Knowledge acquisition of vibrations in high-power transformers using statistical analyses and fuzzy approaches – A case study

• Vibrations in high-power transformers were analyzed.
• Radial graph areas and repeated measure analyses were conducted.
• A fuzzy inference system was constructed regarding status alerts.
• These analyses allow the diagnosis of physical operational states of transformers.

This paper presents the results of a knowledge acquisition of vibrations in high-power transformers in two substations of an Electrical Power System in the Amazon region, Brazil. Radial graph areas were obtained, to enable the analyses of the vibrations at four sites at different measurement points in the transformer bodies. In certain cases, variations in vibrations non-justified by corrective maintenance were observed. The repeated measure analysis showed that the largest mean vibration area occurred in the front of the transformer. With these results, measurements on a sequential processor were made and the vibration behavior due to commutations analyzed. Statistical analyses and the repeated measure analysis were used in the process of knowledge acquisition, establishing the basis for the construction of a fuzzy inference system. Compared to other predictive maintenance methods, vibrations are a physical phenomena that can be used in the construction of non-invasive, low-cost analyses techniques, so this approach may become an important and valuable tool for use in solving problems in environments of uncertainty, as in the programming process of vibration monitoring in high-power transformers, providing a means to diagnose the physical operational states of transformers by the generated inspection alerts.