Water Separation and Formation of Cells with Differential Aeration as Factors Controlling Corrosion of Steel Pipelines in a Crude Oil Storage Facility

Abstract

The mechanism causing the dramatic intensification of the corrosion deterioration of carbon steel pipes in a crude oil storage facility has been investigated. This study considers a number of factors affecting corrosion in crude oil, such as the water content, the corrosivity of the aqueous phase, the kinetics of water–oil separation, the effect of dissolved oxygen, the effect of the crude oil quality, the degree of stagnancy inside of the pipes, the possible contribution of microbially induced corrosion (MIC) and the presence of deposits. The key root of the corrosion intensification was the separation of the water phase, supported by stagnancy, which eventually led to the formation of stable shallow pits surrounded by cathodic areas.

1. Introduction

The purpose of this study was to describe the mechanism that causes the rapid deterioration of carbon steel pipes in a crude oil storage facility. The reported rates exceeded the expected values, resulting in pipe replacement before reaching its projected lifetime. More importantly, the operator reported a gradual increase in the corrosion rate. This occurred only after several years of normal operation. For instance, the corrosion loss measured by the eddy current technique (NDT) 2 years after commissioning equaled, in the worst locations, below 0.2 mm/y; however, during the next inspection, after 5 years, the rates were up to 1.5 mm/y. Due to this increase, some of the pipes were replaced after as few as 8 years, while the projected lifetime was 20 years. This phenomenon occurred repeatedly in pipes used to connect individual above-ground storage tanks, branching off of a main line (Figure 1). The main line is operated more frequently because it is used to distribute crude oil to multiple tanks and does not experience too much corrosion. The corrosion in branched pipes is much more severe, always heavily concentrated around the bottom position (5–7 o’clock) and around the horizontal bends and curvatures.

The corrosivity of crude oil can be classified according to the proximity to the well [1]. Upstream operations, such as drilling and pumping, involve the transport of crude oil mixed with significant amounts of water and gases [2,3]. Midstream operations, such as the facility in question, include transportation and storage prior to refining, and deal with crude oil that has lower concentrations of water and gases [4]. Downstream corrosion problems are primarily caused by the chemical composition of specific fractions and are affected by refining temperatures [5,6].

Corrosivity in midstream operations is generally less severe compared to production; however, this step often spans a much longer duration. In the facility discussed in this paper, crude oil is typically stored for several years. Corrosion inside of the tanks is usually mitigated by the application of organic coatings up to a 1 mm thickness at the tank’s inner bottom and sides. However, uncoated steel is commonly used for piping, connecting valves and pumps due to the challenges of coating their internal surfaces. Corrosion in crude oil is influenced by various factors, such as the water content [7,8], inorganic corrosive compounds, [9,10,11,12,13,14] and microbial activity [15,16,17]. Additionally, the slope and low spots on the pipelines, the corrosion under the deposits, the kinetics of the water–oil separation, the oxygen content and the transport frequency should also be considered.

The corrosion rates of carbon steel reported by other authors during long-term storage are shown in

Table 1. Overview of steel corrosion rates reported from storage facilities.

Reported Corrosion Rate [mm/y]	Conditions	Ref.
0.1	Oman sour crude oil, field C, pipelines	[18]
0.15	Oman sour crude oil, source A, pipelines
0.02–0.62	Oman sour crude oil, field B
0.25–0.6	Oman sour crude oil, source D, pipelines
0.3	Oman sour crude oil, field B, separator
0.05–0.5	Storage tank bottom	[20,21]
0.1–0.5	Storage tank lid	[20]
0.4	Crude oil with 200 ppm water; lab study, threshold water content to form a two-phase system	[22]
1.0	Lab study, 0.1% NaCl water solution	[22]
1.0	Uncoated tank walls, lower part	[23]
1.5	Uncoated tank walls, upper part	[23]
1.5	Oman sour crude oil, field B, process station	[18]

. The choice of the referenced studies is based on similarities in the composition of crude oil (low CO₂ and an increased content of sulfur) and the piping setup. Long-term corrosion rates range from 0.1 to 1.5 mm/y. For comparison, in sweet crude oil, some authors reported rates up to 10 mm/y. This behavior was associated with the disturbance of the semiprotective layer of sulfur corrosion products [18,19].

The motivation for this study was to provide a robust decision matrix to identify the main factors that affect the corrosion of the pipeline. While the typical variables associated with corrosion are known in the community of similar facility operators, the acceleration of corrosion in time was deemed uncommon, without a thorough explanation in the literature, and it was difficult to obtain field experience from other operators. The goal was to implement the findings into realistically applicable preventative measures under a typical crude oil storage operation without extensive design changes and financial cost. Although a number of parameters such as the oxygen, water and chloride content, the microbial activity, the water separation kinetics, the crude oil quality and the area of cathodic reaction have been investigated in laboratory experiments, their primary objective was to identify the principal factors affecting the strong increase in the corrosion rate, not to quantify the role of each parameter in detail. This remains a task for further research.

2. Materials and Methods

2.1. Steel Samples

The branched pipes in the facility of interest (Figure 1) were manufactured from EN P235GH carbon steel with an outer diameter of 700 mm and an 8 mm wall thickness, radially welded (45° or 90 ° to the longitudinal axis). The material was of a ferritic–pearlitic structure, with a nominal ultimate tensile strength of 480 MPa, a yield strength of 235 MPa and an elongation of 23–25%. The pipes were located 0.5 to 1 m above-ground.

