Automated Harmonic Signal Removal-Based Image Feature Extraction Technique: A Comparative Study Using Online Databases

This paper presents an automated harmonic removal technique as an efficient method for identifying and removing the influence of harmonics from the output signal. The method involves disregarding user-defined parameters during system initialization and reconstructing the output signal automatically so that it can be used for system identification. While stochastic subspace-based algorithms (SSI) are generally reliable for modal parameter estimation, applying them to structures with rotating machinery and harmonic excitations presents challenges. Because the SSI method necessitates designating parameters, such as the maximal within-cluster distance, at the outset of each dataset analysis, the issue remains unresolved. In addition to modal identification, the current research concentrates on image-based feature extraction for aggregating and classifying harmonic components and structural poles directly from a stabilization diagram. Utilizing online data sets to validate the algorithm’s efficacy. Using a comparative analysis, the proposed method is compared to existing techniques, namely orthogonal projection-based harmonic signal removal and smoothing techniques based on linear interpolation. The results indicate that the proposed algorithm estimates modal parameters precisely and consistently both before and after the removal of harmonic components from the response signal.