Vibration-Based Diagnostics of Non-Ceramic Insulators: Characterization of Signals

Abstract

This paper presents an experimental method for testing composite insulators based on vibration testing. The method used investigated the propagation, signal shape, and distortion of excited mechanical waves under the influence of defects. The aim of the method was to identify defects in the core of a composite insulator that cannot be economically detected by currently available diagnostic methods in field conditions. Therefore, this experiment aimed to distinguish between the mechanical waves’ characteristics of damaged and intact insulators using inexpensive tools. This article seeks to provide a basis for mechanical vibration diagnostics of composite insulators by demonstrating that damage to the core can result in a perceptible difference in the characteristics of mechanical waves when testing within the frequency range of audible sound.

1. Introduction

This paper presents a diagnostic method based on vibration testing for detecting defects in the core of a composite insulator. The presented technology currently has a pilot status but is a potential solution for the lack of an inexpensive, on-site diagnostic method that is able to detect internal mechanical damage in composite insulators. First, a brief introduction to composite insulators, their currently available diagnostic methods, and the theoretical background and applications of wave propagation and vibration testing are given, as some basic knowledge about these topics is essential for understanding the purpose of this article.

In high-voltage technology, the word “insulator” is not only used as a material property, but also refers to one of the fundamental devices which is called an insulator. As the name suggests, the function of insulators is to insulate parts with earth potential from live parts so that no current flows through them. However, insulators are also responsible for holding or supporting high-voltage devices; thus, depending on the type of application and the voltage level, insulators are manufactured from different materials in different sizes and designs. Three types of insulators are distinguished based on their material: ceramic (porcelain), glass, and composite. The technology of composite insulators has been around for a few decades, and the most common design is the long-rod insulator, which is used to hold and tighten conductors on transmission and railway lines; these are called suspension and strain insulators [1]. Composite insulators are also manufactured in hollow-core and post-insulator designs, which are used, for example, in substations [2]. As of today, due to the characteristics of the technology, composite insulators have been replacing glass and ceramic insulators in overhead line applications, but they are less preferred in substation applications. An example of a transmission line and substation application is shown in Figure 1.

The technology has several advantages over its competitors; one is that it is much more resistant to mechanical stresses, such as vibrations, which are constantly present on transmission lines, but it is also resistant to sudden impacts, such as gunshots [4]. Another advantage is the hydrophobic property of the insulators’ silicone coating, which makes it more difficult for dry-band-arcing to occur in wet weather, as precipitation does not form a continuous layer of water on the insulators’ coating; instead, it accumulates in droplets [5]. Furthermore, as there are no metal parts in the insulator spacing, the electric field strength distribution is more favorable, which makes it possible to use shorter insulators, resulting in more compact transmission lines. This not only serves aesthetic reasons, but often, it is the only way to achieve the required safety distances [6]. Composite insulators are also lighter than their competitors, making them easier to be installed and transported, and the supporting structure has to bear less load during operation. However, composite insulators have the disadvantage of being less resistant to UV radiation and various types of erosion [7]. Thus, they are not used in coastal or desert regions. In addition, there is room for improvement in the diagnostics and asset management of composite insulators [8].

A composite insulator consists of three major parts: the GFRP (glass-fiber-reinforced polymer) composite core, the metal end fittings attached to it, and the sheath on the core, which is most often a liquid silicone rubber cover molded onto the device in one piece [9]. The structure of the insulator is shown in Figure 2. Most of the failures are caused by the deterioration of the sheath of the insulators. The possible failure modes and deterioration processes affecting the sheath of the insulator are discussed in detail and illustrated with pictures in the STRI Guide [10]. Fortunately, these can be easily identified in most cases by visual inspection. It should be noted, however, that visual inspection is usually carried out from a distance, which is less effective compared to a close-up inspection, as major damage can often be hidden in the shadows of sheds when viewed from afar. For this reason, there is growing interest in drone technology for inspections and AI evaluation, as these would allow for a more thorough inspection of insulators without investing significant additional resources [11]. In addition to visual inspection, the most commonly used practice to detect the degradation of the insulator’s sheath is the ultraviolet pulse imaging method [12], which relies on special cameras and hydrophobicity testing. Failure modes affecting the core of the insulator may include, but are not limited to, the charring, cracking, or erosion of the insulator’s core; water ingress; or manufacturing defects in the core [13,14,15]. In practice, the most commonly used on-site technology that can detect defects inside insulators is the infrared thermal imaging method, which also relies on special cameras. The method detects temperature rises caused by higher leakage currents, which usually occur when there is insufficient adhesion between the insulator’s sheath and core. However, the interpretation of the results is not straightforward because there can also be a significant temperature difference between two good insulators [16]. This implies that the method is only reliable in the case of extreme temperature differences resulting from advanced degradation processes. Consequently, the method is unsuitable for detecting problems in the early stages. The method is also not suitable for detecting mechanical damage in the insulator’s core, as the leakage currents will not increase in those cases. There are also other less commonly used technologies; but in summary, there is currently no on-site diagnostic method that can effectively detect the degradation of the insulator’s core, especially if the process is mechanical in nature [17].

