Structures as Sensors: Inferring Within
Structures as Sensors: Inferring Around
Structures as Sensors: Inferring Itself



Indirect sensing, statistical inference, and probabilistic modeling of dynamic structural systems to better understand their behavior and then improve their performance. Realizing that environment around and the users within all have direct physical interaction with the structure, my research will utilize the structure itself as a sensing medium to indirectly sense and infer these information (itself, within, and around). Specifically, my research focuses on 1) itself (structural health monitoring, HVAC operation analysis, etc.), 2) within (tracking and characterization of occupants’ activity, health status, etc.), and 3) around (indirect monitoring of traffic, surrounding infrastructures etc.).




We have been working on the “within” aspect through indoor human tracking and characterization using building as a sensor. We are building a system to sense, identify, localize, and characterize persons’ gait pattern to understand various user information (e.g., indoor human traffic flow, their activity level, their health status, mood, etc.) on a fine-grained level with non-dedicated building vibration monitoring sensors (user-aware).

Our system utilizes high resolution and high frequency vibration sensor data and develops statistical signal processing and machine learning techniques to analyze the data. We are collaborating with elder care facilities in Pittsburgh (Vincentian Collaborative System and Baptist Homes Society), to deploy and carry out the work in the context of elderly healthcare, where the building structure will recognize individual resident’s health status (e.g., balance, fatigue, ability to walk, etc.) through gait-induced structural vibration.

Noh - 2014 Innovation Palooza - Vibration-9082

Picture4Example Measurements

Back to Top




Structural responses are often dependent on surrounding infrastructure (e.g., buildings respond to outside traffic; trains depend on track conditions). For the “around” aspect of the framework, my group has been working on indirect surrounding environment monitoring through vehicles and buildings. The vehicle aspect instruments vehicles with vibration sensors (instead of direct instrumentation of infrastructure) to collect information about surrounding infrastructure (e.g., bridge, railway, etc.) for damage diagnosis purposes (environment-aware).

Collaborating with the Port Authority of Allegheny County, we deployed and validated our system to the Pittsburgh Light Rail System (T-system) over 3+ year period. Various signal processing and data mining techniques are used to extract meaningful information from the collected signal.

Picture1    Picture2

Visit to Port Authority Train Depot

Picture3 Picture1

Train Vibration Measurement and Track Condition Change Detection Result

Back to Top




As part of the “itself” aspect of the structures as sensors idea, we have been working on structural health monitoring (SHM) using wireless sensing units and statistical pattern recognition. These systems allow automated monitoring of structural conditions (self-aware) in an efficient and reliable way to reduce maintenance costs and prevent catastrophic failure. Specifically, we have been working on vibration based damage diagnosis algorithm using wavelet analysis, time-series modeling, and change point detection.

We have been investigating practical challenges in SHM, such as sparse sensing, noisy and nonstationary measurements, online decision making, and varying environmental and operational conditions. My group has also been working on uncertainty modeling and information updating that are necessary to incorporate SHM sensor data to traditional analytical seismic risk analysis approaches, in order to improve the loss estimation of civil structures over their lifetime and allow them to be better prepared for extreme events.

hae young noh - shake table demo - fall 2014 (1 of 1)     

Shake Table Experiments


Information Theoretic Approach for Structural Vibration Analysis

Back to Top