Universitätsbibliothek
Weimar Open Access

Numerical simulation of physical phenomena, like electro-magnetics, structural and fluid mechanics is essential for the cost- and time-efficient development of mechanical products at high quality. It allows to investigate the behavior of a product or a system far before the first prototype of a product is manufactured. This thesis addresses the simulation of contact mechanics. Mechanical contacts appear in nearly every product of mechanical engineering. Gearboxes, roller bearings, valves and pumps are only some examples. Simulating these systems not only for the maximal/minimal stresses and strains but for the stress-distribution in case of tribo-contacts is a challenging task from a numerical point of view. Classical procedures like the Finite Element Method suffer from the nonsmooth representation of contact surfaces with discrete Lagrange elements. On the one hand, an error due to the approximate description of the surface is introduced. On the other hand it is difficult to attain a robust contact search because surface normals can not be described in a unique form at element edges. This thesis introduces therefore a novel approach, the adaptive isogeometric contact formulation based on polynomial Splines over hierarchical T-meshes (PHT-Splines), for the approximate solution of the non-linear contact problem. It provides a more accurate, robust and efficient solution compared to conventional methods. During the development of this method the focus was laid on the solution of static contact problems without friction in 2D and 3D in which the structures undergo small deformations. The mathematical description of the problem entails a system of partial differential equations and boundary conditions which model the linear elastic behaviour of continua. Additionally, it comprises side conditions, the Karush-Kuhn-Tuckerconditions, to prevent the contacting structures from non-physical penetration. The mathematical model must be transformed into its integral form for approximation of the solution. Employing a penalty method, contact constraints are incorporated by adding the resulting equations in weak form to the overall set of equations. For an efficient space discretization of the bulk and especially the contact boundary of the structures, the principle of Isogeometric Analysis (IGA) is applied. Isogeometric Finite Element Methods provide several advantages over conventional Finite Element discretization. Surface approximation with Non-Uniform Rational B-Splines (NURBS) allow a robust numerical solution of the contact problem with high accuracy in terms of an exact geometry description including the surface smoothness. The numerical evaluation of the contact integral is challenging due to generally non-conforming meshes of the contacting structures. In this work the highly accurate Mortar Method is applied in the isogeometric setting for the evaluation of contact contributions. This leads to an algebraic system of equations that is linearized and solved in sequential steps. This procedure is known as the Newton Raphson Method. Based on numerical examples, the advantages of the isogeometric approach with classical refinement strategies, like the p- and h-refinement, are shown and the influence of relevant algorithmic parameters on the approximate solution of the contact problem is verified. One drawback of the Spline approximations of stresses though is that they lack accuracy at the contact edge where the structures change their boundary from contact to no contact and where the solution features a kink. The approximation with smooth Spline functions yields numerical artefacts in the form of non-physical oscillations. This property of the numerical solution is not only a drawback for the simulation of e.g. tribological contacts, it also influences the convergence properties of iterative solution procedures negatively. Hence, the NURBS discretized geometries are transformed to Polynomial Splines over Hierarchical T-meshes (PHT-Splines), for the local refinement along contact edges to reduce the artefact of pressure oscillations. NURBS have a tensor product structure which does not allow to refine only certain parts of the geometrical domain while leaving other parts unchanged. Due to the Bézier Extraction, lying behind the transformation from NURBS to PHT-Splines, the connected mesh structure is broken up into separate elements. This allows an efficient local refinement along the contact edge. Before single elements are refined in a hierarchical form with cross-insertion, existing basis functions must be modified or eliminated. This process of truncation assures local and global linear independence of the refined basis which is needed for a unique approximate solution. The contact boundary is a priori unknown. Local refinement along the contact edge, especially for 3D problems, is for this reason not straight forward. In this work the use of an a posteriori error estimation procedure, the Super Convergent Recovery Solution Based Error Estimation Scheme, together with the Dörfler Marking Method is suggested for the spatial search of the contact edge. Numerical examples show that the developed method improves the quality of solutions along the contact edge significantly compared to NURBS based approximate solutions. Also, the error in maximum contact pressures, which correlates with the pressure artefacts, is minimized by the adaptive local refinement. In a final step the practicability of the developed solution algorithm is verified by an industrial application: The highly loaded mechanical contact between roller and cam in the drive train of a high-pressure fuel pump is considered.

This paper presents numerical analysis of the discrete fundamental solution of the discrete Laplace operator on a rectangular lattice. Additionally, to provide estimates in interior and exterior domains, two different regularisations of the discrete fundamental solution are considered. Estimates for the absolute difference and lp-estimates are constructed for both regularisations. Thus, this work extends the classical results in the discrete potential theory to the case of a rectangular lattice and serves as a basis for future convergence analysis of the method of discrete potentials on rectangular lattices.

This study aims to evaluate a new approach in modeling gully erosion susceptibility (GES) based on a deep learning neural network (DLNN) model and an ensemble particle swarm optimization (PSO) algorithm with DLNN (PSO-DLNN), comparing these approaches with common artificial neural network (ANN) and support vector machine (SVM) models in Shirahan watershed, Iran. For this purpose, 13 independent variables affecting GES in the study area, namely, altitude, slope, aspect, plan curvature, profile curvature, drainage density, distance from a river, land use, soil, lithology, rainfall, stream power index (SPI), and topographic wetness index (TWI), were prepared. A total of 132 gully erosion locations were identified during field visits. To implement the proposed model, the dataset was divided into the two categories of training (70%) and testing (30%). The results indicate that the area under the curve (AUC) value from receiver operating characteristic (ROC) considering the testing datasets of PSO-DLNN is 0.89, which indicates superb accuracy. The rest of the models are associated with optimal accuracy and have similar results to the PSO-DLNN model; the AUC values from ROC of DLNN, SVM, and ANN for the testing datasets are 0.87, 0.85, and 0.84, respectively. The efficiency of the proposed model in terms of prediction of GES was increased. Therefore, it can be concluded that the DLNN model and its ensemble with the PSO algorithm can be used as a novel and practical method to predict gully erosion susceptibility, which can help planners and managers to manage and reduce the risk of this phenomenon.

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.

Wireless sensor networks (WSNs) include large-scale sensor nodes that are densely distributed over a geographical region that is completely randomized for monitoring, identifying, and analyzing physical events. The crucial challenge in wireless sensor networks is the very high dependence of the sensor nodes on limited battery power to exchange information wirelessly as well as the non-rechargeable battery of the wireless sensor nodes, which makes the management and monitoring of these nodes in terms of abnormal changes very difficult. These anomalies appear under faults, including hardware, software, anomalies, and attacks by raiders, all of which affect the comprehensiveness of the data collected by wireless sensor networks. Hence, a crucial contraption should be taken to detect the early faults in the network, despite the limitations of the sensor nodes. Machine learning methods include solutions that can be used to detect the sensor node faults in the network. The purpose of this study is to use several classification methods to compute the fault detection accuracy with different densities under two scenarios in regions of interest such as MB-FLEACH, one-class support vector machine (SVM), fuzzy one-class, or a combination of SVM and FCS-MBFLEACH methods. It should be noted that in the study so far, no super cluster head (SCH) selection has been performed to detect node faults in the network. The simulation outcomes demonstrate that the FCS-MBFLEACH method has the best performance in terms of the accuracy of fault detection, false-positive rate (FPR), average remaining energy, and network lifetime compared to other classification methods.