For the characterization of the extent of corrosion and corrosion products, four bottom sections of the pipes with a minimum dimension of 1000 × 400 mm were selected. Before cutting, the pipes were cleaned with pressurized water and dried with compressed air. The selection was based on the corrosion severity established in a regular nondestructive inspection using the eddy current technique and the likelihood of water accumulation. The following two types of samples experiencing heavy corrosion were sought: those with a high probability of water accumulation at the low points of the pipeline, and those where water was likely to runoff and the surface was in contact with the oil phase. To identify these locations, a spirit level was placed on the pipe and a relative inclination/declination was recorded. By comparing neighboring sections, low spots and high spots were identified. Selected sections were analyzed for wall thickness in cross-sections and surface deposits (scanning electron microscopy, SEM, Carl Zeiss Microscopy; X-ray fluorescence, XRF, Olympus; X-ray diffraction, XRD, Bruker, MA, USA). Also, from the intact parts of these sections, 50 × 50 mm samples were prepared for exposures modeling the artificial pit. These were degreased in ethanol, acetone and n-heptane. A fresh metallic surface was prepared by brushing with a wire wheel followed by 1-min pickling in concentrated HCl.

For laboratory experiments covering the effect of stagnancy and different water contents of water in crude oil mixtures, CR4 steel sheet samples of 20 × 20 × 1 mm were used. The reason for changing the grade was that there was no sheet material available in P235GH steel, and the available plate samples were curved, following the circumference of the pipe. The chemical composition and microstructural differences between the two steel grades were negligible, with a minor expected effect on the corrosion.

2.2. Liquid Sampling and Analysis

The facility is used primarily to store and transport Russian Export Blend (REB) crude. The crude had a 31.4 °API, a falling category of light crudes [24], and a viscosity of 13.5 mPa∙s at 20 °C. To capture the specific fractions of the crude oil, samples were collected from various locations at different time intervals, coinciding with the arrival of new batches. This approach was adopted due to the separation of crude oil fractions occurring during long storage, affecting the crude quality inside of the distribution pipes. Depending on the position of the suction outlet, the transported mixture can have a different composition, since the water is separated at the bottom of the tank and there are lighter oil fractions at the top. To quantify this, crude samples were analyzed for the elemental composition (CHNS analysis, Elementar EL III, Langenfeld, Germany), n-alkane composition (liquid chromatography) and water content (ISO 12937) [25]. The aqueous samples were collected from a water separator and analyzed in view of the elemental composition, pH, total organic carbon and carbonate content.

The kinetics of the water separation was studied using a simplified setup, which is depicted in Figure 2. The crude oil samples were mixed with 0.3, 0.5 and 1 wt. % water, placed in 1 L glass conical separators at 5, 10 and 15 °C and the volume of the water emulsion or homogeneous water phase at the bottom was recorded once a day for 30 days.

2.3. Microbial Analysis

A set of liquid samples and smears from inside of the pipelines for the microbial analysis was collected after draining. Another set of samples was collected on a different occasion directly from the pipe with crude using a flange welded in the selected spot on the pipe. The reason for the latter approach was to avoid the disturbance of the microbial activity on the pipe surface as a consequence of draining. In both cases, the samples were immediately stored in an icebox and transported to a deep-freeze box (–80 °C) within one hour.

The water samples were first centrifugated to concentrate the biomass. An amount of 100 mL of each sample was centrifuged at 16,000× g for 10 min. The oily phase supernatant was discarded and the pellet containing the microbial biomass was stored at −80 °C until the DNA extraction. The first set of samples was analyzed using the FastDNATM SPIN Kit for Soil DNA Extraction (MP Biochemicals, Irvine, CA, USA). The DNA concentration was quantified by fluorimetry using a Qubit^® 2.0 fluorometer (Invitrogen, Life technologies, Carlsbad, CA, USA) according to the producer’s instructions. Functional groups of bacteria were analyzed by specific molecular markers for sulfate-reducing, iron-reducing and denitrification bacteria. The markers were amplified by the primers listed in

Table 2. Primers used for quantitative PCR.

Primer	Sequence 5′-3′	Specificity	Method Ref.
RH1-aps-F	CGCGAAGACCTKATCTTCGAC	SRB 1: gene coding adenosine 5′ phosphosulfate reductase	[26]
RH2-aps-R	ATCATGATCTGCCAGCGGCCGGA
RH1-dsr-F	GCCGTTACTGTGACCAGCC	SRB: gene coding respiratory hydrogensulfite reductase	[26]
RH3-dsr-R	GGTGGAGCCGTGCATGTT
Geo494F	AGGAAGCACCGGCTAACTCC	Geobacteraceae (iron-reducing bacteria) 16S rRNA	[27]
Geo825R	TACCCGCRACACCTAGT		[28]
16SqPCR-F	TCCTACGGGAGGCAGCAGT	All bacteria: universal 16S rRNA primers	[29]
16SqPCR-R	GGACTACCAGGGTATCTAATCCTGTT

¹ SRB: sulfate-reducing bacteria.

.

The second set, collected through a flange, was from three different sections of a single deteriorated pipe. Each sample was analyzed by gene sequencing to study the composition of the microbial community. Bacterial DNA from a sample was isolated according to the manufacturer’s instructions using a commercial PowerWater^® DNA Isolation Kit from MO BIO (Carlsbad, CA, USA). The procedure involved the chemical lysis of cells followed by mechanical lysis and the precipitation of DNA using ethanol. The quantification of the genomic DNA extracted was performed using a Qubit 2.0 fluorometer (Life Technologies, Carlsbad, CA, USA). The universal primer 16S rRNA gene was used for the identification of the total bacterial biomass for the qPCR analysis. A LightCycler ^® 480 Instrument (Roche, Rotkreuz, Switzerland) was used for the qPCR. Library preparation was an initial step for the genomic DNA sequencing. The primers 530F and 802R, as shown in

Table 3. Primers for the amplicon sequencing of the 16S rRNA gene.

