| Author: | Zhang, Linhao |
| Title: | Investigation of moving load responses on structures via local spectral algorithm and machine learning-based method |
| Advisors: | Lai, Siu-kai (CEE) Zhu, Songye (CEE) Ni, Yi-qing (CEE) |
| Degree: | Ph.D. |
| Year: | 2020 |
| Subject: | Live loads High speed trains Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Civil and Environmental Engineering |
| Pages: | xxxv, 258 pages : color illustrations |
| Language: | English |
| Abstract: | Moving load problems are very common in railway engineering applications, such as railway viaducts, trains on the track and pantographs. In structural dynamics, it refers to any structures subjected to loads that move in space and excite the structures into vibration. Investigating such dynamical problems can optimize the structural design and ensure the stability and reliability of structures in real-engineering conditions. With the rapid development of worldwide high-speed rail (HSR) networks, the vehicle-rail-bridge dynamic coupling problem is still a major issue that is being actively discussed in the scientific and engineering communities. This thesis is mainly divided into two parts. The first part focuses on the theoretical model analysis under various case scenarios, e.g., moving speeds, boundary conditions, foundation supports and material properties. Generally, the moving load is regarded as a simple massless force or an inertial force. To consider beam-type moving load problems, a local spectral approach, namely the discrete singular convolution (DSC) algorithm, is used to provide accurate results of vibration responses. To authenticate the accuracy and reliability of the DSC method, a work example for the thermal buckling analysis of cracked plates is given. In addition to the structural design, to enhance the operation safety and ride comfort of trains, the implementation of on-board monitoring systems on in-service trains is a smart trend to make leverage technologies over a rail network infrastructure. This allows timely decision-making for appropriate actions and maintenance, thereby reducing its life-cycle cost and increasing running safety. Due to the massive amount of data generated, an application of artificial intelligence that can offer the ability to automatically learn and improve from experience without being explicitly programmed is highly desired. Therefore, in the second part, an ensemble learning approach, called the random forest (RF) method, is adopted to carry out a predictive analysis for the classification of a huge amount of on-board monitoring data. Furthermore, to realize a self-powered on-board monitoring system on trains, a multi-stable piezomagnetoelastic energy harvesting technique that can work well under low-frequency and low-amplitude ambient sources, is designed according to the vibration responses of train bogies. The DSC algorithm, based on the theory of distribution and wavelet analysis, was firstly proposed for studying the Fokker-Planck equation, and later generalized to the dynamic analysis of various structural members. To compare with conventional numerical approaches (e.g., differential quadrature and finite element methods), the DSC algorithm is the only available numerical method for the prediction of thousands of high-frequency vibration modes in various structure members without encountering much dynamical instability. In addition, the method is capable of providing highly accurate solutions at minimal computational effort. Basically, it is not only a flexible local method to handle complex geometries and boundary conditions, and also performs as a global spectral approach with a high level of accuracy. To demonstrate the flexibility and accuracy of the DSC approach, a work example for the thermal buckling and stability analysis of orthotropic rectangular plates with crack defects is investigated. A long-term influence of dynamic responses on structures can induce potential damages that would reduce the structural stiffness and eventually cause failure. In addition, it is commonly known that engineering structures are usually exposed to thermal environments. The dynamic and stability behavior of plate-type structures can be adversely affected by thermal conditions. Based on the classical plate theory and the line-spring model, the thermal buckling and dynamic behaviors of rectangular isotropic/orthotropic plates with surface part-through cracks under a thermal environment are studied. To evaluate the thermal effect on material properties, a special restrained manner of simply supported conditions for in-plane movements under thermal conditions is also investigated. To go beyond the limitation of the DSC method, a coupling of Taylor's series expansion method is applied to the treatment of free-edge conditions. The DSC results agree well with the existing solutions and newly obtained data can serve as benchmark solutions. After verification, two cases of moving load problems are considered by the DSC method. The first case is to study the dynamic responses of functionally graded (FG) beams under moving loads, while the second one is to investigate the dynamic characteristics of functionally graded graphene platelet-reinforced (FG-GPLR) beams subjected to moving loads. Functionally graded materials (FGMs) are inhomogeneous composites that have gained considerable popularities due to their superior mechanical and thermal properties. In contrast to conventional laminated composites, its intrinsic properties