In this article, we present the initial results of a test method based on vibration testing and wave propagation for detecting mechanical damage in the insulator’s core, such as cracks and cavities that could originate from manufacturing faults or the mishandling of insulators. In engineering, wave phenomena generally refer to the transfer of energy through some medium. In this sense, voltage pulses in a cable, sound waves in the air, or energy transfer in a solid medium due to an impact, which causes the particles to vibrate, are all considered waves. In general, waves have a direction of travel, which can be longitudinal or transverse, based on the direction of energy transfer being the same as or perpendicular to the direction of motion of the particles. For solid bodies, both longitudinal and transverse waves occur, and the propagation of waves changes at medium boundaries or inhomogeneous points in the medium. When a wave reaches an inhomogeneous point, the most common case is that a part of the wave will continue to travel with or without a change of direction, and another part will be reflected. This phenomenon is exploited in the proposed test method. As there are several inhomogeneous points in a damaged insulator, the signal shapes detected by the sensors on the insulators will differ in the case of an intact and a damaged insulator. Mechanical waves and vibrations are similar in that both phenomena involve energy transfer, but at the structural level, the way in which energy is transferred differs. In our case, the easiest way to distinguish between them is to refer to them as mechanical waves between the moment of impact and the moment of detection of the wave and as vibrations after detection. Later in the manuscript, it is described that two sensors are placed on the insulator, one at the point of impact and the other at the far end of the insulator, so that the difference in the detection times of the two sensors gives us the propagation time that will help us distinguish between them.

Ultrasound scanning is a well-known technique in internal medicine and a perfect example of a technology that exploits wave propagation [18]. The vibration testing method could also sound familiar as it is used in many areas of the industry and also in many research areas, such as medical [19] and engineering [20]. One of its applications is in modal analysis [21], where the aim is to analyze and validate the designs of different parts of vehicles, such as aircraft, to ensure they are not sensitive to applied forces that might induce harmful or destructive resonant frequencies without damping. The method is also used to test bridges [22] and buildings [23] for structural damage. To the authors’ knowledge, vibration testing has not been attempted on composite insulators previously. However, a slightly similar study is known. In article [24], vibration testing was conducted on GFRP rods used in building construction. The prominent similarity in our research is that the core of a composite insulator is also a GFRP rod. However, in our case, the core is not tested separately but as part of the insulator. Also, the testing method, frequency range, and method of excitation are different.

2. Materials and Method

2.1. Materials

Four long-rod insulators, designed for 220 kV networks, were examined in detail during the measurements. The insulators have been removed previously from the network for research purposes. They are identical, except that two of the four insulators have one of their end fittings rotated 90° relative to the other two insulators’. This is the sole distinction between Type 1 and Type 2 insulators. Each insulator is 214 cm long and was in good condition before the tests, with no signs of damage. The only tell-tale sign of the time they spent in service is the dirt build-up on the insulators’ covers. The insulators consist of three main parts, each with a distinct material composition: the end fittings are made of cast iron, the cover is silicone rubber, and the core is an epoxy resin-based glass-fiber-reinforced composite.

Various instruments are available on the market to generate, detect, and process mechanical waves, such as impact hammers, shakers, special sensors, and amplifiers. In this study, we used inexpensive and readily available tools where we could. We used a conventional Black & Decker hammer (Hampstead, MD, USA) (175 g) to excite the mechanical waves manually. To detect the excited signals, we used two dynamic microphones as sensors. The microphones are manufactured by Nedis and have a vendor part number of MPWD01BK (NEDIS BV, ’s-Hertogenbosch, The Netherlands). They feature a sensitivity of −75 dB (+/− 3 dB) and are responsive to vibrations within the frequency range of 80 to 12,000 Hz. For the measurements, the microphones were disassembled and only their sensors were used in the measurements, as shown on the right side of Figure 4. To amplify the signals, we also used a commercially available XENYX 1002 FX (Willich, Germany) mixing desk, which offers 130 dB of dynamic range and a bandwidth that extends from below 10 Hz to above 200 kHz. To observe the signals during measurement and save the data for further analysis, we used a Tektronix TBS 1052B-EDU (Beaverton, OR, USA)digital oscilloscope with a vertical resolution of 8 bits, a bandwidth of 50 MHz, and a sampling rate of 1 GS/s. The signals were further processed on a computer using MATLAB (R2023b). Since the tests were performed with readily available, inexpensive equipment, the accuracy of the sensors is presumably different. Furthermore, the excitation was carried out manually. Hence, the excitation magnitude varies between measurements. These should be taken into account during the evaluation of the results.

