This paper deals with a large-scale 3D non-linear magnetostatic analysis using the Hierarchical Domain Decomposition Method (HDDM). To improve the convergence of the interface problem of the HDDM and to reduce the computation time, the solution strategy of Finite Element Analysis (FEA) in subdomains is reexamined. Because our employed equation is singular, it has been solved by the iterative method such as the ICCG method. However, by applying the direct method for solving FEA in subdomains, it is expected that the convergence of the interface problem is improved by improvement of precision in subdomains and the computation time is reduced by storing matrices that are results of the LU decomposition on the main memory. Then, to solve FEA in subdomains by the direct method, the A method with the Lagrange multiplier p is considered. As a result, the convergence of the interface problem is much improved by improvement of precision in subdomain-wise analysis. Then, a non-linear magnetostatic problem with 100 million degrees of freedom is solved. Moreover, the computation time is reduced by 10-20%.

View full abstract

In the design of electromagnetic systems, uncertainty which is ascribed to, for example, production errors or deviations of materials should be taken into account. However, conventional optimization methods, which play a crucial role in the design process, cannot effectively treat the uncertainty. In this paper, two novel robust optimization methods, which require no significant increase in the computational cost, are introduced. Moreover, they are applied to optimization problems of electromagnetic systems to test their usefulness and validity.

View full abstract

To cause effective disturbance to the non dimensional symmetric Lagrange's top equation given in [7], the derivative at an initial point must be independent to the null eigenvalue vector at the equilibrium point. Numerically stable simulation with a regulated central difference method results in good coincidence with the linear analysis.

View full abstract

In this paper, we propose a optimization technique that we call a hierarchical Response Surface Methodology (hRSM). Since it is difficult to approximate a complex parameter space by using a single quadratic polynomial surface, hRSM adopted a recursive sub division technique and used the coefficient of multiple determinations (R²) to represent the response surface in each sub space for the threshold of the sub division. We verified the accuracy of hRSM by using some well-known benchmark functions that a reused in optimization. We confirmed the effectiveness of hRSM by applying it to bio-simulation's parameter estimation.

View full abstract

In this paper, a new balancer configuration for stabilizing an unmanned electric bicycle and a new switching control algorithm are proposed. The balancer can be configured as a flywheel mode or a balancer mode by shifting the center of gravity of the balancer. This balancer configuration is changed according to the situation of the bicycle system, which corresponds to the change of the dimension of the system. The balancer is configured as a flywheel, when disturbances to the system are large, and it will switch to the balancer when the position of the center of the gravity should be shifted. Stabilizing bicycle with the flywheel has better performance than the balancer but it cannot control to shift the bicycle angle to track the desired value, unlike the balancer which can do this motion. The balancing controller is derived based on output-zeroing controller. The parameters of the balancing control are switched based on the configuration of the balancer. Numerical simulation results are shown to verify the effectiveness of the proposed control strategy.

View full abstract

In this paper, we formulate nonlinear system identification problems in the case where output data is incomplete. Firstly, we propose an identification method based on an evolutionary algorithm, which is a fusion between a genetic algorithm (GA) and genetic programming (GP). Next, we give some descriptions on what GA and GP are like. Lastly under the situation when there is incomplete output data we illustrate the effectiveness of the proposed method through some simulations and through an experiment with a cart.

View full abstract