Analysis of deterioration characteristics in oil-immersed insulation pressboard with different durations of aging based on an image-processing method

Abstract

Insulation pressboard samples were obtained by thermal aging (according to Montsinger’s formula, at 130 °C, the pressboard is heated for 0 to 32 days) and discharge experiments. SEM images of samples were analysed. Image segmentation was applied to calculate the fibre width, cross-sectional porosity, and carbon-trace area. Inter-layer fibre models were established to observe fibre morphology using 3-D reconstruction. The initial discharge voltage decreased with age, and the discharge amounts increased. After 16 days of aging, the fibre width had decreased to between 68.1% and 81.8% of unaged pressboard. As the aging increased, cellulose hydrogen bonds were broken, which affected the expansion of interlayer pores, increasing the porosity of the pressboard. After 32 days of aging, the porosity increased to 2.38 times that of a new pressboard. In addition, the longer the aging, the larger the area of carbon marks caused by the discharge breakdown. With the aggravation of thermal aging, the insulating property of pressboard decreased due to the decrease of fibre width and increase of porosity that further accelerated the damage to the fibre structure. It was concluded that the fibre width and porosity could be used as criteria to judge the degradation of pressboard.

Full Article

Analysis of Deterioration Characteristics in Oil-immersed Insulation Pressboard with Different Durations of Aging Based on an Image-processing Method

Yongqiang Wang,* Ruoyu Fei, Changhui Feng, and Jing Shang

Insulation pressboard samples were obtained by thermal aging (according to Montsinger’s formula, at 130 °C, the pressboard is heated for 0 to 32 days) and discharge experiments. SEM images of samples were analysed. Image segmentation was applied to calculate the fibre width, cross-sectional porosity, and carbon-trace area. Inter-layer fibre models were established to observe fibre morphology using 3-D reconstruction. The initial discharge voltage decreased with age, and the discharge amounts increased. After 16 days of aging, the fibre width had decreased to between 68.1% and 81.8% of unaged pressboard. As the aging increased, cellulose hydrogen bonds were broken, which affected the expansion of interlayer pores, increasing the porosity of the pressboard. After 32 days of aging, the porosity increased to 2.38 times that of a new pressboard. In addition, the longer the aging, the larger the area of carbon marks caused by the discharge breakdown. With the aggravation of thermal aging, the insulating property of pressboard decreased due to the decrease of fibre width and increase of porosity that further accelerated the damage to the fibre structure. It was concluded that the fibre width and porosity could be used as criteria to judge the degradation of pressboard.

Keywords: Thermal aging; Oil-immersed insulation pressboard; Microscopic morphology; Image segmentation; Fibre

Contact information: Hebei Provincial Key Laboratory of Power Transmission Equipment Security, Department of Electrical Engineering, North China Electric Power University, Baoding 071003, China;

* Corresponding author: qianghd@126.com

INTRODUCTION

Insulation paper is a solid organic insulation material that is widely used in power transformers. However, during long-term transformer operations, high-temperature exposure causes insulation paper to become thermally cracked, and the paper gradually deteriorates (Darveniza et al. 1992; Lundgaard et al. 2004). To address this concern, extensive research has been conducted (Joshi and Bhanot 2005; Joshi et al. 2006; Wang et al. 2018a). Simultaneously, studies have shown that cellulose, the main component of pressboard, undergoes a depolymerisation reaction and an elimination reaction under the action of heat. These reactions produce many derivatives, and the resulting derivatives will lead to other more complex chemical reactions, resulting in continuous changes in the fibre structure. When judging the aging state of insulating paper, researchers initially measure the insulation paper tensile strength (TS) to determine its condition. Secondly, people also use the degree of polymerisation as a parameter indicative of aging (Emsley et al. 1997). With improvement in the accuracy of observation equipment, researchers have begun to use atomic force microscopy (AFM), scanning electron microscopy (SEM), and other methods to study the aging state of insulating oil paper from a microscopic perspective. Through analysing insulating paper of varying age, Carrascal et al. (2018) obtained information on the kinetics of their aging degradation. Li et al. (2017) studied the changes of moisture, acid value, and furfural in insulating grease paper under thermal aging.

