Search

A novel data-driven method of modal analysis for complex flow dynamics, termed as reduced-order variational mode decomposition (RVMD), has been proposed, combining the idea of the separation of variables and a state-of-the-art nonstationary signal-processing technique -- variational mode decomposition. It enables a low-redundant adaptive extraction of coherent structures in statistically nonstationary flows, with its modes computed by solving an elaborate optimization problem using the block coordinate descent algorithm. Discussion on the intrinsic relations between RVMD and some classic modal decomposition methods demonstrates that RVMD can be reduced into proper orthogonal decomposition (POD) or discrete Fourier transform (DFT) at particular parameter settings. The significant advantages of RVMD for performing time-frequency analysis are highlighted by a signal-processing analogous categorization of the widely-used modal decomposition techniques. It is also confirmed that the combination of RVMD and the Hilbert spectral analysis provides a physically intuitive way to explore the space-time-frequency characteristics of transient dynamics. Finally, all the appealing features of RVMD mentioned above are well verified via two canonical flow problems: the transient cylinder wake and the rectangular turbulent supersonic screeching jet.

Rock bolts have been widely used to reinforce the jointed rock mass. Modelling the mechanical shear behaviors of the bolted rock joints are difficult due to the complex interactions between bolts and rock joints. The applied pretension forces combined with the axial loads developed in the bolt act as the normal forces which are applied to the rock joints. However, these combined normal forces are not considered in the existing analytical models. An analytical model is proposed in this study to predict the shear behavior of the bolted rock joints, by taking into account the pretension forces, the axial forces developed in the bolt, the interfacial bond stress between the bolt and grout, and dowel shear loads acting transversely to the bolt axis. The proposed analytical model is able to provide complete curves of the dowel shear loads, axial loads, and the global shear loads as a function of the joint shear displacement. The analytically predicted axial load vs shear displacement curves and the global shear load vs shear displacement curves are verified by available experimental tests. The validation shows that the proposed model has the capacity to predict the global shear load evolution as well as the axial load evolution. The factors such as the pretension forces, the bolt inclination angles, the concrete strength and the rock joint friction are successfully accounted for in the analytical model.

Ground vibrations from blasting are undesirable and they are very short in duration with high principal frequencies and high amplitude. Structural response to that kind of excitations may differ significantly from that generated by lower-frequency earthquake vibrations. In practice, the damage to nearby structures to generated ground vibrations has been controlled by various codes and regulations available. However, the existing vibrations limits are not always applicable, as they depend on the geological conditions of the site and the dynamic characteristics of the structure. This paper investigated the influence of the rock-soil interface on the response of pile foundation to ground-borne vibration from rock blasting, through numerical simulations using finite element software LS DYNA. It is found that the slope of the rock-soil interface greatly influences the location of damage and the extent of damage on the pile. The bottom of the pile is subjected to a larger blast pressure and stresses are highly concentrated at the pile bottom. Thus, the pile bottom can fail in shear before it reaches its plastic moment capacity. Accepted version

Groundwater seepage related problems are often the No.1 geological hazard in terms of construction expense and safety for subsea rock caverns. In order to excavate caverns safely and effectively, water control should be successfully addressed. Successful water control depends on a set of effective measures, from establishing hydro-geological model to grouting strategy and grout property improvement. In this paper, the authors try to propose a systematic water control methodology during excavation of subsea rock caverns. A reliable hydro-geological model building is the critical step to assess the seepage in the field. Traditional hydro-geological models are often not reliable due to its simplistic assumptions and lack of effective use of measurement data, and the calculated results do not provide useful prediction to guide the cavern design and excavation. To overcome these drawbacks, the information from the geological investigation data, aided by the artificial intelligence method, is used to establish the preliminary hydro-geological model. Based on site monitoring data and geological conditions, an innovation modeling method, back-analysis of the coefficient of hydraulic conductivity of the rock mass, can be used to calibrate the preliminary model. On the basis of established hydro-geological model, groundwater control during excavation mainly relies on accurate water bearing zones detecting ahead of the face, followed by pre-grouting. Two methods are commonly used to detect the water bearing zones: probehole method and geo-physical method. Both methods have their advantages and constraints. An advanced water bearing zone detecting method by integrating probehole method and geo-physical method is proposed. Once water bearing zones are detected, groundwater control is mainly achieved by pre-grouting. Sometimes, pre-grouting quality is not satisfactory due to the unreasonable grouting strategy and unsatisfied property of the grout. According to statistical analysis and comparison of successful grouting cases with similar conditions and relevant literatures, optimized pre-grouting strategy will be proposed, including grouting method selection and workflow determination. For better grouting quality, improvement of property of grout is necessary. To take influences of saline groundwater into account, improvement of grout (especially the rheological behavior) will be mainly achieved by laboratory tests and in-site trials. Based on systematic water management optimization, water control would be more cost-effective.

