This paper describes a study on an inverse analysis method to identify the crack parameters of non-uniform beams by the simple genetic algorithm. The crack is simulated by an equivalent rotational spring model. By using this approximate model, a characteristic equation to calculate the natural frequencies of the system are derived employing the Rayleigh-Ritz method. In order to identify the crack location and the depth of the beam from frequency measurements, objective function that consists of the sum of square errors of frequencies is formed. The genetic algorithm is used to search the minimum value of the objective function. The proposed method is verified by the numerical simulation of some sample problems.

抄録全体を表示

The present paper deals with limit analysis of structures by genetic algorithms (GAs), which are simple GA (SGA), forking GA (fGA) and stochastic schemata exploiter (SSE). Although this problem can be expressed as a linear programming problem, it is formulated as a combinatorial optimization problem of elementary collapse modes in the present paper. A collapse mechanism can be coded as a partial tree graph of a connected graph, which expresses the relationship between elementary modes and common plastic hinges of two elementary modes. Example rigid frames of 2-bay 6-story buildings are analyzed by means of each GA with discussion on efficiency of these methods. As a result, it is illustrated that SSE is superior to the others.

抄録全体を表示

In our previous studies, various optimization methods, such as dynamic programming and neural network, are applied to architectural space-planning problems. As a result, it is realized that the sequential linear programming (SLP) method is superior to others. But because passages are not considered in these studies, several problems are occured in optimal allocations. In this paper, assuming that the passage is one room whose size are variables, the space-planning problem with passages is formulated as an optimization problem. Numerical examples of clinic space planning are optimally designed by the SLP with discussion on practicability of the design results.

抄録全体を表示

The present paper deals with the allocation problem of multi-story shear walls in 3D RC frames. This problem is formulated as a minimum weight design problem under constraints of ultimate strength, relative story displacement and dynamic eccentric ratio. In the present paper, this problem is formulated as two combinatorial problems, which are searching problem of feasible solutions at each floor and minimization problem of total weight of walls. Optimality technique based upon the branch-and-bound method is applied to each problem. Several example of multi-story frames are optimally designed with discussion on influence of the number of stories, scale of plan, and allocation constraint at each floor upon computing efficiency.

抄録全体を表示

In a series of the studies where two-dimensional continua have been optimized topologically by using our own contrived method equipped with genetic algorithm. The method is that shape and topology (i.e. layout) optimization is attained in such a way that optimal layouts of finite element models are obtained by selecting using the layouts of the models as individuals, and by adopting removal and addition parameters, β and γ. of the elements as chromosomes. In the previous paper we discussed that certain relations are certified to exist between the mean values of β and γ. and generations. and then optimal layouts to deterministically obtain by using the relations. Here the similar method is applied to three-dimensonal finite models and its validity is demonstrated in numerical experiments.

抄録全体を表示

In this paper, an adaptive truss structure is a target of design. As in the optimum posture determination problem of adaptive trusses there are many elements which must be treated simultaneously, it is thought that it is difficult to obtain the global optimum solution due to the existence of many local optimum solutions. Then, we propose to use Immune Algorithms (IA) which is superior in the global convergency for solving the optimum posture determination problem.

抄録全体を表示

In this study, the models of parallel distributed genetic algorithms are summarized and a new model of distributed genetic algorithm is proposed. The models of parallel distributed genetic algorithms are classified into three categories; those are a master slave model, a distributed GA model, and a cellular GA model. The models of master slave and distributed GA are suited for PC cluster systems. The characteristics of these models are compared and discussed thorough the numerical examples. From this comparison and discussion, a distributed GA model has higher searching ability and it is much suitable for PC cluster systems. We also proposed a new model of distributed GA; that is called "Dual Individual Genetic Algorithm : DuDGA". In this model, there are only two individuals. Because it is needles to set some parameters of distributed GAs in this model, this model is very useful for users. More than that, DuDGAs have high searching ability. This characteristics are also discussed through the numerical examples.

