Refine
Has Fulltext
- yes (164) (remove)
Document Type
- Article (72)
- Doctoral Thesis (56)
- Conference Proceeding (23)
- Preprint (6)
- Master's Thesis (5)
- Diploma Thesis (1)
- Habilitation (1)
Institute
- Institut für Strukturmechanik (ISM) (164) (remove)
Keywords
- Maschinelles Lernen (27)
- Computerunterstütztes Verfahren (22)
- OA-Publikationsfonds2020 (19)
- Architektur <Informatik> (17)
- Finite-Elemente-Methode (17)
- Machine learning (15)
- Angewandte Mathematik (13)
- Angewandte Informatik (12)
- CAD (10)
- machine learning (10)
A Machine Learning Framework for Assessing Seismic Hazard Safety of Reinforced Concrete Buildings
(2020)
Although averting a seismic disturbance and its physical, social, and economic disruption is practically impossible, using the advancements in computational science and numerical modeling shall equip humanity to predict its severity, understand the outcomes, and equip for post-disaster management. Many buildings exist amidst the developed metropolitan areas, which are senile and still in service. These buildings were also designed before establishing national seismic codes or without the introduction of construction regulations. In that case, risk reduction is significant for developing alternatives and designing suitable models to enhance the existing structure’s performance. Such models will be able to classify risks and casualties related to possible earthquakes through emergency preparation. Thus, it is crucial to recognize structures that are susceptible to earthquake vibrations and need to be prioritized for retrofitting. However, each building’s behavior under seismic actions cannot be studied through performing structural analysis, as it might be unrealistic because of the rigorous computations, long period, and substantial expenditure. Therefore, it calls for a simple, reliable, and accurate process known as Rapid Visual Screening (RVS), which serves as a primary screening platform, including an optimum number of seismic parameters and predetermined performance damage conditions for structures. In this study, the damage classification technique was studied, and the efficacy of the Machine Learning (ML) method in damage prediction via a Support Vector Machine (SVM) model was explored. The ML model is trained and tested separately on damage data from four different earthquakes, namely Ecuador, Haiti, Nepal, and South Korea. Each dataset consists of varying numbers of input data and eight performance modifiers. Based on the study and the results, the ML model using SVM classifies the given input data into the belonging classes and accomplishes the performance on hazard safety evaluation of buildings.
A vast number of existing buildings were constructed before the development and enforcement of seismic design codes, which run into the risk of being severely damaged under the action of seismic excitations. This poses not only a threat to the life of people but also affects the socio-economic stability in the affected area. Therefore, it is necessary to assess such buildings’ present vulnerability to make an educated decision regarding risk mitigation by seismic strengthening techniques such as retrofitting. However, it is economically and timely manner not feasible to inspect, repair, and augment every old building on an urban scale. As a result, a reliable rapid screening methods, namely Rapid Visual Screening (RVS), have garnered increasing interest among researchers and decision-makers alike. In this study, the effectiveness of five different Machine Learning (ML) techniques in vulnerability prediction applications have been investigated. The damage data of four different earthquakes from Ecuador, Haiti, Nepal, and South Korea, have been utilized to train and test the developed models. Eight performance modifiers have been implemented as variables with a supervised ML. The investigations on this paper illustrate that the assessed vulnerability classes by ML techniques were very close to the actual damage levels observed in the buildings.
Rapid Visual Screening (RVS) is a procedure that estimates structural scores for buildings and prioritizes their retrofit and upgrade requirements. Despite the speed and simplicity of RVS, many of the collected parameters are non-commensurable and include subjectivity due to visual observations. This might cause uncertainties in the evaluation, which emphasizes the use of a fuzzy-based method. This study aims to propose a novel RVS methodology based on the interval type-2 fuzzy logic system (IT2FLS) to set the priority of vulnerable building to undergo detailed assessment while covering uncertainties and minimizing their effects during evaluation. The proposed method estimates the vulnerability of a building, in terms of Damage Index, considering the number of stories, age of building, plan irregularity, vertical irregularity, building quality, and peak ground velocity, as inputs with a single output variable. Applicability of the proposed method has been investigated using a post-earthquake damage database of reinforced concrete buildings from the Bingöl and Düzce earthquakes in Turkey.
Earthquake is among the most devastating natural disasters causing severe economical, environmental, and social destruction. Earthquake safety assessment and building hazard monitoring can highly contribute to urban sustainability through identification and insight into optimum materials and structures. While the vulnerability of structures mainly depends on the structural resistance, the safety assessment of buildings can be highly challenging. In this paper, we consider the Rapid Visual Screening (RVS) method, which is a qualitative procedure for estimating structural scores for buildings suitable for medium- to high-seismic cases. This paper presents an overview of the common RVS methods, i.e., FEMA P-154, IITK-GGSDMA, and EMPI. To examine the accuracy and validation, a practical comparison is performed between their assessment and observed damage of reinforced concrete buildings from a street survey in the Bingöl region, Turkey, after the 1 May 2003 earthquake. The results demonstrate that the application of RVS methods for preliminary damage estimation is a vital tool. Furthermore, the comparative analysis showed that FEMA P-154 creates an assessment that overestimates damage states and is not economically viable, while EMPI and IITK-GGSDMA provide more accurate and practical estimation, respectively.
The economic losses from earthquakes tend to hit the national economy considerably; therefore, models that are capable of estimating the vulnerability and losses of future earthquakes are highly consequential for emergency planners with the purpose of risk mitigation. This demands a mass prioritization filtering of structures to identify vulnerable buildings for retrofitting purposes. The application of advanced structural analysis on each building to study the earthquake response is impractical due to complex calculations, long computational time, and exorbitant cost. This exhibits the need for a fast, reliable, and rapid method, commonly known as Rapid Visual Screening (RVS). The method serves as a preliminary screening platform, using an optimum number of seismic parameters of the structure and predefined output damage states. In this study, the efficacy of the Machine Learning (ML) application in damage prediction through a Support Vector Machine (SVM) model as the damage classification technique has been investigated. The developed model was trained and examined based on damage data from the 1999 Düzce Earthquake in Turkey, where the building’s data consists of 22 performance modifiers that have been implemented with supervised machine learning.
The latest earthquakes have proven that several existing buildings, particularly in developing countries, are not secured from damages of earthquake. A variety of statistical and machine-learning approaches have been proposed to identify vulnerable buildings for the prioritization of retrofitting. The present work aims to investigate earthquake susceptibility through the combination of six building performance variables that can be used to obtain an optimal prediction of the damage state of reinforced concrete buildings using artificial neural network (ANN). In this regard, a multi-layer perceptron network is trained and optimized using a database of 484 damaged buildings from the Düzce earthquake in Turkey. The results demonstrate the feasibility and effectiveness of the selected ANN approach to classify concrete structural damage that can be used as a preliminary assessment technique to identify vulnerable buildings in disaster risk-management programs.
Coronary Artery Disease Diagnosis: Ranking the Significant Features Using a Random Trees Model
(2020)
Heart disease is one of the most common diseases in middle-aged citizens. Among the vast number of heart diseases, coronary artery disease (CAD) is considered as a common cardiovascular disease with a high death rate. The most popular tool for diagnosing CAD is the use of medical imaging, e.g., angiography. However, angiography is known for being costly and also associated with a number of side effects. Hence, the purpose of this study is to increase the accuracy of coronary heart disease diagnosis through selecting significant predictive features in order of their ranking. In this study, we propose an integrated method using machine learning. The machine learning methods of random trees (RTs), decision tree of C5.0, support vector machine (SVM), and decision tree of Chi-squared automatic interaction detection (CHAID) are used in this study. The proposed method shows promising results and the study confirms that the RTs model outperforms other models.
Thin-walled cylindrical composite shell structures are often applied in aerospace for lighter and cheaper launcher transport system. These structures exhibit sensitivity to geometrical imperfection and are prone to buckling under axial compression. Today the design is based on NASA guidelines from the 1960’s [1] using a conservative lower bound curve embodying many experimental results of that time. It is well known that the advantages and different characteristics of composites as well as the evolution of manufacturing standards are not considered apporopriately in this outdated approach. The DESICOS project was initiated to provide new design guidelines regarding all the advantages of composites and allow further weight reduction of space structures by guaranteeing a more precise and robust design.
Therefore it is necessary among other things to understand how a cutout with different dimensions affects the buckling load of a thin-walled cylindrical shell structure in combination with initial geometric imperfections. This work is intended to identify a ratio between the cutout characteristic dimension (in this case the cutout diameter) and the structure characteristic dimension (in this case the cylinder radius) that can be used to tell if the buckling structure is dominated by initial imperfections or is dominated by the cutout.