An extensive examination of the discontinuous deformation analysis (DDA) in block dynamic sliding modeling is carried out in this paper. Theoretical solutions for a single block sliding on an arbitrarily inclined plane by applying the horizontal/vertical seismic loadings to the sliding block as acceleration time histories or to the base as constraint displacement time histories are derived. As compared with the theoretical solutions, for a single block sliding, the DDA predicts the sliding displacements and block interaction forces accurately under various base incline angles and friction angles under both the harmonic loadings and a real seismic loading. The vertical seismic component may influence the block sliding displacements to different extent, and the DDA can capture these phenomena successfully and give accurate results. For the calculation of the single block relative sliding, both the theoretical and the DDA solutions indicate that applying the seismic accelerations as constraint displacement time histories (derived by integrating the seismic accelerations twice) to the base is equivalent to applying the seismic accelerations as volume forces to the sliding block in the opposite directions. The DDA modeling also demonstrates that this conclusion still stands for the case of multi-block sliding.

In this thesis, the flat-top partition of unity (PU) based high-order Discontinuous Deformation Analysis (DDA) is developed and applied in simulating different problems. In the proposed method, the continuous blocks in the problem domain are discretized by the flat-top PU mesh, and the interactions between blocks is simulated with the improved contact algorithm. For each flat-top PU unit in the mesh, high-order polynomial is assigned as local approximation, where the coefficients of monomials are denoted as unknowns. The global approximation can be obtained by multiplying the local approximation with the corresponding shape functions. With the principle of virtual work and simplex integration, the global equilibrium equations are established and solved. Thus, the displacement field and stress/strain field can be obtained. In flat-top PU based high-order DDA method, since the flat-top PU mesh and high-order approximation are utilized instead of discretizing the problem domain into numbers of small blocks, only the main discontinuities need to be considered. Therefore, the time-consuming iteration of contact detection, adding and removing stiff springs can be reduced. The construction of regularly patterned flat-top PU mesh is introduced based on the traditional PU mesh, which is easy to understand and straightforward to implement. In order to simulate the stress concentration/singularity and crack propagation problems with high cost-efficiency, the local refinement of flat-top PU mesh is introduced and its property of linear independence is proven from element level based on the equivalent p-version element and rank deficiency counting approach. During the generation of numerical model for discontinuous problem domain, in order to avoid ill-conditioned stiffness matrix in global equilibrium equations, modification for the numerical model is implemented based on the concept of percentage of effective area. During the numerical simulation, the contact algorithm in original DDA method is improved by determining the exact contact position. Therefore, the energy consumption is more accurate, especially for the collision problem. The improved formula for non-reflective boundary is coupled with flat-top PU based high-order DDA method to simulate stress wave propagation within infinite problem domain, where the numerical accuracy can be improved by deriving the analytical velocity of dashpots, instead of calculating the velocity with finite difference method.Based on the crack propagation modes, the tension failure criterion and the Mohr-Coulomb failure criterion are adopted to determine the crack initiation in flat-top PU based high-order DDA method. Within each time step, the criteria are applied at every node in the flat-top PU mesh. In order to avoid generating new unknowns and determining their initial values artificially, it is assumed that the crack propagates along the element sides. Correspondingly, the treatment of PU elements after crack propagation is introduced.Numerical examples are presented and analyzed regarding to each development of the flat-top PU based high-order DDA method. Doctor of Philosophy (CEE)

