Structural Health Monitoring and Maintanence System for Composite Materials. PhD thesis
Diğer yayınlar > Diğer · 2014
Yazarlar: Mahmut PEKEDİS
Diğer, Ulusal, 2014, ss 1-200, Türkiye
Özet
<p>Advances in engineering structures have focused on building structures that effectively integrated with nervous systems that similar to living organisms in order to detect, localize and quantify the damages. Through an increased automation, the human role has shifted from an emphasis on the perceptual-motor skills needed for manual diagnosis process to autonomous decision maker. Since the last two decades, developments in damage detection algorithms have greatly increased the potential for autonomous decisions of structural health. The model based structural health monitoring (SHM) approach is usually applied by building a physics-based model of the structure of interest. Once the model is built, based on a detailed physical description of the system, it is usually updated on the basis of measured data from the structure. The data-based approach is focused on collecting a wide array of measurements from sensors, extracting features from data, comparing data with models to determine the current state of system health. A number of advances have been made in SHM using these approaches. For example, the first case of this research was related to diagnose the delamination in laminated composite beam and plate by using the model based SHM technique. On the other hand, it is still struggling to build autonomous tools that can match the ability of a human to detect and localize a quantity of damage in structures by removing the human cognitive motor ability from these SHM tools. Therefore, there is a need for growing interest in merging the computational and cognitive technologies- techniques and concepts to improve the solution of SHM. The first main object of this research is to explore the efficiency of the developed innovative human-machine cooperative paradigm for SHM applications on human subjects. The idea of the paradigm is that human-machine cooperation always outperforms a machine or a human, acting independently. The cooperation paradigm is including the human nervous system, SHM sensor networks, signal processing techniques, feature extraction methods, multivariate statistical methods, psychophysics procedures, haptic technology to create a harmonious collaboration among humans, neural plasticity and sensory substitution theories. In order to explore the possibility of giving humans a new sense, one that reflects the health of structure by using the similar philosophy, the investigation was performed using a steel tower structure that was constructed with bolted lattices. Damage scenarios in the structure were conducted by releasing some of the bolt loads which has a similar effect of delamination damage that mostly encountered in composite structures. Accelerometers were bonded to various locations of the tower members to measure the dynamic response of the system. Initially, the results obtained from the accelerometers were pro-processed with principal component analysis (PCA). Next, the scores obtained from PCA were encoded in three different ways to represent them as physical haptics stimuli for the human subjects. Results obtained from the human-machine cooperation paradigm demonstrated that subjects were able to recognize the damage with an accuracy of 88 ± 20.21 % and response time of 5.87 ±2.33 seconds when the stimulations are represented in sequence. They were also able to determine the damage with an accuracy of 68.98 ± 30.91 % and response time of 8.51± 3.22 seconds when the encoded stimulations are represented simultaneously to the human subjects. Once the human-machine cooperative paradigm was tested on human subjects, the next task is to detect the damage via support vector machine on SHM measures by using the multi classification module. The SVM results were analyzed and evaluated in detail for linear, polynomial, radial and sigmoid kernels. The results of SVM showed that, the highest accuracy scores were observed in radial basis function kernel with an accuracy of 60.63 % in 6.043914 s of CPU time. As a result, it can be concluded that the human-machine cooperation concept, provides a useful framework to interact with abstract entities such as data from sensor network of SHM. The second main object of this research is to develop an integrated SHM-maintenance algorithm to conduct maintenance process requirements for damages, beside the SHM technique. First, the basic concepts of algorithm were demonstrated. Next, it was applied on a cracked laminated composite beam that subjected to an out of axial plane force. The strain energy method (SEM) was used to detect to damage. Once the damage location was diagnosed, the piezoelectric actuator was activated automatically with an optimized voltage value that was obtained from root mean square error (RMSE) in order to control the integrity of composite beam in a safe manner. In other words, it was investigated that, whether the effect of damage in composite beam could be controlled via piezoelectric actuator. As a result, it can be concluded that, the developed SHM-maintenance coupling algorithm could be used in an effective way to detect the damage in cracked fiber reinforced composite beam that subjected to bending load as well as it can perform a maintain process for this cracked beam to decrease the effect of damage by using the fracture mechanics theories, SEM and RMSE indicators. Keywords: Structural health monitoring, Damage diagnosis, Damage sensation, Human-machine interface, Nociceptors, Haptics systems, Sensory substitution, Machine learning, Fiber reinforced composite structures, Piezoelectric actuators, Damage control, SHM+Maintenance system</p>