In recent years increasingly consideration has been given to the lifetime extension of existing structures. This is based on the fact that a growing percentage of civil infrastructure as well as buildings is threatened by obsolescence and that due to simple monetary reasons this can no longer be countered by simply re-building everything anew. Hence maintenance interventions are required which allow partial or complete structural rehabilitation. However, maintenance interventions have to be economically reasonable, that is, maintenance expenditures have to be outweighed by expected future benefits. Is this not the case, then indeed the structure is obsolete - at least in its current functional, economic, technical, or social configuration - and innovative alternatives have to be evaluated. An optimization formulation for planning maintenance interventions based on cost-benefit criteria is proposed herein. The underlying formulation is as follows: (a) between maintenance interventions structural deterioration is described as a random process; (b) maintenance interventions can take place anytime throughout lifetime and comprise the rehabilitation of all deterioration states above a certain minimum level; and (c) maintenance interventions are optimized by taking into account all expected life-cycle costs (construction, failure, inspection and state-dependent repair costs) as well as state- or time-dependent benefit rates. The optimization is performed by an evolutionary algorithm. The proposed approach also allows to determine optimal lifetimes and acceptable failure rates. Numerical examples demonstrate the importance of defining benefit rates explicitly. It is shown, that the optimal solution to maintenance interventions requires to take action before reaching the acceptable failure rate or the zero expected net benefit rate level. Deferring decisions with respect to maintenance not only results, in general, in higher losses, but also results in overly hazardous structures.
The classical Internet of things routing and wireless sensor networks can provide more precise monitoring of the covered area due to the higher number of utilized nodes. Because of the limitations in shared transfer media, many nodes in the network are prone to the collision in simultaneous transmissions. Medium access control protocols are usually more practical in networks with low traffic, which are not subjected to external noise from adjacent frequencies. There are preventive, detection and control solutions to congestion management in the network which are all the focus of this study. In the congestion prevention phase, the proposed method chooses the next step of the path using the Fuzzy decision-making system to distribute network traffic via optimal paths. In the congestion detection phase, a dynamic approach to queue management was designed to detect congestion in the least amount of time and prevent the collision. In the congestion control phase, the back-pressure method was used based on the quality of the queue to decrease the probability of linking in the pathway from the pre-congested node. The main goals of this study are to balance energy consumption in network nodes, reducing the rate of lost packets and increasing quality of service in routing. Simulation results proved the proposed Congestion Control Fuzzy Decision Making (CCFDM) method was more capable in improving routing parameters as compared to recent algorithms.
This thesis addresses an adaptive higher-order method based on a Geometry Independent Field approximatTion(GIFT) of polynomial/rationals plines over hierarchical T-meshes(PHT/RHT-splines).
In isogeometric analysis, basis functions used for constructing geometric models in computer-aided design(CAD) are also employed to discretize the partial differential equations(PDEs) for numerical analysis. Non-uniform rational B-Splines(NURBS) are the most commonly used basis functions in CAD. However, they may not be ideal for numerical analysis where local refinement is required.
The alternative method GIFT deploys different splines for geometry and numerical analysis. NURBS are utilized for the geometry representation, while for the field solution, PHT/RHT-splines are used. PHT-splines not only inherit the useful properties of B-splines and NURBS, but also possess the capabilities of local refinement and hierarchical structure. The smooth basis function properties of PHT-splines make them suitable for analysis purposes. While most problems considered in isogeometric analysis can be solved efficiently when the solution is smooth, many non-trivial problems have rough solutions. For example, this can be caused by the presence of re-entrant corners in the domain. For such problems, a tensor-product basis (as in the case of NURBS) is less suitable for resolving the singularities that appear since refinement propagates throughout the computational domain. Hierarchical bases and local refinement (as in the case of PHT-splines) allow for a more efficient way to resolve these singularities by adding more degrees of freedom where they are necessary. In order to drive the adaptive refinement, an efficient recovery-based error estimator is proposed in this thesis. The estimator produces a recovery solution which is a more accurate approximation than the computed numerical solution. Several two- and three-dimensional numerical investigations with PHT-splines of higher order and continuity prove that the proposed method is capable of obtaining results with higher accuracy, better convergence, fewer degrees of freedom and less computational cost than NURBS for smooth solution problems. The adaptive GIFT method utilizing PHT-splines with the recovery-based error estimator is used for solutions with discontinuities or singularities where adaptive local refinement in particular domains of interest achieves higher accuracy with fewer degrees of freedom. This method also proves that it can handle complicated multi-patch domains for two- and three-dimensional problems outperforming uniform refinement in terms of degrees of freedom and computational cost.
Rechargeable lithium ion batteries (LIBs) play a very significant role in power supply and storage. In recent decades, LIBs have caught tremendous attention in mobile communication, portable electronics, and electric vehicles. Furthermore, global warming has become a worldwide issue due to the ongoing production of greenhouse gases. It motivates solutions such as renewable sources of energy. Solar and wind energies are the most important ones in renewable energy sources. By technology progress, they will definitely require batteries to store the produced power to make a balance between power generation and consumption. Nowadays,rechargeable batteries such as LIBs are considered as one of the best solutions. They provide high specific energy and high rate performance while their rate of self-discharge is low.
Performance of LIBs can be improved through the modification of battery characteristics. The size of solid particles in electrodes can impact the specific energy and the cyclability of batteries. It can improve the amount of lithium content in the electrode which is a vital parameter in capacity and capability of a battery. There exist diferent sources of heat generation in LIBs such as heat produced during electrochemical reactions, internal resistance in battery. The size of electrode's electroactive particles can directly affect the produced heat in battery. It will be shown that the smaller size of solid particle enhance the thermal characteristics of LIBs.
Thermal issues such as overheating, temperature maldistribution in the battery, and thermal runaway have confined applications of LIBs. Such thermal challenges reduce the Life cycle of LIBs. As well, they may lead to dangerous conditions such as fire or even explosion in batteries. However, recent advances in fabrication of advanced materials such as graphene and carbon nanotubes with extraordinary thermal conductivity and electrical properties propose new opportunities to enhance their performance. Since experimental works are expensive, our objective is to use computational methods to investigate the thermal issues in LIBS. Dissipation of the heat produced in the battery can improve the cyclability and specific capacity of LIBs. In real applications, packs of LIB consist several battery cells that are used as the power source. Therefore, it is worth to investigate thermal characteristic of battery packs under their cycles of charging/discharging operations at different applied current rates. To remove the produced heat in batteries, they can be surrounded by materials with high thermal conductivity. Parafin wax absorbs high energy since it has a high latent heat. Absorption high amounts of energy occurs at constant temperature without phase change. As well, thermal conductivity of parafin can be magnified with nano-materials such as graphene, CNT, and fullerene to form a nano-composite medium. Improving the thermal conductivity of LIBs increase the heat dissipation from batteries which is a vital issue in systems of battery thermal management. The application of two-dimensional (2D) materials has been on the rise since exfoliation the graphene from bulk graphite. 2D materials are single-layered in an order of nanosizes which show superior thermal, mechanical, and optoelectronic properties. They are potential candidates for energy storage and supply, particularly in lithium ion batteries as electrode material. The high thermal conductivity of graphene and graphene-like materials can play a significant role in thermal management of batteries. However, defects always exist in nano-materials since there is no ideal fabrication process. One of the most important defects in materials are nano-crack which can dramatically weaken the mechanical properties of the materials. Newly synthesized crystalline carbon nitride with the stoichiometry of C3N have attracted many attentions due to its extraordinary mechanical and thermal properties. The other nano-material is phagraphene which shows anisotropic mechanical characteristics which is ideal in production of nanocomposite.
It shows ductile fracture behavior when subjected under uniaxial loadings. It is worth to investigate their thermo-mechanical properties in its pristine and defective states. We hope that the findings of our work not only be useful for both experimental and theoretical researches but also help to design advanced electrodes for LIBs.
