LEE- Katı Cisimlerin Mekaniği Lisansüstü Programı
Bu topluluk için Kalıcı Uri
Gözat
Yazar "Mugan, Ata" ile LEE- Katı Cisimlerin Mekaniği Lisansüstü Programı'a göz atma
Sayfa başına sonuç
Sıralama Seçenekleri
-
ÖgeHigh cycle fatigue life estimation in frequency domain using multi input multi output Q-T matrix method(Graduate School, 2019-12-19) Topak Kula, Emel ; Mugan, Ata ; 503171502 ; Solid Mechanics ; Katı Cisimlerin MekaniğiComponent failures due to vibrational fatigue damage is a common problem in automotive industry. There are many studies in literature to predict the fatigue life of vehicle components. It is reported that frequency domain fatigue life estimation methods (FDFEMs) have advantages over time domain approaches. In this study, FDFEMs are applied to vibrational fatigue problems having multi inputs and multi outputs (MIMO). In particular, high cycle fatigue life estimation problem was considered and S-N diagrams were mainly employed in analyses. Input-output relations are obtained in frequency domain using the Q-T matrix method that enables using a reduced order model in frequency domain. The Q-T matrix method for MIMO system representation is convenient to obtain transfer functions between selected input and output degrees of freedom (DOFs) considering the cross-correlations among the DOFs. Following, popular FDFEMs such as Narrow Band Approximation, Tovo- Benasciutti, Zhao and Baker, and Dirlik Methods are applied to some critical components in a truck chassis based on MIMO Q-T matrix models. The FDFEMs considered in this study can be classified as narrow band (namely, Narrow Band Approximation) and wide band (namely, Tovo- Benasciutti, Zhao and Baker, and Dirlik Methods) processes. Comparisons are made with the Ncode solutions and experimental results. It is concluded that the FDFEMs associated with the MIMO Q-T matrix method yield accurate fatigue life estimations and their CPU times are much less than those of full order models. In this thesis, it is analyzed how to get the transfer functions between accelerations of input DOFs and accelerations of critical DOFs of some critical parts. Q-T matrix method is a convenient method to obtain transfer functions of critical locations on a vehicle using some chassis input locations, from inputs' power spectral density (PSD) to outputs' PSD. Acceleration values at each input point are measured by an accelerometer. Output PSDs are obtained by Q-T matrix method and turned into stress values. Q-T matrix method which is entegrated with FDFEMs can be applied to random excitation and MIMO problems in frequency domain. Due to the accuracy of results and consideration of cross-coupling effects, Q-T matrix method was preferred in this study. Due to the fact that knowledge of input force is not neccesary, operational modal analysis can be implemented. Therefore, it is cheaper and faster compared to experimental modal analyses. The Frequency Response Functions (FRFs) used in the formulations of Q-T matrix method were obtained by finite element models. To this end, Nastran software was used to obtain FRFs. Nonetheless, the FRFs could be obtained by experimental modal tests in which multiple shakers should be employed. Applicability of this study is in real-time and on-board for heavy duty truck is investigated. The instrumentation of a heavy duty truck produced by Otosan was completed, accelerometers were attached to the critical locatons of the truck chassis. Then, the Raspberry Pi controller was used for computational tasks. The thesis was a part of a Tubitak Ardeb project focusing on merely vibrational fatigue damage. Besides the truck use case, this study is feasible also for other complicated structures such as airplanes, ships, buses and otomobiles. All algorithms for this study, first, was written in Matlab for Ardunio. Later, this algorithms were converted to Python codes for Raspberry Pi as Python is free software, which means cost reduction. The developed approach was applied to a heavy duty truck chassis. The critical components that were examined in this study were the exhaust muffler bracket and fuel tank bracket. Experimental verifications were made on test roads and vibration test bench excited by shakers. The validation of computational models, mathematical derivations and programs were made by making comparisons with numerical results of Ncode software and experimental tests. It is observed that although the input excitations (namely, forces) were wide band processes, the chassis dynamic behaves like a filter and the PSDs of stresses at selected locations on the chassis were found to be narrow band processes. It was surprising that the fatigue life estimations at critical locations obtained by the Dirlik method which was originally developed for wide band processes yielded larger errors than those of the Narrow Band Approximation which was originally developed for narrow band processes. It is also noteworthy that the computational load of the Narrow Band Approximation is lower than that of the Dirlik method.