Discontinuous Deformation Analysis (DDA) and Finite Element Method (FEM) belong to two different numerical approaches, the discontinuum-based method and the continuous-based method. The Nodal-based Discontinuous Deformation Analysis (NDDA) introduced in this thesis is a coupled method of DDA and FEM. By incorporating the finite element mesh into the discrete block, the unique block kinematics of the DDA is inherited and the stress field within the block is refined. To make the NDDA method a more powerful tool in the rock engineering analysis, a “crack propagation procedure” is implemented into the NDDA program to describe the failure process in a fully automatic way. Also, the rockbolt element is implemented in the NDDA framework to numerically simulated the rock/bolt interaction when the rock mass is reinforced by rockbolts. An alternative contact mechanism, the Augmented Lagrangian Method (ALM) is introduced into the standard DDA method, and the efficiency and the accuracy of the method are discussed. Finally, the potential use of the new method is demonstrated through the slope sliding and topping analysis. Doctor of Philosophy (CEE)

Discontinuous Deformation Analysis (DDA) is selected as an advanced numerical method to study the response of rock mass under dynamic loads. The basic concept and theory of the DDA are studied and the unique features of the DDA method to solve the rock mass problems are discussed. Some new subroutines and modifications have been added to the original DDA code to extend its applications. A stress recovery procedure is proposed to improve the stress accuracy over the block-interface. Simulations of the rock mass with underground blast loading and rock mass in scaled blast load test are carried out respectively. Some preliminary understanding and interpretation to the simulation results are given. MASTER OF ENGINEERING (CEE)

This thesis presents an enhanced nodal-based discontinuous deformation analysis (NDDA) based on the coupling of the discontinuous deformation analysis (DDA) and the finite element method (FEM), for modeling blocky systems, especially for simulating crack propagation in rock mass. The NDDA can provide a more accurate stress and strain distribution in each block and has a higher computational efficiency than the standard DDA in dealing with continuum materials. Furthermore, the enhanced NDDA allows for tensile and shear fracturing happening inside an intact block, which provides a way for material transforming from continuum into discontinuum. A computer program named 2D-NDDA was developed to handle the combination of continuous and discontinuous in large displacement problems, as well as large deformation and failure analysis, under external loads and boundary conditions. Doctor of Philosophy (CEE)

