JOURNAL OF COMPUTATIONAL METHODS IN ENGINEERING (ESTEGHLAL)
Year:2013 | Volume:32 | Issue:1|Papers Count:10

Forecasting is one of the effective tools for planning and establishing the financial strategies. Forecasting accuracy is also one of the most important factors in choosing the forecasting method. Nowadays, despite the numerous forecasting models available, accurate forecasting is not yet a simple task, especially in financial markets. Thus, different models have been combined together in order to achieve more accurate results. Combining different models or using hybrid forecasting models is a common way for overcoming deficiecies of the single models and improving their performance. In the literature, several hybrid models of multilayer perceptrons have been proposed in order to overcome the disadvantages of these models. In this paper, a new hybrid model of multilayer perceptrons is proposed using probabilistic neural classifiers. The proposed model improves the performance of the multilayer perceptrons using the unique advantages of the probabilistic neural classifiers in detecting the break points and better and more complete modeling of the specific patterns in the under-study time series. Empirical results of exchange rate forecasting indicate the efficiency of the proposed model in comparison with other models.

The main goal of this investigation was to study the influence of sheet spacing on the fatigue life of 5083-O aluminium alloy spot-welded joints. The values of notch strength reduction factors were obtained for all kinds of spot-welded joints with different gap distances between sheets based on volumetric approach. The fatigue lives of spot welded joints were then determined using the notch strength reduction factors and the available smooth S-N curve of 5083-O aluminium alloy sheets. The results were compared with the experimental fatigue test data, showing a very good agreement between numerical predictions and experimental results.

Nowadays, the need for tall buildings, substruction of highways, construction of underground trains and consequently excavating operations is essential for modern urban communities. One of the most common types of retaining structures used for varying excavation depths is the metal truss. The aim of this study was to introduce a logical connection between two and three-dimensional analyses in a manner that by using two-dimensional analysis, horizontal and vertical displacements as well as internal forces of the retaining structures with a good approximation in three-dimensional analysis (which is closer to the reality) can be calculated. Therefore, several parametric modelings were simulated in order to investigate the behavior of supported excavations via truss. Also, it was aimed to determine the ratio between maximum horizontal displacement at two-dimensional and three-dimensional analyses (PSR ratio). The PSR ratios for the depth of 4, 7 and 10 meters were 0.7, 0.55 and 0.4, respectively. The internal forces of a critical truss (near the center of the excavation) including axial forces and bending moments were also studied. Comparing the values of axial forces in the main structural members of the critical truss in two-dimensional and three-dimensional analyses shows a good agreement between the two types of modeling. The magnitude of bending moment in three-dimensional to two-dimensional analyses was 0.14, which proves unrealistic in two-dimensional analysis, and can consequently be neglected in structural design procedure.

This paper presents a new boundary element formulation for free and forced vibration analysis of 2D elastic domains. The main idea of the present formulation is to approximate the inertia terms using the inverse multiquadric radial basis functions (IMQ RBFs). The dual reciprocity method (DRM) with the static-type fundamental solution is reconsidered by using the proposed RBFs. The fictitious particular solution kernels of inverse multiquadric RBFs corresponding to displacement and traction, a few terms of which are singular, are explicitly derived. Therefore, a simple mathematical trick is employed to resolve the singularity problem. In addition, the limiting values of the particular solution kernels are evaluated. Several examples are studied to demonstrate the validity and the accuracy of the proposed formulation. The results are compared to the obtained analytical and other RBFs available in the literature.

In this paper, time domain dynamic analysis of dam-reservoir interaction is presented using finite difference method. A semi-implicit scheme is used for the discretization of the governing equation of dam, and an explicit scheme is used for the discretization of the governing equation of fluid. Water is considered compressible but its internal viscosity is neglected. A Sommerfeld’s radiation condition at the infinity boundary of the fluid domain is implemented. To verify the proposed method, numerical examples are given to be compared with the existing solutions. The result shows a good agreement with semi-analytical solution. Finally, the effects of reservoir bottom absorption, the free surface waves of the reservoir and the vertical component of earthquake are analyzed and the results are presented.

In this article, the boundary element method (BEM) is used to analyse the propeller hub and the blades in its vicinity. In solver, the modeling of hub and blades is performed with source and dipole distribution on their surfaces. Iterative procedure is utilized to consider the near body effect. In each step, each body was independently analyzed with the influence of near body considered in inflow velocity. The induced velocity of propeller was studied with and without PBCF (Propeller Boss Cap Fin) downstream. PBCF, as an energy-saving device, causes the induced velocity of propeller to be reduced and uniformed.

In this paper, using Rayleigh-Ritz method (RRM) and based on classical plate theory, thermal buckling of laminated composite rectangular plates is analyzed. Effects of plate aspect ratios, fiber orientations, and the kinds of lamination scheme with and without the foundation on the uni-lateral thermal buckling behavior are analyzed. All studies done in uni-lateral buckling field are specialized for single-layer isotropic plates which are affected by in-plane mechanical loads. The numerical results are validated very satisfactorily for the present method, with its accuracy and reliability confirmed.

In this study, the impact strength of homogenous and intra-ply hybrid composites reinforced with ductile and brittle fibers was predicted based on a finite element method. For this purpose, a four-ply composite with the quasi-isotropic stacking sequence was designed considering the dimensional, physical and mechanical properties. Next, using Abaqus software, the parameters such as contact force and deflection, at 30 J impact energy were determined. In order to validate the simulated model, the theoretical results were compared with the experimental ones. The results revealed that the model can predict the variation of maximum contact force and maximum deflection with changing the content of brittle fiber. In addition, the results can be used to analyze the stress and strain distributions during the impact.

Sparse matrix-vector multiplication (SpMV) is the key operation in the iterative methods for solving linear systems of equations. Almost all numerical methods need to solve such a system in their solution procedure. There have been a lot of researches on this subject and it is still a very hot research area. One of the best methods to increase the performance of this operation is using graphics processing units (GPUs). These processors have had a great improvement in their processing capabilities. In this research, a new method to perform this operation using open computing language (OpenCL) is presented. The results show that by optimizing the parameters of this method one can gain a much higher performance compared to CPUs, even when using the open multi-processing standard (OpenMP). Also, the results show that there is not much sensitivity near the optimal parameters, which paves the way to estimate them from the matrix properties.

In this paper, a novel numerical approach is proposed for determination of a lower bound solution for the bearing capacity of strip footings. In this method, the geometry of problem is constructed by nodes and, there is no need for mesh in the traditional sense. The gradient of stress is smoothed piecemeal by the aid of the stabilized nodal integration technique and, the equilibrium and boundary conditions are fully satisfied at the entire domain consequently. The stress field is discretized by a mesh-free technique called Shepard's method. Due to the individual properties of Shepard's shape functions, the non-yielding condition is just controlled at the nodes. Putting the objective function and the related constraints together forms a mathematical optimization problem which is solved by a linear programming technique. At the end, the accuracy and efficiency of the proposed method is investigated by solving some examples for the cohesive soils with uniform and depth dependent shear strength, and the cohesive-frictional soil.