The importance of modern simulation methods in the mechanical analysis of heterogeneous solids is presented in detail. Thereby the problem is noted that even for small bodies the required high-resolution analysis reaches the limits of today's computational power, in terms of memory demand as well as acceptable computational effort. A further problem is that frequently the accuracy of geometrical modelling of heterogeneous bodies is inadequate. The present work introduces a systematic combination and adaption of grid-based methods for achieving an essentially higher resolution in the numerical analysis of heterogeneous solids. Grid-based methods are as well primely suited for developing efficient and numerically stable algorithms for flexible geometrical modeling. A key aspect is the uniform data management for a grid, which can be utilized to reduce the effort and complexity of almost all concerned methods. A new finite element program, called Mulgrido, was just developed to realize this concept consistently and to test the proposed methods. Several disadvantages which generally result from grid discretizations are selectively corrected by modified methods. The present work is structured into a geometrical model, a mechanical model and a numerical model. The geometrical model includes digital image-based modeling and in particular several methods for the theory-based generation of inclusion-matrix models. Essential contributions refer to variable shape, size distribution, separation checks and placement procedures of inclusions. The mechanical model prepares the fundamentals of continuum mechanics, homogenization and damage modeling for the following numerical methods. The first topic of the numerical model introduces to a special version of B-spline finite elements. These finite elements are entirely variable in the order k of B-splines. For homogeneous bodies this means that the approximation quality can arbitrarily be scaled. In addition, the multiphase finite element concept in combination with transition zones along material interfaces yields a valuable solution for heterogeneous bodies. As the formulation is element-based, the storage of a global stiffness matrix is superseded such that the memory demand can essentially be reduced. This is possible in combination with iterative solver methods which represent the second topic of the numerical model. Here, the focus lies on multigrid methods where the number of required operations to solve a linear equation system only increases linearly with problem size. Moreover, for badly conditioned problems quite an essential improvement is achieved by preconditioning. The third part of the numerical model discusses certain aspects of damage simulation which are closely related to the proposed grid discretization. The strong efficiency of the linear analysis can be maintained for damage simulation. This is achieved by a damage-controlled sequentially linear iteration scheme. Finally a study on the effective material behavior of heterogeneous bodies is presented. Especially the influence of inclusion shapes is examined. By means of altogether more than one hundred thousand random geometrical arrangements, the effective material behavior is statistically analyzed and assessed.
Advanced finite elements are proposed for the mechanical analysis of heterogeneous materials. The approximation quality of these finite elements can be controlled by a variable order of B-spline shape functions. An element-based formulation is developed such that the finite element problem can iteratively be solved without storing a global stiffness matrix. This memory saving allows for an essential increase of problem size. The heterogeneous material is modelled by projection onto a uniform, orthogonal grid of elements. Conventional, strictly grid-based finite element models show severe oscillating defects in the stress solutions at material interfaces. This problem is cured by the extension to multiphase finite elements. This concept enables to define a heterogeneous material distribution within the finite element. This is possible by a variable number of integration points to each of which individual material properties can be assigned. Based on an interpolation of material properties at nodes and further smooth interpolation within the finite elements, a continuous material function is established. With both, continuous B-spline shape function and continuous material function, also the stress solution will be continuous in the domain. The inaccuracy implied by the continuous material field is by far less defective than the prior oscillating behaviour of stresses. One- and two-dimensional numerical examples are presented.
The present paper is part of a comprehensive approach of grid-based modelling. This approach includes geometrical modelling by pixel or voxel models, advanced multiphase B-spline finite elements of variable order and fast iterative solver methods based on the multigrid method. So far, we have only presented these grid-based methods in connection with linear elastic analysis of heterogeneous materials. Damage simulation demands further considerations. The direct stress solution of standard bilinear finite elements is severly defective, especially along material interfaces. Besides achieving objective constitutive modelling, various nonlocal formulations are applied to improve the stress solution. Such a corrective data processing can either refer to input data in terms of Young's modulus or to the attained finite element stress solution, as well as to a combination of both. A damage-controlled sequentially linear analysis is applied in connection with an isotropic damage law. Essentially by a high resolution of the heterogeneous solid, local isotropic damage on the material subscale allows to simulate complex damage topologies such as cracks. Therefore anisotropic degradation of a material sample can be simulated. Based on an effectively secantial global stiffness the analysis is numerically stable. The iteration step size is controlled for an adequate simulation of the damage path. This requires many steps, but in the iterative solution process each new step starts with the solution of the prior step. Therefore this method is quite effective. The present paper provides an introduction of the proposed concept for a stable simulation of damage in heterogeneous solids.
A fast solver method called the multigrid preconditioned conjugate gradient method is proposed for the mechanical analysis of heterogeneous materials on the mesoscale. Even small samples of a heterogeneous material such as concrete show a complex geometry of different phases. These materials can be modelled by projection onto a uniform, orthogonal grid of elements. As one major problem the possible resolution of the concrete specimen is generally restricted due to (a) computation times and even more critical (b) memory demand. Iterative solvers can be based on a local element-based formulation while orthogonal grids consist of geometrical identical elements. The element-based formulation is short and transparent, and therefore efficient in implementation. A variation of the material properties in elements or integration points is possible. The multigrid method is a fast iterative solver method, where ideally the computational effort only increases linear with problem size. This is an optimal property which is almost reached in the implementation presented here. In fact no other method is known which scales better than linear. Therefore the multigrid method gains in importance the larger the problem becomes. But for heterogeneous models with very large ratios of Young's moduli the multigrid method considerably slows down by a constant factor. Such large ratios occur in certain heterogeneous solids, as well as in the damage analysis of solids. As solution to this problem the multigrid preconditioned conjugate gradient method is proposed. A benchmark highlights the multigrid preconditioned conjugate gradient method as the method of choice for very large ratio's of Young's modulus. A proposed modified multigrid cycle shows good results, in the application as stand-alone solver or as preconditioner.
The present article proposes an alternative way to compute the torsional stiffness based on three-dimensional continuum mechanics instead of applying a specific theory of torsion. A thin, representative beam slice is discretized by solid finite elements. Adequate boundary conditions and coupling conditions are integrated into the numerical model to obtain a proper answer on the torsion behaviour, thus on shear center, shear stress and torsional stiffness. This finite element approach only includes general assumptions of beam torsion which are independent of cross-section geometry. These assumptions essentially are: no in-plane deformation, constant torsion and free warping. Thus it is possible to achieve numerical solutions of high accuracy for arbitrary cross-sections. Due to the direct link to three-dimensional continuum mechanics, it is possible to extend the range of torsion analysis to sections which are composed of different materials or even to heterogeneous beams on a high scale of resolution. A brief study follows to validate the implementation and results are compared to analytical solutions.
In this study, an application of evolutionary multi-objective optimization algorithms on the optimization of sandwich structures is presented. The solution strategy is known as Elitist Non-Dominated Sorting Evolution Strategy (ENSES) wherein Evolution Strategies (ES) as Evolutionary Algorithm (EA) in the elitist Non-dominated Sorting Genetic algorithm (NSGA-II) procedure. Evolutionary algorithm seems a compatible approach to resolve multi-objective optimization problems because it is inspired by natural evolution, which closely linked to Artificial Intelligence (AI) techniques and elitism has shown an important factor for improving evolutionary multi-objective search. In order to evaluate the notion of performance by ENSES, the well-known study case of sandwich structures are reconsidered. For Case 1, the goals of the multi-objective optimization are minimization of the deflection and the weight of the sandwich structures. The length, the core and skin thicknesses are the design variables of Case 1. For Case 2, the objective functions are the fabrication cost, the beam weight and the end deflection of the sandwich structures. There are four design variables i.e., the weld height, the weld length, the beam depth and the beam width in Case 2. Numerical results are presented in terms of Paretooptimal solutions for both evaluated cases.
NUMERICAL SIMULATION OF THERMO-HYGRAL ALKALI-SILICA REACTION MODEL IN CONCRETE AT THE MESOSCALE
(2010)
This research aims to model Alkali-Silica Reaction gel expansion in concrete under the influence of hygral and thermal loading, based on experimental results. ASR provokes a heterogeneous expansion in concrete leading to dimensional changes and eventually the premature failure of the concrete structure. This can result in map cracking on the concrete surface which will decrease the concrete stiffness. Factors that influence ASR are parameters such as the cement alkalinity, the number of deleterious silica from the aggregate used, concrete porosity, and external factors like temperature, humidity and external source of alkali from ingression of deicing salts. Uncertainties of the influential factors make ASR a difficult phenomenon to solve; hence my approach to this matter is to solve the problem using stochastic modelling, where a numerical simulation of concrete cross-section with integration of experimental results from Finger-Institute for Building Materials Science at the Bauhaus-Universität Weimar. The problem is formulated as a multi-field problem, combining heat transfer, fluid transfer and the reaction rate model with the mechanical stress field. Simulation is performed as a mesoscale model considering aggregates and mortar matrix. The reaction rate model will be conducted using experimental results from concrete expansions due to ASR gained from concrete prism tests. Expansive strains values for transient environmental conditions due to the reaction rate will be determined from calculation based on the reaction rate model. Results from these models will be able to predict the rate of ASR expansion and the cracking propagation that may arise.
