Latin hypercube sampling global optimization
SA mimics the annealing process used in the metallurgical industry, by which slow cooling is applied to metals to produce better aligned, low energy-state crystallization. This technique falls in the more general category of evolutionary algorithms. GA draws inspiration from the principles and mechanisms of natural selection to perform search and optimization. This study attempts to solve the RBO pr oblem with two well established metaheuristic methods such as Genetic Algorithm (GA) and Simulated Annealing (SA). Traditional methods may suffer from the unaffordable computational costs of global optimization or premature convergence to local optima. Besides interpolation capability, Kriging models have enough flexibility to approximate arbitrary functions with a high level of accuracy, and can assess the level of uncertainty of model predictions. When compared with the polynomial regression models, Kriging interpolation models for structural reliability problems have several competitive features. More recently, attention has been given to alternative approaches based on artificial neural networks, support vector machines and Kriging interpolation models. The first option is to apply first- and second- order polynomial regression models as surrogates for the true limit state function. Several surrogate models suitable for structural reliability design are presently available in the literature. The reliability analysis of complex and realistic structural systems is in general a computationally demanding task. The LHS method will be selected in this research as MCS sampling procedure in view of its capability to cover the whole range of each sampled variable. However, using an appropriated sampling strategy (for example, weighted sampling, first-order sensitivity method, Latin hypercube sampling - LHS) allows the member of MCS sampling points to be significantly reduced yet reaching the target level of accuracy. This makes MCS computationally expensive. It is a non-intrusive, sampling based numerical method, but often requires a large ensemble of sampling points to provide a reliable and stable estimate of uncertainty. Among numerous methods of uncertainty propagation analysis, the most commonly used method is Monte Carlo Simulation. Quantifying and propagating the uncertainty in the simulation or design process as a key component of risk analysis, robustness evaluation or reliability based optimization (RBO) attracts attention of researchers and designer. Uncertainty is an inevitable issue in the process of manufacture, infrastructure, and engineering design. Keywords: Topology and Sizing Optimization of Trusses, Gravitational Search Algorithm, Efficient Member Grouping, Double and Triple Layer Grid Structures The relative efficiency of surrogate models and their relationship with metaheuristic search engine are discussed in the article. In order to avoid premature convergence of the optimization process, the RBO problem is solved with metaheuristic methods such as genetic algorithm and simulated annealing. The Latin Hypercube sampling approach is applied to a structural finite element model to obtain an effective database for building surrogate models. In this paper, we propose an effective method to decouple the loops of reliability assessment analysis and optimization by creating surrogate models. Despite the advantages of RBO, its application to practical engineering problem is still quite challenging. It searches for the best compromise between cost and safety while considering system uncertainties by incorporating reliability measures within the optimization. Abstract : Reliability based optimization (RBO) is one of the most appropriate methods for structural design under uncertainties.