Primer	Sequence 5′-3′	Coverage			Ref.
		Archae	Bacteria	Eukaryotes
530F	GTGCCAGCMGCNGCGG	54.9	96.9	94.0	[31]
802R	TACNVGGGTATCTAATCC	91.8	92.5	0.9	[32]

, were used to amplify the variable V4 region of the 16S rDNA gene for the sequencing of amplicons. The size of the amplicon was kept below 400 bp to cover as much microbial diversity as possible. The PCR conditions and sequencing methods used are described in [30]. Amplicon sequencing was performed on the Ion Torrent PGM using the Ion PGM Hi-Q Sequencing Kit with the Ion 314 Chip v2 (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer’s instructions.

The obtained data were processed with QIIME 2 2021.8 software [33]. Sequences were demultiplexed. Data quality was filtered using the q2-demux plugin followed by denoising with DADA2 (via q2-dada2) [34]. Taxonomy was assigned to ASVs using the q2-feature-classifier [35] and classified by the classify-sklearn naive Bayes against the Silva 138 database [36]. The QIIME 2 outputs were processed using the phyloseq R package [37].

2.4. Laboratory Exposures

Laboratory exposures were performed to qualitatively assess the individual effect of the crude composition, water content, salinity of the water phase, presence of deposits and, most importantly, the frequency of pumping. CR4 steel samples were ground with P120 grit paper, degreased, cleaned with demineralized water and dried with compressed air. For each experimental condition, 4 replicas were placed in a 250 mL glass container (Figure 3).

The corrosivity of the crude admixtures was studied using a pristine sample of crude mixed with demineralized water with a maximum conductivity of 10 µS/cm and about 5 ppm of dissolved oxygen. The resulting composition was 0, 0.1 and 1 wt. % of water in crude oil. The effect of the water salinity was studied by adding 5 wt. % NaCl water solution to crude oil instead of demineralized water. The solution was prepared using the same demineralized water and analytical-grade NaCl. As deposits are commonly found on the inner surface of corroded pipes, their effect on corrosivity was also considered by adding sand to create a layer of 5 mm thickness on the metallic samples before adding the corrosive solution. Sand was selected based on the composition of deposits in the pipeline, which was rich of quartz, but also for practical aspects, including the uniform chemical composition of such a layer, and the permeability and easiness of the sterilization, to avoid microbial contamination.

The effect of oil pumping was studied using a stirred setup (Figure 3). Glass containers with CR4 carbon steel samples were stirred every 1 or 3 weeks to model the impact of the frequency of oil pumping.

The exposures lasted 4 months. An overview of all experimental conditions is provided in

Table 4. Conditions of laboratory experiments.

Experiment	Configuration	Steel	Number of Samples	Mixing Frequency
A	Crude oil	CR4	4	1 or 3 weeks
B	Crude oil + 0.1 wt. % water
C	Crude oil + 1 wt. % water
D	Crude oil + 1 wt. % water + sand
E	Crude oil + 0.1 wt. % NaCl solution (5 wt. %)
F	Crude oil + 1 wt. % NaCl solution (5 wt. %)
G	Crude oil + 1 wt. % NaCl solution (5 wt. %) + sand

.

The corrosion rates were evaluated using mass-loss data after the corrosion product removal according to ISO 8407 [38]. Samples were cleaned with demineralized water and organic remnants were removed using ethanol and acetone. The mass loss was then evaluated by interval pickling in Clarke solution (HCl + 5 wt. % SnCl₂, 2 wt. % Sb₂O₃). The corrosion rate was then calculated using the following formula:

$$v_{corr}={\displaystyle \frac{\Delta m}{\rho \times A\times \tau }}$$

where $\Delta m$ is the mass loss, $\rho$ is the density of iron, A is the exposed area and $\tau$ is the exposure length.

3. Results

3.1. Corrosion Degradation

The elemental composition of a typical pipe segment is given in

Table 5. Elemental composition of pipe steel, in wt. %.

Element	C	Si	Mn	P	S	Cr	Ni	Cu	Al	Fe
EN P235GH (standard) [39] 1	-	≤0.35	≤1.2	-	-	≤0.3	≤0.3	≤0.3	-	-
Pipe sample	0.159	0.23	1.1	0.017	0.007	0.043	0.040	0.14	0.069	bal.
CR4 [40] 2	0.04	0.008	0.202	0.011	0.009	0.033	0.037	0.074	0.034	bal.

¹ Equivalent UNS K02801. ² Equivalent UNS G10080 [41].

. It conformed to the requirements of ISO 11960 [39] for P235GH steel.

The appearance of the pipe segments is shown in Figure 4. The overall observation matched the NDT inspection data collected by the operator. The corrosion attack was strongly localized to discrete spots, all in the 6 o’clock position. Continuous corrosion damage was minimal. The wall thickness loss is indicated in mm and the percentage wall thickness reduction next to the arrows in Figure 4. The heaviest corrosion was observed in sample IIIa, where the wall was corroded to up to 89% of the original thickness of 8 mm. The topography of corrosion degradation is shown in Figure 5. There were no significant differences between the samples. The damage was in the form of pits with shallow borders. Exceptions were the corroded locations around the welds. On multiple samples, a protruding piece of the weld material was found in the middle, with the corrosion concentrated on both sides of the weld; see Figure 5c.

Data on the corrosion severity and the likelihood of water presence, based on the location of low points, are given in

Table 6. Thickness losses on selected pipe sections; comparison of laboratory measurement and NDT data.