The Marmara Region (NW Turkey) has experienced significant earthquakes (M > 7.0) to date. A destructive earthquake is also expected in the region. To determine the effect of the specific design spectrum, eleven provinces located in the region were chosen according to the Turkey Earthquake Building Code updated in 2019. Additionally, the differences between the previous and updated regulations of the country were investigated. Peak Ground Acceleration (PGA) and Peak Ground Velocity (PGV) were obtained for each province by using earthquake ground motion levels with 2%, 10%, 50%, and 68% probability of exceedance in 50-year periods. The PGA values in the region range from 0.16 to 0.7 g for earthquakes with a return period of 475 years. For each province, a sample of a reinforced-concrete building having two different numbers of stories with the same ground and structural characteristics was chosen. Static adaptive pushover analyses were performed for the sample reinforced-concrete building using each province’s design spectrum. The variations in the earthquake and structural parameters were investigated according to different geographical locations. It was determined that the site-specific design spectrum significantly influences target displacements for performance-based assessments of buildings due to seismicity characteristics of the studied geographic location.
A topology optimization method has been developed for structures subjected to multiple load cases (Example of a bridge pier subjected to wind loads, traffic, superstructure...). We formulate the problem as a multi-criterial optimization problem, where the compliance is computed for each load case. Then, the Epsilon constraint method (method proposed by Chankong and Haimes, 1971) is adapted. The strategy of this method is based on the concept of minimizing the maximum compliance resulting from the critical load case while the other remaining compliances are considered in the constraints. In each iteration, the compliances of all load cases are computed and only the maximum one is minimized. The topology optimization process is switching from one load to another according to the variation of the resulting compliance. In this work we will motivate and explain the proposed methodology and provide some numerical examples.
The computational costs of newly developed numerical simulation play a critical role in their acceptance within both academic use and industrial employment. Normally, the refinement of a method in the area of interest reduces the computational cost. This is unfortunately not true for most nonlocal simulation, since refinement typically increases the size of the material point neighborhood. Reducing the discretization size while keep- ing the neighborhood size will often require extra consideration. Peridy- namic (PD) is a newly developed numerical method with nonlocal nature. Its straightforward integral form equation of motion allows simulating dy- namic problems without any extra consideration required. The formation of crack and its propagation is known as natural to peridynamic. This means that discontinuity is a result of the simulation and does not demand any post-processing. As with other nonlocal methods, PD is considered an expensive method. The refinement of the nodal spacing while keeping the neighborhood size (i.e., horizon radius) constant, emerges to several nonphysical phenomena.
This research aims to reduce the peridynamic computational and imple- mentation costs. A novel refinement approach is introduced. The pro- posed approach takes advantage of the PD flexibility in choosing the shape of the horizon by introducing multiple domains (with no intersections) to the nodes of the refinement zone. It will be shown that no ghost forces will be created when changing the horizon sizes in both subdomains. The approach is applied to both bond-based and state-based peridynamic and verified for a simple wave propagation refinement problem illustrating the efficiency of the method. Further development of the method for higher dimensions proves to have a direct relationship with the mesh sensitivity of the PD. A method for solving the mesh sensitivity of the PD is intro- duced. The application of the method will be examined by solving a crack propagation problem similar to those reported in the literature.
New software architecture is proposed considering both academic and in- dustrial use. The available simulation tools for employing PD will be collected, and their advantages and drawbacks will be addressed. The challenges of implementing any node base nonlocal methods while max- imizing the software flexibility to further development and modification
will be discussed and addressed. A software named Relation-Based Sim- ulator (RBS) is developed for examining the proposed architecture. The exceptional capabilities of RBS will be explored by simulating three dis- tinguished models. RBS is available publicly and open to further develop- ment. The industrial acceptance of the RBS will be tested by targeting its performance on one Mac and two Linux distributions.
The computational costs of newly developed numerical simulation play a critical role in their acceptance within both academic use and industrial employment. Normally, the refinement of a method in the area of interest reduces the computational cost. This is unfortunately not true for most nonlocal simulation, since refinement typically increases the size of the material point neighborhood. Reducing the discretization size while keep- ing the neighborhood size will often require extra consideration. Peridynamic (PD) is a newly developed numerical method with nonlocal nature. Its straightforward integral form equation of motion allows simulating dynamic problems without any extra consideration required. The formation of crack and its propagation is known as natural to peridynamic. This means that discontinuity is a result of the simulation and does not demand any post-processing. As with other nonlocal methods, PD is considered an expensive method. The refinement of the nodal spacing while keeping the neighborhood size (i.e., horizon radius) constant, emerges to several nonphysical phenomena.
This research aims to reduce the peridynamic computational and imple- mentation costs. A novel refinement approach is introduced. The pro- posed approach takes advantage of the PD flexibility in choosing the shape of the horizon by introducing multiple domains (with no intersections) to the nodes of the refinement zone. It will be shown that no ghost forces will be created when changing the horizon sizes in both subdomains. The approach is applied to both bond-based and state-based peridynamic and verified for a simple wave propagation refinement problem illustrating the efficiency of the method. Further development of the method for higher dimensions proves to have a direct relationship with the mesh sensitivity of the PD. A method for solving the mesh sensitivity of the PD is intro- duced. The application of the method will be examined by solving a crack propagation problem similar to those reported in the literature.
New software architecture is proposed considering both academic and in- dustrial use. The available simulation tools for employing PD will be collected, and their advantages and drawbacks will be addressed. The challenges of implementing any node base nonlocal methods while max- imizing the software flexibility to further development and modification will be discussed and addressed. A software named Relation-Based Sim- ulator (RBS) is developed for examining the proposed architecture. The exceptional capabilities of RBS will be explored by simulating three distinguished models. RBS is available publicly and open to further develop- ment. The industrial acceptance of the RBS will be tested by targeting its performance on one Mac and two Linux distributions.
Methods based on B-splines for model representation, numerical analysis and image registration
(2015)
The thesis consists of inter-connected parts for modeling and analysis using newly developed isogeometric methods. The main parts are reproducing kernel triangular B-splines, extended isogeometric analysis for solving weakly discontinuous problems, collocation methods using superconvergent points, and B-spline basis in image registration applications.
Each topic is oriented towards application of isogeometric analysis basis functions to ease the process of integrating the modeling and analysis phases of simulation.
First, we develop reproducing a kernel triangular B-spline-based FEM for solving PDEs. We review the triangular B-splines and their properties. By definition, the triangular basis function is very flexible in modeling complicated domains. However, instability results when it is applied for analysis. We modify the triangular B-spline by a reproducing kernel technique, calculating a correction term for the triangular kernel function from the chosen surrounding basis. The improved triangular basis is capable to obtain the results with higher accuracy and almost optimal convergence rates.
Second, we propose an extended isogeometric analysis for dealing with weakly discontinuous problems such as material interfaces. The original IGA is combined with XFEM-like enrichments which are continuous functions themselves but with discontinuous derivatives. Consequently, the resulting solution space can approximate solutions with weak discontinuities. The method is also applied to curved material interfaces, where the inverse mapping and the curved triangular elements are considered.
Third, we develop an IGA collocation method using superconvergent points. The collocation methods are efficient because no numerical integration is needed. In particular when higher polynomial basis applied, the method has a lower computational cost than Galerkin methods. However, the positions of the collocation points are crucial for the accuracy of the method, as they affect the convergent rate significantly. The proposed IGA collocation method uses superconvergent points instead of the traditional Greville abscissae points. The numerical results show the proposed method can have better accuracy and optimal convergence rates, while the traditional IGA collocation has optimal convergence only for even polynomial degrees.
Lastly, we propose a novel dynamic multilevel technique for handling image registration. It is application of the B-spline functions in image processing. The procedure considered aims to align a target image from a reference image by a spatial transformation. The method starts with an energy function which is the same as a FEM-based image registration. However, we simplify the solving procedure, working on the energy function directly. We dynamically solve for control points which are coefficients of B-spline basis functions. The new approach is more simple and fast. Moreover, it is also enhanced by a multilevel technique in order to prevent instabilities. The numerical testing consists of two artificial images, four real bio-medical MRI brain and CT heart images, and they show our registration method is accurate, fast and efficient, especially for large deformation problems.
We investigate the thermal conductivity in the armchair and zigzag MoS2 nanoribbons, by combining the non-equilibrium Green's function approach and the first-principles method. A strong orientation dependence is observed in the thermal conductivity. Particularly, the thermal conductivity for the armchair MoS2 nanoribbon is about 673.6 Wm−1 K−1 in the armchair nanoribbon, and 841.1 Wm−1 K−1 in the zigzag nanoribbon at room temperature. By calculating the Caroli transmission, we disclose the underlying mechanism for this strong orientation dependence to be the fewer phonon transport channels in the armchair MoS2 nanoribbon in the frequency range of [150, 200] cm−1. Through the scaling of the phonon dispersion, we further illustrate that the thermal conductivity calculated for the MoS2 nanoribbon is esentially in consistent with the superior thermal conductivity found for graphene.