Image processing includes image enhancement and image segmentation, including wavelet transform (Kim et al. 2016), threshold method (Otsu et al. 1979), boundary detection method, area method, etc. At present, image processing technology is widely used. Researchers use mature image processing software or mathematical function software for image processing. They are involved in the study of insulators (Pirie et al. 2020) and transformers (Mariprasath and Kirubakaran 2018), but they are rarely used in the analysis of insulating pressboard microstructure. Liao et al. (2008, 2011), by analysing SEM images, found that thermal aging resulted in cracks at the fibre surface, fibre peeling, cracking, pitting during discharge. Yan et al. (2011, 2012) used AFM to analyse the molecular chain changes on the microscopic surface of fibres. It was found that the oil-immersed paper surface consists of droplets and crystalline solids, which develop over time as damage accumulates. These conclusions are mainly through direct observation of SEM images. That is, there is insufficient statistical information available, and more clear data are needed to reach conclusions.

In this study, SEM images of pressboard under different aging conditions were observed using image processing techniques. Meanwhile, a three-dimensional model of the fibre was established to analyse the internal fibre morphology of the insulating pressboard. The microscopic information before and after breakdown of the needle plate discharge model at different thermal aging stages was calculated. The main purpose was to obtain data related to the deterioration of the pressboard. Through the calculation of the data, the deterioration degree of the pressboard can be quantitatively evaluated, and the trends of micro-morphology feature quantity changes, such as fibre width and porosity, can be further analysed. The present work will provide data support and new research ideas for further research on the long-term use ability and value of pressboard.

EXPERIMENTAL

Design of Discharge Experiment of Insulation Pressboard

Samples at these different states were prepared before discharge testing. Standard insulation pressboard, 80 mm × 80 mm × 2 mm, and Karamay 25# transformer oil (Sinopec Lubricants Co., Ltd., Beijing, China) were used for this experiment. The insulating pressboard was made from ultra-high voltage transformer pressboard (Weidmann Electrical, Co., Ltd., Rapperswil-Jona, Switzerland) comprised mainly of wood pulp, which is primarily composed of cellulose, hemicellulose, and lignin. The density of the insulation paper used in this paper was 1.20 g/cm³, the relative dielectric constant (before the experiment) was 3.59, and the insulation resistivity was 6.25 × 1013 Ω·m. The specific preparation process was as follows: an appropriate amount of pressboard and Karamay 25# transformer oil were placed in a vacuum drying oven at 90 ℃ and 50 Pa to dry for 24 h to remove water and air bubbles in the pressboard. Subsequently, the temperature was adjusted to 80 °C, the pressboard was immersed in transformer oil, and vacuum impregnated for 12 h. The pre-treated oil and pressboard were blended at a mass ratio of 10:1 and subjected to accelerated aging at 130 ℃ (Wang et al. 2018b). The aging time was calculated using Montsinger’s rule (Tu et al. 2016; Zhou et al. 2017),

 (1)

where C is the reference operating temperature (80 °C) of the insulating material, T₀ is the insulation life at the reference working temperature, θ is the actual working temperature of the insulating material (130 °C), T is the insulating life at the actual working temperature, and α is the thermal aging coefficient. In Chinese transformers, the thermal aging factor α is generally 0.1155. In this paper, considering the safety and efficiency of the experiment, selected pressboards were aged for 0 days, 2 days, 4 days, 8 days, 16 days, and 32 days, which are equivalent to the normal operation of the transformer for 0 years, 1.77 years, 3.53 years, 7.06 years, 14.12 years, and 28.24 years, respectively. After aging, the insulation pressboard was immersed in filtered and purified new oil for two days. Subsequently, the vacuum was extracted three times to fully aspirate the aged oil in the pressboard.

The discharge device consisted of a needle electrode with a length of 30 mm, a width of 1 mm, a tip curvature of 30°, and a brass plate with a width of 100 mm and a thickness of 20 mm. The whole experimental platform is shown in Fig. 1. It was comprised of a booster console, an experimental transformer, a protective resistor, a coupling capacitor, and a smooth aluminium conductor. Among them, the non-corona testing transformer was a YDTW 50/250-type oil-immersed transformer (Wang et al. 2019a).

