The topology optimization method is the one of powerful tools to improve the performance of electrical machines. Various topology optimization techniques, e.g. the material-density-based method and the level-set method, have been proposed. While the material-density-based method would be able to drastically change the structure of electrical machine, the complexity of converged structure would be higher from the viewpoint of practical design. Generally, the complexity of converged solution derived from the level-set-based method would be lower than that of the material-density-based method. Its characteristic is desirable for the practical engineering process, nevertheless level-set method has two disadvantages. First, the level-set method is not topology optimization but shape optimization. Second, the convergence speed is much slower than the material-density-based method. To overcome these difficulties, the fast topology optimization method using level set function supported by the Method of Moving Asymptotes (MMA) was proposed, further the efficacy was shown in the 2-D magnetic field problems. In this paper, the performance of its method on nonlinear 3-D magnetic field problems is investigated.

Manufacturers have tackled the important issues of speeding up and improving quality of product development in an extremely changeable market. Analysis leads design is very useful approach to solve the issues by using a supercomputer. Optimization technology is very powerful tool for shortening design period. However, a construction of calculation process and an automatic execution of parameter survey become a burden to perform the optimization. Therefore, we developed the parameter survey support analysis system based on OSS (Open Source Software). Users can construct a calculation flow by drag and drop approach and execute parameter survey using a template as which parameter survey flow is defined. The effectiveness of the proposed system was demonstrated by applying it to centrifugal compressor design. We confirmed that the construction of calculation process and execution of parameter survey took a short time.

A combinatorial approach using simulated annealing is presented for optimization of types and placements of braces in steel frames for seismic strengthening of building structures. Machine learning approaches based on decision tree and support vector machine are also utilized to predict the decent and feasible solutions during optimization and to reduce the number of evaluations of mechanical properties of solutions. It is shown in numerical examples that the performances such as story drift, volume and maximum stress in the existing beams and columns can be improved by optimizing the types and placements of braces.

State information is usually assumed to be obtained in typical model predictive control (MPC), but we can not measure system states in many real app lications. In order to overcome this difficulty, we construct a new type of state obs erver for the system, and propose to use the estimated states in MPC system. A s imulation example shows that the MPC systems with observer works well even wh en various constraints and noises are involved.

A three-link and three-joint optimal control model to reproduce human squat movements has already been proposed. The previous model is characterized by sequentially optimizing the half-crouching process and the rising-up process of squat movements. This research proposes another optimal control model (via-point model) characterized by reproducing squat movements as one continuous movement, and compares the effectiveness and the required number of the boundary conditions between the two models. Consequently, the following results are obtained: (1) the proposed model can also reproduce measured squat movement trajectories; (2) there exists the optimum via-point time to minimize the proposed model's criterion function; (3) the proposed model's necessary boundary conditions at the via point are three joint angles and their angular velocities and become smaller in number than the previous model's ones. These results suggest that the proposed model can be a more plausible model of the human squatting mechanism than the previous model and that the via-point time and the joint angles and their angular velocities at the via-point can be factors involved strongly in human squat movements.