Cooling Performance of a Novel Circulatory Flow Concentric Multi-Channel Heat Sink with Nanofluids
(2020)
Heat rejection from electronic devices such as processors necessitates a high heat removal rate. The present study focuses on liquid-cooled novel heat sink geometry made from four channels (width 4 mm and depth 3.5 mm) configured in a concentric shape with alternate flow passages (slot of 3 mm gap). In this study, the cooling performance of the heat sink was tested under simulated controlled conditions.The lower bottom surface of the heat sink was heated at a constant heat flux condition based on dissipated power of 50 W and 70 W. The computations were carried out for different volume fractions of nanoparticles, namely 0.5% to 5%, and water as base fluid at a flow rate of 30 to 180 mL/min. The results showed a higher rate of heat rejection from the nanofluid cooled heat sink compared with water. The enhancement in performance was analyzed with the help of a temperature difference of nanofluid outlet temperature and water outlet temperature under similar operating conditions. The enhancement was ~2% for 0.5% volume fraction nanofluids and ~17% for a 5% volume fraction.
The longitudinal dispersion coefficient (LDC) plays an important role in modeling the transport of pollutants and sediment in natural rivers. As a result of transportation processes, the concentration of pollutants changes along the river. Various studies have been conducted to provide simple equations for estimating LDC. In this study, machine learning methods, namely support vector regression, Gaussian process regression, M5 model tree (M5P) and random forest, and multiple linear regression were examined in predicting the LDC in natural streams. Data sets from 60 rivers around the world with different hydraulic and geometric features were gathered to develop models for LDC estimation. Statistical criteria, including correlation coefficient (CC), root mean squared error (RMSE) and mean absolute error (MAE), were used to scrutinize the models. The LDC values estimated by these models were compared with the corresponding results of common empirical models. The Taylor chart was used to evaluate the models and the results showed that among the machine learning models, M5P had superior performance, with CC of 0.823, RMSE of 454.9 and MAE of 380.9. The model of Sahay and Dutta, with CC of 0.795, RMSE of 460.7 and MAE of 306.1, gave more precise results than the other empirical models. The main advantage of M5P models is their ability to provide practical formulae. In conclusion, the results proved that the developed M5P model with simple formulations was superior to other machine learning models and empirical models; therefore, it can be used as a proper tool for estimating the LDC in rivers.
Tall buildings have become an integral part of cities despite all their pros and cons. Some current tall buildings have several problems because of their unsuitable location; the problems include increasing density, imposing traffic on urban thoroughfares, blocking view corridors, etc. Some of these buildings have destroyed desirable views of the city. In this research, different criteria have been chosen, such as environment, access, social-economic, land-use, and physical context. These criteria and sub-criteria are prioritized and weighted by the analytic network process (ANP) based on experts’ opinions, using Super Decisions V2.8 software. On the other hand, layers corresponding to sub-criteria were made in ArcGIS 10.3 simultaneously, then via a weighted overlay (map algebra), a locating plan was created. In the next step seven hypothetical tall buildings (20 stories), in the best part of the locating plan, were considered to evaluate how much of theses hypothetical buildings would be visible (fuzzy visibility) from the street and open spaces throughout the city. These processes have been modeled by MATLAB software, and the final fuzzy visibility plan was created by ArcGIS. Fuzzy visibility results can help city managers and planners to choose which location is suitable for a tall building and how much visibility may be appropriate. The proposed model can locate tall buildings based on technical and visual criteria in the future development of the city and it can be widely used in any city as long as the criteria and weights are localized.
Matrix-free voxel-based finite element method for materials with heterogeneous microstructures
(2019)
Modern image detection techniques such as micro computer tomography
(μCT), magnetic resonance imaging (MRI) and scanning electron microscopy (SEM) provide us with high resolution images of the microstructure of materials in a non-invasive and convenient way. They form the basis for the geometrical models of high-resolution analysis, so called image-based analysis.
However especially in 3D, discretizations of these models reach easily the size of 100 Mill. degrees of freedoms and require extensive hardware resources in terms of main memory and computing power to solve the numerical model. Consequently, the focus of this work is to combine and adapt numerical solution methods to reduce the memory demand first and then the computation time and therewith enable an execution of the image-based analysis on modern computer desktops. Hence, the numerical model is a straightforward grid discretization of the voxel-based (pixels with a third dimension) geometry which omits the boundary detection algorithms and allows reduced storage of the finite element data structure and a matrix-free solution algorithm.
This in turn reduce the effort of almost all applied grid-based solution techniques and results in memory efficient and numerically stable algorithms for the microstructural models. Two variants of the matrix-free algorithm are presented. The efficient iterative solution method of conjugate gradients is used with matrix-free applicable preconditioners such as the Jacobi and the especially suited multigrid method. The jagged material boundaries of the voxel-based mesh are smoothed through embedded boundary elements which contain different material information at the integration point and are integrated sub-cell wise though without additional boundary detection. The efficiency of the matrix-free methods can be retained.
Accurate prediction of stable alluvial hydraulic geometry, in which erosion and sedimentation are in equilibrium, is one of the most difficult but critical topics in the field of river engineering. Data mining algorithms have been gaining more attention in this field due to their high performance and flexibility. However, an understanding of
the potential for these algorithms to provide fast, cheap, and accurate predictions of hydraulic geometry is lacking. This study provides the first quantification of this potential. Using at-a-station field data, predictions of flow depth, water-surface width and longitudinal water surface slope are made using three standalone data mining techniques -, Instance-based Learning (IBK), KStar, Locally Weighted Learning (LWL) - along with four types of novel hybrid algorithms in which the standalone models are trained with Vote, Attribute Selected
Classifier (ASC), Regression by Discretization (RBD), and Cross-validation Parameter Selection (CVPS) algorithms (Vote-IBK, Vote-Kstar, Vote-LWL, ASC-IBK, ASC-Kstar, ASC-LWL, RBD-IBK, RBD-Kstar, RBD-LWL, CVPSIBK, CVPS-Kstar, CVPS-LWL). Through a comparison of their predictive performance and a sensitivity analysis of the driving variables, the results reveal: (1) Shield stress was the most effective parameter in the prediction of all geometry dimensions; (2) hybrid models had a higher prediction power than standalone data mining models,
empirical equations and traditional machine learning algorithms; (3) Vote-Kstar model had the highest performance in predicting depth and width, and ASC-Kstar in estimating slope, each providing very good prediction performance. Through these algorithms, the hydraulic geometry of any river can potentially be predicted accurately and with ease using just a few, readily available flow and channel parameters. Thus, the results reveal that these models have great potential for use in stable channel design in data poor catchments, especially in developing nations where technical modelling skills and understanding of the hydraulic and sediment processes occurring in the river system may be lacking.
Complex vortex flow patterns around bridge piers, especially during floods, cause scour process that can result in the failure of foundations. Abutment scour is a complex three-dimensional phenomenon that is difficult to predict especially with traditional formulas obtained using empirical approaches such as regressions. This paper presents a test of a standalone Kstar model with five novel hybrid algorithm of bagging (BA-Kstar), dagging (DA-Kstar), random committee (RC-Kstar), random subspace (RS-Kstar), and weighted instance handler wrapper (WIHWKstar) to predict scour depth (ds) for clear water condition. The dataset consists of 99 scour depth data from flume experiments (Dey and Barbhuiya, 2005) using abutment shapes such as vertical, semicircular and 45◦ wing. Four dimensionless parameter of relative flow depth (h/l), excess abutment Froude number (Fe), relative sediment size (d50/l) and relative submergence (d50/h) were considered for the prediction of relative scour depth (ds/l). A portion of the dataset was used for the calibration (70%), and the remaining used for model validation. Pearson correlation coefficients helped deciding relevance of the input parameters combination and finally four different combinations of input parameters were used. The performance of the models was assessed visually and with quantitative metrics. Overall, the best input combination for vertical abutment shape is the combination of Fe, d50/l and h/l, while for semicircular and 45◦ wing the combination of the Fe and d50/l is the most effective input parameter combination. Our results show that incorporating Fe, d50/l and h/l lead to higher performance while involving d50/h reduced the models prediction power for vertical abutment shape and for semicircular and 45◦ wing involving h/l and d50/h lead to more error. The WIHW-Kstar provided the highest performance in scour depth prediction around vertical abutment shape while RC-Kstar model outperform of other models for scour depth prediction around semicircular and 45◦ wing.
