MS17.9 - Structural Health Monitoring

In operational modal analysis, the observability and quality of the extracted modal parameters, apart from the characteristics of structures, loads, and noise, largely depend on the data acquisition system. Especially in large-scale structures, because of the low frequency- and low amplitude of vibration, the configuration and arrangement of the sensors is of particular interest. Milad Tower, with a height of 435m in Tehran, is the sixth tallest telecommunication tower in the world. The authors have previously identified the tower's modal parameters by developing an automated data-driven subspace approach and investigated the long-term variations of modal parameters for structural health monitoring purposes. This paper adopts this identification approach to a Milad Tower finite element model subjected to random and seismic loadings. The identified modal parameters are compared with those obtained based on real acceleration records. The performance of the MEMS accelerometers is investigated through the stabilization of identified natural frequencies for different setups and scenarios. Also, the performance of various accelerometers temporarily mounted on the tower is compared, and a response signal measured under the most recent earthquake in Tehran is presented. The qualitative and quantitative assessment of the sensors are discussed, and required improvements are recommended.

Structural health monitoring (SHM) systems have become a common practice to investigate the integrity and safety of civil infrastructures. By embedding sensor networks on vital structures, raw monitoring data from the physical world could be captured. Due to harsh environments and/or electromagnetic interference, sensors on structures are prone to produce unusual and erroneous readings, often known as outliers. These faulty, erroneous, corrupted outliers are not uncommon, and their existence may greatly influence the results of structural health evaluation and lead to false alarms. Therefore, soon after measurement, there is a high demand for exe-cuting data cleaning for SHM data. In this study, we propose a novel robust gross outlier re-moval method, termed Hankel-structured robust principal component analysis (HRPCA), to effectively remove the unwanted gross outliers embedded in the monitoring data of structural dynamic responses. Different from the deep-learning-based approaches that possess only outli-er identification or anomaly classification ability, HRPCA is a rapid and integrated methodolo-gy for data cleaning, which enables outlier detection, outlier identification, and recovery of fault data. It capitalizes on the fundamental duality between the sparsity of the signal and the rank of the structured matrix. Using of the annihilating filter-based fundamental duality, the originally redundant yet relevant structural dynamic response data could be modeled as lying in a low-dimensional (low-rank) subspace with additional Hankel structure, which allows for better sep-aration of gross outliers embedded in the monitoring data. Then the outlier removal issue turns into a matrix factorization problem, which could be successfully solved by robust principal component analysis (RPCA). To validate the denoising capability of HRPCA, real-world moni-toring data with varying degrees of outliers (e.g., single outlier, multiple outliers, and periodic outliers) is used to illustrate the efficiency of the proposed method.

The study aims at exploring the capabilities and performance limits of advanced robust approaches, such as those based on the novel Multiple Model (MM) and the Hyper-Sphere (HS) principles, for random vibration response-based damage detection and characterization for a population of composite aerostructures under varying operating conditions and uncertainty. The study is based on Monte Carlo simulations using digital, Abaqus based, models thus allowing for maximum flexibility in experimentation. Each aerostructure is a Carbon-Epoxy square hollow beam of length equal to 2570 mm and varying cross section. It consists of two main parts, the web and the skin, which are assembled using adhesive and models the tail boom of a UAV. The population is characterized by material and manufacturing uncertainty, modeled via ±10% variation around the nominal material property values. Additional uncertainty is caused by temperature variability (within the -55…71 ℃ range) as well as force excitation profile uncertainty due to the employment of excitation realizations related to two distinct stochastic profiles within the 0-2 kHz frequency range. Two types of early-stage damage are considered: Debonding and Delamination, each one of two different sizes and at one of two distinct locations; hence a total of 8 damage scenarios. Debonding is simulated via adhesive stiffness degradation and delamination via corresponding degradation on all skin layers. Both damage types induce small effects on the structural dynamics which are largely masked by those due to varying operating conditions and uncertainty, thus leading to a challenging diagnosis problem. Multiple Model (MM) and Hyper-Sphere (HS) based advanced data-driven robust methods employing Multiple Input Single Output (MISO) Transmittance Function representations of the structural dynamics estimated as stochastic MISO AutoRegressive with eXogenous excitation (MISO-TF-ARX) models are employed. This model type is selected as it achieves elimination of excitation effects, while the MISO-TF-ARX model parameter vector, or Principal Component Analysis (PCA) reduced versions, constitute the methods’ feature within which distinct approximations of the healthy subspace are constructed (Baseline Phase). Once a fresh set of vibration response signals is obtained, a corresponding MISO-TF-ARX model corresponding to a point in the feature space is estimated, and damage detection is based on examining a proper distance metric of it from the healthy subspace (Inspection Phase). Once damage is detected, its characterization, in terms of damage type, location, and size, is attempted based on a supervised hierarchical classification scheme that employs an angle-based approach within the feature space. Damage detection is comprehensively assessed with a population of 18 healthy aerostructures (9 employed in the Baseline and 9 in the Inspection Phases) and 72 damaged structures, different from their Baseline counterparts, while a total of 1323 Monte Carlo numerical experiments are run for the population. The methods achieve high detection performance, reaching a 100% correct detection rate for 0% false alarm rate for the best performing PCA-HS-TF method despite the few Baseline experiments. Damage characterization results are also very promising, characterized by 80% correct classification rate.