2.2. Methodology

In general, vibration test setups are divided into two categories. In the first one, the service conditions of the device under test are reproduced. In the second one, the device under test is examined in a so-called “free state”, where external influences are excluded. In the present study, the latter was chosen to filter out the vibrations that would originate from the conductor that is held by the insulator and from the transmission tower that holds the insulator. This arrangement was chosen to facilitate testing of the method’s basic applicability and lay the foundation for a future study that will test the method under field conditions. This “free state” was created by attaching the insulators to the laboratory’s floor using flexible ropes and the crane hook hanging from the laboratory’s ceiling. The insulator was thus positioned vertically during the tests. A sensor was attached to each end fitting of the insulator with cable ties in a fixed position where they were aligned with each other. The sensors are connected to the mixing desk (amplifier), and the outputs of the mixing desk are connected to the oscilloscope (real-time display and data acquisition). A schematic sketch of the measurement is shown in Figure 3.

The measurement procedure after the assembly was as follows: the oscilloscope was set to single-shot mode, with the recording criterion set to the rising edge of the signal detected on sensor 1, which nearly coincides with the point of impact. Then, using the hand hammer, the end fitting of the insulator was tapped (excitation). The insulators were excited on their end fittings from two different angles. In the first case, the impact was on the opposite side of the insulator from the sensor; in the second case, the impact was on the very end of the insulator end fitting. In the first case, the aim was to generate transverse waves and in the second case, longitudinal waves. Following the excitation, the oscilloscope displayed the signal waveforms, which were saved on a pen drive for later analysis on a laptop. The photos taken during the measurement and the laptop that was later used to process the data are shown in Figure 4.

The proposed method was used to test all four insulators in their intact state, using both excitation points. As the core of an insulator can currently only be diagnosed non-destructively at a very high cost, even in a laboratory, we have assumed that the cores of the insulators were originally intact. This was based on the intact appearance of the insulators’ cover and the fact that the insulators had been in service for several years without failure and were only removed from service for research purposes. Therefore, we artificially produced the damaged samples by drilling small holes into the insulators at a distance of 67 cm from the lower end fitting. The choice of spacing was arbitrary, but care was taken to ensure that they were placed in the same location on the different insulators. The position of the hole is likely to affect the results, and the insulators may have had some minor defects in their cores when they arrived at the laboratory. However, their effects are negligible, as the results are evaluated by comparing the different stages of the destruction process. The drilling was carried out in several stages in the hope of obtaining a continuous picture of the progressive degradation of the insulators and an answer to the question of what the minimum damage is that such an inspection method can detect. In the first stage, the insulator was intact; in the second, a hole with a diameter of 3 mm was drilled halfway through the core of the insulator; in the third stage, the previously started hole was drilled through the core; and in the fourth stage, a hole of 5 mm was drilled through the insulator, as shown in Figure 5. During the drilling process, the insulator’s sheath also degrades, but this does not affect our measurement results because the silicone sheath is highly effective at damping mechanical waves and vibrations; therefore, reflections originating from the insulator’s sheath are already negligible.

We investigated the propagation of the excited waves, focusing on the characteristics of the front of the wave and the propagation speed of the waves. But we have also put a great emphasis on investigating the subsequent vibrations after the sensors have detected the waves. The artificial degradation was carried out on two insulators, on which the measurements were repeated after each degradation stage. In every single scenario, the measurements were repeated three times. The signals downloaded from the oscilloscope were recreated using MATLAB and examined in the time and frequency domains. The signal waveforms were examined in terms of propagation time, amplitudes relative to each other, signal phase, and different frequency components. To facilitate this, we used the time domain curve measured by the oscilloscope and the frequency curve obtained from it with FFT (fast Fourier transformation). We also constructed the cumulative energy curve of the signals. The latter was constructed using a simple formula:

$$E\left(t\right)={{\displaystyle \int }}_0^{t_{end}}{U}^2\left(t\right)dt$$

(1)

where $U \left(t\right)$ is the signal’s amplitude at a given time, $t$, and $U^2\left(t\right)$ is its square.