The seismic vulnerability assessment of existing reinforced concrete (RC) buildings is a significant source of disaster mitigation plans and rescue services. Different countries evolved various Rapid Visual Screening (RVS) techniques and methodologies to deal with the devastating consequences of earthquakes on the structural characteristics of buildings and human casualties. Artificial intelligence (AI) methods, such as machine learning (ML) algorithm-based methods, are increasingly used in various scientific and technical applications. The investigation toward using these techniques in civil engineering applications has shown encouraging results and reduced human intervention, including uncertainties and biased judgment. In this study, several known non-parametric algorithms are investigated toward RVS using a dataset employing different earthquakes. Moreover, the methodology encourages the possibility of examining the buildings’ vulnerability based on the factors related to the buildings’ importance and exposure. In addition, a web-based application built on Django is introduced. The interface is designed with the idea to ease the seismic vulnerability investigation in real-time. The concept was validated using two case studies, and the achieved results showed the proposed approach’s potential efficiency
One of the most important subjects of hydraulic engineering is the reliable estimation of the transverse distribution in the rectangular channel of bed and wall shear stresses. This study makes use of the Tsallis entropy, genetic programming (GP) and adaptive neuro-fuzzy inference system (ANFIS) methods to assess the shear stress distribution (SSD) in the rectangular channel.
To evaluate the results of the Tsallis entropy, GP and ANFIS models, laboratory observations were used in which shear stress was measured using an optimized Preston tube. This is then used to measure the SSD in various aspect ratios in the rectangular channel. To investigate the shear stress percentage, 10 data series with a total of 112 different data for were used. The results of the sensitivity analysis show that the most influential parameter for the SSD in smooth rectangular channel is the dimensionless parameter B/H, Where the transverse coordinate is B, and the flow depth is H. With the parameters (b/B), (B/H) for the bed and (z/H), (B/H) for the wall as inputs, the modeling of the GP was better than the other one. Based on the analysis, it can be concluded that the use of GP and ANFIS algorithms is more effective in estimating shear stress in smooth rectangular channels than the Tsallis entropy-based equations.
In recent years, lightweight materials, such as polymer composite materials (PNCs) have been studied and developed due to their excellent physical and chemical properties. Structures composed of these composite materials are widely used in aerospace engineering structures, automotive components, and electrical devices. The excellent and outstanding mechanical, thermal, and electrical properties of Carbon nanotube (CNT) make it an ideal filler to strengthen polymer materials’ comparable properties. The heat transfer of composite materials has very promising engineering applications in many fields, especially in electronic devices and energy storage equipment. It is essential in high-energy density systems since electronic components need heat dissipation functionality. Or in other words, in electronic devices the generated heat should ideally be dissipated by light and small heat sinks.
Polymeric composites consist of fillers embedded in a polymer matrix, the first ones will significantly affect the overall (macroscopic) performance of the material. There are many common carbon-based fillers such as single-walled carbon nanotubes (SWCNT), multi-walled carbon nanotubes (MWCNT), carbon nanobuds (CNB), fullerene, and graphene. Additives inside the matrix have become a popular subject for researchers. Some extraordinary characters, such as high-performance load, lightweight design, excellent chemical resistance, easy processing, and heat transfer, make the design of polymeric nanotube composites (PNCs) flexible. Due to the reinforcing effects with different fillers on composite materials, it has a higher degree of freedom and can be designed for the structure according to specific applications’ needs. As already stated, our research focus will be on SWCNT enhanced PNCs. Since experiments are timeconsuming, sometimes expensive and cannot shed light into phenomena taking place for instance at the interfaces/interphases of composites, they are often complemented through theoretical and computational analysis.
While most studies are based on deterministic approaches, there is a comparatively lower number of stochastic methods accounting for uncertainties in the input parameters. In deterministic models, the output of the model is fully determined by the parameter values and the initial conditions. However, uncertainties in the input parameters such as aspect ratio, volume fraction, thermal properties of fiber and matrix need to be taken into account for reliable predictions. In this research, a stochastic multiscale method is provided to study the influence of numerous uncertain input parameters on the thermal conductivity of the composite. Therefore, a hierarchical multi-scale method based on computational homogenization is presented in to predict the macroscopic thermal conductivity based on the fine-scale structure. In order to study the inner mechanism, we use the finite element method and employ surrogate models to conduct a Global Sensitivity Analysis (GSA). The SA is performed in order to quantify the influence of the conductivity of the fiber, matrix, Kapitza resistance, volume fraction and aspect ratio on the macroscopic conductivity. Therefore, we compute first-order and total-effect sensitivity indices with different surrogate models.
As stochastic multiscale models are computational expensive, surrogate approaches are commonly exploited. With the emergence of high performance computing and artificial intelligence, machine learning has become a popular modeling tool for numerous applications. Machine learning (ML) is commonly used in regression and maps data through specific rules with algorithms to build input and output models. They are particularly useful for nonlinear input-output relationships when sufficient data is available. ML has also been used in the design of new materials and multiscale analysis. For instance, Artificial neural networks and integrated learning seem to be ideally for such a task. They can theoretically simulate any non-linear relationship through the connection of neurons. Mapping relationships are employed to carry out data-driven simulations of inputs and outputs in stochastic modeling.
This research aims to develop a stochastic multi-scale computational models of PNCs in heat transfer. Multi-scale stochastic modeling with uncertainty analysis and machine learning methods consist of the following components:
-Uncertainty Analysis. A surrogate based global sensitivity analysis is coupled with a hierarchical multi-scale method employing computational homogenization. The effect of the conductivity of the fibers and the matrix, the Kapitza resistance, volume fraction and aspect ratio on the ’macroscopic’ conductivity of the composite is systematically studied. All selected surrogate models yield consistently the conclusions that the most influential input parameters are the aspect ratio followed by the volume fraction. The Kapitza Resistance has no significant effect on the thermal conductivity of the PNCs. The most accurate surrogate model in terms of the R2 value is the moving least square (MLS).
-Hybrid Machine Learning Algorithms. A combination of artificial neural network (ANN) and particle swarm optimization (PSO) is applied to estimate the relationship between variable input and output parameters. The ANN is used for modeling the composite while PSO improves the prediction performance through an optimized global minimum search. The thermal conductivity of the fibers and the matrix, the kapitza resistance, volume fraction and aspect ratio are selected as input parameters. The output is the macroscopic (homogenized) thermal conductivity of the composite. The results show that the PSO significantly improves the predictive ability of this hybrid intelligent algorithm, which outperforms traditional neural networks.
-Stochastic Integrated Machine Learning. A stochastic integrated machine learning based multiscale approach for the prediction of the macroscopic thermal conductivity in PNCs is developed. Seven types of machine learning models are exploited in this research, namely Multivariate Adaptive Regression Splines (MARS), Support Vector Machine (SVM), Regression Tree (RT), Bagging Tree (Bag), Random Forest (RF), Gradient Boosting Machine (GBM) and Cubist. They are used as components of stochastic modeling to construct the relationship between the variable of the inputs’ uncertainty and the macroscopic thermal conductivity of PNCs. Particle Swarm Optimization (PSO) is used for hyper-parameter tuning to find the global optimal values leading to a significant reduction in the computational cost. The advantages and disadvantages of various methods are also analyzed in terms of computing time and model complexity to finally give a recommendation for the applicability of different models.
The study presents a Machine Learning (ML)-based framework designed to forecast the stress-strain relationship of arc-direct energy deposited mild steel. Based on microstructural characteristics previously extracted using microscopy and X-ray diffraction, approximately 1000 new parameter sets are generated by applying the Latin Hypercube Sampling Method (LHSM). For each parameter set, a Representative Volume Element (RVE) is synthetically created via Voronoi Tessellation. Input raw data for ML-based algorithms comprises these parameter sets or RVE-images, while output raw data includes their corresponding stress-strain relationships calculated after a Finite Element (FE) procedure. Input data undergoes preprocessing involving standardization, feature selection, and image resizing. Similarly, the stress-strain curves, initially unsuitable for training traditional ML algorithms, are preprocessed using cubic splines and occasionally Principal Component Analysis (PCA). The later part of the study focuses on employing multiple ML algorithms, utilizing two main models. The first model predicts stress-strain curves based on microstructural parameters, while the second model does so solely from RVE images. The most accurate prediction yields a Root Mean Squared Error of around 5 MPa, approximately 1% of the yield stress. This outcome suggests that ML models offer precise and efficient methods for characterizing dual-phase steels, establishing a framework for accurate results in material analysis.
