Ho Chi Minh City University of Architecture

This article presents two diagnostic techniques for assessing the mechanical and electrical condition of oil-immersed power transformers: sweep frequency response analysis (SFRA) and dissolved gas analysis (DGA). The SFRA method detects potential mechanical faults in transformer windings by applying the correlation coefficient (Rxy). For a 63 MVA, 115/23/11 kV transformer, experimental results show that the winding is considered normal when RLF ≥ 2.0, RMF ≥ 1.0, and RHF ≥ 0.6, whereas RLF < 0.6 indicates severe deformation. In contrast, the DGA method identifies internal faults such as partial discharge, low-energy and high-energy discharges, and thermal faults by analyzing gas concentrations in the insulating oil. For a 20 MVA, 110/22 kV transformer, measured gas concentrations (µl/l, ppm) include H₂ (0.00), CH₄ (40.43), C₂H₆ (18.24), C₂H₄ (20.92), C₂H₂ (0.00), CO₂ (8932.66), and CO (2131.18), corresponding to a “thermal fault below 300°C.” Both diagnostic methods are integrated into user-friendly MATLAB-based tools to support testing centers in Vietnam. The combination of SFRA and DGA enhances diagnostic accuracy, enables timely maintenance decisions, and contributes to improving the reliability of power supply systems.

This study predicts the degree of damage in 2D truss members using the Dung Beetle Optimization algorithm (DBO) in conjunction with finite element analysis for searching natural frequencies. The difference in these natural frequencies between the finite element model and actual measurements is known as the objective function. A number of scenarios, from simple to complex, are proposed to assess the performance of this algorithm. The convergence results, which are based on MATLAB software, required only 25~30 iterations and were then displayed to validate the potential practical use of DBO. Moreover, this short study only focused on the application of DBO to predict damage of truss structures, without going into the comparison of advantages and disadvantages with other existing algorithms.

The online version contains supplementary material available at 10.1007/s10068-025-01993-x.