Atmospheric weathering and corrosion, in a tropical country like Brazil, on metallic maintenance of power transmission lines

Atmospheric weathering and corrosion, in a tropical country like Brazil, on metallic maintenance of power transmission lines Fernando DINIZ*¹, Thiago Luiz FERREIRA¹, Euro Pinto de ALMEIDA², Alberto Rodrigues de SOUSA¹, Camila Marçal Gobi PACHER³, Julia Stefany Chagas ALBRECHT³, Mariana D’Orey Gaivão Portella BRAGANÇA³, Kleber Franke PORTELLA³, Juliano de ANDRADE³, Bruno Cougo KOWALCZUK³, Maurício Marlon MAZUR³. ¹ARGO Brazil [email protected], [email protected], [email protected] ²FREELANCE Brazil [email protected] ³LACTEC Brazil camila[email protected]m.br, [email protected], [email protected], [email protected], [email protected], [email protected], [email protected] SUMMARY In Brazil, with its almost 10,000 km of coastal zone, quality controlling the metallic materials needed for energy transmission becomes a major challenge because of the high corrosion risk. The Brazilian electricity sector had about 6,200 km of new lines concluded throughout the territory, in 2020 alone. These lines are specially in shore areas, once almost 27% of the population lives. Several factors interfere in the transmission line design phase, such as specifying metallic materials compatible with the environmental aggressiveness. Furthermore, once the corrosion is an uncontrolled condition, developing different, expedite and assertive techniques to diagnose and accompany the durability loss process is necessary. In the north and northeast regions of Brazil, there is also the aggravating factor that the very typical climates cause great metallic corrosion impacts. Mainly on towers and power substation and their accessories areas, throughout the year, there are high ambient temperatures, with average values close and higher to 30 °C. These conditions are also allied to practically two well-defined climatic seasons, the summer (with lower levels of rainfall) and the winter, where there are constant daily rains, followed by large gusts of wind, usually coming from to the mainland. With these characteristics, higher levels of salinity are added to the atmospheric air, typically containing chloride and sulfate ions, causing high rates of corrosivity, from medium until very high corrosion level (C3 to C5+), according to the international standards. 10778 B2 OVERHEAD LINES PS2: Latest techniques in asset management, capacity enhancement, refurbishment. 2024 Paris Session 1 From this context, ARGO and LACTEC carry out a study about corrosion, to develop an experimental and numerical methodology, with artificial intelligence (AI), for the survey of the environmental conditions on the corrosivity rate of the metallic materials. The study has been conducted on the power substations and transmission towers located in the north and northeast regions of Brazil and the premise is the seasonal survey of expedited data on metallic specimens, for instance: tower trusses and cylindrical coupons of the same materials with and without galvanized surface protection. The comparison between the results of collecting samples installed in the field, at heights, where the biggest rates of metallic materials degradation of the power substation towers were observed, and the results obtained, in controlled environments, such as artificial salt spray and SO2 chambers, determine the condition of exposed metallic material and allow the calibration and the application of the AI diagnostic methodology developed and will be support ARGO enterprise to take best assets management decisions. A corrosivity rate was found in the considered regions, compatible with very high corrosion level (C5+), although, basic study premises had applied medium corrosion level (C3). The distance between basic studies, design premises and real conditions found in Northeast was verified. Comparison among samples collected at the most aggressive environmental condition and controlled environment allowed the establishment of a corrosion pattern, that will be used to train AI to identify real condition of metallic structure installed by image survey. This work is going to show partial results gotten up to be published as well as, expectations and next steps for investigation. KEYWORDS Corrosion, Atmospheric pollution, Transmission lines, Galvanized carbon steel, Atmospheric corrosion, Artificial Salt Spray, Electrochemical tests. 2 1. INTRODUCTION Brazil has an extensive network of approximately 180 thousand kilometres of transmission lines (LTs) as of 2023, and there are ambitious plans to augment this infrastructure to over 215 thousand kilometres by 2027. This expansion is a strategic response to meet the growing demand for electrical energy across a vast territorial expanse spanning roughly 8.5 million square kilometres, inclusive of a coastline stretching over 7 thousand kilometres [ 1 ]. While Brazil generally enjoys a favourable climate, recent decades have witnessed discernible climate shifts. Data up to 2015 reveals that about 17% of Brazilian climate seasons have deviated from the typical Köppen patterns, particularly with a regression of humid tropical climates (Af and Am) and temperate climates (C). This alteration has paved the way for the prevalence of tropical climates (Aw), as well as arid and semi-arid climates (B) [ 2 ]. According to Dubreuil et al., 2018 [2], northeastern Brazil