The numerical simulation of damage using phenomenological models on the macroscale was state of the art for many decades. However, such models are not able to capture the complex nature of damage, which simultaneously proceeds on multiple length scales. Furthermore, these phenomenological models usually contain damage parameters, which are physically not interpretable. Consequently, a reasonable experimental determination of these parameters is often impossible. In the last twenty years, the ongoing advance in computational capacities provided new opportunities for more and more detailed studies of the microstructural damage behavior. Today, multiphase models with several million degrees of freedom enable for the numerical simulation of micro-damage phenomena in naturally heterogeneous materials. Therewith, the application of multiscale concepts for the numerical investigation of the complex nature of damage can be realized. The presented thesis contributes to a hierarchical multiscale strategy for the simulation of brittle intergranular damage in polycrystalline materials, for example aluminum. The numerical investigation of physical damage phenomena on an atomistic microscale and the integration of these physically based information into damage models on the continuum meso- and macroscale is intended. Therefore, numerical methods for the damage analysis on the micro- and mesoscale including the scale transfer are presented and the transition to the macroscale is discussed. The investigation of brittle intergranular damage on the microscale is realized by the application of the nonlocal Quasicontinuum method, which fully describes the material behavior by atomistic potential functions, but reduces the number of atomic degrees of freedom by introducing kinematic couplings. Since this promising method is applied only by a limited group of researchers for special problems, necessary improvements have been realized in an own parallelized implementation of the 3D nonlocal Quasicontinuum method. The aim of this implementation was to develop and combine robust and efficient algorithms for a general use of the Quasicontinuum method, and therewith to allow for the atomistic damage analysis in arbitrary grain boundary configurations. The implementation is applied in analyses of brittle intergranular damage in ideal and nonideal grain boundary models of FCC aluminum, considering arbitrary misorientations. From the microscale simulations traction separation laws are derived, which describe grain boundary decohesion on the mesoscale. Traction separation laws are part of cohesive zone models to simulate the brittle interface decohesion in heterogeneous polycrystal structures. 2D and 3D mesoscale models are presented, which are able to reproduce crack initiation and propagation along cohesive interfaces in polycrystals. An improved Voronoi algorithm is developed in 2D to generate polycrystal material structures based on arbitrary distribution functions of grain size. The new model is more flexible in representing realistic grain size distributions. Further improvements of the 2D model are realized by the implementation and application of an orthotropic material model with Hill plasticity criterion to grains. The 2D and 3D polycrystal models are applied to analyze crack initiation and propagation in statically loaded samples of aluminum on the mesoscale without the necessity of initial damage definition.
The design and application of high performance materials demands extensive knowledge of the materials damage behavior, which significantly depends on the meso- and microstructural complexity. Numerical simulations of crack growth on multiple length scales are promising tools to understand the damage phenomena in complex materials. In polycrystalline materials it has been observed that the grain boundary decohesion is one important mechanism that leads to micro crack initiation. Following this observation the paper presents a polycrystal mesoscale model consisting of grains with orthotropic material behavior and cohesive interfaces along grain boundaries, which is able to reproduce the crack initiation and propagation along grain boundaries in polycrystalline materials. With respect to the importance of modeling the geometry of the grain structure an advanced Voronoi algorithm is proposed to generate realistic polycrystalline material structures based on measured grain size distribution. The polycrystal model is applied to investigate the crack initiation and propagation in statically loaded representative volume elements of aluminum on the mesoscale without the necessity of initial damage definition. Future research work is planned to include the mesoscale model into a multiscale model for the damage analysis in polycrystalline materials.
Structural vibration control of high-speed railway bridges using tuned mass dampers, semi-active tuned mass dampers, fluid viscous dampers and magnetorheological dampers to reduce resonant structural vibrations is studied. In this work, the addressed main issues include modeling of the dynamic interaction of the structures, optimization of the parameters of the dampers and comparison of their efficiency.
A new approach to optimize multiple tuned mass damper systems on an uncertain model is proposed based on the H-infinity optimization criteria and the DK iteration procedure with norm-bounded uncertainties in frequency domain. The parameters of tuned mass dampers are optimized directly and simultaneously on different modes contributing significantly to the multi-resonant peaks to explore the different possible combinations of parameters. The effectiveness of the present method is also evaluated through comparison with a previous method.
In the case of semi-active tuned mass dampers, an optimization algorithm is derived to control the magnetorheological damper in these semi-active damping systems. The use of the proposed algorithm can generate various combinations of control gains and state variables. This can lead to the improvement of the ability of MR dampers to track the desired control forces. An uncertain model to reduce detuning effects is also considered in this work.
Next, for fluid viscous dampers, in order to tune the optimal parameters of fluid viscous dampers to the vicinity of the exact values, analytical formulae which can include structural damping are developed based on the perturbation method. The proposed formulae can also be considered as an improvement of the previous analytical formulae, especially for bridge beams with large structural damping.
Finally, a new combination of magnetorheological dampers and a double-beam system to improve the performance of the primary structure vibration is proposed. An algorithm to control magnetorheological dampers in this system is developed by using standard linear matrix inequality techniques. Weight functions as a loop shaping procedure are also introduced in the feedback controllers to improve the tracking ability of magnetorheological damping forces. To this end, the effectiveness of magnetorheological dampers controlled by the proposed scheme, along with the effects of the uncertain and time-delay parameters on the models, are evaluated through numerical simulations.
Additionally, a comparison of the dampers based on their performance is also considered in this work.
Encapsulation-based self-healing concrete has received a lot of attention nowadays in civil engineering field. These capsules are embedded in the cementitious matrix during concrete mixing. When the cracks appear, the embedded capsules which are placed along the path of incoming crack are fractured and then release of healing agents in the vicinity of damage. The materials of capsules need to be designed in a way that they should be able to break with small deformation, so the internal fluid can be released to seal the crack. This study focuses on computational modeling of fracture in encapsulation-based selfhealing concrete. The numerical model of 2D and 3D with randomly packed aggreates and capsules have been developed to analyze fracture mechanism that plays a significant role in the fracture probability of capsules and consequently the self-healing process. The capsules are assumed to be made of Poly Methyl Methacrylate (PMMA) and the potential cracks are represented by pre-inserted cohesive elements with tension and shear softening laws along the element boundaries of the mortar matrix, aggregates, capsules, and at the interfaces between these phases. The effects of volume fraction, core-wall thickness ratio, and mismatch fracture properties of capsules on the load carrying capacity of self-healing concrete and fracture probability of the capsules are investigated. The output of this study will become valuable tool to assist not only the experimentalists but also the manufacturers in designing an appropriate capsule material for self-healing concrete.
Polylactic acid (PLA) is a highly applicable material that is used in 3D printers due to some significant features such as its deformation property and affordable cost. For improvement of the end-use quality, it is of significant importance to enhance the quality of fused filament fabrication (FFF)-printed objects in PLA. The purpose of this investigation was to boost toughness and to reduce the production cost of the FFF-printed tensile test samples with the desired part thickness. To remove the need for numerous and idle printing samples, the response surface method (RSM) was used. Statistical analysis was performed to deal with this concern by considering extruder temperature (ET), infill percentage (IP), and layer thickness (LT) as controlled factors. The artificial intelligence method of artificial neural network (ANN) and ANN-genetic algorithm (ANN-GA) were further developed to estimate the toughness, part thickness, and production-cost-dependent variables. Results were evaluated by correlation coefficient and RMSE values. According to the modeling results, ANN-GA as a hybrid machine learning (ML) technique could enhance the accuracy of modeling by about 7.5, 11.5, and 4.5% for toughness, part thickness, and production cost, respectively, in comparison with those for the single ANN method. On the other hand, the optimization results confirm that the optimized specimen is cost-effective and able to comparatively undergo deformation, which enables the usability of printed PLA objects.
In this research, an attempt was made to reduce the dimension of wavelet-ANFIS/ANN (artificial neural network/adaptive neuro-fuzzy inference system) models toward reliable forecasts as well as to decrease computational cost. In this regard, the principal component analysis was performed on the input time series decomposed by a discrete wavelet transform to feed the ANN/ANFIS models. The models were applied for dissolved oxygen (DO) forecasting in rivers which is an important variable affecting aquatic life and water quality. The current values of DO, water surface temperature, salinity, and turbidity have been considered as the input variable to forecast DO in a three-time step further. The results of the study revealed that PCA can be employed as a powerful tool for dimension reduction of input variables and also to detect inter-correlation of input variables. Results of the PCA-wavelet-ANN models are compared with those obtained from wavelet-ANN models while the earlier one has the advantage of less computational time than the later models. Dealing with ANFIS models, PCA is more beneficial to avoid wavelet-ANFIS models creating too many rules which deteriorate the efficiency of the ANFIS models. Moreover, manipulating the wavelet-ANFIS models utilizing PCA leads to a significant decreasing in computational time. Finally, it was found that the PCA-wavelet-ANN/ANFIS models can provide reliable forecasts of dissolved oxygen as an important water quality indicator in rivers.
