This research effort explored the possibility of using interwoven conductive and nonconductive fibers in a composite laminate for structural health monitoring (SHM). Traditional SHM systems utilize fiber optics, piezoelectrics, or detect defects by nondestructive test methods by use of sonar graphs or x-rays. However, these approaches are often expensive, time consuming and complicated. The primary objective of this research was to apply a resistance based method of structural health monitoring to a composite structure to determine structural integrity and presence of defects. The conductive properties of fiber such as carbon, copper, or constantan - a copper-nickel alloy - can be utilized as sensors within the structure. This allows the structure to provide feedback via electrical signals to a user which are essential for evaluating the health of the structure. In this research, the conductive fiber was made from constantan wire which was embedded within a composite laminate; whereas prepreg fiberglass, a nonconductive material, serves as the main structural element of the laminate. A composite laminate was constructed from four layers of TenCate 7781 “E” fiberglass and BT250E-1 resin prepreg. Integrating the constantan within the composite laminate provides a sensory element which supplies measurements of structural behavior. Thus, with fiberglass, epoxy, and a constantan conductive element, a three-part composite laminate is developed. Test specimens used in this research were fabricated using a composite air press with the recommended manufacturer cure cycle. A TenCate BT250E-1 Resin System and 7781 "E" impregnated glass-fiber/epoxy weave was used. A constantan wire of 0.01” gauge diameter was integrated into the composite structure. The composite laminate specimen with the integrated SHM system was tested under tensile and flexural loads employing test standards specified by ASTM D3039 and D7264, respectively. These test methods were modified to determine the behavior of the laminate in the elastic range only. A tension and flexural delamination test case was also developed to investigate the sensitivity of the SHM system to inherent defects. Moreover, material characteristic tests were completed to validate manufacturer provided material characteristics. The specimens were tested while varying the constantan configurations, such as the sensor length and orientation. A variety of techniques to integrate the sensor were also investigated. Two different measurement methods were used to determine strain. Strain measurements were made with Instron Bluehill 2 software and correlated to strain obtained by the structural health monitoring system with the use of a data acquisition code written to interact with a micro-ohm-meter. The experimental results showed good agreement between measurements made by the two different methods of measurement. Observations discovered that varying the length of the sensor element improved sensitivity, but resulted in different prediction models when compared to cases with less sensor length. The predictions are based on the gauge factor, which was determined for the each test case. This value provides the essential relationship between resistance and strain. Experiments proved that the measured gauge factor depended greatly on the sensor length and orientation. The correlation was of sufficient accuracy to predict strain values in a composite laminate without the use of any added tools or equipment besides the ohm-meter. Analytical solutions to the loading cases were developed to validate results obtained during experiments. The solutions were in good agreement with the experimental results.

In the field of structural engineering it is of vital importance that the condition of an ageing structure is monitored to detect damages that could possibly lead to failure of the structure. Over the past few years various methods for monitoring the condition of structures have been proposed. With respect to civil and mechanical structures several methods make use of modal parameters such as, natural frequency, damping ratio and mode shapes. In the present work four methods for modal parameter estimation and two methods for have been evaluated for their application to multi degree of freedom structures. The methods evaluated for modal parameter estimation are: Wavelet transform, Hilbert-Huang transform, parametric system identification and peak picking. Through various numerical simulations the effectiveness of these methods is studied. It is found that the simple peak-picking method performs the best and is able to identify modal parameters most accurately in all the simulation cases that were considered in this study. The identified modal parameters are then used for locating the damage. Herein the flexibility and the rotational flexibility approaches are evaluated for damage detection. The approach based on the rotational flexibility is found to be more effective. / Master of Science

Stress and overload are strongly associated with unsafe behaviour, which motivated various studies to detect them automatically in workplaces. This study aims to advance safety research by developing a data-driven stress and overload detection method. An unsupervised deep learning-based anomaly detection method is developed to detect stress. The proposed method performs with convolutional neural network encoder-decoder and long short-term memory equipped with an attention layer. Data from a field experiment with 18 participants was used to train and test the developed method. The field experiment was designed to include a pre-defined sequence of activities triggering mental and physical stress, while a wristband biosensor was used to collect physiological signals. The collected contextual and physiological data were pre-processed and then resampled into correlation matrices of 14 features. Correlation matrices are used as an input to the unsupervised Deep Learning (DL) based anomaly detection method. The developed method is validated, offering accuracy and F-measures close to 0.98. The technique employed captures the input data attributes correlation, promoting higher interpretability of the DL method for easier comprehension. Over-reliance on uncertain absolute truth, the need for a high number of training samples, and the requirement of a threshold for detecting anomalies are identified as shortcomings of the proposed method. To overcome these shortcomings, an Adaptive Neuro-Fuzzy Inference System (ANFIS) was designed and developed. While the ANFIS method did not improve the overall accuracy, it outperformed the DL-based method in detecting anomalies precisely. The overall performance of the ANFIS method is better than the DL-based method for the anomalous class, and the method results in lower false alarms. However, the DL-based method is suitable for circumstances where false alarms are tolerated. / Dissertation / Doctor of Philosophy (PhD)

