PSI - Issue 77

Francisco Afonso et al. / Procedia Structural Integrity 77 (2026) 575–583 F. Afonso et al. / Structural Integrity Procedia 00 (2026) 000–000

576

2

1. Introduction

Industrial transformers are valuable and crucial assets in electrical grids; faults during operation can cause un planned outages and significant monetary losses [1, 2]. Throughout their lifetime, transformers may experience vari able loading states that can stress their structure [3]. Consequently, there is a pressing need to ensure transformers are suitable for operation before deployment. During manufacturing and assembly, several wellness tests are performed, including structural integrity assess ments of the transformer tank under di ff erent environmental pressure states [4]. The tank’s structural integrity is essential for e ffi cient operation and longevity of the equipment [5]. Typically, the transformer tank is modelled and analysed using finite element analysis (FEA), guiding material selection and assessing the structural behaviour of a certain design [6, 7]. FEA can be complemented by experimental validation, as shown by Hasan et al. [5], where a final pressure test on a 120 MVA power transformer was used to detect leaks and measure tank surface deforma tion. In their research, the pressure values were gradually monitored, tracking pressure drops indicative of leaks, and measuring deformation with scales positioned at various tank locations. Non-contact optical methods can provide an interesting approach for structural analysis of industrial transformers, by enabling full-field inspection of the tank surfaces. Digital image correlation (DIC) is widely used for structural monitoring across various applications; initially, a speckle pattern is applied to a surface, the test is recorded with an optical setup, and DIC correlates speckle motion to estimate surface displacement [8, 9]. This method enables high-accuracy, full-field assessment of surfaces, although it can be noise-sensitive and usually requires painting the test surface [9]. Barros et al. [10] used a structure from motion (SFM) algorithm to calibrate a moving camera, and 3D DIC to determine the deformation of a specimen, between a reference and a loaded state. Kalaitzakis et al. [11] implemented stereo DIC using a camera-equipped drone on a railroad tie, both studies implementing DIC with non stationary cameras. Mallya et al. [12] conducted a review of how DIC is implemented in structural health monitoring of components in aerospatial engineering, highlighting the method’s high applicability in this domain. As reviewed by Suchocki et al. [13] terrestrial laser scanners (TLS) are another option for remote defect detection; a widely implemented distance measuring technique in TLS is the phase-shift (PS) principle, where a laser pulse is emitted and reflected back to the sensor from the specimen’s surface, enabling distance estimation from the phase di ff erence between emitted and received laser beam. The output of these sensors is generally a detailed point cloud. In this paper, the surface of an industrial core-type transformer is monitored before and during the vacuum drying process in a factory environment. Two non-contact, optical methods were implemented: a stereo DIC setup and a phase-shifting TLS (FARO FOCUS S). The results show that both DIC and the phase-shifting TLS successfully captured deformation data in an industrial setting, although environmental vibrations from surrounding activities and personnel access through the area where the sensors were installed posed challenges. Since painting the transformer surface was not possible, speckle patterns printed on adhesive paper and composed of magnets were used; the results revealed a slight variation between the measurements obtained by each method, which correspond to small absolute discrepancies, with an order of magnitude between 10 − 3 m for the highest di ff erence, and 10 − 5 m for the lowest. The results could have been impacted by the aforementioned challenges, the alternative speckle implementations, lower contrast between the magnet speckle pattern and the tank wall, non optimal speckle sizes due to uncertainty in available space for DIC setup implementation prior to testing, and potential errors in the simulated data. Despite acceptable results, validating these alternative speckle implementations requires further testing. The aims of this research are to perform non-contact analysis of industrial transformers during vacuum testing in a factory environment, identify advantages and limitations of the implemented optical methods, and compare the results against simulation data. Chapter 2 describes the optical setups, Chapter 3 presents and discusses the results, while Chapter 4 provides concluding remarks and future work suggestions.

Nomenclature

PS Phase-shift TLS Terrestrial Laser Scanner DIC Digital image correlation