Debejyo Chakraborty

Damage Classification Structural Health Monitoring in Bolted Structures Using Time-frequency Techniques

The analysis, detection, and classification of damage in complex bolted structures is an important component of structural health monitoring. In this article, an advanced signal processing and classification method is introduced based on time-frequency techniques. The time-varying signals collected from sensors are decomposed into linear combinations of highly localized Gaussian functions using the matching pursuit decomposition algorithm. These functions are chosen from a dictionary of time-frequency shifted and scaled versions of an elementary Gaussian basis function. The dictionary is also modified to use real measured data as the basis elements in order to obtain a more parsimonious signal representation. Classification is then achieved by matching the extracted damage features in the time-frequency plane. To further improve classification performance, the information collected from multiple sensors is integrated using a Bayesian sensor fusion approach. Results are presented demonstrating the algorithm performance for classifying signals obtained from various types of fastener failure damage in an aluminum plate.

An adaptive learning damage estimation method for structural health monitoring

Structural health monitoring is an important problem of interest in many civil infrastructure and aerospace applications. In the last few decades, many techniques have been investigated to address the detection, estimation, and classification of damage in structural components. One of the key challenges in the development of real-world damage identification systems, however, is variability due to changing environmental and operational conditions. Conventional statistical methods based on static modeling frameworks can prove to be inadequate in a dynamic and fast changing environment, especially when a sufficient amount of data is not available. In this paper, a novel adaptive learning structural damage estimation method is proposed in which the stochastic models are allowed to perpetually change with the time-varying conditions. The adaptive learning framework is based on the use of Dirichlet process (DP) mixture models, which provide the capability of automatically adjusting to structure within the data. Specifically, time–frequency features are extracted from periodically collected structural data (measured sensor signals), that are responses to ultrasonic excitation of the material. These are then modeled using a DP mixture model that allows for a growing, possibly infinite, number of mixture components or latent clusters. Combined with input from physically based damage growth models, the adaptively identified clusters are used in a state-space setting to effectively estimate damage states within the structure under varying external conditions. Additionally, a data selection methodology is implemented to enable judicious selection of informative measurements for maximum performance. The utility of the proposed algorithm is demonstrated by application to the estimation of fatigue-induced damage in an aluminum compact tension sample subjected to variable-amplitude cyclic loading.

Damage Classification for Structural Health Monitoring Using Time-Frequency Feature Extraction and Continuous Hidden Markov Models

We propose an algorithm for the classification of structural damage based on the use of the continuous hidden Markov modeling (HMM) technique. Our approach employs HMMs to model time-frequency damage features extracted from structural data using the matching pursuit decomposition algorithm. We investigate modeling with continuous observation-density HMMs and discuss the trade-offs involved as compared to the discrete HMM case. A variational Bayesian method is employed to automatically estimate the HMM state number and adapt the classifier for real-time use. We present results that classify structural and material (fatigue) damage in a bolted-joint structure.

Time-Frequency based Classification of Structural Damage

The detection and classification of damage in complex materials and structures is essential from both safety and economic perspectives. In this paper, we propose algorithms for the classification of structural damage based on time-frequency techniques. Our approach is based on matching damage features in the time-frequency plane using highly localized Gabor functions and time-varying received signals from real experimental measurements. Example results are presented for the classification of fastener damage in an aluminum plate, demonstrating the utility of the proposed methodology.

Ultrasonic sensing and time-frequency analysis for detecting plastic deformation in an aluminum plate

We investigate the use of low frequency (10-70 MHz) laser ultrasound for the detection of fatigue damage. While high frequency ultrasonics have been utilized in earlier work, unlike contacting transducers, laser-based techniques allow for simultaneous interrogation of the longitudinal and shear moduli of the fatigued material. The differential attenuation changes with the degree of damage, indicating the presence of plasticity. In this paper, we describe a structural damage identification approach based on ultrasonic sensing and time-frequency techniques. A parsimonious representation is first constructed for the ultrasonic signals using the modified matching pursuit decomposition (MMPD) method. This decomposition is then employed to compute projections onto the various damage classes, and classification is performed based on the magnitude of these projections. Results are presented for the detection of fatigue damage in Al-6061 and Al-2024 plates tested under 3-point bending.

