Browsing by Author "Pal, Ashish"

Now showing 1 - 4 of 4
  • Thumbnail Image
    Item
    Deep Learning-Based Subsurface Damage Localization Using Full-Field Surface Strains
    (MDPI, 2023) Pal, Ashish; Meng, Wei; Nagarajaiah, Satish
    Structures in their service life are often damaged as a result of aging or extreme events such as earthquakes or storms. It is essential to detect damage in a timely fashion to ensure the safe operation of the structure. If left unchecked, subsurface damage (SSD) can cause significant internal damage and may result in premature structural failure. In this study, a Convolutional Neural Network (CNN) has been developed for SSD detection using surface strain measurements. The adopted network architecture is capable of pixel-level image segmentation, that is, it classifies each location of strain measurement as damaged or undamaged. The CNN which is fed full-field strain measurements as an input image of size 256 × 256 projects the SSD onto an output image of the same size. The data for network training is generated by numerical simulation of aluminum bars with different damage scenarios, including single damage and double damage cases at a random location, direction, length, and thickness. The trained network achieves an Intersection over Union (IoU) score of 0.790 for the validation set and 0.794 for the testing set. To check the applicability of the trained network on materials other than aluminum, testing is performed on a numerically generated steel dataset. The IoU score is 0.793, the same as the aluminum dataset, affirming the network’s capability to apply to materials exhibiting a similar stress–strain relationship. To check the generalization potential of the network, it is tested on triple damage cases; the IoU score is found to be 0.764, suggesting that the network works well for unseen damage patterns as well. The network was also found to provide accurate predictions for real experimental data obtained from Strain Sensing Smart Skin (S4). This proves the efficacy of the network to work in real-life scenarios utilizing the full potential of the novel full-field strain sensing methods such as S4. The performance of the proposed network affirms that it can be used as a non-destructive testing method for subsurface crack detection and localization.
  • Thumbnail Image
    Item
    Hybrid method for full-field response estimation using sparse measurement data based on inverse analysis and static condensation
    (Elsevier, 2022) Pal, Ashish; Meng, Wei; Nagarajaiah, Satish; Smalley-Curly Institute
    In structural health monitoring, measuring the accurate and spatially dense response near critical locations of the structure can be advantageous to estimate damage to the structure. Due to several physical restrictions or limitations of the sensing method, it may not always be possible to generate reliable data at critical locations. In this study, a hybrid method is presented that makes use of the measured displacement data and finite element (FE) model of the structure to predict dense full-field response. The presented method can incorporate unknown boundary conditions and unknown body forces by applying correction/fictitious forces to match predicted and measured responses. Using static condensation followed by inverse analysis, these additional forces are found by setting up a least square problem. Due to the problem being ill-posed, L2-penalty is used to control the prediction error. Numerical simulation of a plate subjected to body force showed an accurate prediction of full-field response except for a few boundary locations. To handle this, the proposed method is used in conjunction with linear interpolation near boundary locations. The method is validated in a laboratory experiment for a plate with a notch having displacement measured using Digital Image Correlation (DIC). On comparing strains calculated using predicted displacements, FEM, and DIC, the predicted strains show better agreement with the FEM than DIC. This affirms that the proposed hybrid technique can be used at critical locations where DIC fails to provide reliable strain data.
  • Thumbnail Image
    Item
    Next-generation 2D optical strain mapping with strain-sensing smart skin compared to digital image correlation
    (Springer Nature, 2022) Meng, Wei; Pal, Ashish; Bachilo, Sergei M.; Weisman, R. Bruce; Nagarajaiah, Satish
    This study reports next generation optical strain measurement with “strain-sensing smart skin” (S4) and a comparison of its performance against the established digital image correlation (DIC) method. S4 measures strain-induced shifts in the emission wavelengths of single-wall carbon nanotubes embedded in a thin film on the specimen. The new S4 film improves spectral uniformity of the nanotube sensors, avoids the need for annealing at elevated temperatures, and allows for parallel DIC measurements. Noncontact strain maps measured with the S4 films and point-wise scanning were directly compared to those from DIC on acrylic, concrete, and aluminum test specimens, including one with subsurface damage. Strain features were more clearly revealed with S4 than with DIC. Finite element method simulations also showed closer agreement with S4 than with DIC results. These findings highlight the potential of S4 strain measurement technology as a promising alternative or complement to existing technologies, especially when accumulated strains must be detected in structures that are not under constant observation.
  • Thumbnail Image
    ItemEmbargo
    System Identification and Damage Detection Solutions: Sensor Fusion, Algorithms and Physics-Guided Learning
    (2024-08-08) Pal, Ashish; Nagarajaiah, Satish
    The underlying behavior of physical systems is commonly modeled through mathematical equations. Most systems inherently show some amount of nonlinearity in their response. Depending on the system characteristics, surroundings, and application, a variety of nonlinearities can be found. The coefficients of linear and nonlinear functions in the mathematical description of the system provide additional insights. Physical systems such as buildings and bridges change with time because of aging, environmental factors, and extreme events such as earthquakes and storms. Identifying these changes using system identification (SI) and damage detection (DD) methods is crucial. The identification of the correct type of nonlinearity indicates changes in the intrinsic behavior of the system, while the corresponding coefficients help track the changes in the physical properties. The SI and DD methods rely on the system's measured response for inference. The quality of the measurements often guides the quality of the results. Using sensors that provide high-quality data can be expensive and, in some situations, have limited use based on the physical restrictions. Incorporating additional information such as the system's numerical model, governing equation, and data fusion from multiple cheaper/average quality sensors can generate superior-quality data can be used to get reliable and accurate SI and DD results. The advancements in machine learning methods, computational power, and novel full-field strain sensing methods such as strain sensing smart skin can be useful tools for the purpose. In this thesis, several algorithms have been proposed for SI and DD. These include (i) surface DD from fusing the measured full-field strain data and the finite element model of the structure, (ii) deep learning-based subsurface DD from full-field strain data, (iii) fusing coarse displacements and finite element model of the structure to obtain dense accurate displacements, (iv) Kalman filter-based high-quality displacement estimation from data fusion of average-quality displacement and acceleration, (v) Physics-informed SI of nonlinear dynamic systems via deep learning or Kalman filter methods. The proposed algorithms leverage the additional information from the physics of the system or multiple sensors to obtain results that are otherwise not possible to achieve. For example, the finite element model of the structure provides physically consistent information that is used to construct dense displacement data from coarse data, which is far superior to interpolation techniques. The fusion of displacement and acceleration data can find the physical inconsistencies in the measured data, which can be corrected to generate superior-quality displacement and acceleration data. Incorporating physical constraints into the measured data can help identify the unknown governing equation of motion of dynamic systems from field measurements only. Using information from multiple sources in a physically consistent manner can help unfold answers that cannot be pursued individually.