Sample	Water Presence Probability	Corrosion Attack Severity	Maximal Thickness Loss
			Cross Section	Eddy Current Method (NDT)
Ia	High	Low	28%	30%
Ic	High	Low	48%	50–100%
IIb	High	Medium	44%	40%
IIIb	Low	Low	29%	40%
IV	Low	Medium	31%	N/A
IIa	N/A	High	76%	50–100%
IIIa	N/A	High	89%	50–100%

N/A—not analyzed or not available.

. The severity of the corrosion attack (column 3) reflects the graphical representation produced by the automatic eddy current software. This representation was at first used for sample selection; the accuracy was verified on the cross-sections.

The first two rows show the data for sample I, where the presence of water was highly probable. The corrosion losses were medium, reaching a maximum of 48% of the cross-section. As stated before, the attack was heavily localized, with most of the inner surface undamaged. Samples IIIb and IV came from higher sections with a low probability of water accumulation. They showed a similar wall thinning from 29 to 31%. Sample II from the section with both low and high spots also showed corrosion losses irrespective of the probability of water accumulation.

The last column shows the thickness losses recorded in the cross-section. The NDT and cross-sectional analysis are generally in good agreement. For higher losses, the difference is caused by the data interpretation performed by the NDT device, that is, losses from 50 to 100% are characterized as one category without further distinction.

Comparison of thickness losses to the location of low spots showed a lack of correlation between the two phenomena. Severe corrosion was observed in spots with a low probability of the presence of water, and vice versa. Based on the detailed analysis of the cross-sections, it appears that the accumulation of water occurs more locally, for example, around natural obstacles that protrude from the inner wall, such as welds (Figure 5c).

3.2. Crude and Water Composition

The laboratory evaluation of three samples showed that they contained between 0.03 and 0.35 wt. % water, 1.7 wt. % sulfur and 4 to 35 mg/L chloride. The crude was sour due to a relatively high sulfur content. The total chloride content was apparently low; however, as an ionic species, chloride accumulates in the polar aqueous phase. Recalculating the average of 15 mg/L chloride to the average water content of 0.2 wt. % resulted in a value of 7000 mg/L chloride in the water phase.

The analysis of the water from the collector tank showed a chloride content of almost 1200 mg/L (

Table 7. Parameters of water from the water collector.

Parameter	Value	Units
pH at 25 °C	8.0	-
Conductivity at 25 °C	370	mS∙m−1
Total alkalinity (KNK4.5)	6.4	mmol∙L−1
Apparent alkalinity (KNK8.3)	0
COD Cr solution	2540	mg∙L−1
Ca2+	188
Mg2+	10.9
Na+	724
K+	16.6
NH4+	16.5
Fe total	0.7
Mn total	2.0
Cu2+	<0.001
SiO2	11.9
HCO3−	390
Cl−	1186
SO42−	15.6
NO3−	16.7
CaCO3 supersaturation	195

). While not as high as calculated, the chloride distribution is still heavily shifted towards the water phase, and the content in the oil phase is therefore lower. The aqueous phase was slightly alkaline, with high amounts of species forming solid deposits. It is important to note here that a pH of 8 is sufficiently alkaline to shift the CO₂-HCO₃⁻-CO₃²⁻ equilibrium towards HCO₃⁻. This can be illustrated using Bjerrum diagrams for seawater [42,43,44]. All of the dissolved CO₂ gas is therefore converted to HCO₃⁻ and CO₃²⁻. Similar conclusions can be drawn for H₂S [45].

The tendency to deposit formation is quantified in the last row, corresponding to the amount of CaCO₃ that can be precipitated from the solution before reaching equilibrium. The high chemical oxygen demand (COD) values cannot be attributed solely to microbial activity because the water was polluted with organic compounds from the oil phase.

The crude quality from the beginning, middle and end of a batch is shown in Figure 6. The sample collected in the middle of the batch contained the highest amount of low-carbon alkanes, and consequently the lowest representation of alkanes with longer chains. The samples collected at the beginning and at the end of the batch were quite similar, with compositions shifted toward the heavier compounds.

Due to the inhomogeneity of the properties of the crude samples from the beginning and end of the batch, only the crude samples from the middle of the batch were used to prepare mixtures for laboratory exposures.

3.3. Water Separation Kinetics

The results of the water separation kinetics are shown in

Table 8. Volume of separated aqueous phase at 15 °C, in mL (only distinct observations shown).

Day	Water Added to Pristine Crude
	0.3 wt. %	0.5 wt. %	1 wt. %
1	3	5	3
30	3	5	3

,

Table 9. Volume of separated aqueous phase at 10 °C, in mL (only distinct observations shown).

Day	Water Added to Pristine Crude
	0.3 wt. %	0.5 wt. %	1 wt. %
1	0.5e	0.5	0.5
30	0.5e	0.5	0.5

e: Water–oil emulsion formed instead of transparent water fraction.

and

Table 10. Volume of separated aqueous phase at 5 °C, in mL (only distinct observations shown).

Day	Water Added to Pristine Crude
	0.3 wt. %	0.5 wt. %	1 wt. %
1	0	0.2	0.5
3	0	0.4	0.5
4	0	0.4	0.5
5	0.2e	0.4	0.5
30	0.2e	0.4	0.5

. Typically, the separation proceeded as follows: After mixing, the crude oil and the added water formed a single phase. After some time, the second phase formed and a transparent aqueous or lighter brown water–oil emulsion (hinted with ‘e’ in Table 9) formed at the bottom. Above this phase, droplets of the aqueous phase continued to form until they coalesced with the aqueous phase. A similar separation time was reported in studies focused on oil–water separation kinetics [46] and the effect of the flow on corrosion [47].