Prestressed concrete bridges are susceptible to deterioration over time which might significantly affect their capacity and overall performance. In previous decades, infrastructure owners have found that continuous monitoring of these assets is a valuable tool for their management as it facilitates the decision-making process regarding the intervention strategies required. However, as data acquisition and measurement technologies have advanced tremendously in recent years, the amount of information that can be retrieved daily is not easy to manage and analyse. This study presents an evaluation of the effectiveness between different machine learning methods regarding prediction and interpretation of structural responses as well as the feasibility of mapping an independent variable, aspects such as metric performance, learning curves and residual plots was analysed. A comparison was made on the machine learning algorithms performing regression analysis where each model scored over 98% in the R-square metric. This study utilised data collected from a prestressed concrete bridge located in Autio, northern Sweden, that has been continuously monitored for more than three years.

One of the key areas that have attracted attention in the construction industry today is Structural Health Monitoring, more commonly known as SHM. It is a concept developed to monitor the quality and longevity of various engineering structures. The incorporation of such a system would help to continuously track health of the structure, indicate the occurrence/presence of any damage in real time and give us an idea of the number of useful years for the same. Being a recently conceived idea, the state of the art technique in the field is straight forward - populating a given structure with sensors and extracting information from them. In this regard, instrumenting with too many sensors may be inefficient as this could lead to superfluous data that is expensive to capture and process. This research aims to explore an alternate SHM technique that optimizes the data acquisition process by eliminating the amount of redundant data that is sensed and uses this sufficient data to detect and locate the fault present in the structure. Efficient data acquisition requires a mechanism that senses just the necessary amount of data for detection and location of fault. For this reason Compressive Sensing (CS) is explored as a plausible idea. CS claims that signals can be reconstructed from what was previously believed to be incomplete information by Shannon's theorem, taking only a small amount of random and linear non - adaptive measurements. As responses of many physical systems contain a finite basis, CS exploits this feature and determines the sparse solution instead of the traditional least - squares type solution.As a first step, CS is demonstrated by successfully recovering the frequency components of a simple sinusoid. Next, the question of how CS compares with the conventional Fourier transform is analyzed. For this, recovery of temporal frequencies and signal reconstruction is performed using the same number of samples for both the approaches and the errors are compared. On the other hand, the FT error is gradually minimized to match that of CS by increasing the number of regularly placed samples. Once the advantages are established, feasibility of using CS to detect damage in a single degree of freedom system is tested under unforced and forced conditions. In the former scenario, damage is indicated when there is a change in natural frequency of vibration of the system after an impact. In the latter, the system is excited harmonically and damage is detected by a change in amplitude of the system's vibration. As systems in real world applications are predominantly multi-DOF, CS is tested on a 2-DOF system excited with a harmonic forcing. Here again, damage detection is achieved by observing the change in the amplitude of vibration of the system. In order to employ CS for detecting either a change in frequency or amplitude of vibration of a structure subjected to realistic forcing conditions, it would be prudent to explore the reconstruction of a signal which contains multiple frequencies. This is accomplished using CS on a chirp signal. Damage detection is clearly a spatio-temporal problem. Hence it is important to additionally explore the extension of CS to spatial reconstruction. For this reason, mode shape reconstruction of a beam with standard boundary conditions is performed and validated with standard/analytical results from literature. As the final step, the operation deflection shapes (ODS) are reconstructed for a simply supported beam using CS to establish that it is indeed a plausible approach for a less expensive SHM. While experimenting with the idea of spatio-temporal domain, the mode shape as well as the ODS of the given beam are examined under two conditions - undamaged and damaged. Damage in the beam is simulated as a decrease in the stiffness coefficient over a certain number of elements. Although the range of modes to be examined heavily depends on the structure in question, literature suggests that for most practical applications, lower modes are more dominant in indicating damage. For ODS on the other hand, damage is indicated by observing the shift in the recovered spatial frequencies and it is confirmed by the reconstructed response.

No description available.

No description available.

No description available.

No description available.

No description available.