Rock masses are discontinuous, due to the existence rock fractures, like faults, cracks, joints, and other similar geological features, which could not only affect the structural strength, but also provide interconnected channels for water inflow. The discontinuities in rock mass pose a great threat to rock excavation, and could lead to economic loss or even casualties, if they are not properly handled. One way to improve the rock mass quality is grouting, which helps to seal those discontinuities, and strengthen the rock mass simultaneously. However, the grouting cost accounts for a considerable percentage of the total cost for construction of underground facilities, especially for rock masses with complex geological conditions. One reason for the high cost is that the stop criteria cannot be properly determined, which is normally based on either empirical criteria or experience; moreover, the accuracy of existing methods for grouting simulation is poor, due to the lack of effective theoretical models and insufficient consideration of geological conditions. A lot of practical factors should be considered to better model the grouting process, including the influence of fracture deformation (referring to normal deformation in this thesis) and flowing water, the existence of interconnected fracture network, and fracture roughness and tortuosity. Previous work on grouting investigation is summarized in the thesis, introducing fundamental information on grouting technology, including characteristics of rock masses and cement grout; existing theoretical and empirical models used for grouting prediction; and some basic concepts related to grouting practice. For grout flow in rock fractures not directly intersected by the injection hole, channel flow can represent the grout flow, a one-dimensional (1D) model with grout flow within two parallel plates. This model is very useful and popular, due to its simplicity and its application in evaluating laboratory testing data. The governing equation for the 1D grouting model is derived analytically, and numerical algorithm is developed for solving the corresponding equations. Two different methods are defined under two boundary conditions: constant injection pressure or constant injection rate, as the corresponding algorithms for equation solving are significantly different. Thereafter, the characteristics of grout penetration using the two methods are discussed. For rock fractures directly intersected by the injection hole, the distribution of cement grout around the injection hole is in radial direction, which is two-dimensional (2D). Therefore, a governing equation for radial grout penetration initiating from the injection hole is derived. Real time grout penetration within two neighboring fractures directly intersected by the injection hole is calculated based on numerical algorithm, and the influence of fracture deformation (normal deformation) induced by grout pressure on the penetrability of rock fracture with smaller aperture is analyzed. Then, an analytical model is proposed to predict grouting of single rock fracture with flowing water based on the theory of potential flow, which is used for explaining the data obtained from grouting test in the laboratory considering flowing water. A module for predicting grout penetration in a 2D fracture system is developed based on the Discontinuous Deformation Analysis (DDA), which can present good visualization of real-time grout penetration in intersected rock fractures and real-time pressure variation at various intersection nodes. Finally, the influence of fracture roughness on grout penetration and fluid transportation is investigated based on laboratory test and numerical simulation. Most existing models take rock fractures as two smooth parallel plates, neglecting the characteristics of fracture surfaces, like fracture roughness, which, however, do exist and can have substantial impact on fluid flow within real rock fractures. Therefore, with assistance of 3D printing, large scale rock fracture with rough surfaces are generated and used for grouting test in laboratory, which are compared with grouting test using traditional fracture channel with smooth parallel surfaces. It is found from the laboratory test results that the impact from fracture roughness is considerable, so the influence is further investigated with numerical simulations. Doctor of Philosophy

A rockbolt is a reinforcing element installed untensioned or tensioned into the ground to restrict the deformation of the surrounding rock mass in rock engineering. The rockbolt is usually a solid bar or tube element which is installed within the boreholes drilled into the rock. To improve the rockbolt design in complex rock conditions, it is necessary to have a good understanding of the reinforcement mechanism of the rockbolts. In this thesis, the reinforcement mechanism of a rockbolt system under excavation conditions is studied using the two-dimensional discontinuous deformation analysis (2D-DDA) method and other numerical tools. The load transfer capacity of a rockbolt element is controlled by the components of bonds at the rock and rockbolt, such as the rock mass, the mortar and the profile configuration of the rockbolt. 2D-DDA method is used to investigate the pullout performance of a continuously mechanically coupled (CMC) rockbolt element. The flat joint contact model is introduced into the 2D-DDA code to simulate the force versus displacement behaviour of an artificial joint. Parametric studies show that the confining pressure plays an important role in the bond-slip modeling. The obtained key parameters of the trilinear bond-slip model can be used to simulate the bond behaviour between the rockbolt and rock under similar conditions. A rockbolt model is developed and integrated into the 2D-DDA code based on the load transfer theory. Three mechanical behaviours are considered in the model, i.e., the bond, the axial tension and the shear at joint. The proposed rockbolt model could be used to simulate four major failure modes of rockbolt, such as the de-bonding along the interface, tensile failure, faceplate failure and shear failure. The numerical results show that the numerical models have reasonable agreements with the experiments results. The parameter studies show that the load transfer and load capacity should be reconsidered when the ground conditions are changed. The rock/rockbolt interactions are examined by simulating the pullout test in rock mass with systematic joint sets. The effects of the rock properties, joint orientations, rockbolt lengths and in-plane spacing are studied. The reinforcement rock unit (RRU) induced by single rockbolt is normally in a cone shape and restrained by the orientation of the discontinuities. The RRU induced by grouped rockbolts is heavily influenced by the joint orientations and the rockbolt spacing. Further applications of the RRUs on the estimation of the artificial pressure arch in the roof are illustrated. The rockbolt to stabilize roof wedges in underground opening is studied using the joint relaxation method to consider the rock wedge in deformable rock conditions. Parametric studies are carried out to investigate the key parameters that influence the effects of rockbolts when they are used to stabilize the rock wedge. The results show that 2D-DDA could be used to find the critical horizontal pressure to sustain the rock wedges. A cross section in storage gallery in Jurong Rock Caverns (JRC), Singapore, is conducted to show the application of the 2D-DDA models for rockbolt design to support roof wedges. The stress versus vertical displacement curves showed clearly the effect of rockbolt on sustaining the movement of rock wedge. The way to create characteristic diagram could be used by the site engineers to evaluate the rockbolting design to sustain the possible roof wedges. A support design method for horseshoe-shaped rock caverns is proposed with considerations of the progressive damage of the rock mass using the 2D finite element method (FEM) and the artificial neural network (ANN). The performances of the rock cavern during excavation are investigated based on the convergence-confinement method (CCM). The ANN models are built using the numerical results to find the complex relationships among the rock mass condition, the sequential excavation parameters and the cavern performances. An evaluation chart is proposed by integrating the ANN models into the EXCEL software. The proposed evaluation chart provides an effective method to evaluate the support safety, the functions of the patterned rockbolt and the optimization of subdivisions of the excavation cross section. More advanced function is still required to present the relationship between the parameters of sequential excavation method (SEM) and the cavern performance. Doctor of Philosophy