The kinetics were driven by the following two main factors: the temperature and water content. The shortest time to water separation in the setup with the highest amount of added water was reached at 15 °C. Clear separation of the water phase was observed within one day, and no further changes were observed for the remainder of the 30 days. In contrast, the slowest water separation was observed in the experiment at 5 °C for a mixture containing 0.3 wt. % added water. For the first 4 days, no separation was visible; on day 5, the water phase formed and remained stable for the rest of the 30-day experiment. At the lowest water content and temperature of 5 and 10 °C, a water–oil emulsion formed initially and became transparent later. These separation and demulsification times correspond to the measurements by other authors at similar water contents [48,49].

3.4. Corrosion Products and Deposits

The results of the XRF and XRD analyses of a mixture of deposits with corrosion products from eight heavily corroded locations from samples I-IV are shown in

Table 11. Elemental composition of deposits by XRF, in wt. %.

Element	Min.	Max.	Avg.
Fe	66	83	73
Ca	4	17	9
Si	3	13	9
S	2	11	4
Al	0.2	2.2	1.4
Mn	0.6	0.8	0.7
Na	0.4	0.6	0.5
K	0.1	0.7	0.4
Cl	0.1	0.6	0.4
Mg	0.2	0.5	0.3

and

Table 12. Phase composition of deposits by XRD, in wt. %.

Trivial Name	Formula	Average	Maximum
Siderite	FeCO3	19	59
Wustite	FeO	2	10
Pyrrhotite/Mackinawite	FeS	5	12
Melanterite	FeSO4.H2O	2	13
Magnetite	Fe3O4	9	29
Lepidocrocite/Goethite/Akaganite	FeOOH	1	10
Maghemite	Fe2O3	5	21
Quartz	SiO2	19	48
Calcite	CaCO3	6	19
Muscovite	K1Al2(Al0.5Si4)O10(OH)2	7	22
Magnesiumcalcite	MgCaCO3	6	27
	CaMg0.3Fe0.7(CO3)2	2	9
Albite	NaAlSi3O8	7	15
	KAlSi3O8	3	11
Carbonate, total	CO32−	8	31
Silicate, total	SixOy	7	48
Fe2+, total	Fe2+	16	28
Fe3+, total	Fe3+	7	14

. The results show a highly irregular mixture of products, which is not uncommon for samples collected in field exposures, unlike laboratory experiments.

The XRF measurement proved the deposits were rich in iron, which represented more than 66 wt. % in all eight samples (Table 11). Other dominant elements were calcium, silicon and sulfur. Iron originated from steel corrosion. Calcium cations were found in a large quantity in the water sample from the collector (Table 7). Silicon was present there, as well as silica, SiO₂. Sulfur was present in the crude oil (see Section 3.2). It can form corrosion products with iron that are protective to steel [18,19].

The identification of specific inorganic substances by XRD is given in Table 12. It is important to note that XRD can only provide quantitative analysis of the crystalline portion of a specimen. Nevertheless, the compounds can be split into the following two groups: iron corrosion products and deposit-forming compounds. The identified corrosion products consisted of FeCO₃, FeO, FeS, FeSO₄·H2O, Fe₃O₄, FeOOH and Fe₂O₃. Corrosion products typically forming in well-oxygenated aqueous environments contain mainly iron in the oxidation state 3+. In the analyzed mixtures, iron species in the oxidation state 2+ dominated over the former ones by a factor of two, suggesting that a portion of the corrosion process was taking place in anoxic conditions [50,51,52,53].

In addition to iron corrosion products, a significant amount of inorganic deposit was detected, including magnesium and calcium carbonate and other mixed-type carbonates. These deposit-forming species precipitated as a result of the high concentration in water, as discussed in Section 3.2.

3.5. Oxygen Distribution

The oxygen concentration was tested in a setup similar to the one used for the water separation kinetics. An amount of 150 mL of water, deaerated with nitrogen (30 min), was placed in a glass separator; and the rest was filled with crude oil to reach a 1 L total volume, and then it was thoroughly mixed. The water was removed after 1, 3 and 21 days, and the oxygen content was immediately measured using a lead–silver electrode.

After deaeration, the water contained 0.1 ppm of O₂ (Figure 7). After just 1 day in contact with crude oil, the concentration reached 2.2 ppm, and then gradually increased to 2.6 ppm at 3 days and 3.4 ppm on day 21. This experiment showed that the content of oxygen in the crude oil was sufficiently high to continue enriching the water phase with oxygen.

3.6. Microbial Analysis Results

The microbial analysis of the smeared and liquid samples is shown in

Table 13. Results of the MIC evaluation.

Sample/Position		Corrosion Damage Severity	Total Amount of Bacteria	Sulfate-Reducing Bacteria	Geobacteraceae (Iron-Reducing)
Ib	6 o’clock, 30 cm from cut	High	+++	++	++
	6 o’clock, 60 cm from cut		++	+	+
Ic	5 o’clock	High	+++	++	++
	6 o’clock		++	+	+
	7 o’clock		++	+	+
V	3 o’clock	Undamaged	++	++	++
	9 o’clock		++	+	++
	6 o’clock	Open pit	+	+	+
	9 o’clock		++	++	++
	6 o’clock	Undamaged	++	++	++
	6 o’clock		+	+	+
water collector	water		++++	++	+++

++++–+: highest–lowest amount.

. The number of plus signs indicates the levels of microbial activity; the last two columns show the activity for the two most common corrosion-related species. In the water collector, the general and specific microbial activity was at the highest level. The activity on the surface of the collected pipe sections varied.