has experienced a significant increase in aridity until 2015. This transformation is accompanied by a shift towards rainier climates in the southern Amazon region, signalling profound changes in the country's climate dynamics. These alterations bring significant challenges for engineering and related areas, necessitating proactive measures to minimize the impacts on various materials and metallic structures designed for energy conduction. Protecting thousands of tons of steel and other metals from corrosive processes is imperative, given the influence of factors such as water, soil, air, wind, rain, and sea air, along with their ions, notably chlorides and sulfates, some of which emanated also from industrial sources [ 3 ]. The coexistence of these elements, coupled with consistently high average temperatures, elevated relative humidity, less intense rainfall, and dew, creates ideal conditions for surface electrolytes that promote metallic corrosion processes. Addressing these challenges is important for the long-term sustainability and reliability of Brazil's energy transmission infrastructure. The implications for engineering materials, stemming from the level of environmental aggressiveness, are illustrated in Figure 1, which focuses on determined regions in Brazil. In these areas, corrosion rates on carbon steel (CS) and galvanized carbon steel (GCS) and the average chemical composition of ions present in the environment were systematically measured, encompassing atmospheric parameters, microclimatic, and macroclimatic elements. A more extensive analysis reveals corrosion rates spanning from medium to very high, categorized as C2 to C5+, primarily influenced by the proximity of measurements in coastal, and industrial regions. The presence of particulate materials and aerosols, containing chloride ions and sulfur compounds (H2S, SOx, among others), alongside metallic particles such as aluminium observed in a section of an LT tower in the northeast region of Brazil (near an industrial source), significantly contributed to these corrosion indices. Moreover, as substantiated in documented studies, the consistently high average relative humidity prevalent across much of the Brazilian territory promotes the formation of surface electrolytes. This is intensified by extended periods of wet surface conditions, coupled with average local environmental temperatures exceeding 25 °C. Consequently, these environmental conditions collectively intensify the challenges posed to engineering materials, necessitating careful consideration and strategic material selection in the design and construction processes [ 4 -39]. Ensuring prolonged material durability in challenging environments necessitates a comprehensive approach that extends beyond environmental analysis. Particularly, in regions with heightened aggressiveness, the utilization of materials inherently resistant to corrosive processes or the application of dependable coatings becomes imperative. In all cases, achieving superior surface homogeneity and layer thicknesses tailored to the projected lifespan is essential. A practical illustration of this concept is the utilization of a galvanizing layer, as shown in Figure 2, strategically dependent on the prevailing environment. This coating is employed to mitigate corrosive processes, especially during brief exposure periods. To assess the effectiveness of such protective measures, the Preece test is recommended [ 5 ]. The test results in an oxidation-reduction reaction involving copper ions present in the solution and exposed metallic iron, as described in Equation 1 [ 6 ]. 3𝐶𝑢2+ + 2𝐹𝑒← →3𝐶𝑢 + 2𝐹𝑒2+ Equation 1 To additionally guarantee the durability of metal products exposed to long periods of outdoor operation, it is essential to anticipate all the main external influences, as already mentioned, in addition to the 3 mechanical properties of the materials involved, as they can affect their components. In this sense, it is recommended to evaluate their resistance limits after being subjected to metallurgical processes, such as impacts, bends, holes, coatings, among others, in addition to the construction stages of the tower project [ 7 ; 8 ; 9 ]. By the progression of software and algorithms, these advancements may serve as critical benchmarks influencing the acceptance and deployment of metallic materials in specific environments. Complementing the established characterization methods, procedures involving the integration of images and artificial intelligence (AI) have gained prominence due to their heightened objectivity and rapid response capabilities [ 10 ; 11 ; 12 ]. As a guiding framework, the methodology relies on the decomposition of images into three spectra, namely, hue, saturation, and value, which is known as the HSV method. This not only reflects the dynamic nature of material assessment but also underscores the efficiency and precision achieved through the fusion of image processing and AI technologies. Figure 1. Indicative of metal corrosion processes and environmental aggressiveness in different regions of Brazil [4; 13 ; 14 ; 15 ; 16 ; 17 ; 18 ; 19 ; 20 ; 21 ; 22 , 23 ; 24 ; 25 ; 26 ; 27 ; 28 ; 29 ; 30 ; 31 ; 32 ; 33 ; 34 ; 35 ; 36 ; 37 ; 38 ; 39 ). Figure 2. Corrosion processes in structures and fencing in GCS, including: a) support column for electrical equipment at PS Parnaíba in galvanized steel with the surface showing white