An Algorithm to Find Interval Solution Ranges and Its Use to Improve Prediction Accuracy in Multi-Objective Optimal Design

In the design process of electrical products, it is often necessary to find solutions that simultaneously satisfy multi-objective performance requirements that sometimes conflict. Such solutions are expected to be redundant to cover a wider range of design parameters. In this paper, we propose a new algorithm to find the interval solution ranges and calculate the range size. Furthermore, the proposed algorithm was used to find the interval solution ranges from the spectrum prediction through ANN-based machine learning algorithm for EMI filters used for noise suppression in brush motor drive systems. The relationship between recall and the magnitude of the interval solution range based on the confusion matrix indicates that lower recall does not result in a wider interval solution range. This suggests that reducing the number of false negatives (FN) and increasing the number of true positives (TP) will result in a better interval solution range.