Looking at all the cases from different scenarios—the combination of different insulators, excitation points, degradation stages, and duration of the time investigated—gives us a staggering amount of different curves (360) to compare. Naturally, the repeated measurements can be averaged, which reduces this number to 120, and several curves can be plotted on one graph. However, it is still impossible to present all the cases in one article, so the authors have selected the cases where the most significant differences can be observed. For example, no significant differences were found when testing the same type of insulators in their intact conditions, so we focused on the results of only one insulator from each type, which we also constantly damaged. In the following figures, the signal names are called 1, 2, 3, and 4; the numbers represent the insulators’ current stage of the deterioration process, according to Figure 5.

3. Results

3.1. Examination of the Front of the Excited Mechanical Waves

In this section, the propagation of the waves, the front of the wave, and some subsequent oscillations are investigated over a period of 5 ms. The sampling rate of the oscilloscope at this time base allows for the frequency components to be investigated up to 12 kHz with the help of FFT. However, the figures may show smaller ranges, where significant difference is only observed for a small range.

We started these tests by measuring the propagation time of the mechanical waves excited in the insulators. Since sensor 1 almost coincides with the excitation point and sensor 2 is located at the far end of the insulator, the propagation time is effectively the same as the time between the detection times of sensors 1 and 2. Our experience has shown that the propagation speed differs depending on the direction of the excitation. When the excitation coincided with the axis of the insulator, the measured values showed significant variance, ranging between 380 and 500 µs. Additionally, these values deviate significantly from the theoretical value we calculated (ca. 620 µs) using reference [25]. Specifically, we assumed a propagation speed of 3070 ${\displaystyle \frac{m}{s}}$ for the composite rod core and 5600 ${\displaystyle \frac{m}{s}}$ for the cast iron end fittings, considering the full insulator length of 214 cm, with 26 cm on each end occupied by end fittings, leaving a core length of 162 cm. To verify this result, we recalculated the theoretical propagation speed for the composite core based on its mechanical properties using the formula

$$v=\sqrt{{\displaystyle \frac{E}{\rho }}},$$

(2)

where $E$ is the elastic modulus and $\rho$ is the density of the material. For the calculation, we assumed E = 20,000 MPa and $\rho =1980{\displaystyle \frac{\mathrm{kg}}{m^3}}$ based on [26]. The result of (2) is $3178 {\displaystyle \frac{m}{s}}$, which is very close to the previously used 3070 ${\displaystyle \frac{m}{s}}$. It is crucial to note that the actual velocity in composite materials may vary significantly due to factors such as specific composition, microstructure, fiber orientation, porosity, and temperature. Nevertheless, the observed discrepancy is primarily attributed to the fact that the epoxy-based GFRP rod behaves as a dispersive medium. In dispersive media, wave components at different frequencies propagate at different speeds. In cases of normal dispersion, lower-frequency components travel faster. Given that our analysis was conducted within the audible frequency range (limited to 12 kHz by our sensors), dispersion likely contributed to the observed faster propagation speed. In reference [24], a propagation speed of 5171 ${\displaystyle \frac{m}{s}}$ was observed at lower frequencies under laboratory conditions. Using this propagation speed for the calculation, the expected propagation time would be approximately 401 µs, which aligns with our measurements. Due to these factors, as well as the substantial differences in propagation speeds we observed during repeated measurements in identical cases—and notably, the lack of significant differences between intact and damaged insulators—we concluded that propagation speed in this setup is unsuitable as a diagnostic tool for distinguishing between intact and damaged insulators. Consequently, we used the propagation speed solely to time-shift the data of sensor 2, enabling easier graphical comparison with signals detected by sensor 1. When the excitation was perpendicular to the axis of the insulator, we observed an even larger variance in propagation times, but other than that, the conclusions are the same as in the previous case.

Following the investigation of the propagation time, we started to compare the different signal shapes (time domain, spectrum, cumulative energy). During the analysis of the signals measured in the 5 ms timeframe, a significant difference between the signal shapes (beyond the difference in the amplitudes) was found only in the case of the data recorded for axial excitation. Thus, only those cases will be discussed in this part. Later, however, when the entire signal waveform was investigated, we were able to detect differences for perpendicular excitation as well, but those were outside the 5 ms time range that is examined here.