The experiment was conducted using the constant voltage method. When testing with high voltages, pressure measurement was completed at a step rate of 1 kV/s. When the signal of the PD tester was stable, the voltage was stabilised for 5 min, and this voltage value was noted as the initial discharge voltage U₀. It was understood that U = 1.2U₀ and U was applied as a constant voltage on the pressboard until the pressboard broke down.

Fig. 1. Discharge experimental platform

SEM Image Processing

The objective guiding the extraction of the fibre profile from the SEM image was the identification of the dividing point between the fibre profile and the background, and the dividing points between fibres. First, the image was converted into a greyscale image. After calculating the histogram of the greyscale image (Fig. 2b), the OTSU algorithm was used to separate the fibres from the background.

The OTSU (Du et al. 2019) functions by using the characteristics of the grey image to divide the picture into two parts: the background and the foreground. In the binarized image, the segmentation threshold between the fibre and background is T, the ratio of the number of pixels in the fibre N₀ to the entire image is ω₀, the average greyscale is μ, the ratio of the number of pixels N₁ in the background to the total pixels is ω₁, the background average greyscale is μ₁, the classes square error is g, the pixel points of the image length and width are respectively M and N, whereby:

If the g was maximized and the threshold T was the required value, then the first divided image f(x,y) was obtained.

After the fibre and background segmentation were divided, the watershed image segmentation algorithm based on top-hat transformation was selected for the second segmentation, and improved the algorithm using multi-structure element morphology.

The second segmentation was performed using a watershed segmentation algorithm:

where  is the corrosion operation symbol,  is the expansion symbol, and b is composed of three basic structural elements:

Substituting b₁, b₂, and b₃ into Eq. 4 to obtain G’₁, G’₂, and G’₃ and assigning different weights α₁, α₂, and α₃ to obtain G’, gives the final fibre extraction image:

(a)

(b)

(c)

Fig. 2. SEM image processing: (a) SEM original micrograph of insulation pressboard 500x, (b) greyscale histogram, and (c) results of fibre contour image segmentation of insulation pressboard

The segmentation results of the original SEM images (Fig. 2a) using the above segmentation algorithm are shown in Fig. 2c.

Measurement Parameters

The previous study (Wang et al. 2020) described the method of calculating the fibre width, porosity, and other parameters using image processing, which is also applicable to the study of the deterioration of insulating paperboard in different aging states.

Fibre width

The MATLAB™ tool-kit can be used directly to measure fibre width, but this method is inaccurate. The pressboard fibre tends to curve and its thickness is unevenly distributed. Thus, the central axis method was used to measure the fibre width.

Porosity

By extracting the outline of the pores in the cross-sectional image on this side, it was possible to calculate the porosity (the proportion of the pore area in the cross-section), and the internal cross-sectional structure changes of the pressboard was analysed.

Carbonisation mark area

The extent to which the tip discharge was destructive to the pressboard was determined by considering the carbonisation mark area. In this paper, the outline of Carbonisation mark was extracted, and the area was expressed by the number of pixels.

3-D reconstruction of insulation pressboard fibre

In a previous study (Wang et al. 2020), the authors used a three-dimensional model to observe the degradation characteristics of fibres at different discharge stages. In the present study, the authors adopted the same method to extract the cross-sectional images of the fibre sequence to construct a 3-D model of the fibres under different aging conditions.

RESULTS AND DISCUSSION

Partial Discharge Experiment Analysis of Aged Insulation Pressboard

In the present research, the constant voltage method was used to conduct long-term pressurisation experiments on insulation pressboard with different lengths of aging. The initial discharge voltage and tolerance time of each pressboard are shown in Table 1, which demonstrates that the high temperature accelerated the aging of the insulation pressboard. As the aging of the insulation pressboard increased, the tolerance time of the test sample decreased. After 32 days of aging, the initial discharge voltage of the pressboard decreased 8.3%, and the tolerance time decreased 27.7% (compared with untreated pressboard).