Recently, conventional tunnelling methods (or the drill & blasting tunneling methods, the new Austrian tunneling methods) have been frequently applied to tunnels/caverns under severe geological conditions. Under such adverse conditions, it is a concern that tunnel support materials might fail because of the huge ground pressure and large tunnel deformation. As a result, the tunnel may lose its stability due to decreased effectiveness of tunnel supports. Therefore, essential components of tunnel supports (e.g., rock bolts) are becoming more important. According to the previous studies, the interface behaviour between the rock bolt and the bond material (i.e., the debonding process along the interface and crack propagation in the bond material) is recognised as one of the most important factors significantly affecting the reinforcement effects of rock bolts. However, there is little information available on the detailed interface behaviour. This research aims to investigate the interface behaviour via both laboratory tests (shear test) and numerical simulations using the discontinuous deformation analysis (DDA). As a results, it can be found that the DDA is an effective tool reproducing the accurate interface behaviour, and therefore the simulation results enable us to better understand the bolt-grout interface behaviour in detail. Based on the accurate DDA rock bolt model, a new energy-absorbing rock bolt was designed. This rock bolt has high loading capacity and large deformability. Moreover, it can control the ultimate tunnel deformation utilizing the specially designed anchor. By carrying out the DDA-based simulations of pull-out tests and tunnel excavation, the reinforcement effect of the newly-developed rock bolt was investigated. As a result, it can be concluded that the new rock bolt has a significant advantage over the widely-used fully grouted rock bolt when the large tunnel displacement occurs. Doctor of Philosophy

It is well-established that pore pressure built up in discontinuities has a profound effect on the mechanical behaviors of the jointed rock masses. The discontinuous deformation analysis (DDA) is a discontinuum theory which can account for the interactions between pore pressure and rock mechanics. It is also a numerical method or computer program commonly applied to hydro-geomechanical modeling. Research and development of DDA hydro-mechanical model has been thriving over the last two decades. The main objective of this research study is to further contribute to those efforts. Specifically, the following works have been conducted to achieve this objective: (i) a seepage analysis extension is proposed to the DDA method for the modeling of seepage flow within jointed rock masses; (ii) to improve the computational efficiency of the extension in simulating quasi-static hydro-mechanical problems, a new solution scheme featuring a constant hydraulic aperture is introduced; (iii) a stochastic flow modeling extension is proposed to the DDA method for the simulation of heterogeneous hydraulic conductivity and aperture fields; (iv) a hydraulic crack initiation-propagation extension complete with a coupled hydro-mechanical analysis algorithm is introduced to the hybrid DDA-FEM model for the simulation of hydraulic fracturing problems. To investigate the reliability, efficiency and limitation of the proposed DDA extensions, a series of numerical examples are presented and discussed. The capability of the new DDA extensions in modeling realistic engineering problems are also verified through meaningful case studies. Doctor of Philosophy (CEE)