3.1.1. Examination of Insulator Type 1

In the case of insulator Type 1, we found discrepancies at several points. Unfortunately, the correlation between the condition of the insulator and the differences in the waveforms was not always clear. However, partial conclusions could be drawn in such cases by comparing different scenarios. An example of this is the testing of insulator Type 1 in the case of axial excitation. If we look at the cumulative energy curves, shown in Figure 6, it can be seen that the curve of signal 1 is more undulating on sensor 1, and the curve of signal 3 is more undulating on sensor 2. However, we cannot draw a conclusion from this alone. In Figure 7, the signals are shown in the time domain. What is mainly noticeable are the signals with different amplitudes; unfortunately, this is due to differences in the excitations, which also makes it impossible to judge the waviness of the signals.

However, by examining the spectra of the signals, shown in Figure 8, we can see the cause of the ripples on the cumulative energy curves. Signal 1 detected on sensor 1 shows significant frequency components between 8.4 and 9.5 kHz that do not appear in the case of the other signals. This is the most significant development in this series of measurements because we observed similar results when we tested insulator Type 2 in the same scenario. The cause of the ripple of signal 3 can also be seen in the spectral curves of sensor 2 where a significant frequency component appeared between 800 and 1500 Hz. This is also a frequency range that has repeatedly proved interesting during the analysis, but this is the only case where only signal 3 has deviated so much from the others.

3.1.2. Examination of Insulator Type 2

During the analysis, a correlation was found between the insulator’s condition and the waveforms’ deviation when an axially directed mechanical wave was excited on insulator Type 2. In this case, it can be observed on the cumulative energy curves in Figure 9 that on sensor 1 that signal 1 has the lowest energy out of the four cases, while on sensor 2, it is the opposite, suggesting that the inhomogeneity caused by the damage in the insulator refracts the waves associated with signals 2, 3, and 4 and more energy is reflected toward sensor 1 than that which reaches sensor 2.

Only minimal amplitude differences can be observed when the signals are examined in the time domain. However, in the frequency domain, it can be seen that the components between 500 and 1000 Hz differ significantly and that signal 1 has frequency components that the other signals do not. These components fall between 7.5 and 12 kHz, as seen in Figure 10.

3.2. Examination of the Complete Waveform

In this section, the signal shape of the excited wave is investigated up to 0.25 s, which in most cases was enough to allow for the waveform to decay completely. The sampling rate of the oscilloscope at this time base allows for the frequency components to be investigated up to 5 kHz with the help of FFT. However, the figures may show smaller ranges, where a significant difference is only observed for a small range.

3.2.1. Examination of Insulator Type 1

Differences in waveforms of the different states of insulator Type 1 were also observed in the 0.25 s timeframe. For excitation in the axial direction, the spectra of the signals detected by sensor 1 showed differences around 500 Hz, which were seen in several cases. In this case, no clear conclusion can be drawn from the cumulative energy curves or signal waveforms in the time domain.

When tested in the 0.25 s range, the most significant difference between the insulator states was obtained when we tested insulator Type 1 with perpendicular excitation to its axis. The cumulative energy curves, shown in Figure 11, clearly show the difference since the energy of signal 4 detected by sensor 1 (fixed to the excited end fitting) was the highest, while the energy of signal 4 detected by sensor 2 (fixed to the other end fitting of the insulator) was almost the lowest. This is similar to what we have seen before in Section 3.1.2.

This can be explained by the fact that the excited wave is refracted (multiple times) across the damage we caused, and the reflections that travel toward sensor 1 have more energy than those that travel toward sensor 2. Compared to 3.1, considering a larger time range, the difference can also be identified in the time domain. As shown in Figure 12, the signals detected by sensor 1 are almost identical in all four cases initially, but signal 4 starts to deviate significantly from the shape of the other signals around 0.052 s. This occurs at about 0.0025 s after the mechanical wave is detected, which is sufficient time for the wave to pass from one end fitting to the other and back several times (about 0.88 ms). The spectrum of the signals detected by sensor 1 is shown in Figure 13, where it can be noticed that around 500 Hz, signal 4 has a significantly for frequency component than the other signals. The divergence is also observed on signals 2 and 3, but the magnitude is less significant.

3.2.2. Examination of Insulator Type 2

Differences in the waveforms of the different states of insulator Type 2 were also observed in the 0.25 s timeframe. For excitation in the axial direction, the spectra of the signals detected by sensor 1 showed differences around 500 Hz, similar to those observed in the case of insulator Type 1, but no further findings could be made.