No clear correlation was found between the severity of the corrosion and the microbial activity. Samples taken at the 3 and 9 o’clock positions showed significant microbial activity, despite these locations being clearly above even the theoretically present water level, and free of corrosion attack. The microbial activity in the corroded sections Ib and Ic was not too different from that in the undamaged parts of sample V.

Another microbial analysis was performed on a single, strongly damaged pipe in three different sections (VIa, VIb and VIc) spread about 15 m apart. However, this time the pipe was not drained so to minimize the disturbance and to avoid the possible disturbance of microbial species during the cutting and decontamination. The goal was to verify the type of bacteria, focusing on the presence of both typical and less common corrosion-related species. The results are shown in Figure 8. The amount of DNA collected in the VIa and VIb samples was very low (<0.5 ng/µL). The composition of the microbial community differed in all three samples (see

Table 14. List of dominant species in the second collection of microbial samples.

Sample	Dominant Species
VIa	Streptococcus > Methanobacterium > Acetobacterium > Granulicatella
VIb	Methanobacterium > OPB41 > CPR2 > Candidatus Falkowbacteria
VIc	Izemoplasmataceae > Candidatus Falkowbacteria > Smithella

).

The mechanism of a potentially corrosive interaction of microorganisms with metal is summarized in

Table 15. Summary of bacteria present in the samples that can accelerate corrosion of metal surface.

Bacteria	Presence in Sample No.	Effect	Ref.
Bacillus	VIa, VIb	Iron oxidation, contribution to the formation of iron nitride	[54]
Parcubacteria	VIa, VIb, VIc	Can reduce sulfur in combination with fermentation, and thus may contribute to the corrosion through H2S production	[55]
Methanobacterium	VIa, VIb, VIc	Hydrogenotrophic, can use hydrogen on steel surfaces for their metabolism, thereby accelerating steel corrosion	[56]
Candidatus Omnitrophus	VIa, VIb, VIc	Fe oxidation; require Fe for magnetosome biosynthesis	[57]
OPB41 (Actinobacteria)	VIa, VIb, VIc	A common group of bacteria with diverse species reported previously in concrete sewer pipes; however, the role of Actinobacteria in the MIC of concrete and steel remains obscure	[58]
Acetobacterium	VIa, VIb, VIVI	Fe oxidation; hydrogen mediated electron uptake from Fe(0), resulting in corrosion	[59]
Desulfosporosinus	VIa, VIb, VIc	Cathodic depolarization by hydrogen uptake; anodic depolarization by corrosive iron sulfides; precipitation of FeS	[60]
Corynebacterium	VIb	Hydrocarbon-degrading bacteria capable of the degradation of crude oil	[61]
Proteiniphilum	VIa, VIb, VIc	Produces acetic acid, a corrosive agent for the deterioration of metal	[62]
JS1	VIa, VIb, VIc	JS1 have been identified in sulfate-reducing consortia, which are able to mineralize benzene in methane hydrate-bearing sediments, associated with anaerobic methanotrophic communities in sediments with high sulfate reduction rates; they are active in anaerobic sulfate-reducing conditions and are able to utilize glucose (or glucose metabolites) and acetate, which can induce corrosion by the formation of iron sulfide	[63]
Thermovirga	VIa, VIb, VIc	Capable of reducing elemental sulfur, in addition to thiosulfate to sulfide, and thus accelerate corrosion; typically found in oil environments	[64]
Clostridium_ sensu_stricto_13	VIb	Has the ability to direct the uptake of electrons from metallic iron under nitrate reduction conditions and induce corrosion	[65]
Methanocorpusculum	VIa, VIb, VIc	Like SRB, methanogens consume hydrogen, and are thus capable of performing cathodic depolarization, and can aid in the corrosion process	[66]
Streptococcus	VIa, VIb, VIc	Some species of Streptococcus have the capacity to release lactic acid and to grow in acidic environments, becoming powerful corrosive agents	[67]
Caldisericum	VIa, VIb, VIc	Capable of reducing sulfur compounds such as thiosulfate, sulfite and elemental sulfur as electron acceptors, resulting in a production of sulfide, a corrosive agent	[68]

. While all of the main constituents of sample VIa can have an impact on corrosion, the overall microbial activity was low, as illustrated by the total amount of DNA. The same observation applies for sample VIb. The activity in sample VIc was sufficient for the corrosion mechanism to be affected; however, in this case, the dominant species do not interact with metals. Therefore, microbial corrosion was found unlikely to be a major contributor towards the enhanced corrosion degradation.

3.7. Laboratory Exposures Results

The corrosion rates of the steel samples in different setups based on the mass-loss measurements are given in Figure 9. Corrosion losses in pure crude oil were low, with a carbon steel corrosion rate between 0.7 and 1.2 µm/y.

Even a low addition of 0.1 wt. % water strongly affected the corrosivity under stagnant conditions. The frequently stirred mixture caused the steel to corrode at only 2 µm/y, whereas less frequent stirring resulted in corrosion of 32 µm/y, i.e., an increase by a factor of 16. The addition of 1 wt. % of water further increased the corrosivity to steel to 26–59 µm/y. Comparing this to corrosion rates in mixtures with 0.1 wt. % of water, it is obvious that the effect of further water addition was more pronounced in the more-stirred mixture, where the corrosion rate increased from 2.2 to 26 µm/y, i.e., 12×. For the less-stirred mixture, the increase was only twofold, from 32 to 59 µm/y. In other words, the prolongation of the delay between stirring events had a relatively stronger accelerating effect in less corrosive mixtures with less water added. On average, the corrosion rate was two to three times higher under more stagnant conditions.