As a special set of fractures of regular orientation and spacing, en-echelon fractures widely exist in nature. It is worthwhile to understand the interaction among them, and how this interaction finally contributes to a fault. Although relevant studies have acquired instructive results for the interaction of en-echelon flaws, further work is necessary to enhance the understanding about the evolution of en-echelon fractures into a fault in laboratory, especially for brittle natural rocks. In this study, uniaxial compression test is conducted on marble specimens containing en-echelon flaws. The fracturing processes of these specimens leading to the formation of a fault have been generalized. Linking cracks of tensile and shear properties show different development styles at the macro-scale. The flaw-array angle dominates the linking patterns between flaws, geometry of tensile wing cracks (TWCs) developed near the outer flaw tips and the specimen peak stress. Secondly, two series of specimens containing en-echelon flaws are subjected to different loading levels to produce progressive damage zones (linking cracks). After that, microscopic observations have been conducted on these damage zones (linking cracks) to investigate the microscopic characteristics of tensile and shear cracks. Analysis on microcrack density reveals the distinct development of grain boundary cracks and transgranular cracks in tensile and shear cracks. Additionally, abundant cases about the nucleation of transgranular cracks have been examined carefully. Four nucleation mechanisms of transgranular cracks in Carrara marble are thus identified. Finally, a recently reported contact model—flat-joint contact model is applied to the particle flow code (PFC2D) simulation to model the present laboratory experiments. The numerical results provide a quantitative assessment on the interaction of en-echelon flaws. Furthermore, cracking behavior in these synthetic specimens are found to agree well with experimental results obtained in both microscopic and macroscopic observations. This agreement suggests not only the reliability of experimental results but also the capability of the flat-joint contact model of PFC2D in modeling rock fracturing problems. Doctor of Philosophy

Discontinuous Deformation Analysis (DDA) method can be applied to study the static and dynamic behavior of fractured rock masses, where the fractures are represented explicitly and assigned with independent physical parameters. In this thesis, the fracture flow approach and the dual permeability approach are introduced to model the fluid flow behaviors through the fractured rock mass based on DDA method. The fracture networks connected by major fractures are often considered as the only path for the fluid flow through the fractured rock mass. However, there are still lots of minor fractures and also pores inside the rock matrices that can conduct the fluid flow at the same time. Thus, a dual permeable media model is needed for the description of the flow behavior when the porous media exist in the rock mass. The main development work of this thesis focuses on the seepage inside the fractured rock mass with the considerations of fracture flow and the dual permeable media flow under the DDA framework, including: 1. A preliminary coupled DDA hydro-mechanical model is developed for steady flow in two dimensional fracture networks. The case studies are performed to verify the correctness and potential applications of the coupled DDA model. The numerical results show that both the flow rate and the mechanical responses for the rock blocks under the hydraulic pressure can be predicted reliably by this coupled method. 2. Using the coupled DDA hydro-mechanical model, the stability analysis for the underground rock caverns under variable geological conditions is carried out. It shows that the stability for the rock caverns also has a relationship with the block size, block shape and joint dip angles when the ground water exists. The study concludes that the underground water cannot be neglected for the stability analysis. 3. The transient fluid flow is developed and coupled into DDA. Grouting zone is taken into account for the case studies, and the hydraulic head distribution under different grouting zone sizes is studied. 4. Based on the coupled DDA hydro-mechanical model, the dual permeable media approach is developed to take the porous rock flow into consideration. The global solution method is employed to solve the seepage field when both the fracture flow and porous medium flow are considered in the domain. Doctor of Philosophy (CEE)

Rock mass is distinguished from other engineering materials by the existence of discontinuities. The discontinuities ranging in size from micro-defects to macrojoints greatly influence the dynamic behavior of rock mass. Efforts have been made to study the dynamic behavior of rock mass and the effects of discontinuities on wave propagation. Doctor of Philosophy (CEE)