Table 1. Statistical Data: Initial Discharge Voltage and Discharge Tolerance Time of Insulating Pressboard at Different Aging States

Figure 3 demonstrates the curve of the maximum PD quantity of the pressboard at different aging states with the tolerance time under the needle-plate model. In the initial stage (1 to 4 h), the PD amount of each sample was similar. With increased pressing time, the maximum PD quantity of each sample showed an upward trend before breaking down, but the effect differed in each group of specimens. The maximum PD growth rate of samples aged for 0 to 4 days was relatively stable without much fluctuation. The discharge amount of cardboard aged for 8 to 16 days exhibited a large fluctuation in the middle of the discharge. The increase rate of discharge in 6 to 7 h was significantly higher than that in the previous period. Although the overall trend was upward, the discharge amount decreased significantly in the period 7 to 12 h. The maximum PD quantity of 32-day aged pressboard rapidly increased after 4 h, and the growth rate was much higher than that of other samples. Overall, in the initial stage of discharge, the pressboard’s duration of aging had no notable effect on the maximum PD quantity, and the difference in the maximum PD quantity of each sample was small. In the middle stage of testing, the maximum PD quantity of different samples gradually varied more widely. In the later stage, on the one hand, the difference between the maximum PD of each sample increased. However, with increased aging, the maximum PD increased and breakdown occurred earlier.

Fig. 3. Variation of maximum discharge from insulating pressboard at different aging states

Change of Fibre Width Before and After Discharge Testing of Insulation Pressboard at Different Aging States

The microstructure of insulation pressboard with different durations of aging before and after discharge testing was observed by scanning electron microscope (Figs. 4 and 5).

The image processing method introduced in the “Experimental” section was used to extract the fibre contour in different states, and the fibre width was measured. The results are illustrated in Figs. 6 and 8. The fibre width was mainly within the range between 20 μm and 44 μm, and this was divided into 14 intervals (Table 2).

Table 2. Fibre Width Intervals

The following rules were deduced from observation of the data :

(1) There were fibrils on the surface of the unaged pressboard. After 2 to 4 days of aging, the fibrils disappeared and the surface became smoother, but compared with the unaged pressboard, the whole remained neatly aligned, and the fibre width was mainly distributed between 40 and 44 μm. After PD, the rate of decrease in the fibre width of the pressboard underwent a marked change compared with that before discharge. In pressboard aged for 4 days after breakdown, the number of fibres with a width of less than 40 μm increased notably.

(2) The fibre width of 8-day-aged cardboard decreased slightly compared with that aged for 2 to 4 days. The fibres were scattered, but no obvious holes appeared on the fibre surface. After breakdown, the fibres gradually polymerised and fractured, and the width decreased to a narrow range between approximately 32 and 34 μm.

(3) After 16 days of aging, some fibres were bent and a few fibres had exfoliated, causing the roughness to increase. Most fibres were between 30 and 36 μm in width (68.1% to 81.8% of untreated cardboard). After breakdown, the fibre width was mainly distributed in the range between 24 and 26 μm (54.5% to 59.1% of untreated paperboard), and there were obvious vestiges of cauterisation visible.

(4) After 32 days of aging the cardboard, the fibre width was concentrated at approximately 30 μm, a small number of the fibres had broken, and holes appeared on the fibre surface. After the breakdown, there were almost no fibres with a width greater than 40 μm, and the proportion of fibres less than 20 μm increased notably. Fibres with a width of 24 to 26 μm accounted for the highest proportion of the total.

(5) With increased aging, the width of the fibre gradually decreased. However, after 16 to 32 days of aging, the downward trend decelerated. After PD, the fibre deterioration was noticeably worse than that before breakdown.

Fig. 4. Scanning electron microscopy images of insulation pressboard before discharge (500x magnification): (a) SEM image of unaged pressboard, (b) SEM image of pressboard aged 2 days, (c) SEM image of pressboard aged 4 days, (d) SEM image of pressboard aged 8 days, (e) SEM image of pressboard aged 16 days, and (f) SEM image of pressboard aged 32 days

Fig. 5. Scanning electron microscopy images of insulation pressboard after discharge (500x magnification): (a) SEM image of unaged pressboard, (b) SEM image of pressboard aged 2 days, (c) SEM image of pressboard aged 4 days, (d) SEM image of pressboard aged 8 days, (e) SEM image of pressboard aged 16 days, and (f) SEM image of pressboard aged 32 days