This paper shows how to model linear plants using data obtained by simple experiments, such as those performed in data-driven controls. The method estimates the frequency response of the plant and uses it to estimate its transfer function with the prediction error method in the frequency domain. The frequency domain prediction error method allows the selection of the frequency components to be used for modeling easily and suppresses the effects of noise on the modeling results. Numerical experiments of open-loop modeling and actual experiments of closed-loop modeling with velocity control system for a two-inertia system demonstrate the effectiveness of the proposed method.

Improving the initial setup accuracy of various hot strip mill actuators has been an issue for a long time. This study proposes new methods to enhance the accuracy if setup values based on the just-in-time method that uses a large amount of stored operational data. The key to the proposed methods is to extract the appropriate neighborhood data necessary for the setup calculations. The proposed method combines local regression, neighborhood data extracted via clustering based on linear regression, and the setup value calculated through the conventional method. We confirmed that the proposed method better improves the accuracy compared to the conventional method.

In recent years, the development of optimization methods in multi-agent systems has been remarkable. We have proposed a distributed scheduling method using the Alternating Direction Method of Multipliers (ADMM). However, in many cases, the scheduling process oscillates and does not converge. In this study, we present a theorem regarding the stability of distributed scheduling using ADMM. Additionally, we propose a modified ADMM algorithm to ensure that the algorithm reaches a stable state. The results of computer experiments demonstrate the effectiveness of the proposed method.