Ph.D. Proposal

Swapnil Phatak

(Faculty Advisor: Professor Dimitri Mavris)

"A Methodology for Real-time Vibration-based Structural Health Monitoring and Damage Detection"

Monday, March 23

10:00 a.m. - 1:00 p.m.

Weber, CoVE

Abstract:

The advent of advanced design methods, materials, and manufacturing technologies has enabled the development of unconventional aircraft configurations that are projected to exhibit superior performance compared to traditional tube-and-wing aircraft. Due to the lack of operational data for novel aircraft configurations, traditional periodic maintenance procedures that rely on historical performance trends are rendered obsolete. Furthermore, the demand for maximizing aircraft uptime while maintaining high safety standards has sparked research interest in condition-based maintenance and real-time structural damage detection. Vibration-based structural health monitoring has emerged as a promising enabler, but there are several challenges that inhibit the practical application of this technology. This thesis addresses two of these challenges: optimal sensor placement and real-time damage detection.

Aircraft structural health monitoring is a statistical pattern recognition problem. As a result, the probability and accuracy of damage detection largely depend on the quantity and quality of data obtained from onboard sensors. This makes optimal sensor placement one of the most important steps in the overall structural health monitoring process. Redundant sensor placement increases the data volume but does not contribute new information about the monitored structure. The current non-redundant sensor placement methods do not utilize geometric data along with the vibration characteristics of the structure, which leads to suboptimal sensor configurations. This work introduces a novel combined redundancy metric that can be used to find an optimal tradeoff between individual sensor redundancy and the total information content of a sensor configuration. The NASA Common Research Model (CRM) geometry will be used to compare the proposed sensor placement method with existing methods.

According to a recent study by NASA, approximately 4% of all commercial aircraft accidents are caused due to in-fight structural component failure. As periodic inspections do not detect in-flight structural damage, real-time structural health monitoring is needed. Time series analysis methods have proven to be effective in identifying anomalous patterns in sensor data. These methods can be coupled with machine learning to perform data-driven damage detection and localization. However, limited data availability and poor scalability of these methods restrict their large-scale field implementation. This work proposes a substructure vector autoregression-based time series analysis method that can be applied to large sensor networks. This method only relies on real-time sensor data and does not require any simulation model and/or data-driven strategy, which makes it suitable for online structural health monitoring and damage detection. An experimentally validated benchmark dataset will be used to demonstrate the operational effectiveness of the proposed damage detection methodology.

Committee:
Dr. Dimitri Mavris (advisor), School of Aerospace Engineering
Dr. Graeme Kennedy, School of Aerospace Engineering
Dr. Kai James, School of Aerospace Engineering
Dr. Jason Corman, School of Aerospace Engineering