抄録
This paper proposes the method that obtain a reduced-order model using sampled data. The method is similar to the proper orthogonal decomposition (POD). POD determines the base vectors based on the concept of maximizing the variance of the data. On the other hand, this method determines the base vectors based on the concept of minimizing the sum of the distance between the space spanned by the base vectors and the data. In this paper, a relation between a number of sampled data and a accuracy of the obtained model is investigated. The results are as follows: 1) At least in the calculation example of this paper, the number of sampled data is sufficiently the same as the number of dimensions required for solution space approximation. 2) Generalization of conditions to be satisfied by data, such as number, characteristics, etc., is a future task.