Zn corrosion; b) detail of GCS structure with white corrosion and central region with possible red corrosion; and c) galvanized steel fence with widespread red corrosion. To particularize the method for applying a rapid field analysis of installed corrosion, the image recognition criterion therefore corresponds to the definition of appropriate intervals in these aspects that will distinguish the occurrence (or not) of corrosion. Often, establishing the appropriate amplitude for each of the parameters leads to a highly complex task. The decision-making process involves inherent 4 risks, including the potential for subjective choices and a lack of assertiveness in technique application. Addressing these challenges, artificial intelligence (AI) algorithms, exemplified by the "particle swarm optimization" (PSO) method, offer an alternative means for establishing parameters [ 40 ]. Given the continuous variability of values within the Hue, Saturation, and Value (HSV) system, the application of the PSO method proves opportune. Illustrating the effectiveness of this application [ 41 ], a real-case scenario involved the examination of a truss from an electrical transmission line tower in northern Brazil with minimal environmental exposure. The analysis process, facilitated by the PSO method, revealed a local highlight (green tone) and provided an estimation of the affected area with surface corrosion. This application stands as a demonstrative instance, showcasing the potential of the methodology in enhancing precision and assertiveness in image analysis within real-world contexts. The methodology validation resulted in high assertiveness in identifying corrosive processes. A larger area of concentration of greenish colour was identified in portions of the surface, consistent with the appearance of corrosive processes observed. Additionally, despite the occurrence of false positives/negatives (as, for example, evidenced in the through hole on the right side of the corner), in addition to some noise, in the selections provided, represented by the erroneous classification in some regions, the extent to which the test bank data used to calibrate the parameters became more effective with the inclusion of new images, the tendency was to make the appearance of noise less expressive. 2. EXPERIMENTAL Recognizing that there are numerical parameters which can be extracted from the image was useful to better understanding the application of artificial intelligence within the scope of the present study. However, the use of the PSO method to properly identify the surface corrosion ratio of the truss elements was just the starting point. The main goal is far more ambitious. In this context, the use of artificial intelligence techniques, such as the artificial neural network (ANN) method, will play a key role. The ANN method can be regarded as a robust numerical approach on reproducing causality patterns between non-linearly correlated data sets. Since the relation between visual corrosion and HSV parameters of the image was verified, the aim now is to extend this on finding relations between those parameters and the corrosion ratio itself. In other words, one proposes to evaluate if it would be possible to somehow quantify the corrosion through image analysis rather than only visually qualify its occurrence or not. 2.1 Case study and metallic samples In this case study, an electrical transmission line of approximately 1,100 km in length, and power substations (PSs) located between the northeast and north of the country were preliminarily considered (Figure 5). Part of its towers were being georeferenced and inspected for the degree of corrosivity, using a developed systems to expose cylindrical coupons of SAE 1020 carbon steel and of galvanized carbon steel, with a 90-120 µm zinc coating (dimensions: (15 x 80) mm - Figure 4). The samples were preliminarily identified, cleaned with isopropyl alcohol and acetone, washed in an ultrasound bath, and subsequently weighed for initial measurement, before the installation in field. The purpose of this study is to accompany the deterioration process of this samples during a year, attending a complete seasonal cycle. Although, the results presented in this paper comprehends just the first collecting campaign, for which the samples were exposed during the first three months. Laboratory studies were conducted at the same time and over the same kind of samples (Figure 6), with the exposure of metallic test specimens in accordance with the experimental methodologies proposed in the ABNT NBR 8094:1983 [ 42 ] standards, in a BASS model USX chamber; and ABNT NBR 8096:1983 [ 43 ], in a chamber also branded as BASS, model UK-01/2011. The samples were exposed in the artificial salt spray chamber for a period of 750 h, with intermediate evaluation every 250 h. For the sulfate chamber test, the duration of the experiment was 72 h, with evaluation intervals every 24 h, 5 Figure 3. Location and georeferencing of the distance and altitude from the coast of the sampling points on five power substation energy as Bacabeira, Parnaíba III, Tianguá II, Acaraú and Pecém II, obtained with the Google Earth software. Figure 4. Photograph of corrosion test samples of SAE 1020 carbon steel (a) and galvanized carbon steel (b) and its developed system to field exposition (c). Field and laboratory exposed samples were evaluated through gravimetric