Search for: All records

With the recent success of representation learning methods, which includes deep learning as a special case, there has been considerable interest in developing techniques that incorporate known physical constraints into the learned representation. As one example, in many applications that involve a signal propagating through physical media (e.g., optics, acoustics, fluid dynamics, etc.), it is known that the dynamics of the signal must satisfy constraints imposed by the wave equation. Here we propose a matrix factorization technique that decomposes such signals into a sum of components, where each component is regularized to ensure that it nearly satisfies wave equation constraints. Although our proposed formulation is non-convex, we prove that our model can be efficiently solved to global optimality. Through this line of work we establish theoretical connections between wave-informed learning and filtering theory in signal processing. We further demonstrate the application of this work on modal analysis problems commonly arising in structural diagnostics and prognostics. 

Ultrasonic wavefields are widely employed in nondestructive testing and structural health monitoring to detect and evaluate structural damage. However, measuring wavefields continuously throughout space poses challenges and can be costly. To address this, we propose a novel approach that combines the wave equation with computer vision algorithms to visualize wavefields. Our algorithm incorporates the wave equation, which encapsulates our knowledge of wave propagation, to infer the wavefields in regions where direct measurement is not feasible. Specifically, we focus on reconstructing wavefields from partial measurements, where the wavefield data from large continuous regions are missing. The algorithm is tested on experimental data demonstrating its effectiveness in reconstructing the wavefields at unmeasured regions. This also benefits in reducing the need for expensive equipment and enhancing the accuracy of structural health monitoring at a lower cost. The results highlight the potential of our approach to advance ultrasonic wavefield imaging capabilities and open new avenues for Nondestructive testing and structural health monitoring. 

Recent advancements in physics-informed machine learning have contributed to solving partial differential equations through means of a neural network. Following this, several physics-informed neural network works have followed to solve inverse problems arising in structural health monitoring. Other works involving physics-informed neural networks solve the wave equation with partial data and modeling wavefield data generator for efficient sound data generation. While a lot of work has been done to show that partial differential equations can be solved and identified using a neural network, little work has been done the same with more basic machine learning (ML) models. The advantage with basic ML models is that the parameters learned in a simpler model are both more interpretable and extensible. For applications such as ultrasonic nondestructive evaluation, this interpretability is essential for trustworthiness of the methods and characterization of the material system under test. In this work, we show an interpretable, physics-informed representation learning framework that can analyze data across multiple dimensions (e.g., two dimensions of space and one dimension of time). The algorithm comes with convergence guarantees. In addition, our algorithm provides interpretability of the learned model as the parameters correspond to the individual solutions extracted from data. We demonstrate how this algorithm functions with wavefield videos. 

Guided wave testing is a popular approach for monitoring the structural integrity of infrastructures. We focus on the primary task of damage detection, where signal processing techniques are commonly employed. The detection performance is affected by a mismatch between the wave propagation model and experimental wave data. External variations, such as temperature, which are difficult to model, also affect the performance. While deep learning models can be an alternative detection method, there is often a lack of real-world training datasets. In this work, we counter this challenge by training an ensemble of variational autoencoders only on simulation data with a wave physics-guided adversarial component. We set up an experiment with non-uniform temperature variations to test the robustness of the methods. We compare our scheme with existing deep learning detection schemes and observe superior performance on experimental data. 

While guided wave structural health monitoring (SHM) is widely researched for ensuring safety, estimating performance deterioration, and detecting damage in structures, it experiences setbacks in accuracy due to varying environmental, sensor, and material factors. To combat these challenges, environmentally variable guided wave data is often stretched with temperature compensation methods, such as the scale transform and optimal signal stretch, to match a baseline signal and enable accurate damage detection. Yet, these methods fail for large environmental changes. This paper addresses this challenge by demonstrating a machine learning method to predict stretch factors. This is accomplished with feed-forward neural networks that approximate the complex velocity change function. We demonstrate that our machine learning approach outperforms the prior art on simulated Lamb wave data and is robust with extreme velocity variations. While our machine learning models do not conduct temperature compensation, their accurate stretch factor predictions serve as a proof of concept that a better model is plausible.

 

Environmental effects are a significant challenge in guided wave structural health monitoring systems. These effects distort signals and increase the likelihood of false alarms. Many research papers have studied mitigation strategies for common variations in guided wave datasets reproducible in a lab, such as temperature and stress. There are fewer studies and strategies for detecting damage under more unpredictable outdoor conditions. This article proposes a long short-term principal component analysis reconstruction method to detect synthetic damage under highly variational environments, like precipitation, freeze, and other conditions. The method does not require any temperature or other compensation methods and is tested by approximately seven million guided wave measurements collected over 2 years. Results show that our method achieves an area under curve score of near 0.95 when detecting synthetic damage under highly variable environmental conditions. 

The role of anisotropic grain boundary energy in grain growth is investigated using textured microstructures that contain a high proportion of special grain boundaries. Textured and untextured Ca‐doped alumina was prepared by slip casting inside and outside a high magnetic field, respectively. At 1600°C, the textured microstructure exhibits faster growth than the untextured microstructure and its population of low‐angle boundaries increases. Atomic force microscopy (AFM) is employed to measure the geometry of thermal grooves to assess the relative grain boundary energy of these systems before and after growth. In the textured microstructure, the grain boundary energy distribution narrows and shifts to a lower average energy. Conversely, the energy distribution broadens for the untextured microstructure as it grows and exhibits abnormal grain growth. Further analysis of the boundary networks neighboring abnormal grains reveals an energy incentive that facilitates their growth. These results suggest that coarsening is not the only dominant grain growth mechanism and that the system can lower its energy effectively by replacing high energy boundaries with those of low energy. The faster growth of lower energy boundaries suggests that isotropic simulations do not adequately account for anisotropic grain growth mechanisms or anisotropic mobility.

 

This paper studies the effectiveness of joint compression and denoising strategies with realistic, long-term guided wave structural health monitoring data. We leverage the high correlation between nearby collections of guided waves in time to create sparse and low-rank representations. While compression and denoising schemes are not new, they are almost exclusively designed and studied with relatively simple datasets. In contrast, guided wave structural health monitoring datasets have much more complex operational and environmental conditions, such as temperature, that distort data and for which the requirements to achieve effective compression and denoising are not well understood. The paper studies how to optimize our data collection and algorithms to best utilize guided wave data for compression, denoising, and damage detection based on seven million guided wave measurements collected over 2 years.

 

Guided ultrasonic wave localization systems use spatially distributed sensor arrays and wave propagation models to detect and locate damage across a structure. Environmental and operational conditions, such as temperature or stress variations, introduce uncertainty into guided wave data and reduce the effectiveness of these localization systems. These uncertainties cause the models used by each localization algorithm to fail to match with reality. This paper addresses this challenge with an ensemble deep neural network that is trained solely with simulated data. Relative to delay-and-sum and matched field processing strategies, this approach is demonstrated to be more robust to temperature variations in experimental data. As a result, this approach demonstrates superior accuracy with small numbers of sensors and greater resilience to spatially nonhomogeneous temperature variations over time.