KEYWORDS: Sensors, Data modeling, Education and training, Structural health monitoring, Data acquisition, Animal model studies, Machine learning, Deep learning, Systems modeling, Smart sensors
In the realm of structure health monitoring for pressure vessels intended for space habitats, identifying sensor anomalies is of critical importance. The sensor anomalies are data patterns that diverge from anticipated measurement behaviors. To address the multifaceted challenges, we propose a hierarchical mechanism for sensor anomaly detection. This strategic approach not only filters out aberrant data but also subsequently ensures the extraction of reliable results for structure health monitoring, providing a safeguard against potential erroneous decision-making. Furthermore, this approach allows for efficient data handling across multiple sensors and incorporates physical knowledge into the deep learning model to comprehensively detect any sensor anomalies that are physically implausible. As a result, we achieve a more holistic and robust detection of sensor anomalies, ensuring heightened reliability in health monitoring for pressure vessel.
The process of identifying structural damage can be approached as an optimization challenge, where the goal is to bridge the gap between observed experimental data and theoretical model predictions. This way creates the likelihood to apply the metaheuristic algorithms for the inverse damage identification. Through experiments, we can collect vibration data such as acceleration responses, natural frequencies, mode shapes, and piezoelectric impedance, which serve as indicators of damage. However, such data may be tainted with noise or errors. Furthermore, limitations in model accuracy or a lack of comprehensive understanding of experimental boundary conditions can inject uncertainties into the damage detection process. Traditional probabilistic methods have been employed to counter these uncertainties, but they often rely on predefined statistical distributions, typically Gaussian distribution. In real-world applications, the myriad sources of uncertainty and the paucity of specific experimental data can make it difficult to exactly ascertain these distributions. In this regard, the non-probabilistic interval analysis is introduced. This method leans on the defined bounds of uncertainty in data, rather than their probabilistic nature. It assesses structural damage by measuring factors like the nominal reduction in stiffness, the likelihood of damage, and an index that combines the two, which are quantified through the non-probability reliability method. Besides, the reduced order modeling through component mode synthesis is adopted to speed up the optimization iterations. To validate this approach, vibration-based attributes are used for truss structure, ensuring a robust identification of structural damage when faced with uncertainties in data.
Fault parameters in a structure are identified by matching measurements with model predictions in the parametric space. As high frequency measurements are preferred to uncover small-sized damage, piezoelectric impedance/admittance active interrogation has shown promising aspects. Nevertheless, challenges remain. The amount of useful measurement information is generally insufficient to pinpoint damage. The inverse identification is usually underdetermined. In this research, we develop a combinatorial enhancement to tackle these challenges. A tunable piezoelectric impedance sensing procedure is developed in which an adaptive inductor element is integrated with the piezoelectric transducer, which will lead to significantly enriched measurement data for the same damage. Subsequently, an intelligent learning automata-based multi-objective particle swarm optimization framework is synthesized to inversely identify the damage location and severity. Case studies are conducted to highlight the accuracy of the damage identification.
Structural damage identification using the impedance/admittance measurements of a piezoelectric transducer can be converted into a multi-objective optimization framework targeting the minimization of the discrepancy between prediction and experimental measurements, with damage locations and severities as unknown variables. However, the unknowns are usually on a large scale and show sparse characteristics since the damage only occurs at a small area. This places the burden on the optimization algorithms in the identification process. Here, a sparse initialization algorithm is introduced to generate a sparse population for the large-scale variables to tackle the challenge. The algorithm is combined with the particle swarm algorithm to locate and quantify the damage for verification purposes. Several cases are considered, and the results show that the algorithm can generate high-quality damage identification solutions with limited simulation and experimental measurements.
KEYWORDS: Optimization (mathematics), Transducers, Algorithm development, Finite element methods, Structural health monitoring, Detection and tracking algorithms, Process modeling, Stochastic processes, Image segmentation, Analytical research
In this research, we report a new fault identification algorithm utilizing multi-objective optimization. Fault identification problem is commonly under-determined, as measurement information may not be sufficient to facilitate a direct inversion. We formulate an optimization problem, aiming at minimizing the discrepancy between model prediction and measurement. This yields multiple possible fault scenarios, which lays down foundation for further inspection.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.