We conducted extensive molecular dynamics simulations to investigate the thermal conductivity of polycrystalline hexagonal boron-nitride (h-BN) films. To this aim, we constructed large atomistic models of polycrystalline h-BN sheets with random and uniform grain configuration. By performing equilibrium molecular dynamics (EMD) simulations, we investigated the influence of the average grain size on the thermal conductivity of polycrystalline h-BN films at various temperatures. Using the EMD results, we constructed finite element models of polycrystalline h-BN sheets to probe the thermal conductivity of samples with larger grain sizes. Our multiscale investigations not only provide a general viewpoint regarding the heat conduction in h-BN films but also propose that polycrystalline h-BN sheets present high thermal conductivity comparable to monocrystalline sheets.
For this paper, the problem of energy/voltage management in photovoltaic (PV)/battery systems was studied, and a new fractional-order control system on basis of type-3 (T3) fuzzy logic systems (FLSs) was developed. New fractional-order learning rules are derived for tuning of T3-FLSs such that the stability is ensured. In addition, using fractional-order calculus, the robustness was studied versus dynamic uncertainties, perturbation of irradiation, and temperature and abruptly faults in output loads, and, subsequently, new compensators were proposed. In several examinations under difficult operation conditions, such as random temperature, variable irradiation, and abrupt changes in output load, the capability of the schemed controller was verified. In addition, in comparison with other methods, such as proportional-derivative-integral (PID), sliding mode controller (SMC), passivity-based control systems (PBC), and linear quadratic regulator (LQR), the superiority of the suggested method was demonstrated.
In this study, a new approach to basis of intelligent systems and machine learning algorithms is introduced for solving singular multi-pantograph differential equations (SMDEs). For the first time, a type-2 fuzzy logic based approach is formulated to find an approximated solution. The rules of the suggested type-2 fuzzy logic system (T2-FLS) are optimized by the square root cubature Kalman filter (SCKF) such that the proposed fineness function to be minimized. Furthermore, the stability and boundedness of the estimation error is proved by novel approach on basis of Lyapunov theorem. The accuracy and robustness of the suggested algorithm is verified by several statistical examinations. It is shown that the suggested method results in an accurate solution with rapid convergence and a lower computational cost.
Biodiesel, as the main alternative fuel to diesel fuel which is produced from renewable and available resources, improves the engine emissions during combustion in diesel engines. In this study, the biodiesel is produced initially from waste cooking oil (WCO). The fuel samples are applied in a diesel engine and the engine performance has been considered from the viewpoint of exergy and energy approaches. Engine tests are performed at a constant 1500 rpm speed with various loads and fuel samples. The obtained experimental data are also applied to develop an artificial neural network (ANN) model. Response surface methodology (RSM) is employed to optimize the exergy and energy efficiencies. Based on the results of the energy analysis, optimal engine performance is obtained at 80% of full load in presence of B10 and B20 fuels. However, based on the exergy analysis results, optimal engine performance is obtained at 80% of full load in presence of B90 and B100 fuels. The optimum values of exergy and energy efficiencies are in the range of 25–30% of full load, which is the same as the calculated range obtained from mathematical modeling.
This research aims to model soil temperature (ST) using machine learning models of multilayer perceptron (MLP) algorithm and support vector machine (SVM) in hybrid form with the Firefly optimization algorithm, i.e. MLP-FFA and SVM-FFA. In the current study, measured ST and meteorological parameters of Tabriz and Ahar weather stations in a period of 2013–2015 are used for training and testing of the studied models with one and two days as a delay. To ascertain conclusive results for validation of the proposed hybrid models, the error metrics are benchmarked in an independent testing period. Moreover, Taylor diagrams utilized for that purpose. Obtained results showed that, in a case of one day delay, except in predicting ST at 5 cm below the soil surface (ST5cm) at Tabriz station, MLP-FFA produced superior results compared with MLP, SVM, and SVM-FFA models. However, for two days delay, MLP-FFA indicated increased accuracy in predicting ST5cm and ST 20cm of Tabriz station and ST10cm of Ahar station in comparison with SVM-FFA. Additionally, for all of the prescribed models, the performance of the MLP-FFA and SVM-FFA hybrid models in the testing phase was found to be meaningfully superior to the classical MLP and SVM models.
In engineering science the modeling and numerical analysis of complex systems and relations plays an important role. In order to realize such an investigation, for example a stochastic analysis, in a reasonable computational time, approximation procedure have been developed. A very famous approach is the response surface method, where the relation between input and output quantities is represented for example by global polynomials or local interpolation schemes as Moving Least Squares (MLS). In recent years artificial neural networks (ANN) have been applied as well for such purposes. Recently an adaptive response surface approach for reliability analyses was proposed, which is very efficient concerning the number of expensive limit state function evaluations. Due to the applied simplex interpolation the procedure is limited to small dimensions. In this paper this approach is extended for larger dimensions using combined ANN and MLS response surfaces for evaluating the adaptation criterion with only one set of joined limit state points. As adaptation criterion a combination by using the maximum difference in the conditional probabilities of failure and the maximum difference in the approximated radii is applied. Compared to response surfaces on directional samples or to plain directional sampling the failure probability can be estimated with a much smaller number of limit state points.
The Element-free Galerkin Method has become a very popular tool for the simulation of mechanical problems with moving boundaries. The internally applied Moving Least Squares approximation uses in general Gaussian or cubic weighting functions and has compact support. Due to the approximative character of this method the obtained shape functions do not fulfill the interpolation condition, which causes additional numerical effort for the imposition of the essential boundary conditions. The application of a singular weighting function, which leads to singular coefficient matrices at the nodes, can solve this problem, but requires a very careful placement of the integration points. Special procedures for the handling of such singular matrices were proposed in literature, which require additional numerical effort. In this paper a non-singular weighting function is presented, which leads to an exact fulfillment of the interpolation condition. This weighting function leads to regular values of the weights and the coefficient matrices in the whole interpolation domain even at the nodes. Furthermore this function gives much more stable results for varying size of the influence radius and for strongly distorted nodal arrangements than classical weighting function types. Nevertheless, for practical applications the results are similar as these obtained with the regularized weighting type presented by the authors in previous publications. Finally a new concept will be presented, which enables an efficient analysis of systems with strongly varying node density. In this concept the nodal influence domains are adapted depending on the nodal configuration by interpolating the influence radius for each direction from the distances to the natural neighbor nodes. This approach requires a Voronoi diagram of the domain, which is available in this study since Delaunay triangles are used as integration background cells. In the numerical examples it will be shown, that this method leads to a more uniform and reduced number of influencing nodes for systems with varying node density than the classical circular influence domains, which means that the small additional numerical effort for interpolating the influence radius leads to remarkable reduction of the total numerical cost in a linear analysis while obtaining similar results. For nonlinear calculations this advantage would be even more significant.
The modeling of crack propagation in plain and reinforced concrete structures is still a field for many researchers. If a macroscopic description of the cohesive cracking process of concrete is applied, generally the Fictitious Crack Model is utilized, where a force transmission over micro cracks is assumed. In the most applications of this concept the cohesive model represents the relation between the normal crack opening and the normal stress, which is mostly defined as an exponential softening function, independently from the shear stresses in tangential direction. The cohesive forces are then calculated only from the normal stresses. By Carol et al. 1997 an improved model was developed using a coupled relation between the normal and shear damage based on an elasto-plastic constitutive formulation. This model is based on a hyperbolic yield surface depending on the normal and the shear stresses and on the tensile and shear strength. This model also represents the effect of shear traction induced crack opening. Due to the elasto-plastic formulation, where the inelastic crack opening is represented by plastic strains, this model is limited for applications with monotonic loading. In order to enable the application for cases with un- and reloading the existing model is extended in this study using a combined plastic-damage formulation, which enables the modeling of crack opening and crack closure. Furthermore the corresponding algorithmic implementation using a return mapping approach is presented and the model is verified by means of several numerical examples. Finally an investigation concerning the identification of the model parameters by means of neural networks is presented. In this analysis an inverse approximation of the model parameters is performed by using a given set of points of the load displacement curves as input values and the model parameters as output terms. It will be shown, that the elasto-plastic model parameters could be identified well with this approach, but require a huge number of simulations.