This thesis deals with soil-structure interaction effect and structural response analysis, in-structure shock assessment, and shock wave mitigation when a structure is subjected to a blast load. Dynamic media-structure interaction is analytically discussed in the background of an underground structure subjected to a soil-transmitted dynamic load, in which an interfacial damping is incorporated to represent the dynamic soil-structure interaction. The effects of the interaction are analyzed and the aspects affecting the interaction are discussed. With this soil-structure interaction, in-structure shock of a typical underground structure subjected to a soil-transmitted blast load induced by a subsurface detonation is analyzed with a simplified beam model and a rigid-body-motion included plate model, respectively. With acceleration time history of the derived structural member as excitation, shock response spectra are established to assess the in-structure shock level of the equipment attached to the buried structure. To mitigate the in-structure shock, a new design of underground structures is proposed by adding an isolation slab inside the structure. The excitation mechanism for the equipment within the structure is altered and the vertical shock level is effectively reduced. In addition, in-structure shock induced by a soil blast with non-zero rise time is also analyzed. A small scale test is designed and conducted to validate the prediction. Tests results indicate that the predictions are favorably comparable with the experiment. When subjected to a close range spherical airburst, the response of a blast mitigation cladding with metal foam core is determined by energy method. Shock theory and rigid-perfectly-plastic-locking model is adopted to obtain the response of metal foam under high velocity crushing. This prediction has practical significance since it delineates the situation of a cladding subjected to a close range detonation event with realistic boundary. Further, density gradient metal foam as core of a blast mitigation cladding is theoretically investigated with shock theory and rigid-perfectly-plastic-locking model. Doctor of Philosophy (CEE)

The architecture of neural networks (NNs) has a significant impact on a network’s generalization ability. Ensemble neural networks (ENNs) are commonly used networks in many engineering applications due to their better generalization properties. An ENN usually includes several back-propagation networks in its structure, where the back-propagation network is a single feed-forward network trained with the back-propagation learning rule. In this thesis, the Akaike information criterion (AIC) and the entropy are used as the automating design tools for balancing the generalization against the parameters and finding the best combining weights of the ENNs. Two ENNs, namely, the AIC based ENN and the entropy based ENN are developed first. Since the AIC and entropy have their own merits for solving different problems, a new AIC-entropy based ENN is developed. Two analytical functions – the peak function and Friedman function are used first to assess the accuracy of the proposed ensemble approaches. The verified approaches are then applied to the civil engineering applications. Doctor of Philosophy (CEE)

The present research aims to implement the Symmetric Galerkin Boundary Element Method (SGBEM) for 2-dimensional elastodynamic problems. The SGBEM employs both the displacement integral equation and the traction integral equation which lead to a matrix equation with completely symmetric coefficients. This research focuses the effective techniques in dealing with the strong and hypersingular integrals embedded in the SGBEM formulations. To achieve this goal, two versions of the SGBEM implementation are developed where different formulations are derived and various new schemes are proposed. Doctor of Philosophy (CEE)

The aim of this study was to build a fuzzy rule-based inference system to satisfy the above mentioned demand, inaddition, to provide bridge designers with valuable information about the impacts of design factors on bridge deterioration. Master of Engineering (CSE)

The box girder is the most widely used superstructure in concrete bridge construction. Because of its desirable characteristics, concrete box girder bridges have been widely used as an economic and aesthetic solution for overcrossings, separation structures and viaducts found in today's modern highway systems. Master of Engineering (CSE)

The adaptive boundary element method is a process for which computer can be used to justify and improve the accuracy of boundary element solutions automatically. Although adaptive processes are very important to ensure the effectiveness and correctness of a numerical analysis, there are many arguments with the current schemes used in adaptive boundary element analysis. This thesis aimed to improve those schemes and develop a more effective adaptive computer program. Master of Engineering (CSE)

With the wide use of the external wall tiling as the finishing system of buildings, system failures such as tile falling, buckling, cracking and adhesive delamination have been reported in many countries, including Singapore. These problems lead to the controversy as to whether it is necessary to install tiles for external wall surfaces. Among the important influence factors to the system performance, this thesis concentrates on the tiling workmanship to study its reduction effect to tile bonding strength by the pull-off test. Master of Engineering (CSE)