method to determine the partial corrosion rate, following the precepts of ASTM G1-03 [ 44 ]. Specifically, to the GCS samples, electrochemical tests of open circuit potential (OCP), electrochemical impedance spectrometry (EIS) and cyclic voltammetry (CV) were conducted to obtained the corrosion data, according to international standards of ASTM G5-14, ASTM G59-97 and ASTM G102-89 [ 45 , 46 e 47 ]. For that, it was used a Gamry potentiostat, model Reference 3000, an Ag/AgCl/ KCl 3 M reference electrode, and a graphite bar, as a counter electrode. The electrolyte was a NaCl aqueous solution, at 3.5% by mass. The OCP was evaluated during ten minutes, before the EIS test, which were managed at an amplitude of 10 mV, frequency sweep from 10 kHz to 100 mHz. The CV test was also preceded to another OCP measurement, during five minutes and it was developed to -30 mV (vs. OCP) to 0.0 mV (vs. OCP), with a scanning velocity of 1 mV/s. GCS samples were also evaluated by visual inspection, using a camera Nikon D5300 to obtain images before and after each exposure; and by morphological analysis, with scanning electron microscopy (SEM) and with an analytical X-ray microprobe, EDS. The microscopy tests were conducted with a TESCAN equipment, model VEGA 3. Secondary scattered and backscattered electron detectors (SE/BSE) and an analytical X-ray microprobe, EDS type, Oxford brand, were used. The application of artificial intelligence techniques has been divided into two distinct phases. Initially, a preliminary investigation was performed to elucidate the visual identification capacity of corrosion through numerical parameters linked to the image. The initial objective was to select some numerical technique to identify if there were appropriate range values in the HSV associated with the presence and detection of superficial corrosion. The PSO metaheuristic method was implemented via Python software to obtain the appropriate ranges of hue, saturation, and value. From the algorithm iterations, the superposition of local and global gradients of the so-called “particles” was obtained, which converged to the solution of the parameter ranges in the HSV system. From the validated reference responses 6 constructed in ImageJ software, the IA algorithm was implemented to iteratively determine the subrange in the HSV system of the reference standard. The HSV system was used to identify the processes that differed in shades of grey (complete and corroded parts), highlighted in the saturation spectrum. The methodology was defined as the direct identification of the corroded surface, by quantifying the portion of the non-intact surface. The data was obtained by calibration of the tool from photographs of different corrosion patterns with their respective percentages of corroded surface area on degraded specimens in artificial weathering chambers (salt spray chamber). Using the algorithm, it was possible to use AI with the PSO method to adjust the correct values of value ranges in the HSV system to replicate the defined selection pattern. The effectiveness of the preliminary methodology in visually identifying the presence of corrosion on the specimen (shown in Figure 4) allowed the current study to investigate more quantitative aspects related to galvanized steel corrosion. In parallel, another search algorithm was developed and applied to images of artificially aged specimens exposed to salt spray to investigate the range of parameters potentially correlated with higher levels of mass loss sample's mass loss. 3. RESULTS AND DISCUSSION 3.1 Materials and corrosion range characterization A loss of thickness in the coupon samples installed in the power substations of ARGO I in the northeastern region of Brazil, between the months of August and November 2023 (3 months), a period locally characterized as winter, is presented in Figure 8. In the figure were also included the average temperature, relative humidity, prevailing winds for the continent and their average speed, and the average distance to the coastline. During this sampling period, the region experiences a prevailing dry climate, but as can be observed, it has an average relative humidity greater than 65 %RH, which favour the formation of nocturnal electrolyte and the corrosion process. The winds blowing towards the continent have varying average speeds, with intense air masses facilitating the transport of sea salt over long distances. The average corrosion rate during this period for CS was more intense for power substations closer to the coastline, decreasing with the distance of separation and possibly from the shelter and breaking zones of the waters, in the following decreasing order of damage: PS Acaraú, PS AUT, (244 ± 54) µm.year-1 - 11 km> PS Pecém II, PS PED, (111 ± 10) µm.year-1 - 7 km> PS Bacabeira, PS BCB, (79 ± 7) µm.year1 - 17 km> PS Parnaíba III, PS PBT, (56 ± 6) µm.year-1 - 30 km> PS Tianguá II, PS TGD > (51 ± 2) µm.year-1 - 97 km. The galvanization process (GCS samples) showed its effectiveness in controlling local corrosion during this period for the samples used, resulting in only two substations with corrosive substrate processes, in decreasing order: PS Tianguá II, PS TGD > (16 ± 1) µm.year-1 - 97 km> PS Bacabeira, PS BCB, (9 ± 7) µm.year-1 - 17 km= PS Pecém II, PS PED, (0 ± 0) µm.year-1 - 7 km= PS Acaraú, PS AUT, (0 ± 0) µm.year-1 - 11 km= PS Paranaíba III, PS PBT, (0 ± 0) µm.year-1 - 30 km. PS