The corrosivity of the aqueous NaCl phase differed with the water content. At 0.1 wt. % water, the addition of NaCl resulted in a 4-fold increase under frequent stirring. A decrease of more than 50% was found with less frequent stirring. In mixtures with 1 wt. % water, the NaCl addition had a weakly beneficial effect. It may be linked to a lower oxygen solubility in the NaCl solution compared to water [69], which overlapped with the accelerating effect of chloride ions under these already corrosive conditions. The addition of sand, which was aimed to model the effect of silicate deposits, did not cause a significant increase in the corrosion rate.

To sum up, three important observations can be made here. First, comparing the exposure in pure crude oil to the exposure in crude oil with the 1 wt. % water under predominantly stagnant conditions resulted in a 60-fold difference in the corrosion rate. Second, the presence of water in oil can dramatically increase the corrosivity. Third, the level of stagnancy can have an important effect. The difference between 1- and 3-week periodic stirring is best shown for the mixtures with the 0.1 wt. water content. Stirring every week allowed us to keep the corrosivity low, while the 3-week period was sufficient for inducing a significant increase in the corrosivity.

Since the difference between the corrosion rates observed in laboratory experiments (maximum 59 µm/y) and on-site (1.5 mm/y) was still high, further possibilities for corrosion intensification were explored. One of the known mechanisms which can lead to gradual acceleration over several orders of magnitude is corrosion in confined zones. An area around such a confined space with good oxygen access serves as a cathode, while the anodic dissolution of metal predominates in the confined space with limited oxygen access [69]. For passivating metals, this mechanism is known as crevice corrosion and is linked to a loss of passivity inside of the crevice [70,71,72]. Although, without the depassivation step, similar corrosion acceleration in confined zones such as under deposits or in construction crevices is well described and also regularly observed for metals corroding in activity [73,74]. Since there was a layer of corrosion products and solid deposits on the surface of the inspected pipes, in some cases forming a lid over shallow pits, it could be argued that this mechanism can also play a role in the observed corrosion acceleration.

This hypothesis was studied in an experimental setup, depicted in Figure 10. A tubular orifice (⌀ 8 × 6 mm) was machined into a P235GH steel plate (50 × 50 × 8 mm) and filled with demineralized water. The level of water was set to cover the outer surface of the orifice. The exposure area on the surface was limited by a polypropylene pipe of a 23 mm inner diameter, glued to the surface with epoxy. The orifice was covered with a single sheet of filtration paper to simulate a layer of deposits and corrosion products. The outer surface of the orifice was two times larger than the inside. The total duration of exposure was 3 months, with nine replicas in each experiment. The results of the experiment are shown in Figure 11.

The mass loss of the samples with the open pits was 0.20 ± 0.08 g/m²·y. When the pits were covered, the mass loss increased 13-fold to 2.50 ± 0.92 g/m²·y. The flat reference samples showed a mass loss of 0.18 ± 0.07 g/m²·y, similar to the open pit configuration. Recalculation of the mass loss for the samples with the covered pits to the pit area (the bottom and side wall) gave the corrosion rate of 1.6 mm/y. Indeed, it should be noted that the use of the thickness-loss calculation is typically dissuaded for localized corrosion. It is used here only to allow for the comparison with previous results and with the NDT data recorded on site.

4. Discussion

In general, the corrosion rates observed in the facility on the piping connecting the above-ground storage tanks from 0.2 to 1.5 mm/y fit well within the corrosion rates described in the literature. In most cases, the corrosion was strongly localized at the 5–7 o’clock position. The facility operator reported an acceleration of the thinning rate over time. The low spots in the long sections of the piping did not correlate with the locations of the most severe damages. However, detail analysis showed that some strongly corroded areas were in the vicinity of obstacles that formed local low spots, such as radial welds. It can be expected that such locations served as sites where the water phase deposited from the oil–water mixture.

Liquid phase analyses revealed a low concentration of chloride in the crude, suggesting low corrosivity. However, due to the preferential separation of inorganic compounds to the polar aqueous phase, the actual chloride concentration in the aqueous electrolyte was several orders of magnitude higher. This has been confirmed by the analysis of water from a separator, where there was up to 1.2 g/L chloride. This uneven distribution of polar species needs to be considered in further studies. Besides chloride, the water contained a high concentration of deposit-forming compounds, such as Ca²⁺, Mg²⁺, carbonates and silicates. This corresponded well with the analysis of the solid deposits mixed with corrosion products, where carbonates and silicates dominated. The effect of dissolved gasses was evaluated as negligible due to the high pH of the water phase. At some point, likely during downstream operations, the crude mixture was clearly in contact with CO₂ and H₂S gasses, resulting in high concentrations of their ionized forms. Bjerrum diagrams clearly suggest that, at a pH of 8, virtually all of the available CO₂ is converted to HCO₃⁻ and the H₂S is converted to HS⁻. The last significant dissolved gas, oxygen, was at levels corresponding to the water saturation. It is likely that the crude mixture was in partial contact with atmosphere and became saturated with oxygen during the prolonged storage.

The iron corrosion products consisted predominantly of ferrous ions, which are not commonly found in oxygenated conditions. Since the lack of oxygen was ruled out due to its abundance in the oil phase and the easiness of the oxygen transport to the aqueous phase, two other mechanisms of oxygen depletion mechanisms were suggested. Microbial colonies often form tubercles or slime layers, limiting oxygen access to the metal surface. While the microbial activity in the system was non-negligible, two separate analyses proved that it was independent of the location of the most severe damage. Moreover, the second analysis focused on the identification of microbial species showed that the vast majority of the present microbial species were benign strains unrelated to any acceleration of corrosion. Desulfosporosinus, a species of sulfate-reducing bacteria, was present in all three samples. It is known to induce the formation of mackinawite, FeS, which formed on average 10 wt.% of crystalline substances in the deposits. Still, Desulfosporosinus was a very minor constituent within the microbial species. Therefore, although microorganisms probably had a certain corrosive effect, microbially induced corrosion is not considered to be the primary cause of the corrosion acceleration.