lead to the reduction of stress concentration and thermal stress. To investigate the dynamic performance of FG beams under a traveling force, a classical Euler-Bernoulli beam model subjected to the elastically restrained and typical boundary conditions (simply-supported and clamped supports) is analyzed. In general, a point load moving on a beam structure can be represented by the Dirac-delta function. Due to the special characteristics of this time-dependent singular function, it is hard to directly apply a strong-form based method. To circumvent this issue, this work proposes a DSC regularized Dirac-delta approach to couple with the Newmark-β integration scheme for the dynamic analysis of beams carrying a moving load. A parametric analysis for moving load velocity, material behavior and spring stiffness on the dynamic behavior of FG beams is extensively investigated. More importantly, the major finding of this work indicates that this approach is a good candidate for moving load problems since the equally spaced grid system adopted in the DSC scheme can achieve a preferable representation of moving load sources. In addition to FGMs, graphene, the first isolated 2D material with hybridized carbon atoms, possesses exceptionally high elastic modulus, tensile strength and very large specific surface area (numerous times larger than that of carbon nanotubes). The first experimental isolation of graphene was discovered in 2004. This versatile material is more than 100 times stronger than steel but several times lighter. The strength of graphene can be used in composites and coatings for reinforced applications. Rapidly developed manufacturing technologies have made this material commercially available and significantly promoted its engineering applications, e.g., foam cores in beams. To fill the gap of the forced vibration analysis of FG-GPLR beams under a moving load, the present work constructs a shear deformable Timoshenko beam model that is reinforced by graphene nanoplatelets and is rested on an elastic (Winkler-Pasternak) foundation. The elastic modulus of nanocomposites is obtained by the Halpin-Tsai model. A comprehensive parametric study is presented, with a particular focus on the influence of moving loads, foundation supports and material properties (i.e., weight fraction, distribution pattern and geometry size of graphene reinforcement). The aforementioned work aims to characterize the interaction of moving loads on beam structures. For moving load problems in railway engineering, it is still desired to provide an intelligent on-board condition assessment that can help operators to take timely actions before extreme events occur. With the emergence of advanced sensing technologies, the use of data-driven (also called "model-free") methods based on measured ambient and forced responses of structures for condition monitoring paves an efficient way. It is observed that there still exist some deficiencies of the commonly used intelligent fault diagnosis techniques such as massive parameter adjustments and low efficiency in high-dimensional input data. Therefore, the RF method is adopted for the condition assessment of a real case (i.e., on-board monitoring data obtained from an in-service high-speed train). The merits of this method include the capability of handling high-dimensional problems, high level of accuracy, stability in classifications with fewer parameter adjustments, and low-sensitivity to noise without any over-training problems. There are four basic steps in the RF method, namely: (i) data acquisition and pre-possessing; (ii) feature extraction; (iii) feature selection; and (iv) training and forecasting. An optimization dataset is first fed to an RF model for training and the refined model is then employed to classify the dynamic status based on the newly collected data. To evaluate the health conditions of train wheels (i.e., well-behaved situation and out-of-round (OOR) state), the classification results indicate that the RF method can identify the OOR condition with an accuracy of 98.21%, and the average recognition accuracy is up to 98.91%. This study is beneficial for making a rational lathing decision of high-speed trains. Making use of advanced high-precision wireless communications is the best solution for data acquisition, dynamic response monitoring and information transmission. Nevertheless, the power supply of a wireless sensing system installed on the bogie structure of trains to monitor its operational status is a critical but unresolved issue. Conventional wired solutions and batteries for numerous wireless sensors are not practical. To enhance its autonomy, a feasible way is to directly harness vibration energy from trains and convert it to electrical power to support such wireless sensing networks. The vibration features of high-speed trains (bogie frames) are generally broadband (about 5-30 Hz), low-amplitude (about 0.5-2 g, where g is the gravitational acceleration), time-varying and speed-dependent. For this problem, a new multi-stable piezomagnetoelastic energy harvesting technique is proposed. To examine the working efficiency, both theoretical and experimental studies are conducted. The results show that the experimental prototype can scavenge energy from low-amplitude (< 3 m/s2) and low-frequency (< 20 Hz) vibration sources. This is a promising potential application to realize a self-powered wireless sensing system for train vehicles. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/11053