However, in the case of perpendicular excitation, we had a different observation. On sensor 1, the waveforms in the time domain were almost identical, with minor amplitude variations. However, in contrast, for sensor 2, the signal waveforms were significantly different, as shown in Figure 14. The difference also appears on the spectra of the signals, mainly within the range of 80–180 Hz, as seen in Figure 15.

4. Discussion

In total, 28 cases were examined, varying in terms of time intervals, insulator types, measured parameters, and excitation methods. A total of 40 unique measurement sets were conducted, each repeated three times, resulting in 120 measurement series and 360 data curves suitable for comparison. The most significant cases are summarized in four points below, categorized by the combination of examined time intervals and the excitation methods.

When the front of the mechanical wave was investigated in the 5 ms time range, no significant difference was found for excitation perpendicular to the insulator’s axis. For a similar excitation, when tested in the 0.25 s time range, a difference was found, but outside the 5 ms time range. This suggests that although the initial mechanical wave is less affected by defects in the case of perpendicular excitation, they become more significant during the complete decay of the vibration phenomenon.

Nevertheless, in the 5 ms range, when the mechanical waves were excited in the direction of the axis, the results of insulator Type 2 were correlating with the state of the insulators. This could be seen from the variation in the cumulative energy curves and the variation of the frequency components of the spectrum. In the case of insulator Type 1, the amplitudes of the excitations were very different, so no correlation between the cumulative energy curves and the deteriorating state of the insulators could be found, but the spectra of sensor 1 show that the same components are missing in signals 2, 3, and 4 as in the case of insulator Type 2 related to signal 1. Here, however, a component appeared in signal 3 that was not observed elsewhere. It can also be concluded from the 5 ms tests that the propagation time is not affected significantly by the damage done to the insulator.

When we examined the waveforms in the 0.25 s time interval, the spectrum could only be produced up to 5 kHz due to the sampling limitations of the oscilloscope. This limits the range of the frequency domain test compared to before. However, more significant differences could be found by examining the signals over the time domain. In the 0.25 s range, in contrast to the 5 ms range, more significant differences were found when we induced the waves perpendicular to the insulator’s axis compared to when we induced them in the same direction of the insulator’s axis. When the waves were induced in the same direction as the axis, we observed differences in the cumulative energy and time domain curves for both insulator Types 1 and 2, but we could not clearly relate them to the state of the insulators. However, the spectra showed that the frequency components of the measured signals changed around 500 Hz after the degradation of the insulators in both cases.

When the waves on the insulators were excited perpendicular to their axis, we found that the signals on sensor 1 were the same regardless of the insulator’s condition, while on sensor 2, the signals were initially the same but became more divergent over time. This naturally affected the cumulative energy curves as well as the spectra. In the case of the spectra, we, again, found differences in the components around 500 Hz for insulator Type 1, and between 80 and 180 Hz for insulator Type 2. The deviations were noticeable on sensor 1 for insulator Type 1 and sensor 2 for insulator Type 2, possibly due to the different end fittings.

5. Conclusions

This study explored a diagnostic method for detecting defects in the cores of composite insulators based on the analysis of mechanical waves and vibrations. By investigating the propagation, signal shape, and distortion of excited mechanical waves and subsequent vibrations, the study aimed to distinguish between damaged and intact insulators using inexpensive tools.

It was possible to distinguish between intact and damaged insulators by the cumulative energy curves, the signal waveforms in the time domain, and, most often, by the spectrum of the signals. Based on the results, the most critical frequency range is between 500 and 1500 Hz, since the deviations in the spectra are most significant here. However, there were also examples of deviations in other frequency ranges. Although cumulative energy curves and time domain signals have been used in many cases to distinguish between damaged and intact insulators, these curves are more influenced by the magnitude of the excitation, which, in our case, poses a problem as we induced the mechanical waves manually. However, the resulting inaccuracy can be eliminated if the excitation is left to a device.

It should also be remembered that the insulators were roughly damaged during the tests, and the aim was to detect more minor defects in the field. Fortunately, the tests were carried out with very low-end instruments, so the method’s sensitivity can be easily increased. However, for this reason, the limits of the method are still unknown.

Nevertheless, our findings indicate that this method could lead to a practical, on-site, non-destructive tool for detecting core damage in composite insulators under real-world conditions. However, further testing is needed, especially in field conditions, to fine-tune the method and confirm its effectiveness across different defects and insulator types.