The second considered mechanism of oxygen depletion, i.e., the formation of solid deposits on the surface, the limitation of oxygen transport to the anodic areas beneath the sites and the formation of cells with differential aeration between such areas and their surrounding where oxygen is abundant, has been tested in the artificial pit setup. Equipping the pit with a lid resulted in an 11-fold increase in the corrosivity of a mixture of crude oil and demineralized water. No such effect was observed in experiments with added sand. Obviously, the presence of an easily permeable layer of deposits did not initiate the differential aeration mechanism.

Laboratory experiments proved that the tendency to form a distinct water phase at the bottom of the mixture increased with the water content and temperature. Only when the water phase formed, the corrosion rate of the steel accelerated. This is supported by the observed effect of periodical mixing. The results showed that the most aggressive conditions were achieved by keeping the crude oil–water mixture containing an elevated amount of water relatively still. This set of conditions promoted a rapid separation of the water and jump-started the corrosion process. The three-times more frequently stirred solutions were roughly 50% less corrosive towards steel. It is likely that more frequent mixing would result in a further reduction in the corrosion rate.

Based on the results presented above, the corrosion mechanism summarized in Figure 12 can be proposed. In Step I, no distinct aqueous phase is present at the surface, and the corrosion rate is very low. In Step II, the water begins to separate. The exact time to separation depends on the water content, temperature and crude oil movement. If the mixture remains stagnant for a prolonged period, the corrosion rate further increases (Step III). This still relatively safe corrosion rate remains stable until a pit with significant depth is formed in Step IV. Between Steps IV and V, the corrosion rate is likely to progress gradually until an enclosed pit, depicted in Step V, forms. The dissolution of the steel leads to the saturation of the aqueous phase with metallic species, which precipitate mainly as carbonates, oxides and sulfates together with calcium carbonate and other deposits. The precipitation is likely to start in cathodic areas, outside of the pit, due to the high pH originating from the oxygen reduction reaction. In the final step, the ‘lid’ causes the formation of a differential aeration cell, with the predominant anodic dissolution resulting inside of the pit and the cathodic reaction taking place on the outside surface. The anodic dissolution of iron causes a drop in the pH by the hydrolysis of the corrosion products [75] and the migration of anions inside of the pit, further increasing the corrosion rate in the enclosure.

This mechanism explains both the strong localization of the corrosion attack and the recorded corrosion rates at the discussed facility. Reaching the highest acceleration is apparently controlled primarily by the formation of a stable pit. This process heavily depends on local conditions such as the surface relief, resulting in long-term water accumulation.

Since all of the laboratory experiments were performed in mixtures prepared from the relatively pristine middle-batch crude, further acceleration could be expected in lower-quality crude mixtures. It has been shown that the crude composition differed significantly at the beginning, middle, and end of a batch, with potentially severe consequences on corrosivity.

Based on the knowledge of the degradation mechanism, it is possible to design methods to limit the corrosion of steel pipelines. Since the water cannot be completely removed without altering the piping setup and adding additional drainage, other methods were sought. The simplest one employs periodic pumping as a source of disturbance, preventing complete stagnancy and resetting the timer on the water separation process. A model calculation suggested that, while halving the corrosion rates, and thus prolonging the lifetime of the installation by a factor of two, the costs of periodic pumping would be in the range of tens of percent compared to the complete replacement of the piping sections over the span of two to three decades. The exact economy would obviously differ greatly in different facilities. Nevertheless, for compositions with water contents greater than 0.1 wt. %, it seems to be a feasible solution. Alternatively, protecting the inner side of the pipeline with organic paint was proposed and tested. It proved to be technically feasible and highly efficient in corrosion mitigation, although more intense in terms of initial investment. The performance report of the coating in this setup will be presented elsewhere.

Summarizing these significant findings, it should be noted that further investigation into the effect of individual parameters is required. The study focused on the effect of operational parameters such as the water concentration, pumping frequency and the presence of local low spots on the inner surface of the pipe. The reason is that these parameters can be altered or affected by the operator, whereas others, such as the chloride and oxygen concentration or crude oil quality, are difficult to tackle on an industrial scale. For example, the chloride content can be reduced by desalination; however, this process is used only in downstream processes. Nevertheless, the chloride concentration is certainly an important variable and its effect on corrosion in crude oil needs to be addressed. Similarly, the oxygenation of the aqueous phase has a strong consequence on the corrosion process that needs to be quantified.

5. Conclusions

This paper presents range of non-traditional approaches and mechanisms showcasing the sequency of processes leading to the intensification of corrosion up to over three orders of magnitude. Among the key factors are the kinetics of water separation and the transport of oxygen, both of which are controlled by the degree of stagnancy. The data suggest that pipes operated daily are virtually free of risk of significant corrosion. Operation every week or so can already lead to water separation and an increase in corrosion by a factor of 10–20×; however, the absolute corrosion rates are still technically feasible. Stagnancy up to 3 weeks results in further intensification, resulting in corrosion rates of up to 0.1 mm/y. Even further stagnancy changes the mechanisms and promotes the formation of pits. Corrosion rates in the pits exceed 1 mm/y, corresponding to the quick penetration of the pipe walls observed in the facility investigated in this study.