抄録全体を表示

The plate and shell structures with stiffeners are frequently adopted to acheive large and ligh structures. A new design technique called "growth method", which acheives a minimum compliance design, is suggested to obtain optimal layout pattern of the stiffeners for the thin-walled structures. The growth method starts from seed points, and grows gradually the stiffeners along the grid nodes of the finite element in every interation. The method is applied to decide optimum layouts of stiffeners for square plates undr distributed and concentrated loads.

抄録全体を表示

Optimum design of expensive calculation model has been required. Response surface methodology is developed to solve the expensive calculation model by approximation models. Although the approximation models constructed by the response surface methodology are usually used to get the optimal designs, it is thought that the design accuracy is dependent on the type of activate functions and the design region of interest. In this paper, techniques to search all the local optimal designs within the feasible design region, and techniques for finding more accurate approximation using the proposed Most Probable Optimal Design (MPOD) method is introduced. The application example of the Hybrid III dummy model shows that the MPOD technique is effective in saving the computational cost.

抄録全体を表示

In this paper, the analytical model is developed for active minimization of interior noise within a three-dimensional car cabin enclosure using distributed piezoelectric actuators. Two issures about the interior noise active control are discussed. One is the modeling technique. Modal synthesis method is used to establish the governing motion equations of the fully coupled acoustics-structure-piezoelectric patch system using modal frequencies, damping and mode shapes of a structural and an acoustical system which appear in the frequency range of interest. The second issue is the optimization of peizo actuator's positions. A new response surface methodology MPOD (most probable optimal design) based on holographic neural network, proposed by the authors is applied to solve the global optimal positions of piezoelectric patches. Numerical investigations to the effect of different number of actuators, error sensors as well as disturbing point forces on SPLs are presented.

抄録全体を表示

The theoretical method is first introduced for clarifying the quality and level of the optimized results, obtained by using genetic algorithms (GAs) and evolutional strategies (ESs). The method includes the quasi-intuition, instinct, and plasticity. We apply this method to the problems minimizing the drag force of airfoil. Drag forces are obtained by the numerical calculation code of two-dimensional flows including an instability theory on the transition to turbulence in the boundary layer. It should be stressed that GAs and ESs are relatively easy to get the best-adapted solution for subsonic flow problems. Finally, we will discuss the robustness of the optimized airfoil shapes for small disturbances such as the attachment of insect on airfoil.

抄録全体を表示

This paper proposes an automated search method for multidimensional design window (DW), which is defined as an area of satisfactory design solutions in a design parameter space, using both continuous evolutionary algorithms (CEA) and improved K-means clustering. This method consists of two phases : (a) CEA-based search for satisfactory design solutions in an ill-structured design space, and (b) clustering-based identification of DWs from the solutions. To demonstrate practical performance of the method, it is applied to an automated structural design of micro electrostatic actuator for optical memory.

抄録全体を表示

New inverse analysis method with the finite element method was developed for optical computed tomography (optical C.T.). In optical C.T., near-inferred light is sourced into human body and scattered light is measured to detect an oxygenated part. This inverse problem is reduced to estimate the location of a part with inhomogenous absorption coefficient from scattered light on a surface of human body. The fluence rate φ in human body was assumed to obey the photon diffusion eqation. The finite element method was used to discretize the equation. Non-steady problem was treated by applying Laplace transformation. Because the observation equation is a highly non-linear function of location of inhomogeneiry conventional iterative method requires huge amount of computational time. To overcome this difficulty, we introduced a vector which is easily obtained from the location of inhomogeneiry, and is also related to observed value. This vector enables us to solve the inverse problem without any iterative calculation. A simple 3D example problem was solves to demonstrate the usefulness of this method.

抄録全体を表示

The subject of structure defective diagnosis is an active research area in nondestructive testing (NDT). In this paper, a new structure damage detecting approach is proposed based on the combination of the bispectrum feature extracting method and the Learning Vector Quantization (LVQ) identification method. Since that bispectrum analysis possesses the capability of Gaussian noise restraining, so it is available to be employed to enhance the performance of feature extracting. In simulation, by using the method proposed, it has shown that relatively very higher accuracy of structure damage identification can be obtained comparing with the MAC (formed from the Model parameters) method, especially in case of low signal to noise ratio environment.

抄録全体を表示