Catálogo de publicaciones - libros

The paper describes a new conjugate gradient algorithm for large scale nonconvex problems with box constraints. In order to speed up the convergence the algorithm employs a scaling matrix which transforms the space of original variables into the space in which Hessian matrices of functionals describing the problems have more clustered eigenvalues. This is done efficiently by applying limited memory BFGS updating matrices. Once the scaling matrix is calculated, the next few iterations of the conjugate gradient algorithms are performed in the transformed space. The box constraints are treated by the projection as previously used in [R. Pytlak, The efficient algorithm for large-scale problems with simple bounds on the variables, SIAM J. on Optimization, Vol. 8, 532–560, 1998]. We believe that the preconditioned conjugate gradient algorithm gives more flexibility in achieving balance between the computing time and the number of function evaluations in comparison to a limited memory BFGS algorithm. The numerical results show that the proposed method is competitive to L-BFGS-B procedure.

Reliability and durability of civil infrastructure systems such as highway bridges play a very important role in sustainable economic growth and social development of any country. The bridge infrastructure has been undergoing severe safety and condition deterioration due to gradual aging, aggressive environmental stressors, and increasing traffic loads. Maintenance needs for deteriorating highway bridges, however, have far outpaced available scarce funds highway agencies can provide. Bridge management systems (BMSs) are thus critical to cost-effectively allocate limited maintenance resources to bridges for achieving satisfactory lifetime safety and performance. In existing BMSs, however, visual inspections are the most widely adopted practice to quantify and assess bridge conditions, which are unable to faithfully reflect structural capacity deterioration. Failure to detect structural deficiency due to, for example, corrosion and fatigue, and inability to accurately assess real bridge health states may lead to unreliable bridge management decisions and even enormous safety and economic consequences. In this paper, recent advances in risk-based life-cycle maintenance management of deteriorating civil infrastructure systems with emphasis on high-way bridges are reviewed. Methods of predicting lifetime safety and performance of highway bridges with and without maintenance are discussed. Treatment of various uncertainties associated with the complex deterioration processes due to time-dependent loading, environmental stressors, structural resistances, and maintenance actions are emphasized. The bridge maintenance management is formulated as a nonlinear, discrete, combinatorial optimization problem with simultaneous consideration of multiple and conflicting objectives, which address bridge safety and performance as well as long-term economic consequences. The effectiveness of genetic algorithms as a numerical multiobjective optimizer for producing Pareto-optimal tradeoff solutions is demonstrated. The proposed probabilistic multiobjective optimization BMS is applied at project-level for similar bridges and at network-level for a group of different bridges that form a highway network.

In order to establish a rational bridge management program, it is necessary to develop a cost-effective decision-support system for the maintenance of bridges. In this paper, an attempt is made to develop a new multi-objective genetic algorithm for the bridge management system that can provide practical maintenance plans. Several numerical examples are presented to demonstrate the applicability and efficiency of the proposed method.

An approach for the Mixed Discrete Non-Linear Problems (MDNLP) by Particle Swarm Optimization is proposed. The penalty function to handle the discrete design variables is employed, in which the discrete design variables are treated as the continuous design variables by penalizing at the intervals. By using the penalty function, it is possible to handle all design variables as the continuous design variables. Through typical benchmark problem, the validity of proposed approach for MDNLP is examined.

In a plastic injection molding, design of cooling pipe system is one of the important problems to reduce internal residual stresses of molded products. If plastic materials in the injection molding die are cooled down uniformly and slowly, generation of the residual stresses can be reduced. In this paper, a new method to design a cooling pipe system in the plastic injection molding die taking account of coolant flow in the pipe are presented. To consider the effect of the coolant flow, two kinds of models assumed plastic injection die are prepared. And shape optimization techniques are applied to design the cooling pipe system in the models. To evaluate the optimality, two kinds of evaluation functions, one is to obtain uniform temperature distribution and the other is to control cooling rate, are defined.

This paper suggests an innovative design methodology of heat transfer system based on a so-called adaptive growth law, which is an essential optimum growth rule of branch systems in nature. The branch systems in nature can grow by adapting themselves automatically to the growth environments in order to achieve better global functional performances, such as the maximal absorption of nutrition or sunlight in plants and the intelligent blood delivery of a vascular system in animal body. Thus, it can be expected that an optimum layout of heat transfer system would be obtained by the generation method based on the growth mechanism of branch systems in nature. First, the emergent process of branch systems in nature is reproduced in computer model by studying their common growth mechanisms. The branch systems are grown by the control of a so-called nutrient density so as to make it possible that the distribution of branches is dependent on the nutrient distribution. Then, the generation method is applied to the layout design problem for heat transfer systems. Both the conductive heat transfer system and the convective heat transfer system are designed by utilizing the generation method based on the growth mechanisms of branch systems in nature. The effectiveness of the suggested design method is validated by the FEM analysis and by the comparison with other conventional optimum design methods.

In order to obtain maximally innovative and successful product designs, the utilization of optimization strategies at the product design planning stage is of prime importance, and the methods proposed in this paper enable multiobjective optimization technologies to be effectively applied. The necessity and effectiveness of utilizing optimization techniques at the product design planning stage are first explained, and the features that this requires are then clarified. Optimization solutions provided at the product design planning stage, while far from final, can nevertheless be used to obtain guidelines for more preferable product designs. For this purpose, even if characteristics evaluated at the product design planning stages are simplified and/or idealized, the interrelationships among all related characteristics should be simultaneously and thoroughly explored. The successful application of optimization techniques at the product design planning stage requires the rapid presentation and evaluation of a variety of alternative designs, a deeper understanding of the reasons for the optimized designs that are developed, and breakthrough of the initial optimized design solutions, so that the most effective design can ultimately be implemented in a manufactured product. This paper proposes multiobjective design optimization methodologies and procedures, utilized at the product design planning stage, to achieve these goals.

We introduce the stochastic linear programming (SLP) model classes, which will be considered in this paper, on the basis of a small-scale linear programming problem. The solutions for the various problem formulations are discussed in a comparative fashion. We point out the need for model and solution analysis. Subsequently, we outline the basic ideas of selected major algorithms for two classes of SLP problems: two-stage recourse problems and problems with chance constraints. Finally, we illustrate the computational behavior of two algorithms for large-scale SLP problems.

In this paper we study the parameter estimation problem for stochastic distributed parameter systems by using the modified maximum likelihood method. More specifically, by using the US treasury bond data, the parameter estimation is performed for the stochastic hyperbolic and parabolic models describing the behavior of the term-structure of the US bond. From the prediction results, we can show that the parabolic factor models work better than the hyperbolic ones.

In this paper we analyze the electricity portfolio problem of a big consumer in a multi-stage stochastic programming framework. Stochasticity enters the model via the uncertain spot price process and is represented by a scenario tree. The decision that has to be taken is how much energy should be bought in advance, and how large the exposition to the uncertain spot market, as well as the relatively expensive production with an own power plant should be. The risk is modeled using an Average Value-at-Risk (AVaR) term in the objective function. The results of the stochastic programming model are compared with classical fix mix strategies, which are outperformed. Furthermore, the influence of risk parameters is shown.