Session S10

Session S10: Structural Health Monitoring

Time:                 Wednesday, May 14, 10:30-12:00
Chair:                 Martin P. DeSimio, University of Dayton Research Institute
Co-Chair:          Daniel Huff, Boeing Company

S10-1: Damage Identification in Composite Sandwich Helicopter Blades using Full-Field Laser Displacement and Point Acceleration Frequency Response Function Measurements
Brandon Zwink, David Koester and Douglas Adams Purdue University, West Lafayette IN
Composite helicopter blades and fuselage structures are currently inspected visually or by using an audible coin tap method to identify the presence of sub-surface damage. If damage is detected, local ultrasonic inspections are performed to assess the severity of the damage (depth, area). Global damage identification methods are needed to reduce the time required to identify damage in wide areas of the blade and fuselage. A vibration-based methodology for identifying damage is discussed in this paper. A piezo actuator is attached to a simulated composite blade and is used to stimulate the blade with broadband random vibrations. A scanning laser vibrometer is then used to measure the full-field displacement of the blade. The frequency response functions between the drive voltage and the displacement measurements are used to identify damage due to debonding between the face sheet and honeycomb core. It is shown that the full-field measurements are effective at characterizing the damaged area of the blade over a wide frequency range; however, for damage located far from the actuator, the lower frequency portion of the measured data is most effective at identifying the damage. Damage due to cutting of the sandwich panels for testing is also identified along the panel edges. Point acceleration measurements are also used for the purpose of identifying the damaged region of the blade based on the used of modal deflection shapes in addition to the frequency response function measurements. The feasibility of global inspection for damage identification in wide area composite structures is demonstrated with these local measurements.

S10-2: Co-Located Triangulation: Single Node Damage Position Identification
Seth S. Kessler and Ajay Raghavan Metis Design Corporation, Cambridge MA
A fundamental limitation for current Structural Health Monitoring (SHM) systems is the need for distributed synchronous sensors to determine precise damage location using traditional triangulation methods. Accuracy is dictated by sensor, which drives complexity, weight and cost to resolve reliable position. This paper introduces a patent-pending method to predict accurate damage location from a single SHM node. This co-located triangulation method consists of novel sensors, algorithms and hardware to achieve a significantly more efficient means of localizing damage. Results are presented for proof-of-concept experiment on an aluminum plate using guided waves. Finally, a description of a prototype under development is presented.

S10-3: Impact Localization in a Composite Wing Structure
Steven E. Olson, Martin P. DeSimio University of Dayton Research Institute, Dayton OH, Kevin S. Brown and Mark M. Derriso Air Force Research Laboratory, Wright-Patterson Air Force Base OH
This paper presents a method for estimating impact locations in composite material. Analytical and experimental studies are performed on a composite graphite/epoxy wing structure. The Gauss-Newton method has been widely used to solve source localization problems in isotropic media where the speed is uniform over all propagation directions. Initial experiments are conducted to demonstrate the localization method in an isotropic metal panel. Piezoelectric transducers sense the elastic waves which propagate from the point-of-impact. Assuming known sensor locations and propagation speed, measured times of wave arrival at a set of sensors can be used to estimate the impact location. However, because composite materials are anisotropic, the equation relating wave times of arrival to impact location must be adjusted to account for speed variation as a function of propagation direction. In subsequent experiments on the composite wing structure, the method incorporating directionally dependent velocity is applied to the localization task.

S10-4: Selected Signal Processing Methods for Subsystem Fault Detection within an Integrated Systems Health Management Architecture
Michael J. Roemer, Carl S. Byington and Michael S. Schoeller Impact Technologies LLC, Rochester NY
The work presented herein will highlight selected signal processing-based approaches as applied within subsystems of an Integrated System Health Management (ISHM) architecture for the purposes of incipient fault detection. The selected vehicle subsystem areas to be discussed will include electro-mechanical actuators (EMAs), propulsion system vibration, vehicle structural integrity and general signal anomaly detection. Signal processing and artificial intelligence methods including adaptable signal demodulation, stochastic variable decomposition and probabilistic networks are described within the context of incipient fault detection that can provide a longer time horizon for prognostic implementations. In addition, discussion on individual subsystem event indicators as applied within an intelligent, model-based reasoning approach will also be presented that examines how the fault detection results are used within the overall vehicle health management architecture. The signal processing based implementations described will illustrate the integration of advanced fault detection and reasoning capabilities that can be applied across critical subsystems on a vehicle platform. The examples provided will describe the details of the selected technologies and specific implementation issues that often arise with computationally intensive based approaches.

S10-5: Signal Processing for Lamb Wave Signal Excited by Piezoelectric Patches
Haiying Huang University of Texas at Arlington, Arlington TX and Thierry Pamphile Air Force Research Laboratory, Wright-Patterson Air Force Base OH
Piezoelectric (piezo) patches has become popular for Lamb wave-based structural health monitoring because it can be utilized as both an actuator and a sensor. In addition, different Lamb wave modes can be selected excited by frequency tuning. In this paper, we will discuss employing signal processing technique to obtain the tuning curves of a circular piezo patch actuator. Based on the theoretical model of Lamb wave propagation, the signal processing procedure performs preconditioning, windowing, enveloping, and feature extraction of the sensing signals automatically. The tuning curves of the piezo patch actuator measured at different directions indicated that the Lamb wave field is controlled by the bending dynamics of the piezo actuator.

S10-6: Approaches for Rapid Assessment of Corrosion Protection Condition
Daniel Huff Boeing Company, Mesa AZ
Impedance spectra of exposed aluminum panels were transformed into the time-frequency plane and characterized with respect to the degree of corrosion and type of protective treatment. Analyses were performed in combination with results obtained by equivalent circuit modeling of surface interfaces as well as results obtained from image analysis of the exposed panels.

S10-7: Time-Frequency Methods For Structural Health Monitoring
Debejyo Chakraborty, Wenfan Zhou, Donna Simon, Narayan Kovvali, Antonia Papandreou-Suppappola, Douglas Cochran and Aditi Chattopadhyay Arizona State University, Tempe AZ
The ability to effectively detect and classify damage in complex materials and structures is an important problem in the area of structural health monitoring (SHM). The goal is to provide indicators about the presence, location, size, or severity of damage in a structure of interest. In this paper, we review two stochastic damage classification schemes based on the use of time-frequency techniques. The first method utilizes the matching pursuit decomposition (MPD) to construct cross-term free time-frequency representations (TFRs) of the structural data, with classification performed based on correlations in the time-frequency plane. The second method relies on using hidden Markov models (HMMs) to model time-frequency damage features extracted from structural data using the MPD, and classification is performed in a Bayesian framework. In both cases, the MPD is employed with time-frequency-scale dictionaries composed of highly localized Gaussian functions. Results are presented from an example application to the classification of fatigue-induced crack damage in an aluminum lug-joint specimen, and the utility of the techniques is discussed.