portfolios. Existing research has also predominantly focused on energy outcomes, while giving far less attention to indoor air quality (IAQ). This dissertation addresses these gaps by reframing FIA as a causal inference problem centered on the missing counterfactual: how the building would have performed had the fault not occurred. It develops a scalable, data-driven FIA framework that integrates building automation system (BAS) time-series data, fault records, indoor CO₂ measurements, smart meter data, and weather data to estimate the impacts of operational faults on indoor air quality and energy consumption.

Drawing on more than 100 million operational records across 58 commercial buildings, the study combines causal inference with machine learning to quantify how specific faults affect indoor CO₂ concentrations and whole-building energy use. The results provide large-scale empirical evidence of the real-world consequences of HVAC faults and support the comparison and prioritization of faults across building portfolios. By demonstrating how building operational data can be used to estimate fault impacts at scale, this dissertation advances the shift from monitoring-based to evidence-based commissioning, providing a foundation for healthier, more sustainable, and higher-performance buildings.