Physics based modeling for time-frequency damage classification

We have recently proposed a method for classifying waveforms from healthy and damaged structures in a structural health monitoring framework. This method is based on the use of hidden Markov models with preselected feature vectors obtained from the time-frequency based matching pursuit decomposition. In order to investigate the performance of the classifier for different signal-to-noise ratios (SNR), we simulate the response of a lug joint sample with different crack lengths using finite element modeling (FEM). Unlike experimental noisy data, the modeled data is noise free. As a result, different levels of noise can be added to the modeled data in order to obtain the true performance of the classifier under additive white Gaussian noise. We use the finite element package ABAQUS to simulate a lug joint sample with different crack lengths and piezoelectric sensor signals. A mesoscale internal state variable damage model defines the progressive damage and is incorporated in the macroscale model. We furthermore use a hybrid method (boundary element-finite element method) to model wave reflection as well as mode conversion of the Lamb waves from the free edges and scattering of the waves from the internal defects. The hybrid method simplifies the modeling problem and provides better performance in the analysis of high stress gradient problems.

Big Data-Driven Manufacturing—Process-Monitoring-for-Quality Philosophy

Discussion of big data has been about data, software, and methods with an emphasis on retail and personalization of services and products. Big data also has impacted engineering and manufacturing and has resulted in better and more efficient manufacturing operations, improved quality, and more personalized products. A less apparent effect is that big data has changed problem solving: the problems we choose to solve, the strategy we seek, and the tools we employ. This paper illustrates this point by showing how the big data style of thinking enabled the development of a new quality assurance philosophy called process monitoring for quality (PMQ). PMQ is a blend of process monitoring and quality control that is founded on big data and big model, which are catalysts for the next step in the evolution of the quality movement. Process monitoring for quality was used to evaluate the performance of the ultrasonically welded battery tabs in the new Chevrolet Volt, an extended range electric vehicle. Index terms — Manufacturing, big data, big models, problem solving strategy, process monitoring for quality, acsensorization, quality control Nomenclature α rate of type I error β rate of type II error BD big data BDBM big data – big models BM big models D3M data-drive discovery of models DFSS design for six sigma DLL dynamic-link library GP genetic programming ICE internal combustion engine LVDT linear variable differential transformer MCS multiple classifier systems MPCD maximum probability of correct decision PMQ process monitoring for quality

Active learning data selection for adaptive online structural damage estimation

Adaptive learning techniques have recently been considered for structural health monitoring applications due to their flexibility and effectiveness in addressing real-world challenges such as variability in the monitoring of environmental and operating conditions. In this paper, an active learning data selection procedure is proposed that adaptively selects the most informative measurements to include, from multiple available measurements, in estimating structural damage. This is important, since not all the measurements may provide useful information and could reduce performance when processed. Within the adaptive learning framework, the data selection problem is formulated to choose those measurements which are most representative of the diversity within a damage state. This is achieved by extracting time-frequency analysis based statistical similarity features from the measurements, and selecting uniformly distributed subsets to build representative reference sets. The utility of the proposed method is demonstrated by improvements in adaptive learning performance for the estimation of fatigue damage in an aluminum compact tension sample.

A comparative study of fatigue damage sensing in aluminum alloys using electrical impedance and laser ultrasonic methods

Fatigue damage sensing and measurement in aluminum alloys is critical to estimating the residual useful lifetime of a range of aircraft structural components. In this work, we present electrical impedance and ultrasonic measurements in aluminum alloy 2024 that has been fatigued under high cycle conditions. While ultrasonic measurements can indicate fatigue-induced damage through changes in stiffness, the primary indicator is ultrasonic attenuation. We have used laser ultrasonic methods to investigate changes in ultrasonic attenuation since simultaneous measurement of longitudinal and shear properties provides opportunities to develop classification algorithms that can estimate the degree of damage. Electrical impedance measurements are sensitive to changes in the conductivity and permittivity of materials - both are affected by the microstructural damage processes related to fatigue. By employing spectral analysis of impedance over a range of frequencies, resonance peaks can be identified that directly reflect the damage state in the material. In order to compare the impedance and ultrasonic measurements for samples subjected to tension testing, we use processing and classification tools that are matched to the time-varying spectral nature of the measurements. Specifically, we process the measurements to extract time-frequency features and estimate stochastic variation properties to be used in robust classification algorithms. Results are presented for fatigue damage identification in aluminum lug joint specimens.

Adaptive learning for damage classification in structural health monitoring

A key challenge in real-world structural health monitoring (SHM) is diversity of damage phenomena and variability in environmental and operational conditions. Conventional learning techniques, while adequate for moderately complex inference tasks, can be limiting in highly complex and rapidly changing environments, especially when insufficient data is available. We present an adaptive learning methodology where stochastic models continuously evolve with the time-varying environment and Dirichlet process mixture models are utilized to self-adapt to structure within the data. Coupled with appropriate physics-based phenomenology, the approach provides an adaptive and effective framework for online SHM. The proposed technique is demonstrated for the detection of progressive fatigue damage in a metallic structure under variable-amplitude loading.