Tianguá should be further explored regarding local corrosion processes over time since it showed the highest corrosion rate and is located at a greater distance from the coastline. Table I presents the average corrosion rates for the samples exposed in laboratory, under chloride and sulfate chambers. As it could be seen, the observed losses of thickness for artificial exposure were, even for the first period of analysis, higher than the ones observed in field, for both kind of samples. Specifically for SAE 1020, it was verified mass losses up to four times bigger than the AUT PS, which presented the biggest corrosion under the 3-month study. Based in that, it was concluded that this small portion of time (3 months), used for the first analysis in field, could not represent the complete seasonal cycle, despite some preliminary observations that could help a comprehension of the local atmospheric aggressiveness and support the maintenance of power transmission lines. Additional observations must be addressed for GCS. The SO2 exposition developed in chamber demonstrated a higher corrosion rate for the material, with results comparable to CS. It indicated that the presence of SOx pollutants in the atmosphere, such as verified in urban or industrial environments, could represent a bigger challenge to ensure the project life-cycle for the materials exposed, than the one verified in marine areas. 7 Figure 5. Average thickness losses of CS and GCS coupons, installed in the region, shown on a map, with local parameters of distance from the seafront and wind arrangement (predominant direction) and air mass towards the continent (Google Earth). Table I. Average corrosion rate of CS (SAE 1020) and GCS coupons, exposed to chloride and sulfate in chamber. 3.2 Electrochemical tests Figure 6 presents the obtained results for the electrochemical tests developed over the samples collected from the five PS analysed. The OCP measurement (Figure 6a) indicated that the materials presented a similar behaviour, despite some corrosion rate observed in TGD and BCB. The average value for OCP was -0,94 V (vs. Ag|AgCl), which corroborated with the literature [ 48 ]. Although, the Nyquist plot obtained from the EIS experiment (Figure 6b) showed a significative reduction on the semi-circle radius, which was directly related to the corrosion rate verified in field – a small diameter was observed to TGD, followed by BCB, AUT, PED and PBT. As presented in [ 49 ], the semi-circle radius is related to the electrical resistance of the formed corrosion product film, which tend to be bigger to the samples that presents a more stable and compacted oxide layer formed in the surface. The same was observed in CV graph (Figure 6c), that presented an ascendent behaviour specially for TGD, with higher corrosion rates even for similar potential parameters. BCB presented the second higher corrosion current, followed by the three others PS, AUT, PED and PBT. The electrochemical results corroborated with the mass loss measurement, even for the PS that did not present corrosion rate for GCS (following the aggressiveness pattern verified through de CS analysis). 8 Figure 6. Obtained electrochemical parameters to GCS exposed in field. (a) OCP vs. Ag|AgCl; (b) EIS – Nyquist plot; and (c) CV. 3.3 Scanning electron microscopy sample analysis (SEM) The microscopic and elemental chemical composition analysis by SEM-EDS of GCS test specimens exposed in the field revealed, in certain environments within their installations, the presence on their surfaces of chemical elements possibly originating from salt spray due to their proximity to the maritime edge, such as chlorine, sulfur, calcium, and magnesium. In this context, the Acaraú power substation, located approximately 11 km from the maritime edge, experienced a higher corrosion rate for CS compared to the other four analysed locations (section 3.1). Despite not observing, during this period, a relative thickness loss related to the corrosion process of the zinc coating, iron was also identified on the GCS surface, as shown in Figure 7. Additionally, with a more extensive analysis area, a differentiated morphology or chemical phase (with acicular crystallization) was detected, as highlighted in Figure 8 (arrow region), where a higher concentration of sulfur was also observed. Similar processes were observed in Pecém power station (Figure 9). The area is also located near from the maritime edge (7 km), experienced the second higher corrosion rate for CS, despite the galvanized steel did not present any loss of mass. Figure 7. Microscopic and elemental chemical analysis, using SEM-EDS, of a GCS test specimen exposed for approximately 3 months in the Acaraú power substation (ARGO I), located about 11 km from the maritime edge in the northeastern region of Brazil. The presence of the chemical phase with acicular crystallization on the GCS surface was also more effectively detected in specimens installed for the same 3-month period at the Bacabeira power substation, located approximately 17 km from the maritime edge. In this case, such chemical changes in the GCS material may have contributed to the initiation of corrosion processes in the exposed material with a corrosion rate of (9 ± 7) µm.year-1, as previously mentioned. Microscopic images by SEM-EDS of part of one of the analysed test specimens were shown in Figure 11. Additionally, silicon and