In this paper, we propose a conceptual design method of disc brake considering the variation of natural frequency. By minimizing the maximum value of brake squeal within the variation range of natural frequency, brake squeal is reduced. Three methods are considered when considering the variation of natural frequency. The first method is a way to make it unconstrained. The second and third are extended versions of the algorithms used in the reliability analysis method. Through the numerical example, we show the effectiveness of the proposed method by comparing the method that does not take into consideration the variation of natural frequency, which is the conventional method. Then, we show that the method based on the numerical analysis method is effective among the three proposed methods.

Deformable mirrors are essential components in fundus imaging modality incorporating adaptive optics systems. Appropriate shape control of the mirror surface by actuators is necessary so that optical aberrations caused by a patient’s eye can be nullified. Although much number of actuators is required for precise shape control, the number of actuators should be minimized in terms of the manufacturing cost. In this scenario, layout optimization is an effective and powerful approach. In this paper, we present an optimization method of actuators’ layout which consists of the minimum number of actuators, while offering the minimum error between a target mirror shape and the actual deformed mirror shape by optimizing the electric current values applied to each actuator. The proposed method introduces design variables for the layout optimization problem as variables controlling the weighting coefficients of the electric current minimization which enables to reduce the number of actuators. This formulation allows us to use mathematical programming for solving the layout optimization problem. A numerical example is provided to confirm the validity of the proposed method.

In the design of mechanical products, the optimum design method called multiobjective topology optimization attracts attention. In this optimum design method, a Pareto optimal solution is obtained. The Pareto optimal solution is obtained as a three-dimensional optimal structure, and a clustering method corresponding to its complicated structure is required, but this method is not shown at the present time. Therefore, by using topological data analysis, we show clustering method which can consider the number of holes in the structure and analyze the obtained result. A numerical example for confirming the effectiveness of the proposed method is shown.

Radial basis function (RBF) network, which is one of the neural network (NN), is used for short-term prediction of time series data. In this paper, the affine transformation is used to consider the direction of time series data. Therefore, the coordinate system of time series data is transformed. The detailed numerical procedure is provided in this paper. To examine the validity of the proposed approach, several numerical examples are handled. Through the numerical results, the validity of the proposed approach has been confirmed.

We propose a new method that evaluates the correlativity between factors in natural phenomena by using multi-objective optimization with a genetic algorithm (GA). The effectiveness of this method was tested by applying GAs to determine which hemodynamic factors correlate with artery bifurcation shapes. In the proposed method, first, shapes of each case were optimized by supposing several combinations of hemodynamic factors as objectives using GA. Those factors were a) to minimize maximum Wall Shear Stress (WSS), b) to maximize minimum WSS, c) to minimize WSS gradient, d) to minimize WSS temporary gradient, and e) to minimize inner surface area. We set six combinations in which two of these factors were in a trade-off relationship. Then, we checked how much each optimized shape differed from its original one. After finishing all optimizations, the degree of difference of each combination factor in 14 actual cases was obtained. From the results, the smallest degree of difference was obtained in the combination of minimizing maximum WSS and minimizing inner surface area. The most correlative combination of factors among the six was evaluated by those degrees. We found our method can accurately estimate the correlativity of hemodynamic factors to artery bifurcation shapes.

Vehicle body frames that affect the crashworthiness are required to design efficiently in the early phase of automobile development. Although structural optimization is often utilized for the development in a short period, it has not been studied for crashworthiness vigorously because it is difficult to calculate sensitivity in non-linear dynamic problem represented by impact analysis. Impact absorption members absorb impact energy by deforming plastically such as bending deformation and crushing along the axis continuously. So, load-displacement curve is one of the indicator in the development of them and utilizing approximation of it leads to the development in a short period. In this study, we obtain impact analysis results of one-dimensional frame, approximate load-displacement curves utilizing layered neural network and select optimal curve.

We formulate a pathway layout problem based on an analogy of electrical circuit. A recursive formulation is adopted to obtain an optimal pathway. On the basis of the obtained pathway layout, the electrical circuit analogy approach is also applied for path-planning, corresponding to given starting and destination points. An example pathway layout and a planned path are demonstrated.

Our aim is to detect hyperacute ischaemic stroke lesions in the CT images using the variational auto-encoder, one of a recently developing machine learning algorithms. In its installation process, several requirements such as segmentation, dataset structure, image quality, pre-processing, and verification method were identified. Then we discussed the possibility of using pattern feature models as an alternative method to alleviate the constraints.

In this paper, I show the usefulness of Genetic Programming, which is extensively studied in recent years, and the necessity of considering the result. I explain why I study this method by chapter 1and explain Genetic Programming and Genetic Algorithms by chapter 2. In chapter 3, I describe the function estimation problem discussed for simplicity and the solution method for this problem. Since it is necessary to perform multi-objective optimization. I explain the k-means methods by chapter 4 in order to cluster optimal solutions for analyzing the results. In chapter 5, I show the result and examine the relationship between the structure of the optimal solution and the objective function value. I describe the conclusion and future prospect in chapter 6. In the future, I want to apply this method to various problems.